Aeroacoustics Acoustic Wave Propagation - Aircraft Noise Prediction - Aeroacoustic Instrumentation

Downloaded by Boeing Company on July 15, 2013 | http://arc.aiaa.org | DOI: 10.2514/4.
865206
AEROACOUSTICS:
ACOUSTIC WAVE PROPAGATION;
AIRCRAFT NOISE PREDICTION;
AEROACOUSTIC INSTRUMENTATION
Downloaded by Boeing Company on July 15, 2013 | http://arc.aiaa.org | DOI: 10.2514/4.865206
Edited by
Ira R. Schwartz
NASA Ames Research Center
Moffett Field, California
Assistant Editors:
Henry T. Nagamatsu
General Electric Research and Development Center
Schenectady, New York
Warren C. Strahle
Georgia Institute of Technology
Atlanta, Georgia
Volume 46
PROGRESS IN
ASTRONAUTICS AND AERONAUTICS
Martin Summer-field, Series Editor-in-Chief

Princeton University, Princeton, New Jersey
Technical papers from AIAA 2nd Aero-Acoustics Conference,

March 1975, subsequently revised for this volume
Published by the American Institute of Aeronautics and

Astronautics in cooperation with The MIT Press.
American Institute of Aeronautics and Astronautics

New York, New York
The MIT Press

Cambridge, Massachusetts and London, England
Library of Congress Cataloging in Publication Data

Main entry under title:
AIAA Aero-acoustics Specialists Conference, 2d,

Hampton, Va., 1975.
Aeroacoustics: technical papers from AIAA 2nd Aeroacoustics
Conference, March 1975, subsequently revised for this volume.
(Progress in astronautics and aeronautics; v. 4346)

Includes bibliographical references and index.
CONTENTS: [1] Jet noise and core engine noise. [2] Duct acoustics,
fan noise and control, rotor noise. [3] STOL, airframe, and airfoil
noise. [4] Instrumentation; acoustic wave propagation; aircraft noise
prediction.
1. Jet planesNoiseCongresses. I. Schwartz, Ira R. II. Nagamatsu,
Henry T. III. Strahle, Warren. IV. American Institute of Aeronautics and
Astronautics. V. Title. VI. Series.
TL507.P75 vol. 43-46 [TL671.65] 629.1'08s [629.132'3] 76-3650
ISBN 0-915928-10-8 (v. 1)
Copyright 1976 by
American Institute of Aeronautics and Astronautics
All rights reserved. No part of this book may be reproduced in any

form or by any means, electronic or mechanical, including photo-
copying, recording, or by any information storage and retrieval
system, without permission in writing from the publisher.
Table of Contents
Volume 46
Preface x/x
Conference and Technical Program Committees xxxiv
Chapter I Acoustic Wave Propagation through the Atmosphere 1
Propagation of Aircraft Noise over Long Distances

through the Lower Atmosphere 3
A. L. ABRAHAMSON
Measured Effects of Turbulence on the Rise

Time of a Weak Shock 17
P. E. TUBB
Sound Scattering from Atmospheric Turbulence 35

MING NAN HUANG
Scattering of Coherent Sound Waves by

Atmospheric Turbulence 51
P. L. CHOW, C. H. LID, AND L. MAESTRELLO
Saturation Effects Associated with Sound

Propagation in a Turbulent Medium 67
ALAN R. WENZEL
Computer Model of the Lightnings-Thunder

Process, with Audible Demonstration 77
H. S. RI8NER, F. LAM, K. A. LEUNG,
D. KURTZ, AND N. D. ELLIS
Panel Discussion: Acoustic Wave Propagation

through Atmosphere 89
Chapter II Aircraft Noise Prediction 91
Development of a New Computer System for

Aircraft Noise Prediction 93
JOHN P. RANEY
Aircraft Flyover Noise Measurements 103

E. L. ZWIEBACK
Review of Theory and Methods for the Prediction

of Ground Effects on Aircraft Noise Propagation 123
PAUL B. ONCLEY
Methods for the Prediction of Airframe

Aerodynamic Noise 139
JAMES D. REVELL, GERALD J. HEALY, AND
JOHN S. GIBSON
Review of Theory and Methods for

Turbine Noise Prediction 155
D. C. MATHEWS, R. T. NAGEL, AND J. D. KESTER
Review of Theory and Methods for Combustion
Noise Prediction 177
R. E. MOTSINGER AND J. J. EMMERLING
Panel Discussion: Aircraft Noise Prediction 195
Chapter III Aeroacoustic Instrumentation 201
The Use of Hartmann Generators as Sources of

High-Intensity Sound in a Large Absorption
Flow-Duct Facility 203
D. L. MARTLEW
Outdoor Jet Noise Facility: A Unique Approach 223

R. A. KANTOLA
Factors in the Design and Performance of Free-Jet

Acoustic Wind Tunnels 247

Y. KADMAN AND R. E. HAYDEN
Correction of Open-Jet Wind-Tunnel Measurements

for Shear Layer Refraction 259
ROY K. AMIET
Use of a Laser Shadowgraph for Jet

Noise Diagnosis 281
MICHAEL J. RUDD
Effects of Transducer Flushness on Fluctuating

Surface Pressure Measurements 291
RICHARD D. HANLY
Panel Discussion: Aeroacoustic Instrumentation 303
Index to Contributors to Volume 46 307
Tables of Contents
(Companion Volumes 43, 44, 45)
Volume 43
Preface xix
Chapter I Jet Noise 1
Tests of a Theoretical Model of Subsonic Jet Noise 3

N. S. M. NOSSEIR AND H. S. RIBNER
New Evidence of Subsonic Jet Noise Mechanisms 27

S. PAUL PAO AND LUCIO MAESTRELLO
Simulation by Vortex Rings of the Unsteady
Pressure Field near a Jet 47
C. H. LIU, L. MAESTRELLO, AND M. D. GUNZBURGER
Bernoulli Enthalpy: A Fundamental Concept

in the Theory of Sound 65
JOHN E. YATES AND GUIDO SANDRI
A Potential Flow Model for Calculation of Jet Noise 91

P. O. A. L. DAVIES, J. C. HARDIN, A. V. J. EDWARDS
AND J. P. MASON
Jet Noise Source Location by Cross-Correlation

of Far Field Microphone Signals 107
S. P. PARTHASARATHY
An Experimental Study of the Structure and Acoustic

Field of a Jet in a Cross Stream 121
IVAN CAMELIER AND K. KARAMCHETI
Measurement and Prediction of Jet Noise in Flight 137

K. W. BUSHELL
Jet Noise Modeling: Theoretical Predictions and

Comparisons with Measured Data 159
B. J. TESTER AND C. L. MORFEY
Sound Pressures of a Choked Jet Oscillating

in the Spinning Mode 185
R. WESTLEY AND J. H. WOOLLEY
An Experimental Study of Shock-Free

Supersonic Jet Noise 203
H. K. TANNA AND P. D. DEAN
Periodic "Mach Wave" Radiations from a Small

Supersonic Helium Jet 237
R. WESTLEY AND J. H. WOOLLEY
Ambient and Induced Pressure Fluctuations in

Supersonic Jet Flows 257
V. BARRA, S. SLUTSKY, AND S. PANUNZIO
Density Fluctuations and Radiated Noise for a

High-Temperature Supersonic Jet 283
S. P. PARTHASARATHY, P. F. MASSIER, R. F. CUFFEL,
AND J. R. RADBILL
Acoustic Far Field of a Point Source in Parallel

Flowing Fluid Layers 307
H. Y. LU
Supersonic Jet Noise Suppression with

Multitube Nozzle/Ejectors 329
J. ATVARS, C. P. WRIGHT, AND C. D. SIMCOX
Some Recent Developments in Supersonic
Jet Noise Reduction 353
D. S. DOSANJH, K. K. AHUJA, M. R. BASSIOUNI,
AND P. K. BHUTIANI
Minimization of Jet and Core Noise of a Turbojet

Engine by Swirling the Exhaust Flow 379
IRA R. SCHWARTZ
Jet and Suppressor Correlation Measurements 399

R. A. KANTOLA
Cross-Correlation of Noise Produced inside a Hot

Turbojet Exhaust with and without Suppression 423
W. C. MEECHAM AND D. R. REGAN
The Effects of Forward Speed on a Number of

Turbojet Exhaust Silencers 439

J. R. BROOKS AND R. J. WOODROW
Panel Discussion: Jet Noise 458
Chapter II Combustions and Core Engine Noise 465
Convergence of Theory and Experiment in Direct

Combustion-Generated Noise 467
WARREN C. STRAHLE
Further Experimental Results on the Structure

and Acoustics of Turbulent Jet Flames 483
R. N. KUMAR
Combustion Intensity and Distribution Relation

to Noise Generation 509
E. G. PLETT, M. D. LESHNER, AND M. SUMMERFIELD
Acoustic Motion Induced by a Diffusion Flame 531

MAURICE L. RASMUSSEN
An Experimental Investigation of the Core Engine

Noise of a Turbofan Engine 555
R. G. HOCH, P. THOMAS, AND E. WEISS
Noise Generation and Transmission in

Duct Combustors 579
DAVID W. LINDLEY AND H. A. HASSAN
Core Engine Noise due to Temperature Fluctuations

Convecting through Turbine Blade Rows 589
G. F. PICKETT
Panel Discussion: Combustion and Core Engine Noise 609

Volume 44
Preface x/x
Conference and Technical Program Committees xxx/'v
Chapter I Fan Noise and Control 1
Theoretical and Experimental Studies of Discrete-Tone

Rotor-Stator Interaction Noise 3
G. F. HOMICZ, G. R. LUDWIG, AND J. A. LORDI
Fan Aeroacoustics: The Effect of Stator Blade Number

and Spacing on In-Duct Noise Signatures 23
G. KRISHNAPPA
Wake Cutting by a Cascade of Cambered Blades 43

C. M. HO, AND L. S. G. KOVASZNAY
Noncompact Source Effect on the Prediction of

Tone Noise from a Fan Rotor 55
S. KAJl
The Role of Rotor Blade Blockage in the Propagation

of Fan Noise Interaction Tones 83
M. G. PHILPOT
Turbine Noise Generation, Reduction, and Prediction 109

S. B. KAZIN AND R. K. MATTA
Simulation of Flight Effects on

Aero-Engine Fan Noise 139
B. w. LOWRIE
Effect of Forward Motion on Fan Noise 159

J. E. MERRIMAN AND R. C. GOOD
Model and Full-Scale Test Results Relating to

Fan Noise In-Flight Effects 181
J. P. ROUNDHILL AND L. A. SCHAUT
A Study of Subsonic Fan Noise Sources 209

DONALD B. HANSON
Aerodynamic Sound Generation due to the Interaction

of an Unsteady Wake with a Rigid Surface 233
C. A. SMITH AND K. KARAMCHETI
Panel Discussion: Fan Noise and Control 251
Chapter II Duct Acoustics 257
Effect of Grazing Flow on the Steady-State

Resistance of Square-Edged Orifices 259
T. ROGERS AND A. S. HERSH
Influence of Grazing Flow on Duct Wall
Normal Impedances 289
P. MUNGER AND J. L. WHITESIDES
Fluid-Mechanical Model of the Acoustic

Impedance of Small Orifices 307
A. S. HERSH AND T. ROGERS
Attenuation of the Sound Associated with a Plane

Wave in a Multisectional Duct 331
DENNIS W. QUINN
Sound Propagation in Curved Ducts 347

M. K. MYERS AND P. MUNGUR
Duct Acoustics and Acoustic Finite Element Method 363

A. KAPUR AND P. MUNGUR
Suppression of Multiple Pure Tones 371

ALI HASAN NAYFEH AND JOHN E. KAISER
Sparse Matrix Techniques Applied to Modal

Analysis of Multisection Duct Liners 385
W. R. ARNOLD
Acoustic Wave Propagation in a Lined Duct

with Nonuniform Admittance 397
J. C. YU, C. D. SMITH, AND P. MUNGUR
Computational Methods for Acoustic Radiation

from Circular Ducts 415
R. J. BECKEMEYER, D. T. SAWDY, AND P. GARNER
Effects of a Conical Segment on Sound Radiation

from a Circular Duct 433
D. T. SAWDY, R. J. BECKEMEYER, AND P. GARNER
Wave Envelope Analysis of Sound Propagation in

Ducts with Variable Axial Impedance 451
KENNETH J. BAUMEISTER
Spinning Mode Sound Propagation in Ducts with

Acoustic Treatment and Sheared Flow 475
EDWARD J. RICE
Effects of Friction and Heat Conduction on

Sound Propagation in Ducts 507
P. HUERRE AND K. KARAMCHETI
Panel Discussion: Duct Acoustics 533
Chapter III Rotor Noise 537
Tone Noise of High-Speed Rotors 539

D. L. HAWKINGS AND M. V. LOWSON
Discrete Frequency Rotor Noise 559
JOHN W. LEVERTON
V/STOL Rotor and Propeller Noise: Its Prediction

and Analysis of Its Aural Characteristics 583
RICHARD P. WHITE JR.
Thickness Noise of Helicopter Rotors at

High Tip Speeds 601
F. FARASSAT, R. J. PEGG, AND D. A. HILTON
Development of Experimental Techniques for the

Study of Helicopter Rotor Noise 615
WESLEY L. HARRIS AND ALBERT LEE
Panel Discussion: Rotor Noise 631

Volume 45
Preface xix
Chapter I STOL Noise 1
Scrubbing Noise of Externally Blown Flaps 3

MARTIN R. FINK
Propulsive-Lift Noise of an Upper-Surface-Blown

Flap Configuration 27
N. N. REDDY
Experimental Investigation of the Aeroacoustic

Characteristics of Model Slot Nozzles with
Straight Flaps 41
GRANT T. PATTERSON, M. C. JOSHI, AND
JAMES R. MAUS
Fluctuating Pressures on Aircraft Wing and Flap

Surfaces Associated with Powered-Lift Systems 59
JOHN S. MIXSON, JAMES A. SCHOENSTER, AND
CONRAD M. WILLIS
Acoustic Characteristics of a Large Upper-Surface-

Blown Configuration with Turbofan Engines 83
JOHN S. PREISSER AND DAVID J. FRATELLO
Noise Shielding Effects for Engine-Over-Wing

Installations 103
V. M. CONTICELLI, A. Dl BLASI, AND J. V. O'KEEFE
Effect of Forward Speed on Jet Wing/Flap
Interaction Noise 127
WAMAN V. BHAT AND DAVIDE GALLO-ROSSO
Forward Velocity Effects on Under-the-Wing

Externally Blown Flap Noise 147
J. GOODYKOONTZ, U. VON GLAHN, AND R. DORSCH
Panel Discussion: STOL Noise 175
Chapter II Airframe and Airfoil Noise 177
Diagnostic Calculations of Airframe-Radiated Noise 179

R. E. HAYDEN, Y. KADMAN, D. B. BLISS,
AND S. A. AFRICK
Aircraft Far-Field Aerodynamic Noise: Its

Measurement and Prediction 203
GERALD J. HEALY
Induced Drag Effect on Airframe Noise 221

JAMES D. REVELL
Measurements of Discrete Vortex Noise in a

Closed-Throat Wind Tunnel 237
SANFORDS. DAVIS
Trailing Edge Noise 259

CHRISTOPHER K. W. JAM AND J. C. YU
Flow-Induced Pressure Fluctuations in Cavities

and Concepts for their Suppression 281
HANNO H. HELLER AND DONALD B. BLISS
Effect of Geometry on Open Cavity Flow-Induced

Pressure Oscillations 297
M. E. FRANKE AND D. L. CARR
Noise Radiation from Turbulent Flows

over Compliant Surfaces 315
Y. S. PAN
The Dynamic Response of an Acoustically

Coupled Panel 335
D. H. Y. YEN, L. MAESTRELLO, AND S. L. PADULA
Preliminary Airframe Noise Measurements on a

Transport Model in a Quiet Flow Facility 351
J. G. SHEARIN AND P. J. BLOCK
Measurements and Analysis of Aircraft Airframe Noise 363

TERRILL W. PUTNAM, PAUL L. LASAGNA, AND
KENNETH C. WHITE
An Experimental Study of Airframe Self-Noise 379
P. FETHNEY
A Preliminary Investigation of Remotely Piloted

Vehicles for Airframe Noise Research 405
DAVID J. FRATELLO AND JOHN G. SHEARIN
Experimental Studies of Noise-Shielding Effects

for a Delta-Winged Aircraft 419
R. W. JEFFERY AND T. A. HOLBECHE
Panel Discussion: Airframe and Airfoil Noise 441

This page intentionally left blank

Progress in Martin Summerfield,
Astronautics and Aeronautics Series Editor
PRINCETON UNIVERSITY
VOLUMES EDITORS
1. Solid Propellant Rocket Martin Summerfield
Research. 1960
2. Liquid Rockets and Loren E. Bollinger

THE OHIO STATE UNIVERSITY
Propellants. 1960
Martin Goldsmith
THE RAND CORPORATION
Alexis W. Lemmon Jr.

BATTELLE MEMORIAL INSTITUTE
3. Energy Conversion for Nathan W. Snyder

INSTITUTE FOR DEFENSE ANALYSES
Space Power. 1961
4. Space Power Systems. 1961 Nathan W. Snyder

5. Electrostatic Propulsion. 1961 David B. Langmuir

SPACE TECHNOLOGY LABORATORIES, INC.
Ernst Stuhlinger
NASA GEORGE C. MARSHALL SPACE
FLIGHT CENTER
J. M. Sellen Jr.
SPACE TECHNOLOGY LABORATORIES
6. Detonation and Two-Phase S. S. Penner

C A L I F O R N I A INSTITUTE OF TECHNOLOGY
Flow. 1962
F. A. Williams
H A R V A R D UNIVERSITY
7. Hypersonic Flow Research. Frederick R. Riddell

AVCO CORPORATION
1962
8. Guidance and Control. 1962 Robert E. Roberson

CONSULTANT
James S. Farrior
LOCKHEED MISSILES AND SPACE
COMPANY
9. Electric Propulsion Ernst Stuhlinger

Development. 1963
FLIGHT CENTER
10. Technology of Lunar Clifford I. Cummings and
Exploration. 1963 Harold R. Lawrence
JET PROPULSION LABORATORY
11. Power Systems for Space Morris A. Zipkin and

Flight. 1963 Russell N. Edwards
G E N E R A L ELECTRIC COMPANY
12. lonization in High- Kurt E. Shuler, Editor

NATIONAL BUREAU OF STANDARDS
Temperature Gases. 1963
John B. Fenn, Associate Editor
13. Guidance and Control - Robert C. Langford

GENERAL PRECISION INC.
1964
Charles J. Mundo
INSTITUTE OF NAVAL STUDIES
14. Celestial Mechanics and Victor G. Szebehely

YALE UNIVERSITY OBSERVATORY
Astrodynamics. 1964
15. Heterogeneous Combustion. Hans G. Wolfhard

1964
Irvin Glassman
Leon Green Jr.

AIR FORCE SYSTEMS COMMAND
16. Space Power Systems George C. Szego

Engineering. 1966
J. Edward Taylor
TRW INC.
17. Methods in Astrodynamics Raynor L. Duncombe

U.S. NAVAL OBSERVATORY
and Celestial Mechanics.
1966 Victor G. Szebehely
YALE UNIVERSITY OBSERVATORY
18. Thermophysics and Gerhard B. Heller

Temperature Control of FLIGHT CENTER
Spacecraft and Entry
Vehicles. 1966
19. Communication Satellite Richard B. Marsten

RADIO CORPORATION OF AMERICA
Systems Technology. 1966
20. Thermophysics of Spacecraft Gerhard B. Heller
and Planetary Bodies FLIGHT CENTER
Radiation Properties of Solids
and the Electromagnetic
Radiation Environment in
Space. 1967
21. Thermal Design Principles of Jerry T. Bevans

TRW SYSTEMS
Spacecraft and Entry Bodies.
1969
22. Stratospheric Circulation. Willis L. Webb

ATMOSPHERIC SCIENCES LABORATORY,
1969
WHITE SANDS, AND UNIVERSITY OF

TEXAS AT EL PASO
23. Thermophysics: Applications Jerry T. Bevans

TRW SYSTEMS
to Thermal Design of
Spacecraft. 1970
24. Heat Transfer and Spacecraft John W. Lucas

Thermal Control. 1971
25. Communication Satellites for Nathaniel E. Feldman

the 70's: Technology. 1971
Charles M. Kelly
THE AEROSPACE CORPORATION
26. Communication Satellites Nathaniel E. Feldman

for the 70's: Systems. 1971
Charles M. Kelly
THE AEROSPACE CORPORATION
27. Thermospheric Circulation. Willis L. Webb

ATMOSPHERIC SCIENCES LABORATORY,
1972
WHITE SANDS, AND UNIVERSITY OF T E X A S
AT E L PASO
28. Thermal Characteristics John W. Lucas

of the Moon. 1972
29. Fundamentals of Spacecraft John W. Lucas

Thermal Design. 1972
30. Solar Activity Observations Patrick S. Mclntosh and
and Predictions. 1972 Murray Dryer
ENVIRONMENTAL RESEARCH
LABORATORIES, NATIONAL
OCEANIC AND ATMOSPHERIC
ADMINISTRATION
31. Thermal Control and Chang-Lin Tien

UNIVERSITY OF CALIFORNIA,
Radiation. 1973 BERKELEY
32. Communications P. L. Bargellini

COMSAT LABORATORIES
Satellite Systems. 1974
33. Communications P. L. Bargellini

COMSAT LABORATORIES
Satellite Technology.
1974
34. Instrumentation for Alien E. Fuhs

NAVAL POSTGRADUATE SCHOOL
Airbreathing Propulsion.
1974 Marshall Kingery
ARNOLD ENGINEERING
DEVELOPMENT CENTER
35. Thermophysics and Robert G. Hering

UNIVERSITY OF IOWA
Spacecraft Thermal
Control. 1974
36. Thermal Pollution Analysis. Joseph A. Schetz

VIRGINIA POLYTECHNIC INSTITUTE
1975
37. Aeroacoustics: Jet and Henry T. Nagamatsu, Editor

GENERAL ELECTRIC RESEARCH
Combustion Noise; AND DEVELOPMENT CENTER
Duct Acoustics. 1975
Jack V. O'Keefe, Associate Editor
THE BOEING COMPANY
Ira R. Schwartz, Associate Editor

NASA A M E S R E S E A R C H CENTER
38. Aeroacoustics: Fan, STOL, Henry T. Nagamatsu, Editor

GENERAL ELECTRIC RESEARCH
and Boundary Layer Noise;
AND DEVELOPMENT CENTER
Sonic Boom; Aeroacoustic
Instrumentation. 1975 Jack V. O'Keefe, Associate Editor
THE BOEING COMPANY
Ira R. Schwartz, Associate Editor

NASA AMES RESEARCH CENTER
39. Heat Transfer with Thermal M. Michael Yovanovich
UNIVERSITY OF WATERLOO
Control Applications. 1975
40. Aerodynamics of Base S. N. B. Murthy

PURDUE UNIVERSITY
Combustion. 1976
41. Communications Satellite Gilbert E. LaVean

DEFENSE COMMUNICATIONS
Developments: Systems. 1976 ENGINEERING CENTER
William G. Schmidt
CML SATELLITE CORPORATION
42. Communications Satellite William G. Schmidt

CML SATELLITE CORPORATION

Developments: Technology. 1976
Gilbert E. LaVean
DEFENSE COMMUNICATIONS
ENGINEERING CENTER
43. Aeroacoustics: Jet Noise, Ira R. Schwartz

Combustion and Core Engine
Noise. 1976 Henry T. Nagamatsu
GENERAL ELECTRIC RESEARCH AND
DEVELOPMENT CENTER
Warren C. Strahle
GEORGIA INSTITUTE OF TECHNOLOGY
44. Aeroacoustics: Fan Noise and Ira R. Schwartz

Control; Duct Acoustics:
Rotor Noise. 1976 Henry T. Nagamatsu
DEVELOPMENT CENTER
Warren C. Strahle
45. Aeroacoustics: STOL Noise; Ira R. Schwartz

Airframe and Airfoil Noise. 1976
Henry T. Nagamatsu
DEVELOPMENT CENTER
Warren C. Strahle
46. Aeroacoustics: Acoustic Wave Ira R. Schwartz

Propagation; Aircraft Noise NASA AMES RESEARCH CENTER
Prediction; Aeroacoustic Henry T. Nagamatsu
Instrumentation. 1976 DEVELOPMENT CENTER
Warren C. Strahle
(Other volumes are planned.)


PREFACE
Aviation requirements of modern aircraft have dictated the construction
of larger aircraft structures and jet engines. The advent of larger and more
powerful engines has produced a serious aircraft problem to communities
near airports. The great concern of the Federal Government and the com-
munities about the noise problem has been reflected in more stringent flight
regulations and in severe limitations on the design and development of air-
craft technology. Thus, it is difficult to overemphasize the importance of
aircraft noise, which includes propulsion and airframe structure noise and
sonic boom. The increasing demands on industry to satisfy Federal
regulations concerning aircraft takeoff and landing noise are compromising
the economical performance and growth of present aircraft. Also, several

countries have either banned the supersonic transports or restricted their
flights to subsonic operations over populated areas to prevent public reac-
tion to their sonic booms and takeoff noise levels. The design of future air-
craft technology must be guided by the ever-increasing constraints of
Federal regulations and community response. Regardless of the degree of
aerodynamic and propulsion efficiency of the aircraft components, the air-
plane will not be acceptable if it cannot satisfy the requirements of Federal
Aircraft Regulations (FAR). Therefore, solutions to the noise problems are
currently, and will be in the future, pacing elements in the design and
development of aircraft technology.
Until recently, little progress had been made towards finding solutions to
aeroacoustic problems, partly because of their many complexities and partly
because of the limited research programs to support technological develop-
ment. As the emphasis on aeroacoustics research and noise suppression
technology has increased during the past several years, significant progress
has been made towards achieving the goals of the aircraft industry and the
rural communities. Recent advancements in technology have made it
possible for jet engine noise to be reduced to levels that will permit many of
the present large commercial aircraft to satisfy the FAR-36 noise
requirements. The goals for future commercial aircraft, however, are to ad-
vance technology in order to permit the reduction of the FAR-36 noise
requirement by approximately 10 EPNdB per decade. Much work,
therefore, still must be done to develop the scientific and technological tools
required for substantial advancements in aerodynamic noise abatement.
More emphasis must be placed in areas of research that will provide a better
understanding of the mechanisms of engine and airframe noise generation,
propagation, and attenuation, including flight effects. However,
technological breakthroughs in noise abatement need not be paced by com-
plete scientific understanding of the mechanisms of the phenomenon in-
volved. We may find and implement very successfully an efficient method of
suppressing jet and core engine noise and yet not understand completely the
very complex mechanisms that modify and suppress the turbulence modes
of noise-producing sources. Such an event, for example, would be com-
parable to the discovery of successful methods of controlling boundary-
layer flows by the use of vortex generators and other devices over 30 years
ago without having a complete understanding at that time of the boundary-
layer flow mechanisms involved in the aerodynamic phenomenon.Thus, this
recently increased effort in aeroacoustics research and technology must be
continued to meet present and future environmental noise requirements
being demanded by Congress, the FAA, and a critical public. Accordingly,
the Second AIAA Aeroacoustics Specialists Conference was held at Hamp-
ton, Va.,March 24-26, 1975, to assess the most recent accomplishments in
aeroacoustic research and technology for different flight vehicles and to
stimulate further the growth of new ideas and understanding. A broad spec-
trum of aeroacoustic research and engineering programs is being performed
in many countries, universities, government laboratories, and industries, as
reflected by the technical sessions of the Conference, at which 108 papers
were presented. Of these, 98 papers are published in four volumes of the
AIAA Progress in Astronautics and Aeronautics series (Volumes 43-46).

They cover 10 major areas: 1) jet noise, 2) core engine noise, 3) fan noise
and control, 4) duct acoustics, 5) rotor noise, 6) STOL noise, 7) airframe
and airfoil noise, 8) aircraft noise prediction, 9) acoustic wave propagation
through the atmosphere, and 10) aeroacoustic instrumentation and test
techniques. The subjects discussed in these 10 areas could be categorized in-
to two principal topics which serve to answer the following two questions:
How far have we proceeded in developing methods of analysis and un-
derstanding of aircraft noise generation, propagation, and radiation? How
far have we proceeded in devising methods of reducing aircraft noise?
In Volume 43 the papers are grouped into two sections: Jet noise and core
engine noise. The jet noise papers cover subsonic and supersonic jets; jet
noise suppression techniques, including basic studies of suppression
phenomena; flight effects on jet noise and refraction; and scattering and
shielding studies of noise sources in jets. The core engine noise papers cover
combustion noise and entropy noise from combustion, and turbine
coupling.
Volume 44 contains three sections on fan noise and control, duct

acoustics, and rotor noise. The papers in the fan noise and control section
cover turbomachinery noise problems, such as combination tone noise,
rotor wake and rotor wake-stator interaction noise, coupling of inlet distur-
bances with rotor-stator interactions, flight effects or forward motion on
the acoustic characteristics of fan noise, and turbine noise generation and
reduction. The duct acoustic papers include sound propagation in and
radiation from treated and untreated ducts with flow, nonlinear acoustic
phenomena, computational methods in duct acoustics, and variable
geometry effects. The rotor noise papers include V/STOL and helicopter
rotor noise and propeller noise problem areas, such as blade slap and high-
speed rotor noise, and discrete frequency noise.
Volume 45 contains sections on STOL noise and on airframe and airfoil

noise. Volume 46 has sections on acoustic wave propagation through the at-
mosphere, aircraft noise prediction, and aeroacoustic instrumentation. The
STOL vehicle noise papers include the areas of jet-lift surface-interaction
noise, noise shielding effects, and forward velocity effects on jet-lift surface
noise. The airframe and airfoil noise papers are essentially related to the air-
craft nonpropulsive noise sources that are generated from flow surface in-
teractions. The sonic boom and acoustic wave propagation papers essen-
tially cover sonic boom generation and propagation, including turbulence
effects and caustic focusing, and the propagation of aircraft noise through
the atmosphere, including the effects of turbulence, stratification, and
ground absorption.
In addition to the papers in these volumes, summaries of panel

discussions are included at the end of each section to summarize the status
of the knowledge and the future research required on the subject.
The progress made during the past 1 Vi to 2 years, as reported in the four
volumes, represents significant advances in developing methods of analysis

and understanding of aeroacoustics phenomena and in devising methods,
particularly aerodynamic techniques, of reducing propulsion noise. In
regard to the analysis and understanding of jet noise phenomena (Volume
43), a number of methods have been developed recently which have
provided insights into various aspects of jet noise generation and
propagation; however, numerical modeling of the jet noise sources, noise
source distribution, type, and number has not reached the point of being
definitive and leaves much to be desired. A new formalism by Yates and
Sandri for aerodynamic sound generation in subsonic jets must be con-
sidered a strong competitor to the Lighthill formalism. All the necessary
ingredients for the analysis of full-scale jets appear to be included in their
formulation. This theory will be tested in the near future, and hence its
usefulness as a basic concept will be determined then.
Also contained in Volume 43 are recent notable achievements in jet noise
analysis, particularly concerning computed analytical results for overall jet
noise, using different models made by Ribner et al. and also by Tester and
Morfey. Ribner's analytical results are based upon his formalism, or the so-
called dilatation theory with experimental data. The other approach, taken
by Tester and Morfey, compensates for the inadequacy of the Lighthill
equation,which is not a convected wave equation, to deal properly with the
refraction of sound by the flow in the jet. They advocate using a convective
wave equation, the Lilley equation, which allows for the various shrouding
effects of the flow, refraction along the jet axis, and changes in effective im-
pedance as seen by the sources. Their results correlate favorably with ex-
perimental data.
Further progress on the understanding of noise source strength

distribution in jets was reported by Pao and Maestrello, who used a unique
technique to trace the sound waves to their origin. It appeared that the
sound came from narrowly defined ray bundles, and hence the source
strength was determined.
Some interesting new results of three investigations on the numerical

modeling of subsonic jet noise have become available, as reported in
XXII
Volume 43. Two of these studies used vortex ring approaches, and the other
was concerned with the Navier-Stokes equations as a starting point. All of
these approaches have provided additional knowledge and insights into
various aspects of jet noise. They definitely should be continued.
With regard to supersonic jet noise, studies of the mechanisms of noise
source generation and transmission have provided useful data to our reser-
voir of knowledge and understanding. Further work in this area is en-
couraged.
Several investigations dealing with flight effects which were conducted in
wind tunnels, and also by utilizing aircraft in flight, have indicated the im-
portance of separating the aircraft component noise sources and hence
isolating the jet noise. Comparison of static and flight noise levels indicated
a forward arc amplification of the flight noise levels. The question that must
be answered is whether the amplification of noise is due to jet noise
phenomenon or perhaps an interaction of airframe and core engine noise.
In the area of jet noise suppression, various promising suppressor con-
cepts were investigated, the results of which are discussed in Volume 43. The
Swirling Flow Jet Noise Suppressor that was investigated by Schwartz, using
a full-scale turbojet engine, produced significant noise reduction with
minimal thrust losses. Results of previous studies of the swirling flow jet
noise suppressor, using a full-scale turbofan engine as reported in Volume
37 of this Series, indicated that even greater beneficial suppression effects
could be obtained by swirling the jet in a turbofan engine. The potential ad-
vantages that this swirl jet noise suppressor offers in comparison with other
known suppressors are very impressive. Further research on full-scale,
higher thrust jet engines are continuing.
The aircraft engine core noise problem has been receiving greater em-
phasis by researchers, as reflected in seven research type papers in Volume
43. An additional paper in Volume 46 deals with prediction methods of core
noise. Since engine core noise includes combustion noise sources, and tur-
bine and duct noise sources, the identification of the mechanisms of noise
generation and transmission becomes very complex. The increased interest
in engine core noise is due to the fact that core and jet noise have become the
threshold of dominant aircraft noise in present transport engines. The
coupling of upstream core noise sources with jet noise sources is known to
produce significant increases in jet noise levels. One or two possible
mechanisms could cause core noise, and herein is the difficult problem. It is
the opinion of most researchers that one of the potential mechanisms is the
direct noise that is radiated from the turbulent combustion in the primary
combustor and transmitted through the turbine, passing out the nozzle into
the far field. The other mechanism that is considered to cause core noise is
the noise that is emitted from hot spots being convected through the tur-
bine. The question that arises is: Which one of these mechanisms, or
perhaps both mechanisms, is responsible for engine core noise? The other
question confronting researchers is: What are the coupling mechanisms of
core engine noise and jet noise?
In order to solve any of the important jet noise or core engine noise
problems, experiments must be conducted in engine-type apparatus so that
all the important aerodynamic and thermodynamic parameters, and hence
the resultant interactions, will be present in the flow model. Although the
research emphasis has increased in both core engine noise and jet noise areas
during the last few years, additional emphasis must be placed in these areas
so as ultimately to attain practical solutions.
In an assessment of the status of fan noise research, as presented at the
Conference and in Volume 44, it was concluded that one of the most
significant achievements in recent years that could potentially advance our
knowledge in this area was the observation that flight noise from a fan was
lower than those levels expected or obtained from static testing. The cause
of this discrepancy has been attributed to the poor inflow conditions at the
inlet of the engine on static stands. Therefore, rotor-stator interaction noise
data previously taken in static testing are considered suspect and apparently
incorrect. Experiments with screens to remove turbulence appeared to be
successful. The consensus is to test the configuration in a wind tunnel where
the forward velocity will provide a sufficiently clean inflow to simulate
flight conditions. Once the inflows are smooth, the rotor-stator interaction
noise, which apparently was not the dominant source in most of the
previous tests, can be evaluated.
The recent progress in duct acoustic research is reflected by fourteen
papers in Volume 44 which cover essentially three areaslinear duct
acoustics, the characteristics of absorbent materials, and nonlinear duct
acoustics. Most of the emphasis in the area of the propagation in ducts was
placed on computational methods to attain effective and economical
modeling of the propagation in variable geometry and in hardwall or soft-
wall ducts. New computation methods were developed to reduce computer
processing and storage requirements. Baumeister suggests a wave envelope
technique to reduce the processing and storage needs and successfully ap-
plies it to the optimum, segmented, acoustic duct liner design. Quinn uses a
finite difference technique in combination with conformal mapping to com-
pute the optimum multi-section duct liner design similar to Baumeister's ap-
proach. Kapur and Mungur propose acoustic finite element approaches as a
promising new approach to solve acoustic duct problems. Although similar
methods have been succesfully applied to structures of complex design, their
applicability to acoustic problems requires development of so-called func-
tional equivalents to acoustics problems.
With regard to the research progress on acoustically absorbent materials

for duct acoustic treatment, the papers contained in Volume 44 provide new
insight into the fluid-mechanical characteristics of perforates. Papers by
Hersh and Rogers demonstrate that the utilization of steady-state hydro-
dynamics could provide a better understanding of the characteristics of per-
forate facings. The developed flow model showed the mutual relationship
between steady-state hydrodynamic and acoustic resonance.
Optimization of segmented multiple-section liner design to increase at-
tenuation of the sound field in ducts has been progressing satisfactorily,
with significant results, as indicated by several papers in Volume 44. Further
work in this area has been suggested by several researchers to emphasize
specifically the extended reaction problem since most of the papers analyze
the acoustic treatment as a locally reacting type of material.
In the field of nonlinear duct acoustics, several interesting investigations

have been carried out in which the behavior of the nonlinear duct liner was
determined. The paper by Nayfeh and Kaiser analyzes the development of
the sound field for entering, combination tones into a lined duct. This
analysis determines the liner property that is responsible for the dispersion
of the waves which inhibit nonlinear transfer of energy into the lower har-
monics.
It was concluded at the Conference, and summarized by G. Reethof in his

panel session discussions, Volume 44, that further research must be con-
ducted in the three principal areas of duct acoustics: Linear duct acoustics;
acoustic treatment of duct elements; and nonlinear duct acoustics. For
example, emphasis should be placed on computational methods that are
related to the transmission of complex modes in ducts with axially variable
flow properties and variable geometry, including boundary-layer effects. In
the area of acoustic treatment of duct elements, there is an urgent need for
additional research to determine the mechanisms that are instrumental in the
reduction of noise in segmented ducts. Further work should be done to
determine the effect of high-amplitude sound waves and shock waves on
wall impedance in a duct for various types of wall-treatment materials. In
the nonlinear duct acoustics area, research should be directed first towards
the development of analytical formulation of the wave propagation in
hardwall ducts and next in softwall ducts. A determination must be made of
the relative effects of nonlinear gas dynamics and nonlinear properties for
perforates and porous surfaces. Finally, the panel of acoustic experts at the
Conference concluded (Volume 44) that additional research emphasis is
needed on the radiation problems from ducts with high intensities and with
complex modes or high-order modes.
In an assessment of the status of rotor noise research (Volume 44) it was

concluded that there is an apparent need for a more definitive high-speed
rotor noise modeling prediction and measurement capability. It appears that
theory is ahead of experiments and is getting closer to the prediction of high-
speed rotor noise phenomena. As a basis for this conclusion, reference is
made specifically to four papers that are presented in Volume 44, two of
which are theoretical and two of which are experimental. The research by
Hawkings and Lowson attempts to provide a unified description of the
discrete tone acoustic radiation for high tip speed blades. In the other
theoretical paper on rotor noise, the objective of Farassat, Pegg, and Hilton
is to model high-speed blade slap due to thickness alone. Whereas the
Hawkings and Lowson model is limited to discrete tone sources, the for-
malism by Farassat et al. is more general and not limited to the far field. The
results of both theories are essentially the same. It is claimed that the blade
noise due to thickness is a valid model for high-speed blade slap. In the
Farassat et al. analysis, the effects of noncompact sources have not been in-
cluded. Their predictions,however, are considered to be useful in design
work. It was concluded that extensions of these models should start with
considerations of compact sources. As for the Hawkings and Lowson
theory, any effort to extend this theory should include more than discrete
sources.
The two experimental papers on rotor noise describe work at speeds less
than the high tip speed limit but approaching it. For example, a research
paper by Leverton investigates rotor noise for the case of hover, which is
limited in the tip Mach number at a maximum of 0.68. His results indicate
that the rotational noise increases as approximately a V 2 law. Also, the
rotational noise is essentially independent of the thrust except for the higher
harmonics. The author concludes that profile drag or blade thickness is the
principal mechanism as the noise source and not the fluctuating loads.
The other experimental paper on rotor noise by Harris and Lee also
describes a low blade tip speed Mach number investigation in which
rotational noise and vortex noise are studied for a range of parameters. Em-
phasis in this investigation was placed on studying the effects of four
parameters: 1) the blade pitch setting (variation of thrust and blade
loading), 2) the number of blades, 3) the rotor shaft tilt angle, and 4) the tip
speed. The authors contend that, within some of the limits of the
parameters, their results agree well with theory; however, for others there is
no agreement. They emphasize the fact that there is a critical need to extend
the experiments from the low tip Mach number regime to the higher tip
Mach numbers. These experiments may provide better modeling for the
theories to determine whether it is thickness that generates the blade slap or
whether it is vortex interaction with the blades that produces the noise
source mechanisms.
Another major area of aeroacoustics research which was reviewed and
assessed at the Conference was VTOL or V/STOL Aircraft Noise. The
recent progress in this area was reflected by eight papers, presented in
Volume 45, which cover a broad spectrum of topics: 1) scrub-
bing/interaction noise, 2) upper surface blown flap turbulence and noise, 3)
acoustic pressures on aircraft surfaces related to powered-lift surfaces, 4)
shielding effects, and 5) forward velocity effects on jet wing/flap interaction
noise and also on under-the-wing externally blown flap noise. Most of this
research was directed at developing better methods for the prediction of
noise generation and propagation, at determining the effects of pressure
fluctuation on surfaces and interior noise, and at developing methods for
suppressing aircraft noise. Specific problems associated with VTOL or
STOL, which also exist in other aircraft and engine noise control programs,
include: Inlet design and wall treatment for noise suppression, prediction of
fan noise and development of methods for control, engine core and com-
bustion noise prediction and control, flap-interaction noise prediction and
control, and prediction of airframe noise.
It is apparent that the various experimental research programs conducted
in these V/STOL problem areas have improved in quality by utilizing
sophisticated test techniques, instrumentation, facilities, and imaginative
models. The results of these tests should provide the basis for improved
modeling of the flow mechanisms and hence allow the development of better
methods for the prediction and calculation of VTOL and STOL noise. For
example, the experimental investigation by Fink (Volume 45) which
examined the aeroacoustic mechanism that produces externally blown flap
(EBF) scrubbing noise, has provided concepts of the noise-producing
mechanisms that will be useful to develop prediction methods for EBF noise
estimation. Scrubbing noise appeared to come from weak surface loading
fluctuations that were coherent along the scrubbed span. The opinion is that
these loading fluctuations were induced by converted large-scale vortex
structure of the impinged jet.

In another experimental investigation by Patterson, Joshi, and Maus in
Volume 45, the aeroacoustic characteristics of the upper surface blown flap,
a high-lift device contender for V/STOL aircraft were studied by three ex-
perimental approaches: free-field acoustic measurements, reverberation
chamber acoustic measurements, and flow visualization and hot-film
anemometry measurements of the involved fluid dynamic phenomena.
These experimental techniques revealed significant mixing, turbulence, and
interaction phenomena that will be very useful in the development of noise
prediction criteria for design purposes.
The increased activity by researchers in the airframe and airfoil noise area
is illustrated by the fact that fourteen papers were presented at the Con-
ference, all of which are included in Volume 45. Airframe noise is defined as
the noise generated by an aircraft's nonpropulsive components in flight.
Thus, airframe noise is attributed to turbulent flows interacting with aircraft
surfaces. For advanced aircraft, airframe (nonpropulsive) noise could be
nearly equal to noise from propulsion systems during the final stages of lan-
ding approach. Future noise regulations are estimated at 10 PNdB below
Federal Air Regualtion 36 (FAR-36). Thus, if airframe noise sources exceed
the FAR 36-10-dB level, this federal regulation noise level cannot be
satisfied by reducing propulsion noise sources alone. In fact, a number of
recent measurement programs on different types of aircraft have indeed
determined that airframe noise attains a level within 10 dB of the overall air-
craft power-on noise levels for landing approach conditions. Therefore, air-
frame noise must be considered as a potential obstacle for commercial air-
craft to overcome in order to satisfy proposed government noise regulations
successfully.
Significant progress has been made during the two years since the first
AIAA Aeroacoustics Conference to identify acoustically important air-
frame components and to develop noise prediction methods for each com-
ponent. It became apparent from these research programs that there was a
great dependence of the radiated noise spectrum on the aerodynamic con-
figuration of the aircraft. Further, these investigations established a need to
develop a better or more detailed understanding of the different parameters
and aeroacoustic mechanisms responsible for airframe noise and the
feasibility of reducing principal sources of this nonpropulsive noise. Ac-
cordingly, an assessment of recent research progress and state-of-the-art in
the airframe noise area, as reflected by the fourteen papers mentioned in the
preceding paragraph, revealed significant progress on most aspects of this
aeroacoustics problem.
Noteworthy is the progress being made to develop prediction methods for
the calculation of airframe radiation noise into the far field. Noise
generation by the interaction of airflows and surfaces has been researched
over many years; however, studies of the far-field noise generated by an air-
frame in assembled aircraft have been initiated recently. For example, a

paper by Hayden, Kadman, Bliss, and Africk in Volume 45 identified the
most acoustically important airframe components and developed noise-
prediction procedures for each component. By utilizing semiempirical
techniques, the relative noise contributions of flow over wing, flap, and con-
trol surface edges, landing gear, and open cavities were assessed. The
relative importance of each source was found to be dependent upon aircraft,
configuration details, and flight parameters, such as flap setting, airspeed,
etc. These prediction procedures were applied to a typical CTOL passenger
jet using actual aircraft parameters. The noise levels calculated for such
CTOL aircraft confirmed the existing concern over the airframe noise levels
because these levels were not much below measurements of total aircraft
power-on levels.
An experimental investigation of airframe noise in the far field by

Fethney utilized four aircraft covering a range of geometric and flight
parameters. In this paper (Volume 45) the experimental and analytical
techniques are described, the measured airframe noise levels and spectra are
presented, and the relative contributions of some of the airframe com-
ponents are identified. The principal portion of the airframe noise appears
to be at frequencies between 50 Hz and 1 kHz and is increased substantially
by extension of flaps, slots, and landing gear, comparisons of experimental
data with semiempirical estimation methods show only moderately good
agreement for the cruise configuration and do not support the hypothesis
that the dominant noise source is a vertically aligned dipole.
Another experimental and analytical investigation of aircraft airframe

noise by Putnam, Lasagna, and White (Volume 45) has provided overall
sound pressure levels and airframe noise spectra data from flight tests of
four different aircraft: Aero Commander, Jet Star, CV-990, and B-747. The
aircraft were tested in both the clean (cruise) and landing configurations.
These test results indicated that the airframe noise level for each aircraft in
the clean configuration varied with the fifth power of the airplane's velocity.
The results also showed that, although the airframe noise for the landing
configuration varied as the fifth power of the velocity, the airframe noise
level for aircraft in the landing configuration appeared to depend primarily
on the detailed aerodynamic configuration of each'airplane. It also was con-
eluded in this investigation, as in other investigations, that it is much more
difficult to describe and predict airframe noise for aircraft in the landing
configuration compared to that of aircraft in the cruise configuration.
Modifying the configuration of an airplane from cruise to landing
significantly increases the airframe noise level. This is attributed to the ad-
ditional noise sources in the landing configuration, such as: the landing
gear, landing gear cavities, wing-leading edge, high-lift devices, and
multiple segmented wing trailing edge flaps. An adequate assessment of the
additional noise sources has not yet been made, primarily because no
adequate physical model of each component has been developed. Thus, it is
apparent from the results of these investigations and several others by Revell
and Healy that the semiempirical methods being used to predict airframe
far-field noise have serious limitations, particularly for the landing con-
figuration. Therefore, an urgent need exists to advance our knowledge con-

cerning the mechanisms of airframe noise generation and radiation into the
far-field for each component and to obtain additional noise data from a
broad spectrum of aircraft configurations in the landing configuration. As
more knowledge and data are obtained, a more refined methodology for the
prediction of airframe noise radiation into the far-field will be developed,
especially for the landing configuration. In fact, recent progress of basic
research investigations associated with airframe component noise is
described in seven papers in Volume 45. These studies have provided some
significant insights into the generation mechanisms of airfoil trailing edge
noise, noise generation mechanisms due to flow-induced pressure fluc-
tuations in cavities, and noise radiation from turbulent flows over compliant
surfaces.
Trailing edge noise has been identified as an important component of air-

foil and jet flap noise. A paper by Tarn and Yu (Volume 45) describes their
investigation of trailing edge noise which had as its objectives: 1) to un-
derstand the generation mechanism of trailing edge noise, and 2) to in-
vestigate the interaction between trailing edge noise and the flap surface
especially in relation to the far-field noise directivity pattern. Tarn and Yu
have observed experimentally that, at a jet exit Mach number of 0.3 to 0.8,
an orderly, large oscillatory flow structure existed. The investigators believe
that these large-scale disturbances are the result of flow instabilities. They
also believe that these orderly disturbances are responsible for generating
the dominant part of trailing edge noise, either directly or indirectly. This
finding is considered new and significant. It provides a new understanding
of the generation mechanism of trailing edge noise, and it also will be very
useful in efforts to suppress the trailing edge noise. The interaction of noise
generated by these coherent oscillatory flow disturbances and the flap down-
stream of the trailing edge also was investigated by Tarn and Yu. It was
found that the far-field noise directivity is influenced strongly by diffraction
of sound at the leading edge of the plate.
Flow over cavities is another source of airframe noise that has received in-
creasing attention by researchers during the past several years. It is well
known that high-speed flow over cavities or cutouts in the structural sur-
faces of aircraft usually generates intense tonal pressure fluctuations. The
consequences of such a condition can jeopardize the integrity of nearby
structural components or sensitive instrumentation and also affect the air-
frame noise radiation to the far field. Thus, it is of major importance for the
scientific and engineering communities to understand the mechanisms that
produce cavity oscillations and to develop practical methods for suppressing
these oscillations.
A paper by Heller and Bliss (Volume 45) describes an experimental and
analytical research program that has provided significant fundamental data
and a better understanding of the physical mechanisms governing the
generation of pressure fluctuations in shallow rectangular cavities exposed
to high-speed flow. An analytical model was developed which describes and

characterizes various, features of the phenomena. Experimental in-
vestigations conducted in large-scale wind tunnels provided information
concerning mode shapes, mode levels, and frequencies for cavity depth-to-
length ratios of 2.3 to 5.5 and Mach numbers of 0.8 to 2.0. Small-scale tests
provided low Mach number data. A study of several oscillation suppression
devices indicated that some of these concepts provide substantial reductions
in unsteady pressure levels.
Another interesting paper in Volume 45 is that by Franke and Carr on
studies of open-cavity, flow-induced oscillations. It provides additional data
on the mechanisms of cavity oscillations and potential geometric
modifications for reducing the pressure oscillations in cavities. Studies of
cavity length-to-depth ratios from 1 to 414 were made. Preliminary
qualitative screening studies of various cavity configurations were first
made on the water table at Froude numbers of 0.6 to 3.3. Cavity con-
figurations demonstrating good oscillation suppression characteristics in
water were investigated further in air-flow apparatus by comparing band-
width cavity pressure measurements at flow Mach numbers of 1.6. A
significant observation was the fact that the cavity configurations demon-
strating the greatest oscillation suppression were those in which the shear
layer did not enter the cavity. Ramps at both the leading and trailing edges
of the rectangular cavities were observed to be most effective in reducing
pressure oscillation amplitudes under some conditions.
A recent investigation was conducted by Shearin and Block (Volume 45)

to determine the feasibility of measuring airframe noise on small models of
transports in flow facilities. If such test techniques prove feasible,then the
prohibitive costs and difficulties of obtaining airframe noise data from full-
scale overflight tests could be substantially reduced. Accordingly, the paper
contains descriptions of the facility and equipment used, the problems en-
countered in making airframe noise measurements, and sample acoustic
data.
The use of remotely piloted vehicles to obtain airframe noise data is

another interesting technique that recently was studied. The results of this
investigation are described by Fratello and Shearin in Volume 45.
"Acoustic Wave Propagation through the Atmosphere'' is an important
aeroacoustic problem that was covered to some extent by six papers at the
Conference (Volume 46). Researchers in this area are attempting to deter-
mine what happens to noise after it leaves the generator or source on its path
through the atmosphere to the observer on the ground. Although the general
problem of wave propagation in turbulent media has been studied for many
years, the analytical techniques that have been used extensively are now con-
sidered to be inadequate from an aeroacoustics standpoint. Two additional
considerations that previously were not included in the theories must be
taken into account in the application of analytical techniques to wave
propagation in turbulent media. It is encouraging that these papers included
the two additional effects: 1) cumulative effects due to wave propagation
over long distances in a turbulent media, and 2) relaxation effects.
The cumulative effects include scattering of coherent sound waves by at-

mospheric turbulence and also random wave fluctuations. This is because,
as an acoustic wave propagates through a moving inhomogeneous medium
containing random velocity and temperature fluctuations, the acoustic
energy is scattered, and wave amplitude and phase demonstrate random
fluctuations. Previous, work on scattering theory used ideal and often
unrealistic models with regard to wave amplitude and phase fluctuations as
well as expressions for scattering sound intensity.
A paper by Huang (Volume 46) describes some significant advances in

sound scattering theory. The purpose of this study was to relax some of the
physically unrealistic limitations of previous analyses. In this investigation,
the sound propagation through a turbulent flow field with nonvanishing
mean-flow velocity was studied. By utilizing an analysis which describes
moving sources in a fixed frame, results were obtained for the far-field scat-
tered sound intensity due to uniformly moving turbulence confined in a
finite scattering volume. This analysis was extended to include wave
propagation through mean flow with constant gradient.
Another recent analytical study of the scattering of coherent sound waves

by atmospheric turbulence (Chow, Liu, and Maestrello, Volume 46)
produced a more general theory in comparison with previous work on this
phenomenon. This improved theory can be applied to various problems,
such as a fixed source, a moving source, or a beam of acoustic waves. The
general equations were applied to two special cases, corresponding to
uniform and shear mean flow. The results showed that the propagation of
coherent acoustic distrubances are substantially affected by mean shear and
turbulence. Experimental measurements associated with atmospheric scat-
tering problems are presently needed. The general theory is complex and
requires the knowledge of all elements in the covariance matrix which, un-
fortunately, is not all readily available nor measurable physically.
Therefore, the general equation must be simplified to make it useful for
practical applications.
The other important cumulative effect of acoustic waves is connected with
the fluctuations of the waves. As a wave propagates through a turbulent
medium, it is continually distorted because of diffraction and scattering by
the inhomogeneities of the medium. Consequently, the wave develops ran-
dom fluctuations in its amplitude, phase, intensity, and other parameters.
The fluctuations of the wave usually increase with propagation distance,
and therefore this effect is considered cumulative. A recent theoretical
analysis by Wenzel (Volume 46) of the acoustic wave field radiated into a
turbulent medium showed that the fluctuations of the wave increase initially
in proportion to the propagation distance but that at greater distances the
fluctuations of the waves approach a limiting, or saturation, value. These
results agree well with observations of wave propagation in real media.
An interesting possible effect on acoustic wave propagation through the

atmosphere, which researchers are presently taking into account, is that
aspect of real gases known as relaxation effects. Aerodynamicists have

known for a long time that relaxation effects can be important with strong
shocks and high temperatures. Thus, it is surprising to find real gas effects
important even at weak shock levels. This potential effect arises in the
analysis of the observed shock thickening which, incidentally, has been a
controversial topic for many years. One school of thought was that the
shock thickening is caused by cumulative effects of turbulence over great
distances. Others thought that shock thickening could be attributed to the
relaxation effects. To resolve this controversial issue of what mechanism
causes weak shock thickening in the atmosphere, an experimental in-
vestigation was conducted in the laboratory to measure the effects of tur-
bulence on the rise time of a weak shock. A paper by Tubb in Volume 46 on
the results of this investigation showed that the transmitted shocks through
turbulence demostrated rise times double that of the original shock
propagated without turbulence but fell far short of the 100- to 1000-fold rise
time anomaly of sonic boom. The author claimed that his experiments were
too limited to be conclusive. Thus, this interesting and important issue has
not yet been resolved and therefore warrants further research. Apparently
both the relaxation effects and the cumulative turbulence effects are im-
portant, as was also pointed out in another paper by Abrahamson (Volume
46) but which is the dominant effect and at what stage of the acoustic wave
propagation through the atmosphere does it occur? Hopefully, our
knowledge of this problem will be advanced during the near future by fur-
ther testing in the field as well as by conducting tests in appropriate
laboratory facilities.
The importance of aeroacoustic instrumentation, facilities, and test

techniques cannot be overemphasized. The ability to relate acoustic
measurements in the near and far fields with the noise generating source
mechanisms located in or near the aircraft depends upon the sophistication
of the instrumentation and test techniques being used. Six papers presented
at the Conference and included in Volume 46 reflect significant progress in
aeroacoustic instrumentation and in test techniques. A broad spectrum of
different types of instrumentation have been used to characterize the noise
sources in the flow and the radiated noise to the far field. The investigators
who were attempting to advance the knowledge of noise source mechanisms
in the flow used essentially three different approaches: 1) instruments placed
within the flow or jet stream to measure fluctuating and mean quantities, 2)
utilization of noninvasive techniques to measure fluctuating and mean
quantities, and 3) instrumentation used outside of the flow to measure
acoustic characteristics. With proper orientation of the instruments the
noise source distributions can be predicted. Fortunately, all three methods
were covered in these papers.
Examples of in-flow instrumentation that were discussed at the Con-

ference included a remote sensing microphone probe used in hot jets. In the
noninvasive method category, various methods to measure fluctuating
quantities were being proposed such as the sodium D Line Reversal Method
for the determination of both average and fluctuating temperatures. The

crossed beam laser Schlieren method was developed to obtain flow density
fluctuations. Experimental results using this method were shown for a high-
temperature supersonic jet. The calculated noise field, using the beam-
measured density fluctuations combined with experimentally evaluated
mean shear of the jet, agreed well with that obtained by using microphones
outside of the jet. In another paper, it was shown analytically that the laser
shadowgraph could be used to obtain quantitative data of the radiated
sound in a particular direction and at a specific frequency.
With regard to test techniques, several interesting papers in Volume 46

discuss various techniques, including the hole-in-the-wall method of deter-
mining the axial distribution of the noise sources in a jet. In this method, the
jet flows through an aperture, thus dividing the jet noise measurements into
two parts, upstream and downstream. Either the jet or the aperture can be
moved axially. Another paper described an acoustic wind tunnel that could
be used to simulate forward speed effects and to study the generation of
sound where flow interacts with solid surfaces. In another paper, the
Hartman Generator was proposed as a high-intensity sound source that
would have application in the evaluation of duct liner materials. Our efforts
to develop better aeroacoustic instrumentation and test techniques must be
continued vigorously since the achievements in this field will be the pacing
elements of aeroacoustics research and technology of the future.
We wish to express sincere appreciation and thanks to everyone involved.

In particular, we would like to acknowledge the generous efforts and
assistance of those members of the AIAA Technical Committee on Aero-
Acoustics who served on the Technical Program committee and assisted in
reviewing and selecting papers for the Conference. Their names appear on
page XXXIV. We also are grateful to the persons who served as panelists
and who used their expertise in their respective fields to provide the guidance
for discussion of the current state-of-the-art and recommendations for fu-
ture research. Their names appear in the Panel Discussions. Acknowledge-
ment also is due Dr. Martin Summerfield, Series Editor-in-Chief; Miss Ruth
F. Bryans, AIAA Director, Scientific Publications; Mrs. Jeanne Godette,
XXXIII
Administrative Assistant, AIAA Scientific Publications; Dr. John W.

Lucas, AIAA Vice President-Technical Activities; Mr. Walter J. Brunke,
Administrator, Technical Activities; Miss Bobbi Chifos, Director, Meetings
Operations; and Mrs. Inna Glass, Coordinator Technical Papers.
Ira R. Schwartz
Editor
Henry T. Nagamatsu and Warren Strahle

Associate Editors
February 1976
AIAA 2nd AEROACOUSTICS CONFERENCE
March 24-26,1975 Hampton, Va.
Conference Committee
General Chairman
Ira R. Schwartz
NASA Ames Research Center
Assistant General Chairman

Homer G. Morgan
NASA Langley Research Center
Administrative Chairman
William E. Zorumski
NASA Langley Research Center
Technical Program Committee

General Chairman: Mr. Ira R. Schwartz
Assistant Chairman: Mr. Homer G. Morgan
Dr. Darshan S. Dosanjh Dr. Henry T. Nagamatsu

Dr. Charles E. Feiler Mr. Robert E. Pendley
Dr. Antonio Ferri Mr. Harry E. Plumblee Jr.
Dr. William M. Foley Dr. Gerhard Reethof
Dr. Morton B. Friedman Dr. Herbert S. Ribner
Dr. Krishnamurty Karamcheti Mr. John Schettino
Dr. Jack L. Kerrebrock Dr. Craig D. Simcox
Dr. John Laufer Mr. David L. Smith
Dr. Frank E. Marble Dr. Warren C. Strahle
Mr. Paul F. Massier Mr. Emanuel J. Stringas
Dr. William C. Meecham Dr. Martin Summerfield


PROPAGATION OF AIRCRAFT NOISE OVER LONG DISTANCES
THROUGH THE LOWER ATMOSPHERE
A. L. Abrahamson*
Wyle Laboratories, Hampton, Va.
Abstract
Propagation of sound through the lower atmosphere is influenced

by numerous factors that are difficult to measure and more difficult
to predict. In addition to the well-known laboratory observable loss
mechanisms of heat conduction, gas transport, and molecular ab-
sorption, inhomogeneities in a real atmosphere have significant in-
fluence on a propagating sound wave. This study presents a quali-
tative discussion of different categories of atmospheric inhomogeneity
and their individual and combined effects on sound propagation. Sub-
sequently, a field test involving the propagation of aircraft noise over
distances up to 10 miles is described, and a simplified empirical
model for "excess" atmospheric attenuation due to inhomogeneities
in the atmosphere is derived from the data.
I. Introduction
The decay of sound intensity over a propagation path in the lower

atmosphere is attributable to many mechanisms. These may be di-
vided into two classes: 1) those observable in a still, homogeneous
Presented as Paper 75-542 at the AIAA 2nd Aero-Acoustics Con-

ference, Hampton, Va. , March 24-26, 1975. This work was spon-
sored by the Eustis Directorate of the U. S. Army Air Mobility Re-
search and Development Laboratory, and by the Langley Research
Center and Wallops Flight Station of NASA.
*Senior Research Specialist.
4 A. L. ABRAHAMSON
environment; and 2) those which are dependent upon inhomogeneities

in the atmosphere and the presence of partially absorbent boundaries.
For the purpose of this paper, these two classes will be referred to
as atmospheric absorption and excess attenuation, respectively.
Atmospheric absorption has been investigated thoroughly both

theoretically and in the laboratory*"3. Established models ' exist
which give good agreement with experiments under most conditions.
Atmospheric absorption occurs broadly in two ways: 1) "classical
absorption, n or losses associated with the change of acoustic or
ordered energy of air molecules into heat, or disordered energy of
air molecules through the dual mechanisms of conduction and viscos-

ity; and 2) "molecular relaxation" losses, which occur when poly-
atomic molecules of different gases constituting air commute a por-
tion of their acoustic translational kinetic energy into internal vibra-
tional or rotational energy within the molecules themselves.
For many years, workers in outdoor acoustics have attempted

to apply only atmospheric absorption to calculations of propagation
losses in the lower atmosphere, and a substantial fund of evidence
has accumulated on the errors of this approach. Sound propagating
in Earth f s boundary layer is subject to wind-velocity gradients, tem-
perature gradients, and atmospheric turbulence, which in combina-
tion may produce results not readily observable in a laboratory.
This paper is concerned specifically with sound propagation over
distances up to 10 miles at small angles of incidence to the ground.
II. Sound Propagation in the Lower Atmosphere:

Individual Effects
Wind-Velocity Gradients
When air flows uniformly over a large, solid, and flat surface,
there is a smooth transition from zero velocity at the surface of the
solid to the freestream flow velocity some distance away. A sound
wave in such a boundary layer propagates so that its rays follow
curved instead of straight paths. In the case of sound propagating
in an upwind direction, as depicted in Fig. 1, this gives rise to a
shadow zone some distance from the source. From Huygen ? s
principle (that all points on a wavefront act as secondary sources),
SOUND PROPAGATION OVER LONG DISTANCES
Shadow Zone
llllli:!0pll;li!l:
Receiver i;
RMS Sound Wind

Pressure Velocity
Level Profile
at Surface Spherical
Spreading and Diffraction
.Atmospheric
Âbsorption DISTANCE
Fig. 1 Effect of wind gradient on low-altitude sound propagation:
upwind.
it may be seen that diffraction occurs into the shadow zone, and,
instead of a sharp cutoff, the sound intensity decays exponentially
into the shadow. In the case of sound propagation downwind, a
downward bending will occur, and, apart from following a slightly
longer path, the effect on the intensity of direct sound reaching the
receiver will be slight. Sound directed toward the surface will be
reflected, however, and contained in multiple hops to augment direct
sound at the receiver. In the presence of a nonmonotonic gradient,
focusing may occur, and direct sound intensity downwind from the
source may be amplified substantially.
Temperature Gradients
The temperature of the surface of Earth undergoes cyclic heating

and cooling due to direct radiation from the sun and other climatic
effects. Thus a differential generally exists between measurements
of air temperature at different altitudes. Since the speed of sound
increases with temperature, temperature gradients cause refraction
of sound rays similar to wind velocity gradients. A negative tem-
perature gradient (temperature decreasing with altitude) will result
6 A. L. ABRAHAMSON
in sound waves being bent upward (Fig. 1), whereas a positive tem-
perature gradient will result in sound waves being bent downward.
Turbulence
Generally, uniform laminar flow and a stationary temperature

gradient do not exist. Therefore, the presence of a time-dependent
temperature gradient in air causes convection, whereas flow over a
realistically rough terrain with vegetation and topographic irregu-
larities results in turbulent flow. A sound wave encountering tur-
bulence is scattered. This is a rather loose term for a combina-

tion of reflection, refraction, and diffraction. Thus, if an eddy
were well defined and large compared with the wavelength of sound
under consideration, the principle effects would be reflection and
refraction: if the eddy were small compared with the wavelength,
diffraction would predominate.
In the absence of gradients, consider the sound field of a direc-

tional source imbedded in a thin sheet of turbulence over a partially
absorbent surface (Fig. 2):
1) The directional sound radiation pattern will undergo general

broadening with distance from the source, and measurements will
become relatively independent of azimuth and elevation.
Nonturbulent Region
Turbulent Region
Receiver
Fig. 2 Effect of atmospheric turbulence on low-level sound propa-
gation.
SOUND PROPAGATION OVER LONG DISTANCES !
2) Sound energy, initially directed parallel to the surface, will

be redirected out of the upper boundary of the sheet of turbulence.
Such sound energy will be scattered no more and is lost effectively
to a ground-based observer.
3) Similarly, sound energy, initially traveling parallel to the

surface, will be scattered toward the surface, where it is partially
reflected and partially absorbed, depending upon surface hardness
and density of vegetation. This process may take place many times
over long transmission paths.
Superimposed on the preceding effects will be a "twinkle" or

"flicker" in sound monitored along the propagation path due to the
random nature of turbulence and to the widely different paths that
arriving sound may have traveled.
III. Sound Propagation in the Lower Atmosphere
All of the effects just described are relatively straightforward

in concept, and it is clear which will result in increased sound in-
tensity, and which will result in a decreased sound intensity at the
receiver station. In combination, however, the result is not so
apparent. Suppose that scattering from turbulence is added to the
condition depicted in Fig. 1, as shown in Fig. 3. The effect is to
Nonturbulent Region
Turbulent Region
Receiver
RMS Sound errain Wind
Pressure With Scattering and Velocity
Level at
Surface ____Ground Absorption Profile
- _ _ _ No_Sjcatteringj3r
____Ground Absorption
DISTANCE
Fig. 3 Effects of wind refraction, scattering, and ground absorption
on sound propagation: downwind approach.
8 A. L. ABRAHAMSON
increase the intensity of sound in the shadow zone, and, indeed, if

the turbulence is sufficiently intense, a shadow zone may not even
exist. For downwind sound propagation, adding turbulent scattering
is likely to increase sound attenuation over the case of refraction
alone. Some sound energy will escape from the turbulent layer, and
more will be absorbed partially in a greater number of reflections
at the surface.
IV. Field Measurements

The field measurements reported in this paper were part of a

study on helicopter aural detectability^, and were carried out at
NASA Wallops Flight Station over a period of two weeks. Tests
were conducted in the early morning and late afternoon to minimize
interruptions and errors caused by uncontrolled noise in the area,
and to facilitate propagation measurements over long distances.
Two flight paths were chosen, from due east and due west, to
coincide with the direction of the prevailing wind. Three noise
measurement stations were used for each flight path, with distances
between stations ranging from a minimum of 7300 ft to a maximum
of 12, 700 ft. Terrain in each direction was reasonably flat and
homogeneous, with grassland and tidal mudflats to the east, and
ploughed fields to the west.
The purpose of this aspect of the study was to provide engineer-

ing data on excess atmospheric attenuation resulting from inhomo-
geneities in the atmosphere. In order to construct a thorough ana-
lytical model with which to compare the data, it is necessary to make
extensive measurements of atmospheric parameters. Wind velocity
gradients, turbulence, temperature gradients, terrain roughness,
and large-scale meteorological events are all interrelated, and con-
tinuous knowledge of these parameters is required at all points along
the propagation path.
The detailed atmospheric and ground surface measurements

described previously, coupled with a computer-based solution
scheme, are of considerable interest but were rejected for this study
on the following grounds: 1) unavailability of these data for anyone
wishing to use the model for specific cases due to the vast number
of measurements required, and their nonstationary character; and
2) atmospheric turbulence results in significant modulation of sound
intensity, the magnitude of these fluctuations being such that an ex-
10 Frequency 63 Hz /
Sound Propagation Upwind /
o
o
Best-Fit Line
(for this point set)
Model Line
(from Figure 6)
0 I
0 1 2 3 4
Excess Atmospheric -Attenuation (dB)/lOOO ft
Fig. 4 Effect of wind velocity on excess atmospheric attenuation
from measured one-third-octave data: upwind.
Frequency 63 Hz
10
Sound Propagation Downwind
o
<D
CQ
<w-l
>>
CJ
Best-Fit Line
(for this point set)
Model Line
(from Figure 7)
0 \
0 1 2 3 4
Excess Atmospheric Attenuation fdB)/lOOO ft
Fig. 5 Effect of wind velocity on excess atmospheric attenuation
from measured one-third-octave data: downwind.
10 A. L. ABRAHAMSON
Slope (A) v s Frequency (f)
5 0 *
0>
a
0 Mean A = 0. 06
0 G
100 1000 10,000
Frequency in Hertz
1.0
Intercept (I3) vs Frequency (f)
o o
ex
(L)
O Mean B = 0. 36
0
100 1000 10,000
Frequency in Hertz
Fig. 6 Results of regression analysis on excess atmospheric ab-
sorption for sound propagation: upwind.
haustive study would, at best, yield only a probability distribution

for atmospheric attenuation.
Thus, measurements of atmospheric parameters were made

only at one location along the flight path. These consisted of contin-
uous recordings of temperature, humidity, and wind vector, at a
height of 15 ft, with periodic balloon measurements of temperature,
humidity, and wind gradients. Range measurements were made
using the station radar tracking facilities with a transponder
.4
Slope (A) vs Frequency (f)
.2
100 1000 10,000

Frequency in Hertz
2 . 0 Intercept (B) vs Frequency (f)
100 1000 10,000

Frequency in Hertz
Fig. 7 Results of regression analysis on excess atmospheric ab-
sorption for sound propagation: downwind.
mounted in the aircraft for improved accuracy at low angles of in-

cidence. Helicopter position was sampled 10 times/sec and digi-
tally smoothed in the on-line computer.
V. Analysis of Results
All acoustic measurements were related back to time-of-emis-

sion, and comparisons were made on this basis using retarded dis-
12 A. L. ABRAHAMSON
tances. Since the test was concerned with propagation over very
long distances, noise levels only marginally higher than the pre-
vailing ambient noise were involved. Extraction of acoustic propa-
gation data under these conditions required the assumption of a
moderately stationary natural ambient.
One-third-octave spectral analyses were carried out on the

data and corrections applied for spherical spreading and atmo-
spheric absorption. Resulting values of "excess" attenuation were
correlated in the following manner:
1) For elevation angles less than 2 , a linear regression law of

excess attenuation in each one-third-octave band vs wind speed was
assumed. Separate evaluations were carried out for sound propa-
gation upwind and downwind, e. g. , Figs. 4 and 5, respectively. A
linear regression with frequency was applied to these results as
shown in Figs. 6 and 7, with a cosine interpolation used to relate the
two cases.
2) For elevation angles between 2 and 10 , an empirical linear

decrease in excess attenuation from the value given in the foregoing
by the regression to zero at 10 was indicated.
The model derived from the regression analysis for angles less
than 2 reduced to the following:
E = Au+ B
E = Cu + D
where A = 0. 033 log f + 0. 1
B = -0.23 log f - F 1.6
C = 0.06
D = 0.36
SOUND PROPAGATION OVER LONG DISTANCES 13
and where E , E , E = excess atmospheric attenuation upwind,

downwind, at "0 , " dB/1000 ft
u = mean wind velocity minus standard

deviation, fps
0 = angle between wind and sound propagation

(=0 for propagation into wind)
f = frequency, Hz
A linear weighting factor \|f should be applied to these values for

excess attenuation (E ) at other angles of incidence (9):
where i|f = 1 for 9 < 2
t|f= -0.125 arc sin (a/r) - 1~1 for 2 < 9 < 10
i|f= 0 for 6 > 10
a = aircraft altitude at time of emission
r = aircraft slant range at time of emission
VI. Discussion and Conclusion
Figures 4 - 7 show that substantial scatter exists in measure-

ments of excess attenuation when correlated in the manner just de-
scribed. This was expected and was predictable from the nature"
of the physical phenomena involved. The usefulness of these results
does lie, however, in an indication of trends:
1) Throughout tests in the presence of a more-or-less steady

negative temperature gradient (lapse with no inversion), significant
differences were observed in excess attenuation of sound propagating
with or against the wind.
2) In the far field, at 1 mile or greater from a source, at an

altitude of less than 200 ft, and when the angle of incidence is about
14 A. L. ABRAHAMSON
2 or less, a simple power law with distance from the source was
found to represent "excess" sound attenuation reasonably.
3) In the range over which reliable measurements could be made

(25 Hz to less than 1000 Hz), relatively weak frequency dependence
was observed for propagation in an upwind direction. No frequency
dependence for propagation in a downwind direction was detected.
Although little other data exist for sound propagation over dis-
tances in the range 8000 to 50, 000 ft, a cautionary note should be
sounded in the indiscriminate application of the model contained
herein. Both the overall level and spectrum of atmospheric tur-

bulence are expected to change over more topographically irregular
terrain. This fact, possibly coupled with a deep covering of soft
vegetation, is likely to produce substantially different results for
excess sound attenuation. Furthermore, it should be emphasized
that the great majority of measurements were made at angles of in-
cidence less than 2, over distances greater than 8000 ft, at a source
distance of several thousand feet to the first measuring station. The
use of the model for propagation over shorter distances thus may
produce different results.
References
Kneser, H. O , "Interpretation of the Anamolous Sound Absorption

in Air and Oxygen in Terms of Molecular Collisions, M The Journal
of the Acoustical Society of America, Vol. 5, No. 2, October 1933,
pp. 122-126.
2
Harris, C. M. , "Absorption of Sound in Air in the Audio-Frequency
Range," The Journal of the Acoustical Society of America, Vol. 35,
No. 1, January 1963, pp. 11-17.
3
Evans, L. B. , Bass, H. E. , and Sutherland, L. C. , "Atmosphere
Absorption of Sound: Theoretical Predictions, " The Journal of the
Acoustical Society of America, Vol. 51, No. 5, May 1972, pp. 1565-
1575.
4
"Standard Values of Atmosphere Absorption as a Function of Tem-
perature and Humidity for Use in Evaluating Aircraft Flyover
Noise," Aerospace Recommended Practice 866, August 1964,
Society of Automotive Engineers.
SOUND PROPAGATION OVER LONG DISTANCES 15
Sutherland, L. C . , "Supporting Background Document for Draft

Standard on Atmospheric Absorption Losses in Still, Homogeneous
Air, M Rept. WCR 74-11, 1974, submitted to S1-S7 working group on
Air Absorption, Acoustical Society of America, Wyle Laboratories.
n
Abrahamson, A, L. , "Correlation of Actual and Analytical
Helicopter Aural Detection Criteria," Kept. USAAMRDL-TR-74-
102A, 1974, U. S. Army Air Mobility Research and Development
Laboratory.

MEASURED EFFECTS OF TURBULENCE
ON THE RISE TIME OF A WEAK SHOCK
P. E. Tubb*
University of Toronto, Toronto, Ontario, Canada

Abstract
It has been speculated that passage through fine-grain

atmospheric turbulence causes large increases in sonic boom
rise time: that the turbulence may be responsible for the
observed rise time being some 100 to 1000 times larger than
the theoretical (Taylor) values. A series of ad hoc labora-
tory experiments was undertaken to explore this possibility.
Weak shocks were passed at right angles through the turbulent
flow produced by a coarse grid at the exit of an 8- x 12-in.
duct. The transmitted shocks exhibited rise times double that
of the original shock, on a statistical average. This falls
far short of the 100- 1000-fold rise time anomaly of sonic
boom, but the very much shorter path length may be a factor.
Nomenclature
p - pressure above ambient

t - time
to = time to peak overpressure (conventional rise time) for
a shock wave (Fig. l)
T = time between shocks (Fig. l)
At = time between O.lAp and 0.9Ap (Fig. 2)
T = rise time (as measured experimentally)
= 1.25At (Fig. 2)
Tr = rise time in steady atmosphere (reference)
Presented as Paper 75-5^3 at the AIAA 2nd Aero-Acoustics

Conference, Hampton, Va., March 2^-26, 1975.
This research was sponsored by the National Research
Council (Canada), the Transportation Development Agency
(Canada), and the U.S. Air Force Office of Scientific Re-
search, under Grant AF-AFOSR 70-1885.
^Research Fellow, Institute for Aerospace Studies.
17
18 P. E. TUBB
AD
0.940
0.1 ^p
Fig. 1 Pressure signature of an idealized sonic

boom.
Fig. 2 Experimentally measured rise time of a shock

T.
T-p = rise time after passage through nominally steady flow
Tt = rise time after passage through turbulent flow
x,y,z= Cartesian coordinates (Fig. 3)
U = mean flow velocity in x direction (Fig. 3)
u = x component of turbulent velocity (Fig. 3)
U' = root-mean-square x component of turbulent velocity
U0 = mean flow velocity at the unobstructed duct exit
= Reynolds number of grid Uob/v
b = bar width
v = kinematic viscosity = j-i/p
M- = dynamic viscosity
P = density
S = solidity ratio of turbulence generator
= l-(A0/At)
THE RISE TIME OF A WEAK SHOCK 19
UL
Fig. 3 Coordinate system.
A0 = open area of turbulence generator

At = duct exit area (total)
Introduction
Overflights of supersonic aircraft give rise to transient
pressure waveforms at ground level known as sonic booms. The
predicted pressure signature at ground level for a steady at-
mosphere has a characteristic N shape as shown in Fig. 1. The
main parameters characterizing the waveform are its peak over-
pressure Ap, its rise time to, and its duration T, which for a
supersonic transport (SST) have typical measured values of
2 psf, 300 msec, and 1 msec, respectively.1 Experimental data
such as those presented in Ref. 2 indicate that the predicted
signatures are essentially correct with respect to gross fea-
tures of shape and duration; however, measured rise times are
typically 1 to 10 msec, which is of the order of 1000 times
greater than the theoretical Taylor value of 2.7 jasec for a
shock with an overpressure of 1 psf. 3 [in Taylorfs theory,
the steepness of a shock front is governed by a balance between
nonlinear steepening and diffusion due to viscosity and heat
conduction. He derived the formula Ji ~ (Ui-Us)"1 cm, where Ji
is the shock front thickness in centimeters, Ui is the velocity
of the gas entering the shock, and (Ui-U2) is the particle
velocity in centimeters per second behind the shock in a shock-
fixed reference frame. Shock thickness and rise time are
equivalent concepts in that the rise time is the time required
20 P. E. TUBB
in a stationary reference frame for the shock thickness to

pass. Thus., to - $/Ui sec.] Additionally, the measured N
waves frequently exhibit "spiking" or "rounding" of the sig-
nature, along with numerous irregularities. A rise time in-
crease of at least 100-fold compared with theory always is
observed, even under the calmest atmospheric conditions, where-
as the spiking or rounding occur during disturbed atmospheric
conditions and alternate in a random fashion.
Overpressure and rise time are both important factors in

evaluating potential annoyance and damage from sonic booms.
This has motivated a number of theoretical attempts to explain
the anomalously long rise time and the distorted shape of sonic
booms. Crow^?5 and Pierce^ have attributed the spiking and
rounding to turbulence in the planetary boundary layer. The
mechanism proposed by Crow was the interaction of the weak
shocks with turbulence to produce high-frequency scattered
waves that arrived behind the shocks and distorted the wave
structure. Pierce, on the other hand, attributed the distor-
tion to rippling of the wavefront due to local focusing and
defocusing by large turbulent eddies. He argues that focus-
ing will produce folding of the wavefront, resulting in a
double shock (spiking), whereas defocusing will reduce the
strength of the shock (rounding). Ribner et al.7 independently
put forth a similar argument to explain the spiked and rounded
booms based on earlier experiments^ on focusing and defocusing
of sound rays by jet flows. They argued that local regions in
large-scale (e.g., 200 ft) eddies of atmospheric turbulence
were somewhat jetlike: the experiments of Ref. 8 suggested a
focusing effect (spiked boom) when the sonic boom propagated
counter to the jet flow, and a defocusing effect (rounded boom)
when the boom propagated with the jet flow. Very realistic
spiked and rounded booms were produced when these notions were
tested in the- laboratory with air jets and simulated sonic
booms from a shock tube.7
Later, attempts were made to develop these theories fur-

ther to explain the increased rise times. George and Plot-
kin9?10 and George^- have extended the first-order theory of
Crow to include terms of second order. They attribute the
shock thickening to the notion that the scattered waves will
carry high-frequency energy out of the shock front, thus in-
creasing the rise time. Piercel2 and Pierce and Maglieri^-3
have argued that small-scale turbulence causes multiple folding
of the wavefront, resulting in a succession of microshocks
having different arrival times; when smoothed out by viscosity
and thermal conductivity, these are expected to give rise to
an apparent single thickened shock.
Ffowcs Williams and Howe, on the other hand, although

agreeing that turbulence is the likely cause of spiking and
rounding, do not support the hypothesis that turbulence is the
shock-thickening mechanism. They argue that scattering will
produce at most a twofold increase in rise time and that wave-
front folding is much too slow a mechanism. Instead, they
propose that nonequilibrium gas effects such as those investi-
gated by Hodgson and Johannesenl-5 and Hodgson^-6 are responsible
for the shock thickening.
Thus the theoretical side of the question of shock thick-
ening due to turbulence is considerably unsettled. For this
reason, a series of ad hoc experiments to investigate the
effects of turbulence on the rise time of a weak shock were

undertaken. These experiments and their results are the
subject of this paper.
Experiment
Method
Two main elements were necessary for this investigation:
l) turbulence, and 2) a shock wave. Turbulent flow was gener-
^- Oscilloscope
Ai
Fig. h Schematic of experimental setup
22 P. E. TUBE
Section A-A ^ /O 4Mi -/O

4
i
i-
Fig. 5 Relation of shock tube, duct, and micro-
phones (typical measurements).
ated by placing obstructions at the exit of a wind-tunnel duct,
and a cons tant-temperature anemometer was used to record typi-
cal values of the mean flow velocity U and the root-mean-square
axial component of the turbulent velocity u f . A weak shock,
produced by a shock tube, was passed at right angles through
the turbulent flow and the resulting pressure signature re-
corded using two microphones connected to an oscilloscope
equipped with a Polaroid camera. The rise times of these
shocks then were measured and compared to the rise times of
shocks recorded in the absence of turbulent flow. A schematic
of the experimental arrangement is shown in Fig. k, and some
typical measurements relating the positions of the duct, shock
tube, and microphones are given in Fig. 5.
Experimental Considerations
Mean Flow and Turbulent Velocities, Turbulent velocities

in the planetary boundary layer usually do not exceed 10 fps,
and typical mean flow velocities in the upper region are of
the order of 100 fps.^7 However, since the path length of
turbulence which the shock encounters in the experiment is on
the order of 10~3 that traversed by a sonic boom, it was felt
that a turbulent velocity in excess of 10 fps would be neces-
sary if one were to have a reasonable chance of seeing a sig-
nificant effect. Thus it was decided to operate the wind
tunnel at maximum speed, with bars or grids placed at the
duct exit to produce turbulence of as great an intensity as

possible.
The Shock. Two characteristics of the shock were con-

sidered to be of paramount importance: l) small rise time
(on the order of 10 [_isec or less), and 2) low overpressure.
The small rise time was desirable in order that we would be
starting with an approximation to the Taylor thickness. The
shock tube will not yield shocks of this small thickness auto-
matically because of the noninstantaneous rupture of the dia-
phragm coupled with the limited length for shock development.
As can be seen from Figs. 6 and 7 we were successful in achiev-
ing the desired rise time.
Nonlinear steepening, which refers to the way compression

waves eventually coalesce and steepen into a shock, is a well-
known concept of gasdynamics. It is this mechanism that
causes an originally thick shock to sharpen as it propagates.
Moreover, strong shocks tend to steepen more rapidly than weak
ones. It was judged unlikely that nonlinear steepening would
be appreciable in the few milliseconds required for the wave
to reach the microphones; nevertheless, it was felt that Ap
should be as small as possible to insure that the shock, if it
had been thickened by turbulence, would not steepen again
before arriving at the microphone. Accordingly, an attempt
was made to construct a shock tube that would produce inherent-
ly weak shocks. A maximum value of 2 to 3 psf was set as the
upper limit of desirable shock overpressure, since this is
typical of N waves observed at ground level.
Measurement of Rise Time. The conventional definition

of rise time is the time to peak overpressure to. However, if
the waveform is rounded at either the base or peak then a
determination of t0 may be difficult. To simplify the measure-
ments, "rise time," for the purposes of this experiment, was
redefined as 1.25At, where At is the time taken for the wave-
24 P. E. TUBE
Fig. 6 Typical oscillographs for shocks

produced using carbon paper diaphragms.
a) reference shock (still air); b) shock
after passage through turbulence. Verti-
cal scale: top trace (mic. l) 0.5^4 psf/
div; bottom trace (mic. 2) 0.63 psf/div.
Horizontal scale: a) and b) 20
form to rise from 10f0 to 90fo of its peak overpressure (Fig. 2).
Thus, the experimentally measured rise time r will be equal to
t0 only if the pressure rise is linear, but the error will
usually be less than that involved in trying to measure t0
directly.
Effects of Mean Flow. As mentioned earlier, the experi-
ments of Ref. 7 demonstrated that an N wave passing through
the velocity gradients of a jet blowing counter to or with
1
I I
/Y
1 I
Fig. 7 Typical oscillographs for shocks

produced using "red zip" diaphragms, a)
reference shock (still air); b) shock
after passage through turbulence. Verti-
cal scale: top trace (mic. l) 0.99 psf/
div; bottom trace (mic. 2) 1.17 psf/div.
Horizontal scale: a) and b) 10 jasec/div.
the direction of shock propagation exhibits spiking or round-

ing. Thus, in order to eliminate (or at least minimize) these
effects> it was decided to fire the boom across the flow at an
angle of 90 to the mean flow direction.
Experimental Apparatus
The Wind Tunnel. The flow facility used in the experiment

was the UTIAS Air Duct Facility.1^ Originally constructed as
26 P. E. TUBB
an open-cycle wind tunnel, it now has been converted to closed-

cycle operation. A 5 ft section of ducting was removed in
order to allow access to the flow. The duct cross section had
a depth of 8 in and a width of 12 in; this meant that, with
the shock tube fired across the flow as shown in Figs. 4 and
5, the turbulent path length was approximately 1 ft.
Turbulence Generators and Velocity Measurements. In order

to cause the flow from the duct to become highly turbulent, a
variety of bars and grids (referred to as turbulence genera-
tors) were placed at the duct exit. Their characteristic
width varied from 1/2 to 3 in., with both round and square
cross sections. Details of the six turbulence generators used
are given in Fig. 8.
Because of the ad hoc nature of these experiments, a de-

tailed investigation of the flow produced by the turbulence
generators was not undertaken. However, a TSI hot-film
anemometer system was used to measure the axial mean velocity
U, and the rms turbulent velocity component in the axial
direction, u', at the point (9 x 1^ x in.) with the wind-
tunnel speed control at maximum in order to obtain typical
values.
The anemometer system consisted of a TSI type 1031 power

supply and monitor, type 103^4A constant-temperature anemometer
and linearizer, and type 1210AG hot-film probe (0.006-in. dia).
The d.c. voltage component of the linearizer output was meas-
ured with a Hewlett-Packard type 970A probe multimeter and the
rms a.c. component with a Bruel and Kjaer type 2^417 random
noise voltmeter. A calibration curve of velocity vs voltage
for the hot-film probe was obtained using the probe calibrator
of the Subsonic Aerodynamics group of UTIAS. The curve was
used to convert d.c. and rms a.c. voltage readings to U and
u', respectively.
The mean flow velocity from the unobstructed duct exit

UQ "was Il8 fps, and u 1 was about 3 fps. With the turbulence
generators at the duct exit, the flows were altered as indi-
cated in Table 1. U and u 1 were of the order of 100 and 20
fps, respectively, indicating a substantial increase in the
intensity of the turbulence. Table 1 also contains the Rey-
nolds number for each of the turbulence generators based on
the mean flow velocity of Il8 fps and a characteristic length
equal to the bar width. All are at least a factor of k greater
than 5000, indicating that we were assured of turbulent mixing
conditions in the flow.^9
I
Duct
3" 4
(a.) 3 0. D. Tu-tin
O. D. A l u m i n u m Tubin
I- Wooden Dowels
binj
m
7_ 3 O - 0 . A l u m i n u m Tutbin V___l" x l" Pine Strapping
Fig. 8 a) Turbulence generator 1, b) turbulence

generator 2, c) turbulence generator 3, d) turbu-
lence generator k, e) turbulence generator 5, f)
turbulence generator 6.
28 P. E. TUBB
Table 1 Characteristic parameters for turbulence generators

Turbulence
generator no. 1 2 3 u 5 6
b, in. 1.25 3 3 3 1.25 0.5
Cross section 0 O O O 0 D
U, fps a
9^ 91 91 106 99 105
u' , fpsa 18 18 20 20 20 15
7.3xl04 l.SxlO5 l.SxlO5 l.SxlO5 7-3xl04 2.9xl04
S 0.10^ 0.25 0.25 0.5 0.^47 0,6^43
^Typical values measured at 9 x l|- x -| in; wind-tunnel speed

control at maximum.
b
Based on UQ = 118 fps.
The Shock Tube. The simplest type of shock tube to build

is one of constant area. As a starting point, two lengths of
3-in. o.d. aluminum tubing, 6-in. and 12-in. long, were fasten-
ed to brass clamping plates. The 6-in. section was sealed
with an endplate at the other end, and inlets for compressed
air and a pressure tap were drilled. With a diaphragm clamped
between the two plates, the 6-in. (or driver) section could be
pressurized using compressed air controlled by a needle valve.
A Wallace and Tiernan model 62B-Â-00030 pressure gage was
used to measure the driver pressure. The diaphragm was rup-
tured by passing a metal rod through a hole in the 12-in. (or
development) section, resulting in a shock wave that steepened
under constant-area conditions in the development section. The
shock so generated had a rise time of less than 10 (asec as
desired, but the overpressure was of the order of 10 psf or
greater. Overpressure can be reduced by lowering the driver
pressure, but the effect of this was an unwanted increase in
rise time due to poor diaphragm rupture. Thus it was decided
to reduce overpressure by decreasing the exit area of the
shock tube, using the configuration of Fig. 9. The small
copper tube allows only a small portion of an essentially plane
shock wave to exit and undergo quasispherical diffractive
spreading with no noticeable increase in rise time.
Two diaphragm materials were used in the tests, carbon
paper and "red zip," a cellophane about h jam thick. The carbon
paper would burst at lower driver pressures but had the dis-
Fig. 9 Details of shock tube.
advantage that it -was necessary to self-burst these diaphragms,

since they "would not rupture properly when pierced. The re-
sult was that the reference rise time Tr of the shocks produced
using carbon paper exhibited scatter. Red zip, on the other
hand, ruptured well when pierced and produced consistent ref-
erence rise times. The required driver pressure was higher,
which resulted in higher overpressures; however, they were
still about 3 psf, which was satisfactory. Because both dia-
phragms had advantages, they both were used-in the experiment.
Pressure Waveform Measurements. The pressure waveforms
produced in the experiment were recorded using two Bruel and
Kjaer type Ul35 -J--in. condenser microphones, type 2801 micro-
phone power supplies, and type 2615 cathode follower preampli-
fiers connected to a Tektronix type 555 dual-beam oscillo-
scope. A trigger signal was provided by a crystal microphone
placed an appropriate distance in front of the recording micro-
phones. Microphone 1 was placed directly across from the
shock tube, and microphone 2 was located 3 in. downstream, as
shown in Fig. 5. The second microphone was used to check for
30 P. E. TUBB
possible convective effects due to mean flow. The free-field

open-circuit response of the i^-in. microphones is flat within
2 dB from ^4 Hz to 100 kHz. When coupled to the preamplifier
and supply, the response will "be slightly poorer but still
adequate to resolve a rise time of the order of 10 p,sec.
Results and Discussion
1) Figures 6 and 7 are oscillographs for typical shocks

after passage through undisturbed air (reference case, rise
time TT) and through the turbulent flow (rise time Tt). Many
of the latter exhibit increased rise times. Figures 10 and 11
present a statistical treatment of all of the data, without
regard to the turbulence generators used, in the form of a

plot of frequency of occurrence vs rise time. Although the
reference rise times for carbon paper diaphragms (Fig. 10)
show a wide scatter, nevertheless one can discern that on the
whole rise times with turbulent flow are higher. The results
for nred zip" diaphragms (Fig. 11) are much more clear cut; in
this case, a very consistent reference rise time rr is altered
by the turbulence into a widely scattered TJ., indicating shock
thickening in a large number of events. Note that in both
cases the frequency of occurrence of the thinnest shocks de-
creased with turbulent flow, and no instance of a shock thinner
than the thinnest of the reference shocks was observed. This
would seem to indicate that turbulence is a shock-thickening
mechanism but not a shock-thinning mechanism. However, this
lack of rise time decrease may be because the shock thickness
is within less than an order of magnitude of the Taylor value;
that value, which is dictated by viscosity and thermal con-
ductivity, must represent a lower limit to the possible thin-
ning.
2) To make sure that the shock thickening was due to

turbulence in the flow rather than to some mean flow effect,
a number of shocks using "red zip" diaphragms were fired
through the flow from an unobstructed duct exit (no turbulence
generator). The resultant rise time T^ had an average and a
root-mean-square value of 7.6 |j,sec. This indicates a slight
but not significant increase over the average and root-mean-
square reference rise times of 6.9 fisec, and it is suggested
that the increase is due not to the mean flow but to the resi-
dual turbulence in the flow.
3) No evidence of convective effects due to mean flow

were observed. Shocks arriving at both microphones exhibited
thickening, although not necessarily simultaneously, due to
THE RISE TIME OF A WEAK SHOCK
/a Occurrence
2O
3O
Fig. 10 Distribution of reference rise time (Tr)

and rise time after passage through turbulence
(r-t) for shocks produced using carbon paper dia-
phragms.
36 Occurrence
90
85
8O
J-I
: :: 5
,
2O rWD
35 40 45 SO 55 6(.
15
10 *o i\?,
\
\
' I
5 I f>
VE
WI ^
c> 5 IO 15 20 25 '30" 35
TûsecJ
Fig. 11 Distribution of reference rise time (rr)
and rise time after passage through turbulence
(T^) for shocks produced using "red zip" diaphragms<
32 P. E. TUBE
Table 2 Average and root-mean-square rise time increases

Diaphragm type Carbon paper Red zip
Driver pressure range, psi 1.2 to 2.1 U . 6 to ^.9
Reference shock overpressure,

psf (typical) 0.8 2.5
Total number of reference shocks ^9 ^-2
Rise time of Average 10.1 6.9

reference shock
rr, |asec rms 10.7 6.9

Total number of shocks
through turbulence 82 13^
Turbulent rise Average 13-3 13.8

time rt, jasec rms 13.9 16.3
% increase in average
rise time 32 100
% increase in rms
rise time 30 136
the different turbulent conditions along the different paths

from shock tube to microphone.
U) Table 2 gives average and root-mean-square Tr and T^
for both diaphragms based on Figs. 10 and 11. The strong tur-
bulence increased the average rise time by a factor on the
order of 1.3 for shocks produced by carbon paper diaphragms,
whereas the'average rise time for "red zip" diaphragms was
increased by a factor of 2. The lower percentage rise time
increase for carbon paper appears largely due to the scatter
in rr; this results in a higher average and root-mean-square
reference rise time, resulting in a lessening of the percentage
increase. Although the mean rise time or thickness increased,
not all of the shocks exhibited thickening. This may be due
to the shortness of the turbulent path length; thus, not every
shock may have encountered a significant number of eddy zones
serving as effective "acoustic-lenses.""
Conclusion
Weak shocks passing through turbulence in the laboratory
were found to have a tendency to thicken. However, the mean
shock thickness increased at most by a factor of about 2,

essentially as predicted by Ffowcs Williams and Howe.1^ It
would appear to support their conclusion that atmospheric tur-
bulence cannot account for the 100-fold minimum shock thicken-
ing found in sonic booms. However, the application to sonic
booms is of uncertain validity in view of the relatively short
path length through the turbulence, (it seems impermissible to
apply a scaling factor to the path length, since the reference
Taylor thickness could not be scaled down in the experiments.)
The possibility exists that, over longer distances, even with
lower intensity turbulence, the effect may be cumulative, thus
producing even greater shock thickening. Thus the question
remains unresolved. A more fundamental and comprehensive in-

vestigation is now underway, with a view to providing further
clarification.
References
Gottlieb, J. J., "Simulation of a Travelling Sonic Boom in a
Pyramidal Horn," Rev. 196, 197^, University of Toronto Insti-
tute for Aerospace Studies.
2
Maglieri, D. J. and Henderson, H. R., "Sonic Boom Measure-
ments for SR-71 Aircraft Operating at Mach Numbers to 3*0 and
Altitudes to 2^38^ Meters," TKD-6823, 1972, NASA.
Taylor, G. I., "The Conditions Necessary for Discontinuous
Motion in Gases," Proceedings of the Royal Society (London),
Vol. AQk, 1910, pp. 371-377-
1+
Crow, S. C., "Distortion of Sonic Bangs by Atmospheric Tur-
bulence," Aero. Rep. 1260, 1968, National Physical Laboratory.
Crow, S. C., "Distortion of Sonic Bangs by Atmospheric Tur-
bulence," Journal of Fluid Mechanics, Vol. 37, Part 3, July,
1969, PP. 529-563-
Pierce, A. D., "Spikes on Sonic-Boom Pressure Waveforms,"
The Journal of the Acoustical Society of America, Vol. kh, No.
4, October, 1966, pp. 1052-1061.
<-?
Ribner, H. S., Morris, P. J., and Chu, W. H., "Laboratory
Simulation of Development of Superbooms by Atmospheric Tur-
bulence," The Journal of the Acoustical Society of America,
Vol. 53, No. 3, March, 1973, pp. 926-926.
Ribner, H. S., Wang, M. E., and Leung, K. Y., "Air Jets as

Acoustic Lens or Waveguide," Proceedings of the 7th Inter-
34 P. E. TUBB
national Congress on Acoustics, 1971, Paper 24N9, Budapest,

pp. 461-464.
q
^George, A. R. and Plotkin, K. J., ''Propagation of Sonic
Booms and Other Weak Nonlinear Waves Through Turbulence , "
Physics of Fluids, Vol. I1* , No. 3, March, 1971, pp.
Plotkin, K. J. and George, A. R., "Propagation of Weak
Shock Waves Through Turbulence," Journal of Fluid Mechanics,
Vol. 5^, Part 3, August, 1972, pp. 449-467.
George, A. R., "Interaction of Sound and Sonic Booms with
Turbulence," Interagency Symposium of University Research in
Transportatio~Noise, 1973 , Stanford, pp. 288a-288g.

12
Pierce, A. D., "Statistical Theory of Atmospheric Turbulence
Effects on Sonic-Boom Rise Times," The Journal of the Acoust-
ical Society of America, Vol. ^9, No. 3, March, 1971, pp.
906-924.
Pierce, A. D. and Maglieri, D. J., "Effects of Atmospheric

Irregularities on Sonic-Boom Propagation," The Journal of th e
Acoustical Society of America, Vol. 51, No. 2, February, 1972,
pp. 702-721.
14
Ffowcs-Williams, J. E. and Howe, M. S., "On the Possibility
of Turbulent Thickening of Weak Shock Waves," Journal of Fluid
Mechanics, Vol. 58, Part 3, May, 1973, pp. ^6l-480.
Hodgson, J. P. and Johannesen, N. H., "Real-Gas Effects in
Very Weak Shock Waves in the Atmosphere and the Structure of
Sonic Bangs," Journal of Fluid Mechanics, Vol. 50, Part 1,
November, 1971, pp. 17-20.
Hodgson, J. P., "Vibrational Relaxation Effects in Weak
Shock Waves in Air and tne Structure of Sonic Bangs," Journal
of Fluid Mechanics, Vol. 58, Part 1, March, 1973, pp. 187-196.
17Teunissen, H. W. , "Characteristics of the Mean Wind and
Turbulence in the Planetary Boundary Layer," Rev. 32, 1970,
University of Toronto Institute for Aerospace Studies.
18
Maestrello, L., "UTIA Air Duct Facility for Investigation
of Vibration Noise Induced by Turbulent Flow Past a Panel
(Boundary Layer Noise)," TN 20, 1958, University of Toronto
Institute for Aerospace Studies.
19Schlicting, H., Boundary- Layer Theory, McGraw-Hill, New York,
1968.
SOUND SCATTERING FROM ATMOSPHERIC TURBULENCE
Ming Nan Huang*
Wyle Laboratories, Hampton, Va.

Abstract
Sound propagation through a turbulent flowfield with nonvanishing

mean-flow velocity has been studied. The effects of compressibility
on sound scattering are discussed. Using an analysis describing
moving sources in a fixed frame, particular results are obtained for
the far-field scattered sound intensity due to uniformly moving tur-
bulence confined in a finite scattering volume. These results show
the scattered sound intensity to be amplified by a factor proportional
to {l - M cos(9- 0) }~* due to the motion of the turbulence, where
M is the Mach number of the mean flow, 8 is the scattering angle,
and 0 is the launch angle of the initial acoustic beam. In addition,
this analysis was extended to include wave propagating through mean
flow with constant gradient.
Introduction
When an acoustic wave propagates through a moving inhomoge-
neous medium containing random velocity and temperature fluctu-
ations, the acoustic energy is scattered, and wave amplitude and
phase show random fluctuations. Theoretical descriptions of this
phenomenon have been derived by a number of investigators over the
past 30 years.
Presented as Paper 75-544 at the AIAA 2nd Aero-Acoustics Con-

ference, Hampton, Va., March 24-26, 1975. This work was sup-
ported by NASA-Langley Research Center under Contract NAS1-
12841.
Senior Research Engineer.
35
36 M. N. HUAN6
Using ideal and often unrealistic flow models, expressions for

amplitude and phase fluctuations (in the short-propagation-distance,
geometrical acoustics limit) and scattered sound intensity (in the
long-propagation-distance, Fraunhofer diffraction limit) have been
derived. The purpose of this work was to relax some of the physi-
cally unrealistic limitations of previous analyses of the Fraunhofer
diffraction case. The regions of short-and long-distance propagation
have been defined by the conditions (\r/L 2 )<^ 1, (Xr/L ) 1, respec-
tively, where \ is the wavelength of the unscattered wave, L is a
representative dimension of the scattering volume, and r is the
propagation distance of the scattered wave.
Recently, Clifford and Brown presented a sound scattering

theory for a turbulent medium moving with a mean velocity. Their
theory extended the work of Tatarski for the case of no mean motion
of the turbulent layer. In this work, one has a physical picture of a
scattering media within a finite scattering volume moving as a whole
with the mean wind. This analysis starts with a complete set of mass
and momentum equations for adiabatic propagation of an acoustic
wave in a moving, fluctuating, and turbulent medium. Each of the
turbulent fluctuating quantities is set into an appropriate order. An
inhomogeneous wave equation is derived with nonvanishing mean-flow
quantities.
Solutions are considered in two cases. First, when the mean

physical variables are constant, the sound-scattered pressure is
obtained by following Lighthill's concept** using the retarded time
formulation, a differential volume contributing to the scattering
integral is displaced from its position at time t to the earlier posi-
tion when the scattering process took place. The result is compa-
rable with that of Ref. 1. Besides the compressibility effects, an
additional multiplication factor 1 - M cos (9 -0) appears in the
scattered pressure. This factor shows significant contribution to
the scattered intensity. For isotropic, homogeneous turbulence, no
evidence is found that the compressibility can change the scattered
power. The second case considered is that of sound propagation
through a constant mean velocity gradient.
Derivation of Equations of Sound Waves

in the Turbulent Atmosphere
The motion of a fluid can be described by the balance equations

SOUND SCATTERING FROM TURBULENCE 37
of mass, momentum, and energy. Neglecting the effects of body

forces, these equations may be written as follows:
Continuity Equation
+ (S/dx.Mpu.) - 0 (1)
Momentum Equation
(c)/St)(pu.)
1 + (d/dx.)(pu.u.)
l = t.. . (2)
3 3 i 3 >3
Energy Equation
p ( d e / d t ) + u. (de/dx.)l = -X T .. + t..u. . (3)

w , l l Ij J , I
For a Stokesian fluid as air or water, the stress tensor may be

written as
t.. = -P + f k - (2/3)7] "| d . ) 6.. -f- 2ri d (4)

ij I o y oJ kkJ ij o ij
where
P(x.,t) density at time t and positions x.

l
u
i velocity
T thermodynamic temperature
internal energy, a function of p, and T
heat conductivity
stress tensor
thermodynamic pressure
bulk viscosity
shear viscosity
(u.
V . + u. . ) /2
'
There are six unknown variables (p, u., P, T), but Eqs. (1-3)
represent only five equations. [Equation (2) represents three equa-
tions for i = 1, 2, 3).] The sixth equation is the equation of state
for an ideal gas:
P = pRT (5)
38 M. N. HUANG
where R is the gas constant.
An alternative form of Eq. (3) can be written as
(l/R)(DP/Dt) - ( YP/pR)(Dp/Dt) = -(\JC )T .. +

O V , 11
X { I" k - (2/3)T) 1 u2 . + 2r| d..u. .} (6)

1- L o o J 1,1 o ij j, iJ
where Cp, C^ are the specific heats of air at constant pressure and
volume, respectively. ^^Cp/Cy is the ratio of the specific heats.
Equation (6) makes uses of the substantial derivative notation:
D/Dt = (d/dt) + u.(d/dx.) (7)
For the present analysis, viscosity and heat-conductivity effects

are neglected. Making use of the gas law in the form C^ = 7RT =
7 P/p, Eq. (6) becomes
DP/Dt = C 2 (Dp/Dt) (8)
Using Eq. (8) and the gas law to eliminate p from Eq. (1), the
modified continuity equation becomes
(DP/Dt)+ u.(dP/dx.) +7Pu. . = 0 (9)

i i i, i
and the momentum equation becomes
u.(&u./tex.) + (C 2 /rp)(dp/dx.) = o (10)

The speed of sound in Eq. (10) depends on the temperature. For
a given temperature distribution, Eqs. (9) and (10) are enough to
describe the flowfield.
The basic flow quantities can be written for the combined sound
and turbulent field in the following form:
p = p + p! + p", u. = u. + u! + uV
T = T + T T + T", P = P + PT + P" (11)
where an overbar denotes mean-flow values. Zero-mean turbulent

fluctuation quantities are denoted by single primes, and incident
sound-wave variables are represented by double primes. Let the
velocity turbulence amplitude be represented by a smaller parameter

&l= U{/C0 = Mj., the turbulent Mach number, and temperature tur-
bulence amplitude be represented by 62- Then the turbulent, fluctu-
ating quantities in Eq. (11) can be ordered such that
T'~T&2, P ' ~ P & (12)
where u is a random function with zero mean and rms value of 1.

i
'2
Since in the atmosphere P T ~ pu ( , the ratio of turbulent, fluctu-
ating pressure to the mean pressure is second order in turbulent
Mach number. Density fluctuations may result from both velocity
and temperature fluctuations independently of each other. Plotkin
showed that density fluctuations due to velocity variations are
second order in 61 . Those due to thermal turbulence are first order
in &2' In any given situation, it is difficult to determine which mech-
anism is producing the density perturbation. In this study, the
larger-order, thermal effect always is assumed. Pressure tur-
bulence cannot exist independently from velocity turbulence and is
characterized by M|. For the present case, where ii| ^ 0, pressure
turbulence also may be characterized by MM^-, where M is the Mach
number of the mean flow. This follows from computing (u^ + u| ) %.
An additional condition required for neglecting pressure turbulence is
thus that M be small.
For a turbulent field in the absence of sound waves, the balance

equations have the following form:
P f )/&t + (u. 4- u|)(P+ P') .

1 1 ,i
+ T (P + P')(u. + u!) . = 0 (13)
i i ,i
(u + u')(u + u f ) + C2 (1 + T T /T)
5 j i i ,j o
X ( P + Pf) , / y ( P + Pf) = 0 (14)

9 I
The equations of motion for a sound wave in a turbulent media

are obtained by subtracting Eqs. (13) and (14) from Eqs. (9) and (10),
in which the physical variables are replaced by Eq, (11), i. e.,
40 M. N. HUANG
dTT/dt + (u. + u |)TT . + u'.! P . + uf.T . = 0 (15)
SuM /dt + (u. + u!)u. . + u'Yu. . + u! ) = C2

i 3 J M J M i,J o
X (1 + T'/f )TT . - (C 2 / VP)(T"/T)P . (16)

, 1 0 ,i
2 _
where C =YRT, and 17 is defined as
o
TTs (l/y)(p"/p) (17)
In writing Eqs. (15) and (16), terms of second or higher orders

in 6 (where 6 is the larger of 6 and 6 ) and terms involving pro-
ducts of acoustic variables are neglectea. In addition to their de-
pendence on mean quantities, the coefficients of the acoustic vari-
ables in the preceding equations contain only turbulent velocity and
temperature variations and not pressure fluctuations. This results
because turbulent pressure fluctuations in the atmosphere are sec-
ond order in 6 .
This simplification of the acoustic coefficients is a direct re-

sult of using the modified continuity equation (9) rather than Eq, (1)
in the derivation of Eqs. (15) and (16). Had Eq. (1) been used, the
coefficients also would have contained terms in the density fluctu-
ations p f . The appearance of these extra terms would complicate the
solutions of Eqs. (15) and (16) for any real situation, because infor-
mation on density fluctuations would not be available either in real
time or in a spectral or correlation form. Information on velocity
and temperature fluctuation would be available from direct mea-
surements. Thus, assuming that the mean flowfield is given (either
from measurements or from solution of the mean-flow equations),
with boundary and initial conditions being known, Eqs. (15) and (16)
theoretically are solvable for the acoustic pressure and velocity.
Rather than attempt to solve this set of coupled differential equa-
tions, however, a more appropriate way to find a solution is to
derive an inhomogeneous wave equation for IT. This is done in the
next section.
Scattering of Sound by Turbulence
Applying d/dx. and D/Dt to Eqs. (15)and (16) and combining the
results, which reduce to
C2 IT - D 2 Tf/Dt 2 = -C2 (TTT /T) - 2u. .UT; . - u"u ..

o ,ii o ,1 ,1 1,3 j , i j i,ij
-(ufun + unu' ) + D (U T IT + u"P )/Dt - TT C

j i,j 3 i,j ,i i ,i i ,i ,i o ,i
- (C 2 T" P . /TPT) . (18)

o ,i ,i
where
D/Dt = (d/dt) + u (3/Sx ) (19)

Equation (18) is an inhomogeneous wave equation for sound propa-
gating through a turbulent anisotropic medium and accounts for
first-order interaction effects between the sound wave and the
flowfield. As such, it includes effects due to both scattering by
velocity and temperature turbulence, and refraction due to mean-
flow gradients. The difficulties involved in obtaining a complete
solution to this equation would be enormous, but a considerable
simplification is obtained if a uniform mean flowfield is assumed.
For laboratory-scale flows, this assumption may be invalid, but for
sound propagation in the lower atmosphere, it is very reasonable.
The resulting wave equation becomes
TT .. - (D2TT/Dt2)/C2 = -(T f TT ./T) . - f~ 2(uT.' .u!) .

ii o ,i i L- i J J i
+ u"u f 1 /C 2 (20)
J i,ijJ o
In writing Eq. (20), the results of assuming uniform mean flow and
neglecting higher-order terms of 6 in Eqs. (13) and (14) have been
used. Further simplification is achieved if the turbulent flowfield
is assumed incompressible (i.e., u? . = 0). In this analysis, the
term involved in u{ ^ is retained so that the eventual solution dis-
plays the effects of compressibility.
Using a Galilean coordinate transformation on Eq. (20), with
x. = x. - u.t, =t (21)
1 1 1 '
gives
42 M. N. HUANG
u"c>u'/3xc>x
... _ ... , _. , / d x i -f u"r (22)
J ^ 3 J i i jJ
To derive an expression for the scattered sound field, we expand
TT and u'/into series of the parameter 6, i. e. ,
1T= TT 4- 6TT -f 62TT + (23)

o x
u"= V + 6V + 62V + (24)

i 01 It 2i
Here, TT , V . correspond to wave propagation in the mean
flowfield, and fti , V- . are associated with the scattered wave due to
the turbulent flowfieicL To be consistent with the wave equation (22),
which is valid only to the first order of 6, terms containing the sec-
ond or higher order of 5 in Eqs. (23) and (24) are dropped. With
the neglect of higher-order terms, this method of perturbation be-
comes the single scattering or "Born" approximation.
Substituting the series expansions of TT and u[T into the wave

equation (22) and equating terms of equal order produces the zero-
and first-order wave equations:
c>2TT /3x 2 - (32TT A2)/C2 = 0 v(25)

;
o i o o
(26)
In the fixed coordinate system, the preceding equations reduce

to those derived by Tatarskir They also are obtained by substi-
tuting the series expansion of IT and u[f into Eqs. (15) and (16) (after
assuming that mean gradients are zero). Equating terms of equal
order in 6 yields
(STT /St) + V . . - 0 (27)

o o j, j
oi
/at) -f c2o (Siro/&xi > = o (28
+ V = -uf (SIT /ax ) (29)

11* 1 i o i
2 2
(3V (T f /T)TT + V u'
li o ,i oj i, j
+ uf V (30)
J 01, j
These equations may be used to relate pressure and velocity

in both the undisturbed wave and the scattered wave. This analysis
differs from Tatarski^ and others in that here all terms in Eq. (30)
are regarded as of equal order, whereas in previous work the right-
hand side of Eq. (30) vanish identically in the far field, where no
turbulent fluctuations occurred, and the scattered pressure is re-
lated to the acoustic velocity in the classical way by PC^. The same
result can be found by eliminating Tf from Eqs. (29) and (30) and
solving for V . in the far-field position.
The solution of the wave equation in this section is restricted

to the case as shown in Fig. 1. When a monochromatic plane wave
with an amplitude A, radian frequency ou , and wave number vector
K is forced to propagate into the atmosphere with an angle 0 from
the horizontal direction, then the incidence wave in the moving frame,
i. e. , the solution of the zero-order wave equation, is written as
Fig. 1 Sound scattering diagram, with turbulent fluctuations con-

fined in the scattering volume located at y > 0, which moves with
a velocity u in the horizontal direction.
44 M. N. HUANG
IT (x ,tV A expfif U) t - K x. ") 1 (31)

ov i / L \ o oii/J
where A is the amplitude, uu0 is the radian frequency, and K Qi is the

wave number component measured with respect to the moving coor-
dinates. In the fixed coordinates system, Eq. (31) is written as
TT (x.,t) = A expfiut) ( I + M cos0 f N )t - iK x 1 (32)

o\ i / L o\ / oi iJ
where M = M u. /C , and 0 T is the angle between u and
o
To see the relations between the wave number and frequency in
the fixed or moving frame, let
TT ( x.,
x.,0 -
= A eexp^
x p i o ) t -- i K .x
. .1 (33)
0\ I / ! O 01 iJ
Comparing Eqs. (32) and (33) yields
U) = (jo (1 + M cos0 f ) (34)
In terms of wave number K) measured in fixed coordinates, Eq. (34)

becomes
K = K /(I -f Mcos0 T ) (35)
Therefore, Eq. (33) becomes
TT o (x.,t) = A expf ioj t - iK e .x./(l + Mcos0 f Y](36)
i
where eA . = K
^ /
./K .
Ol Ol O
-a
The relation between KQ and K^ is derived by utilizing SnelPs
law and using the boundary condition, i. e. ,
K cos0 = K cos0T (37)

V;
0 0
Eliminating K and K from Eqs. (35) and (37) gives
0 = cos" |~ cos0f/(l - M cos0!)l (38)

Let % and ey be the unit vectors of the horizontal and^vertical

direction in the moving frame. Then the expression for K is
. 2 2
K = (K cos0) e + K sin0 1(1 -M cos0/sin 0)
0 0 X 0 I
4 12
- Mcos0/sin 0) e (39)
J y
where the value inside the braces must be nonnegative, i. e., 0 >sec"
(1 + M).
Substituting Eq. (31) into Eq. (28), the unscattered acoustic

velocity takes the form
V . ( , tWexpfif fitf-K x.^lc a , = C * e . (40)

01 \ i / L \ 01 i/ J o 01 o o 01 *
With these relations established, the source term in the first-

order wave equation is calculated and the solution for the scattered
wave written down in terms of the retarded time integral in the fixed
coordinate system:
TT^r.t) = (1/41T) C J Q [?,t- (| f -?'l/ c 0 ]/[ I? - ? T I
- M (f - r T )] | d 3 r T (41)
where Q(r,t) represents the terms on the right-hand side of Eq. (26).
The factor | r - F T | - M ( r - r T ) is due to changing from a
moving frame to a fixed frame.
Substituting appropriate expressions for IT and V . into

O1
Eq. (26) yields
11^,1;)= -(1/4TT) C J A expi[oi o t- K Q |f - f f
- M (r - r f )l \ x (sTiK ( T f / T + 2e u f /C )exp(-iK .x. r )l/&x T

J J I L 01 oj j o 01 i J i
- F 02u!/^x!dx!) e . exp(-iK .x!)l ) d 3 r' (42)

L i i j o j oiiJJ
For the far-field approximations |r | |r T |, the equations are

46 M. N. HUANG
|r - r f | - M (r - r')) |r - r f | x [1 - Mcos(9 -0)] (43)
r /\ /\
A
r - rf iK r - r f
A A A .<">
e o e o <L__ e
iK X
si i (44)
r - rr r - rr
where 9 is the angle between the incident wave K Q and the scattered
wave vector K g . Inserting expressions Eqs, (43) and (44) into Eq.
(42) and performing integration by parts in the limiting case of
large r and small scattering volume, the following equation is
obtained:
TT (f,t) = -{[ A exp i ( w Q t - K Q r ) ] /[ 47Tr(l - Mcos(9 - 0 ))]}
P r A A ,_ A / ^ AAA A2
f
X \ I (K K )(T /T + 2e .u!/C ;) + v(K . - K .)(K: .K .- K 7)
jv ! L v s M
O oj j oM
si OI oj sj o
/ -A- ~l r ^ A A ~ | SA
x u r /C 1C exp i(K
v - K )xM d r T (45)
i o oJ L si 01 iJ
Applying a similar analysis for the incident wave vector, the

relation between K and K has the form
s s
( 9 - 0 ) ) e - K sin(9 - 0){T 1 - Mcos(9-0)/sin

/ X S ll_
x (9 -0)] 2 - M2cos4(9 -0)/sin4(9 -0)| "' " e (46)
Equation (46) is valid for (9 -0) > sec" (1 + M).
From Eq. (17), the sound scattered pressure P and If are re-
lated b y s i
VPC <47>
Therefore, Eq. (45) gives the scattered pressure in the fixed
frame system, and the underlined term is due to the compressibility
of turbulent flow. By the definition of acoustic intensity as shown
in Tatarski,2 Eq. (48) is written:
I =(pCjV2)(Tn7*) (48)
The preceding equation is valid only when the receiver is far

away from the turbulent scattering volume. Since
T f (r, t) = \ di|r (i?) exp i| K r - ait]

J T L ..:
and
U!(?,t) = \ dt.(K) exp il K - r - oj"]
where dfc and d\|f. are the spectral densities of temperature and ve-
locity fluctuations, changing to the fixed coordinate and introducing

the correct retarded time (see Brown 6 ) gives
T T (r, t) = d\|r (K) exp
j T'
and
U!(r, t) = ^ dt.(K) exp i f K . r - K ut - uu(t - |r - r T | /C )~| (49)
Using the above relations, and inserting Eq. (45) into Eq. (48)
produces
I = { TT pC 3 A2V/[ 4r^l - Mcos(9 -0) } 1 } (K K )2
L o L \ / J J o s
X
[ n^ + ôiViJ^b] (50)
where K = | K - 1c I , V is the scattered turbulent volume, and .

A ^ S O ' 01
= ôi/ ô * ^ n ê P^^ceding equation, isotropic frozen turbulent flow
is assumed; i. e. , the temperature and velocity fluctuations are
statistically independent. Furthermore, there is no contribution to
the scattered intensity due to compressible turbulent flow. The power
spectra for the temperature and velocity in the inertial subrange have
the forms
2 1
TK (51)
$..(K) = (0. 76C 2 e 2/3 /4tt)(&.. - K.K./K 2 )K~ 11/3 (52)
where Crp is the temperature structure factor, e is the rate of

dissipation of turbulent energy, and C^ & 1. 9. The scattering cross
section is defined as the ratio of the energy scattered by a set of
48 M. N. HUANG
scatterers to the energy received bv the same set per unit volume.
That is a - tr2/! V, where V =fy d^r = V [1 - Mcos(9 -fl)]""1*.
_ 3 2
If the incident acoustic intensity is I0 = pC 0 A /2, the effective
scattering cross section per unit scattering volume per unit solid
angle becomes
a = 0. 03 KQ1/3[ (1 - IQ. ls)/2] -11/6(fo. l/[ 0. 13C2/T2
+ (1 + e e )C 2 e 2/3 /2C 2 1 (53)

x
O S O J '
where
e ' e = cos0 cos(0 - 0) - sin0 sin(0 - 0 ) ( 1 - Mcos0/sin20)2
- M cos 0/sin (0 -0)J [ (l - Mcos(9 -0)) 2 /sin 4 (0 -0)

2 4 4 " 1 ^ 1 /9x
- M cos (0 -0)/sin (0 -0)1 | (54)
In writing Eq. (54), Eqs. (39) and (46) are used.
For small Mach number M such that cos0 (1 - M) and cos

(0 .0) (1 - M), then Eq. (53) reduces to that of Clifford and
Brown?-
Sound Propagation in the Shear Mean Flow
As pointed out in the last section, the governing wave equation

(18) is difficult to solve due to the algebraic complexity. To avoid
these difficulties, the assumption of constant mean-flow velocity was
used in the last section. Here, one can relieve this strong and non-
realistic restriction slightly by letting the mean velocity u.(x.) be a
function of x and its gradient be constant. In this case, the zero-
order wave equation becomes
D2TT /Dt2 - C 2 ff .. = 2Q V _ (55)

1 o o 0,11 1,2 o2,11
We are interested in solutions of Eq. (55) of the form
From private communication with Dr. E. H. Brown.

v
oi(xi' *> = Gi<x2)exp i
where the amplitude F and G are related by Eq. (28), i. e. ,
dF/dx 2 = -il - Mieol (K o /C o )G 2 (58)
Substituting these expressions into Eq. (55) yields
dVdx222 + 2[ a i( 2 K 0 /C o (K o - MI K OI) ] dF/dx2

The solution of F with constant mean velocity gradient has been

discussed by Pridmore-Brown? His discussion is not repeated here.
G 2 can be found from Eq. (58) after F is solved. The scattered
wave has the form
D2 TT /Dt2 - C 2 TT .. = C 2 (T'TT
v ./f)7 . + 2!" v(uJV . .)7 .
1 o ,11 o o.i ,i L j 01, j ,1
+ u' 5 0TT 1 (60)
2 1, 2 o,iJ
The incompressibility of turbulent flow is assumed in the preceding
equation. Again using Eq. (41), the scattered pressure IT is ob-
tained by introducing expressions (56) and (57) into the right-hand
side of Eq. (60).
Discussion
The structure of atmospheric turbulence can be investigated by
using sound scattering as a tool. Consequently, the more precisely
the calculations of scattered sound intensity are obtained, the more
valuable the available information to find the turbulent character-
istics becomes. The principal objective of this analysis is to pro-
vide accurate prediction of scattered sound energy. Although the
scattering cross section in this work gives the same result as that of
Ref. 1, a different approaching method is used. The restrictions
on the incident-wave propagation angle and receiver angle in Ref. 1
are sin20 > > 2Mcos0 and sin2(9 -0)> > 2Mcos(0 -0). In this
analysis, the angles are less confining, i. e. , 0 > sec""*(l + M) and
and (9 -0) > sec -1 (1 + M).
The results of Eqs. (29) and (30) on the acoustic scattered

velocity provide a valuable step toward development of a scattering
50 M . ' N . HUANG
model when scattered sound energy is measured and recorded by a

receiver immersed in a turbulent region. In such a case, the
acoustic velocity is obtained by eliminating TT from Eqs. (29) and
(30). By taking the divergence, the reduced equation becomes
i I _
2 2 2
[2
L d(u! . . V .c) u!/^x l /C \ v(61)
;
3 01 3 i 03 j iJ o J
where $ = V . The preceding equation can be solved for $ by
using Eq. (41)' again. Finally, the acoustic velocity can then be de-
rived by using Poisson's integral equation.
When sound propagates in a mean flow with nonzero velocity

gradient, as expected, more energy will be scattered. Assuming
incompressible turbulent flow, Eq. (60) is identical to Eq. (26)
except for the additional term appearing on the right-hand side of
Eq. (60), because of the mean velocity gradient. A similar pro-
cedure can be carried out for the case including the mean temper-
ature gradient.
References
Clifford, S. F. , and Brown, E. H. , "Acoustic Scattering from a
Moving Turbulent Medium, " Journal of the Acoustical Society of
America , Vol 55, No. 5, May 1974, pp. 929-933.
2
Tatarski, V. I. , The Effects of the Turbulent Atmosphere on Wave
Propagation , National Technical Information Service, Springfield,
Va. , 1971.
3
Lighthill, M. J. , M On Sound Generated Aerodynamically, "
Proceedings of the Royal Society of London, Series A, Vol. 211,
March 1952, pp. 564-587.
4
Plotkin, K. J. , "The Effect of Atmospheric Inhomogeneities on
the Sonic Boom," Ph. D Thesis, 1971, Cornell University.
Pridmore-Brown, D. C. , "Sound Propagation in a Fluid Flowing

through an Attenuating Duct, " Journal of Fluid Mechanics , Vol. 4,
No. 4, August 1958, pp. 393-406.

Brown, E. H. , "Acoustic-Doppler-Radar Scattering Equation and

General Solution, " Journal of the Acoustical Society of America,
Vol. 52, No. 5, November 1972, pp. 1391-1396.
SCATTERING OF COHERENT SOUND WAVES BY ATMOSPHERIC TURBULENCE
P. L. Chow
Wayne State University, Detroit, Mich.
and
C. H. Liu and L. Maestrello'

NASA Langley Research Center, Hampton, Va.
Abstract
An analytical study of the propagation of coherent sound

waves through an atmosphere containing both mean and fluctuat-
ing flow variables is presented. The general flow problem is
formulated as a time-dependent wave propagation in a half-space
containing the turbulent medium. The coherent acoustic waves
are analyzed by a smoothing technique, assuming that mean flow
variables vary with the height only. The general equations for
the coherent waves are derived and then applied to two special
cases, corresponding to uniform and shear mean flow, respec-
tively. The results show that mean shear and turbulence intro-
duce pronounced effects on the propagation of coherent acoustic
disturbances.
Introduction
Wave propagation in turbulent media has been subjected to

intensive study by many workers in the past 20 years. Most are
concerned with radio wave propagation.1 Mathematical techni-
ques for treating such problems can be found in the expository
articles by Frisch^ and Chow.3 The scattering of acoustic
waves by turbulence has been investigated by Obukoff,^
Tatarski,^ Kraichnan,^ Monin,? and others.^ So far, authors
Presented as Paper 75-545 at the AIAA 2nd Aero-Acoustics

Conference, Hampton, Va., March 24-26, 1975.
Âssociate Professor of Mathematics.
Space Technologist, Acoustics and Noise Reduction
Division.
THead, Aeroacoustics Section, Noise Control Branch.
51
52 P. L. CHOW, C. H. LIU AND L. MAESTRELLO
treated the propagation of coherent waves in an inviscid turbu-

lent fluid, which was assumed to be homogeneous. The effect of
turbulent scattering on the propagation of sound waves has
attracted the attention of workers in aeroacoustics following
the experimental discovery of Maglieri,^ who observed large
fluctuations of sonic-boom waveforms. Sound propagation
through a turbulent jet flow was studied by Maestrello et al.10
The present paper is concerned with an analytical study of

the propagation of coherent sound waves through the atmosphere
containing both mean and fluctuating flow variables. The gen-
eral problem is formulated as a time-dependent wave propagation
due to a source in a half-space containing the turbulent med-

ium. The problem is governed mathematically by a system of
partial differential equations for the acoustical variables,
with the turbulent flow quantities appearing in the coeffi-
cients. The coherent acoustic waves are analyzed by means of a
smoothing perturbation technique (or a multiple scattering
theory), under the assumption that mean flow variables vary
with the height alone. The general equations for the coherent
waves are derived from the basic fluid-dynamics equation of
motion. These form a coupled system of integrodifferential
equations in which the coefficients involve the mean and covar-
iance functions for the flow variables. The significance of
studying the coherent waves is that, if one wishes to replace
the turbulent medium by an effective medium, the characteris-
tics of the coherent waves describe the properties of such a
medium. Besides, the expected value is the most essential
information to know for any random function. The incoherent
waves will be treated in our subsequent work.
Governing Equations
For acoustic wave propagation in a turbulent atmosphere,

the total flowfield (R, P, 0, U) consists of the acoustic field
_(p r , p f , 9 T , u r ) and the basic turbulent flowfield (p, p, 6,
u), so that
(R, P, 0, U) = (p, p, 0, u) + (p', p', 0', u') (1)
where R,U,P, and 0 denote the density, velocity, pressure,

and temperature, respectively. As usual, the acoustic quanti-
ties are taken as a small perturbation about the basic flow.
To derive the acoustic equations, it is necessary to assume
that the viscous and thermal dissipation terms are negligible
in comparison with the turbulent transfer of momentum and
energy. However, no such assumptions were made in the basic
flowfield. In view of Eq. (4) of Ref. 11, the energy equation
in 0 is uncoupled from the remaining three. Therefore, it is
SCATTERING OF COHERENT SOUND WAVES 53
found convenient to introduce the vector $ = (p, p, u) , where

$X is the transpose of $. Then, the acoustic equations can
be represented by a single equation in the matrix form:
, V) ' = 0 (2)
where
fu V 0 ypV \
, V) = 0 u V pV .
V-
\ V 0 u V/
/Y(V u) 0 (vP) A
0 V u (Vp)
V 0
7*
(vG) -y
and y is the ratio of constant specific heat.
Equation (1) constitutes a first-order system of partial

differential equations for the acoustic field $' with turbu-
lent flow variables as coefficients, whose statistics, e.g.,
means and covariances, are assumed to be given. In the follow
ing section, the mathematical problem described by Eq. (2) is
formulated.
Formulation of Problems
Let r = (x,y,z) = (p,z), and consider the acoustic Eq. (5)

in the half-space z > 0, with a source of disturbance F,
O$'/3t) + AOisV)^1 + B(<2>)T = F t > 0, z > 0 (3)
which is subject to the initial and boundary conditions:
(r); (t,p) (4)
For a reflecting boundary, the normal derivative B$ f /3z

will be given at z = 0. To construct the solution to the
problem (3) and (4), one needs to determine the Green's matrix
G, which satisfies the following adjoining equation:
-(9G+/9t) +A*($,V)G++B*($)G+ = 6(t-s)6(r"-r')I t

'S >
, >n
Z, Z ^ U
(5)
and the homogeneous conditions
_ = 0; 3G+/3z z=0. = 0 (6)

t=0n
where A and B are the adjoint operators of A and B,6 is

the Dirac delta function, and I stands for an identity matrix.
In terms of G+, the solution $ f can be expressed as
$f(t,r) = /Z!>Q G+(t,0;r",r')<2>o(r"') dr'
+
f n G (t,s;r,p ' )<!> M(p') ds dp'
J/n J/
0 z -0 ^ 1 ^
+
/O ^z'>0 G4"(t,s,r,r')F(s,r') ds dr' (7)
Furthermore, it is known that the half-space Green's func-

tion G+ is obtainable from the full-space Green's function G
by the method of images. It is necessary, therefore, to extend
the domain of the functions $ and F symmetrically about the
plane z = 0, and solve the following problem:
- (3G/9t) + A*($,V)G + B*()G
= 6(t-s)6(r-r'), t , s > 0 , |r|, |r'| < ~
t=0 ~ ^
where G is required to be bounded as |r| , |r'| -> . Then

G"^" is related to G by the simple formula
G+(t,x; r,r') = G(t,T; r,r') - G(t,T; r,^') (9)
where r-^1 is the image point of r' with respect to the

plane z = 0. Therefore, to study the problem (3) and (4), it
suffices to treat the simpler problem (8).
Since $ represents the turbulent state, it is a random

vector field. Hence, the solution $ T to Eq. (3) is a random
solution. The interest is in seeking the mean or average
solution f >, where the angular bracket denotes the statisti-
cal average of 0 f . When $T is interpreted as an acoustic
wave, its mean <$ T > is known as a coherent acoustic wave.
Although other statistics, such as the covariance, of the field
$! are also important, the work of this paper is confined only
to the mean waves. In the sequel, an approximate integro-
differential equation for the coherent waves will be derived by
the method of smoothing3 (or a multiple scattering theory) .
The basic turbulent flowfield is decomposed into the

mean <$> and the fluctuating part , so that
(10)
Byâssumption, the mean <$> and the covariance matrix

R = |> <>,p> are known. The explicit form of R is given in
the Appendix. Now let
L = (3/8t) + A(I, V) + B(2) (11)
V = A($,V) + B S ) (12)
where the operators A and B are defined by Eqs. (3) and (4),
respectively. A^($) is similar to A($) except for changing
the element (l/p)V to (-p/2) , and B^($) is the same as
B($) with the element (-l/p2)Vp replaced by (2p/3)V
- (l/2)Vp. Then an application of the smoothing method
yields the following matrix integrodifferential equation:
<$'> - <VL"1V> (13)
In view of (11) and (12), Eq. (13) can be written more explic-
itly as
{9/3t (t,r)
K(R,V)(t,s;r,r') (s,r') ds dr' = F (14)

where the kernal K involves the covariance R and the

Green's matrix G0 in the following manner:
K(R,V)(t,s;r,r') = <A($,V) (t,r)GQ(t,s;r ,r ')A( ,V) (s,r ' )>
+ <A($,V)(t,r")Go(t,s;r,r"')B($)(s,r"')>
+ <B($)(t,r)Go(t,s;r",r')A($,V)(s,r"')>
+ <B($)(t,r)Go(t,s;r,r')B(i)(s,r')> (16)
Here the average quantities can be computed in terms of the

elements of R as shown in the Appendix of Ref . 11, and the

Green's matrix Go satisfies the adjoint problem
- (3/3t) Go + A*(2,V) Go + B*(X<2) Go
= 6(t - s) 6(r - rT)I, t,s > 0, z,z! > 0
= G = 0 (17)
t=0 z=0
The symbols A*, B* designate the adjoint operators of A
and B, respectively. As demonstrated previously, the half-
space problem [Eqs. (3) and (4)] can be reduced to a full-space
problem [Eq. (8)]. For each elemented solution, the averaged
problem [Eq. (7)] in a half-space is also reducible to a full-
space problem by a symmetrical extension of the functions A,
B, F, R, G0 about z = 0 and the method of images. For the
sake of brevity, however, the details are not presented.
Applications to Atmospheric Scattering Problems
Theory developed in the previous section can be used to

study the propagation of coherent waves in the atmospheric
boundary layer or outside the boundary, where the flow is nearly
uniform. Assume that the mean wind profile and the thermody-
namic variables are functions of height z only, and the tur-
bulence is stationary and horizontally homogeneous and isotopic.
Also, it is assumed that the mean velocity has only a horizon-
tal component, say, in the x direction. Then we have
 = P 0 ( z ) > = P 0 < z ) > <u

\ =
(uQ(z),0,0) (18)
and
R ( t , s ; r , r f ) = R(t - s; |p|- p ! | , z , z ' ) (19)

Now, in view of (18), the coefficients of Eq. (17) depend on

the variable z alone, and hence, the Green1s matrix Go is
stationary and transversely homogeneous and isotropic, that is,
Go(t,s;r,rf) = GQ(t - s; |p - p'|, z,zf) (20)
Equations (19) and (20) imply that the kernel K must be of

the form
K(t,s;r,r') = K(t - s; |p - p'| , z,z?)

It follows from (18-20) that the coefficients of Eq. (14)

depend only on z, and the integral there is of a convolutional
type. Therefore, to study the coherent wave, the Fourier
transform can be applied to the problem (14) and (15) to reduce
it to a one-dimensional problem in z. For any smooth function
F of t and p, its Fourier transform can be denoted by f,
which is
f (A,k) = /Q /f (t,p) exp UAt - ik p} dt dp (21)
In the preceding integral, the integration in p is over
An application of Fourier transform to (14) and (15)

yields the following transformed problem:
- IX
(A,k,zT) dzf = F(A,k,z) + Q(k,z)
(22)
z=0
which may be solved more readily. Then the Fourier inversion

of $ f gives the wave field <2>f in the physical domain. By
the same token, the Green's matrix Go can be determined by
this technique.
For a time-harmonic source F = Fo(r)e~"1^t and a homoge-

neous boundary condition, a time-harmonic wave solution,
<$ T > = ^/(r)e""i^t, to Eq. (14) can be sought. In this case, the
variables t and r in Eq. (14) can be separated to obtain

the time-reduced equation for Y :
- /Q JK(T,|P - p'| ,z,zf)exp(iîr) ^(rT) di dr ? = FQ(r)
(23)
where use has been made of (21) . An equation similar to (22)
can be obtained after a Fourier transform in p.
The realistic mean flow profile and other thermodynamic

quantities as well as the structure of turbulence in the atmos-

pheric boundary layer are discussed in the book by Panof sky and
Lumleyl2 and other sources (see, e.g., Ref . 13). For the pur-
pose of illustration, two relatively simple problems are worked
out, although the theory can be applied to much more general
situations.
Special Cases
Case 1: Uniform Mean Flow
When the coherent waves propagate outside the boundary

layer, the mean flow is assumed to be uniform, and the problem
is treated in an unbounded space. This problem has been
examined carefully in Ref 11. For brevity, a detailed
analysis will not be carried out here.
Case 2; Shear Mean Flow
Here, the same assumption is made as in the previous case

except that the mean flow profile is nonuniform (Fig. 1). By
similar steps, which lead to Eqs. (41-43) of Ref. 11, the fol-
lowing equation for in the absence of continuous source
distributions results:

C- 0), r - r f ) R(r - r') V 1 (r"')dr"f = 0

(24)
u (z) - + V
z p
- 2 /[Vi_(- 03, r - r 1 ) : R(r - r!)] V! (r ' )d?' =0

33
( 2 5)
-UQ (z)- -EARTH BOUNDARY LAYER
TURBULENCE
Fig. 1 Coordinate system and geometry of the application.
where, in contrast with Eq. (38) of Ref. 11, gy are elements

of the Green matrix g0(-oo?r) corresponding to the shear mean
flow. Since the coefficients and gjM in Eqs. (24) and (25)
depend on uo(z), analytical solutions are inaccessible.
Instead, a numerical integration will be employed. To simplify
the computation, only the effects of the mean flow uo and the
vertical velocity fluctuations uz on the propagation of the
coherent wave are examined. Then the correlation func-
tion_ Rzz(r - r f ) = <uz(r)uz(rt)> is the only nonzero element
in R. Consequently, Eqs. (24) and (25) are reduced to
1- yp V
dz o
- 3/R (r - r') $- (r') dr' = 0 (26)

ZZ dz
- V
dz
- 2 r}
3P" 0 (27)
where
d/dz = u (z) (3/3z) - io

o
Applying the integrodifferential operator

60 P. L. CHOW, C. H. LID AND L. MAESTRELLO
8
1- - Z9j fp L 8 7
dz zz 3z 55 3I
to Eq. (26) and taking Eq. (27) into account, the following
approximate equation for is obtained:
!% c V < Pr> - i dz
, 2 + o
L /JRzz 3z
a g 11 3az T <rp > d ; .
dz
2
- /R & ^ d? ' = (28)
In the derivation of Eq. (28), use is made of the fact that the
flow variables are slowly varying and of integration by parts.
As a numerical example, a homogeneous boundary condition

is imposed on z = 0, and the acoustic pressure is maintained
constant on the plane x = 0, so that
(x,0) = 0 (0,z) = 1 (29)
Also, the correlation function Rzz and the velocity profile

are taken to be of the form
R (r) = a2 exp {- ar} (30)

zz
u (z)/c = 3 In (1 + z), 0 < z < 10

~ ~ (31)
= B In (11), z _> 10
Finally, expressions (A8) and (A13) of Ref. 11 for gn and

g55 are used in Eq. (28), for simplicity, where u0 is taken
to be (1/2) c0 3 In 11.
A numerical integration of Eq. (28) subject to condition

(29) was carried out. The results for the amplitude A and
phase (j) of are displayed in Figs. 2-7, where A, <j) are
defined as
(r) = A(r) exp {i$(r)} (32)
Discussion and Numerical Results
Because of the complexity of the problem in the case of

shear flow, the results are obtained by numerical method. For
the case in question (case 2) , the turbulence is transversely
homogeneous and isotropic. The Fourier transform technique has
been used to reduce the system to a set of ordinary integro-
differential equations in the vertical plane alone.
2r
A= 1/2
Z=
a = l, (3 = 0.01, k = 10, o = 0
0 10
x.
Fig. 2 Amplitude of wave in shear flow.
2r
ko = 1
A=
a/2
a = l, (3 = 0.1, k = 10,
0 10
x .
Fig. 3 Amplitude of wave in shear flow.
2r
AA = (pp
/ *
)
a = l, (3 = 0.01, k = 10
0 1 2 3 4 5
x
Fig. 4 Amplitude of wave in shear flow with turbulence.
2r
ko = 1
A = (p'ff*)1/2
a = l, 3 = 0.1, k = 10,
0 10
Fig. 5 Amplitude of wave in shear flow with turbulence.
0- kx
(rad)
a = l, (3 = 0.01, k = 10, 0 =
-3
2 4 6 8 10
x
Fig. 6 Deviation of phase from plane wave in shear flow.
ka T
z=3
0 = 1, p = o.01, k = 10,
0 - kx z=2
(rad)
0 2 4 6 8 10
x
Fig. 7 Deviation of phase from plane wave in shear flow with
turbulence.
For the purpose of comparison, the amplitude for the
acoustic wave in the mean flow is plotted in Figs. 2 and 3 for
3 = 0.01 and 0.1, respectively. The effects of turbulence on
the wave amplitude A at different heights are shown in
Figs. 4 and 5. Near the boundary, there exists a strongly
damped layer because of the attenuation effect by turbulent
scattering. Away from the wall, the wall intensity is ampli-
fied by the mean flow and reaches a maximum and then decays
along the downstream as a consequence of random scattering,
which also shifts the points of maxima along the curves. In
Figs. 6 and 7, the deviations of phase from that of the plane
wave e - are shown at various altitudes. In contrast with

the mean flow case, the turbulent scattering introduces phase
oscillations for the mean wave. As seen from these figures,
the combination of turbulence and mean flow can produce some
pronounced effects on a propagating wave. Therefore, these
factors should not be ignored in any realistic modeling of
atmospheric propagation problems.
Concluding Remarks
In contrast with previous works on this subject, this

paper presents a more general theory, which takes into account
the fact that atmospheric turbulence is confined to a half-

space. The theory is applicable to various problems, such as
a fixed source, a moving source, or a beam of acoustic waves.
Applications of incoherent scattering for the acoustic beam
problem are now under investigation. The results may suggest
experimental measurements to be performed in connection with
atmospheric scattering problems. The general theory requires
the knowledge of all elements in the covariance matrix which
are not all measurable physically. In practice, simplification
of the general equation depending on the physical situation is
necessary to make the theory useful, as shown by the examples.
In each case, the result yields a qualitative picture portray-
ing an effective feature of the scattered waves. A general
correlation theory of the sound wave propagation in turbulent
medium also can be developed. However, the complexity in the
correlation equations, which are not presented here, makes its
practical applications very difficult if not impossible.
Appendix
Correlation Matrix R
By definition, $ = (p,p,u), and the correlation matrix
/R R R -\
/ P1P2 P P
1 2 P
1 U 2\
R R R - (Al)
P P P
12 1P2 P
1U2
\ R-
K- R- R- -
where
are correlation vectors, and RU-IUO ^-s tîe velocity correlation

matrix, and so on. Since some element of R may be a matrix,

R is actually a 5 x 5 matrix with scalar elements. In partic-
ular, R-- is a 3 x 3 symmetric matrix of the form
R
xz A\
R
/ XX xy
R R R R (A2)
uu yx yy yz
\ R R R
\ zx zy zz //
wherein R = , R = , with u_ = (u ,u ,u ) .

xx x x xy x y T x y' z
References
Barabanenkov, Y. N., Rytov, Y. A., and Tatarski, V. I.,

"Status of the Thoery of Propagation of Waves in a Randomly
Inhomogeneous Medium," Soviet Physics Uspekhi, Vol. 13, No. 5,
March-April 1971, pp. 551-680.
2
Frisch, U., "Wave Propagation in Random Media," Probablistic
Methods in Applied Mathematics, edited by A. T. Bharucha-Reid,
Academic Press, New York, 1968, Vol. 1, pp. 551-575.
3
Chow, P. L., "Perturbation Methods in Stochastic Wave Propaga-
tion," SIAM Review, Vol. 17, No. 1, January 1975, pp. 57-81.
Obukoff, A. M., "On the Scattering of Sound in a Turbulent

Flow," Doklady Akademii Nauk, SSSR, Vol. 30, 1941, p. 611.
Tatarski, V. I., Wave Propagation in a Turbulent Medium,

Translated by R. A. Silverman, Dover, New York, 1967.
Kraichnan, R. H., "The Scattering of Sound in a Turbulent

Medium,11 The Journal of The Acoustical Society of America,
Vol. 25, No. 6, November 1953, pp. 1096-1104.
Monin, A. S., "Characteristics of the Scattering of Sound in a

Turbulent Atmosphere," Soviet Physics Acoustics, Vol. 7, No. 4,
April-June 1962, pp. 370-373.
O
Wenzel, A. R. and Keller, J. B., "Propagation of Acoustic
Waves in a Turbulent Medium," The Journal of The Acoustical
Society of America, Vol. 50, No. 4, Part 1, October 1971,
pp. 911-920.
9
Maglieri, D. J., "Some Effect of Airplane Operations and
Atmosphere on Sonic Boom Signature," The Journal of The Acous-
tical Society of America, Vol. 39, No. 3, March 1966, pp. 536-
542.
Maestrello, L., Liu, C. H., Gunzburger, M., and Ting, L.,

"Sound Propagation Through a Real Jet Flow Field With Scatter-
ing Due to Interaction With Turbulence," AIAA Paper 74-551,
Presented at the 7th Fluid and Plasma Dynamics Conference,
Palo Alto, California, June 1974.
Chow, P. L., Liu, C. H., and Maestrello, L., "Scattering of

Coherent Sound Waves by Atmospheric Turbulence," AIAA Paper
75-545, Presented at the 2nd AIAA Aeroacoustic Conference,
Hampton, Virginia, March 1975.
12
Lumley, J. and Panofsky, H., The Structure of Atmospheric
Turbulence, Wiley, New York, 1964.
Monin, A. S., "The Structure of Atmospheric Turbulence,"

Theory of Probability and its Applications, Vol. 3, 1958,
pp. 266-296.
SATURATION EFFECTS ASSOCIATED WITH SOUND
PROPAGATION IN A TURBULENT MEDIUM
Alan R. Wenzel
NASA Ames Research Center,

Moffett Field, Calif.
Abstract
A theoretical analysis of the acoustic wave field

radiated by a time-harmonic point source in a homogeneous,
isotropic turbulent medium is presented. The smoothing method
is used to study the incoherent, or randomly fluctuating, com-
ponent of the wave field. The analysis considers the effect on
the wave of the velocity fluctuations, as well as the index-of-
refraction fluctuations, of the medium. An approximate expres-
sion for the second moment of the incoherent wave is obtained
for the case in which the wavelength is much less than the
minimum correlation length of the medium. This expression
shows that the fluctuations of the wave increase initially in
proportion to the propagation distance, but that at larger
distances they tend to a limiting, or saturation, value. These
results agree with observations of waves propagating in real
media. It also is found that the mean square of the total
(i.e., coherent plus incoherent) acoustic pressure is unaffect-
ed by the randomness of the medium.
Introduction
When a wave propagates through a random medium, it is

continually distorted as a result of diffraction and scatter-

Conference, Hampton, Virginia", March 24-26, 1975. This
research was supported by the National Research Council, and
was done while the author held an NRC Resident Research
Associateship at NASA, Ames Research Center.
*Research Scientist
67
68 A. R. WENZEL
ing by the inhomogeneities of the medium. As a consequence, it

develops random fluctuations in its amplitude, phase, inten-
sity, and other parameters. This effect is cumulative, i.e.,
the fluctuations of the wave generally increase with propaga-
tion distance. It is therefore possible that a wave, after
propagating a sufficient distance, may exhibit large fluctua-
tions despite having propagated in a medium which is only
slightly inhomogeneous. An example is the twinkling of stars,
in which the fluctuations in the intensity of the light are of
the order of the mean intensity, whereas the fluctuations in
the atmospheric optical index of refraction are typically of
the order of one part in one million.
Provided that the medium is only weakly inhomogeneous and

that the wave has propagated such a short distance that the
fluctuations in its parameters are at most a small fraction of
their respective mean values, the phenomenon can be analyzed
adequately by conventional perturbation techniques such as the
Born approximation. For these propagation distances, both
theory and experiment show that the fluctuations of the wave
increase as some power of the distance. At larger distances,
however, observations indicate that the fluctuations approach
a limiting, or saturation, value.1>2 in this so-called satu-
ration region, conventional perturbation solutions are no
longer valid. This is because they contain secular terms and
hence predict an indefinite growth of the fluctuations with
distance.
To study propagation in the saturated region, improved

perturbation methods must be used. Among those which have
been tried with some success are modified geometrical
optics,-^-'3 selective summation of diagrams,^>^ and the smooth-
ing method. Although these approaches differ considerably in
detail, they all account, to some extent, for higher-order
terms in the perturbation expansion of the solution and thus
yield results that are more accurate than those obtained with
first-order methods.
The saturation phenomenon is particularly important in

atmospheric acoustics since, under typical experimental con-
ditions, sound waves in the atmosphere may saturate in a
relatively short distance. For example, Embleton et al.2
measured fluctuations of acoustic signals propagating over
ground and found that, for frequencies of from 150 to 5000 Hz,
saturation occurred at propagation distances of between 150
and 400 ft, with the shorter distances corresponding to the
higher frequencies.
SOUND PROPAGATION IN A TURBULENT MEDIUM 69
The approach described here is based on the smoothing

method, which has been discussed by Frisch.' In this formula-
tion, the total wave field is expressed as the sum of a coher-
ent, or mean, wave and an incoherent, or randomly fluctuating,
wave. The present study is concerned primarily with the
incoherent wave.
Basic Equations
The starting point of the analysis is Eq. 60 of Ref. 8,

which is written here with a nonzero source term in the form
(c~2D2 - V2)p = g . (1)
Equation 1 is a convected wave equation which governs the

propagation of acoustic disturbances in a moving, inhomogen-
eous fluid medium. In this equation, p is the acoustic pres-
sure, g is the acoustic source term, and c is the sound
speed of the medium. Also, Dt = 8t + u V, where u [=(ui,
U
2>U3)] is t*ie fluid velocity, and V = (81,82*93)- Since the
basic flow is assumed here to be turbulent, both c and u
are to be regarded as random functions of position r [=(ri,r2,
r^)] and time t. (The source term g is assumed to be non-
random.)
Note that, when u = 0, Eq. 1 is just the ordinary scalar

wave equation. As a consequence, the corresponding results for
the case in which propagation is governed by that equation can
be obtained as a special case of the present results simply by
setting u = 0 throughout the analysis.
Equation 1 is valid provided that both the characteristic

time and length scales of the acoustic disturbance are much
smaller than those associated with the basic flow. In the
following analysis, we shall assume not only that these con-
ditions are satisfied, but also that the basic flow varies so
slowly with time that it can be regarded as time-independent.
Thus we write c = c(r) and u = u(r). This allows us to
consider time-harmonic solutions of Eq. 1, i.e.,
g(r,t) = f(r)exp(-iu>t) , (2)
p(r,t) = q(r)exp(-iu>t) , (3)
with a) > 0. Upon substituting Eqs. 2 and 3 into Eq. 1 and

expanding the operator Dt, we find that q must satisfy
L2 , -
2io)u j 8 j - u i u j 8 i 8j|q = -f . (4)
70 A. R. WENZEL
In accordance with the assumptions previously mentioned, we

have neglected derivatives of basic flow quantities in deriv-
ing Eq. 4.
We now assume that the basic flow represents a small per-

turbation of a uniform fluid at rest. This allows Eq. 4 to be
solved by a perturbation method. Accordingly, we write
c(r) = cQ[l +' ey(r)] , (5)
u(r) = ecQu(r) , (6)

where y(r) and u(r) are dimensionless random functions

with zero mean, CQ is the average sound speed of the medium,
and e is a small parameter measuring the deviation of the
medium from a uniform, motionless state. Inserting Eqs . 5 and
6 into Eq. 4 and expanding in powers of e yields
JL 0 + el^ + A2 + 0(e3)]q = -f , (7)
where the operators LQ, L-^, and L.2 are given by
2 2
L = V +k ,
L
i = -2(kou -lkoVj)
and kg = (JO/CQ.
Keller^ has shown that the ensemble-averaged solution of

Eq. 7, i.e., the coherent wave, satisfies
} + 2(<L2> - <^1LQ1L1>) + 0(e3)] <q> = -f . (8)
(The angular brackets denote an ensemble average.) The corres-

ponding incoherent wave q (= q - <q>), is determined by the
relation
q - -eL~1L1<q> 4- 0(e2) . (9)
In deriving Eqs. 8 and 9, it has been assumed that <L^> = 0,

which is the case here.
The procedure by which Eqs. 8 and 9 are obtained from Eq.

7 is referred to as the smoothing method by Frisch.^ This
method has been used previously to study coherent waves in
various kinds of random media. 8-10 it also has been used to

study incoherent scalar waves. 6
Coherent Wave
We wish to calculate the second moment of the incoherent

wave, using Eq. 9, for the case in which the source term f is
a delta function and the medium is statistically homogeneous
and isotropic. To do this, we need the solution of Eq. 8 for
the coherent wave for this case since that quantity appears in
Eq. 9. That solution was obtained previously H and can be
written, in the present notation, as
<q(r)> = l 4- o e k Q + 2 ( 4 7 r r ) " " e x p ( i k 1 r ) , (10)
where
kx = kQ[l + 2e2(v2 + v2)] + iA2^2^ + v22) . (11)
Here we have introduced the notation v9z = <y 9> and

v2 = u2^>=: <(uv>- Also, the correlation lengths &i
are
and &2 defined by
defined by
CO
in = p(s)ds ,
1
0
, = r t(s)ds ,
/
where the normalized correlation functions p and T are

given by
r\
v p(s) =
V2
l(6ij "s"2sisj)ST'(s)
(The prime denotes differentiation.) In introducing this nota-
tion, we have assumed that the turbulence velocity field is
incompressible as well as isotropic. ^-2 The function T(S) is
the normalized longitudinal velocity correlation of the turbu-
lence. In the derivation of Eqs. 10 and 11, terms of order
e^, as well as higher-order terms in (kQ&i)~l and (kQ^)"-'-*
were neglected.
Equation 10 shows that the coherent wave has the form of

an outgoing spherical wave. Equation 11 shows that Re k^ > k0
and also that Im ki > 0. Thus, the randomness of the medium
72 A. R. WENZEL
results in both a reduction in the propagation speed and an

attenuation (in addition to the attenuation caused by spheri-
cal spreading) of the coherent wave.
In the following analysis, we neglect the slight reduc-

tion in propagation speed of the coherent wave. We also
assume that the term of order ^^(^ + ^2) in Eq* ^ can be
neglected. Then Eq. 10 can be written as
> = (4^r)"1exp(ik1r) , (12)
where now
kQ + ia (13)
and
e\^
* ^-n (o^
I^ î + vv^
-ô /) C14}
v-1
-^/
Incoherent Wave
An expression for the incoherent wave can now be obtained

by substituting Eq. 12 into Eq. 9 and using the known formula
for L!. This yields
q(r) -
+(r!/rt)u.(rt)](rT)~1exp(-art)drt . (15)
Here the integral is taken over all of three-dimensional space.

In deriving Eq. 15, we have neglected higher -order terms in
k01 as well as some terms of order e . Also, we have sub-
stituted for k-j^ from Eq. 13,
We can simplify Eq. 15 by using the method of stationary

phase with kg as the large parameter. It is convenient first
to let r = (0,0, r) (this entails no loss 0f generality since
the medium is isotropic) with r > 0. Then, by applying the
two-dimensional stationary-phase method-^ to the integral over
rj^ and r in Eq. 15, we obtain
00
q(r) -(iek0/4ir) f (|r^-r| + |r | )~1[p(0,0,r^)
|r-r| + jr|)]dr. (16)

(Although a is proportional to kjj, it was not assumed large

in the derivation of Eq. 16 since it is also proportional to
Since kg has been assumed to be large, the function

exp[iko(|r3-r|+ r^j)] will oscillate rapidly for r^ < 0 and
for r^ > r, compared to the other terms in the integrand in
Eq. 16. As a consequence, the contribution from these ranges
of r^ to the integral of Eq. 16 may be neglected. That
equation can then be written
r
q(r) ^ -iek^(47Tr) exp(ik.r) / [y(0,0,r')

^0
(17)
The procedure by which Eq. 17 is obtained from Eq. 15 is

equivalent to a small-angle scattering approximation and is
valid "provided that kg^ 2 >> 1 an^ r/kg&J 2 << !
An expression for the second moment of the incoherent

wave can be obtained now by squaring the absolute value of
both sides of Eq. 17 and averaging. This yields
2k2(4Trr)~2 f lfv2p(|r' - r"

J
|)
^0
xp[-a(r^+rp]dr^dr^ . (18)
(In deriving Eq. 18 we have made use of the fact that correla-
tions between scalar and velocity fields vanish for isotropic,
incompressible turbulence.) The double integral in Eq. 18 can
be evaluated approximately by making the coordinate transfor-
mation 5 = r^ + r^1, n = r^ - r^', and noting that, for
r >>
^1 2> tîe resulting n integration can be extended to
infinity without introducing significant error. The trans-
formed integral is then easily evaluated, after which we obtain
|q(r)|2= (4Trr)"2[l - exp(-2ar)] . (19)
When ar 1, Eq. 19 can be written as
" 2 2ar . (20)
This result is the same as that obtained from the Born approx-
imation and shows that, when the propagation distance is small,
74 A. R. WENZEL
the mean square fluctuations of the wave increase (apart from

the spherical spreading term) in proportion to the propagation
distance. At larger distances, however, the fluctuations tend to
a saturation value, as shown by Eq. 19. Our results are thus in
agreement with observations of waves propagating in real media.
Equations 12 and 19 yield the relation
<|q|2>= |<q>|2 +<|q|2>= (4^r)~2 , (21)

which shows that the total mean square acoustic pressure is
unaffected by the turbulence. In this sense, then> there is no

overall attenuation of the acoustic field by the turbulence.
We can use the above results to obtain an order-

of-magnitude estimate of the saturation distance (i.e., the
propagation distance over which the wave becomes effectively
saturated) for acoustic propagation in the atmosphere. Follow
ing Plotkin and George, we take, for this calculation,
e ~ 10~3 and ^ ~ &2 ~ 1 m > as being typical of daytime con-
ditions in the atmospheric boundary layer . If we now define
the saturation distance rg by writing arg = 1, then Eq. 14
yields, for a 1-kHz wave, rs ~ 100 m. This estimate compares
favorably with the experimental results of Embleton et al.^
(as discussed in the Introduction) .
References
Clifford, S. F., Ochs, G., and Lawrence, R., "Saturation of

Optical Scintillation by Strong Turbulence," Journal of the
Optical Society of America, Vol. 64, Feb. 1974, pp. 148-154.
2
Embleton, T. F. W., Olson, N., Piercy, J., and Rollin, D.,
"Fluctuations in the Propagation of Sound near the Ground,"
The Journal of the Acoustical Society of America^ Vol. 55,
Feb. 1974, p. 485 (abstract).
3
Yura, H. T., "Physical Model for Strong Optical - Amplitude
Fluctuations in a Turbulent Medium," Journal of the Optical
Society of America^ Vol. 64, Jan. 1974, pp. 59-67.
4
de Wolf, D. A., "Saturation of Irradiance Fluctuations Due to
Turbulent Atmosphere," Journal of the Optical Society of Amer-
ica3 Vol. 58, April 1968, pp. 461-466.
5
Brown, W. P., Jr., "Fourth Moment of a Wave Propagating in a
Random Medium," Journal of the Optical Society of America,
Vol. 62, Aug. 1972, pp. 966-971.
6
Sancer, M. I. and Varvatsis, A., "Saturation Calculation for
Light Propagation in the Turbulent Atmosphere," Journal of the
Optical Society of America, Vol. 60, May 1970, pp. 654-659.
7
Frisch, U., "Wave Propagation in Random Media," in Probabil-
istic Methods in Applied Mathematics,, A. T. Bharucha-Reid,
editor, Academic Press, New York, 1968, p. 114.
8
Wenzel, A. R. and Keller, J., "Propagation of Acoustic Waves
in a Turbulent Medium," The Journal of the Acoustical Society
of America3 Vol. 50, Sept. 1971, pt. 2, pp. 911-920.
^Keller, J. B., "Stochastic Equations and Wave Propagation in

Random Media," Proceedings of the Symposium on Applied Mathe-
matics, Vol. 16, 1964, pp. 145-170.
10
Karal, F. C. and Keller, J. B., "Elastic, Electromagnetic,
and Other Waves in a Random Medium," Journal of Mathematical
Physics, Vol. 5, April 1964, pp. 537-549.
l:L
Wenzel, A. R., "Propagation of Acoustic Waves in a Turbulent
Medium," Ph.D. dissertation, New York Univ., New York, June
1970, Eq. 5.14.
12
Batchelor, G. K., The Theory of Homogeneous Turbulence,
Cambridge University Press, London, 1960, p. 46.
13
Born, M. and Wolf, E., Principles of Optics^ Pergamon Press,
New York, 1970, p. 753.
14
Plotkin, K. J. and George, A., "Propagation of Weak Shock
Waves through Turbulence," Journal of Fluid Mechanics3 Vol.
54, pt. 3, 1972, pp. 449-467.

COMPUTER MODEL OF THE LIGHTNING - THUNDER PROCESS,
WITH AUDIBLE DEMONSTRATION
T
H. S. Ribner,* F. Lam,t K. A. Leung,'
D. Kurtz, and N. D. Ellis (
University of Toronto, Toronto, Canada
Abstract
A lightning stroke is modeled as an irregular line distri-

bution of point impulsive sources; an elemental N-wave (sonic
boom) evolves by nonlinear distortion from each point. These
notions are embodied in a computer program to yield the thunder
pressure signature in time for a given lightning signature in
space. Previous results appeared as x-y plots, but now a fast
D-A output yields the computed thunder signature as a varying
voltage in real time. This is fed into an amplifier and loud
speaker to produce the synthetic thunder. (A short tape
recording was demonstrated at the Conference.)
Introduction
There has been a considerable body of research into the

physics of lightning (e.g., Ref. l) but much less on the
mechanism of reasonably distant thunder (e.g., Refs. 2-6). In
the early years, in fact, much of the published work on thunder
$ Presented as Paper 75-5^-8 at the AIAA 2nd Aero-Acoustics

Conference, Hampton, Va., March 2^-26, 1975.
Supported by National Research Council (Canada) and U.S.
Air Force Office of Scientific Research, Grant AF-AFOSR 70-
1885.
* Professor.
t Research Assistant, Institute for Aerospace Studies; now
at Virginia Polytechnic Institute, Blacksburg, Va.
T Research Assistant, Institute for Aerospace Studies; now
at Stanford Research Institute, Palo Alto, Calif.
Research Assistant, Institute for Aerospace Studies; now
at Dept. of Electrical Engineering, University of Toronto.
Cf| Research Engineer, Institute for Aerospace Studies; now
at de Havilland Aircraft of Canada, Downsview, Ontario.
77
78 RIBNER, LAM, LEUNG, KURTZ, AND ELLIS
generation was purely speculative. A number of serious re-

searches on the physics of thunder have modeled the lightning
channel as a straight cylindrical column, ignoring its pro-
nounced tortuosity (e.g., Refs. 2-6). This model probably is
adequate, close to the channel, and it is almost a mathematical
necessity to delineate the strongly nonlinear evolution of the
near-field pressure. On the other hand, this rectilinear
model of lightning is totally inadequate to predict thunder
at any substantial distance from the stroke; it "will, of
course, predict a single sharp crack with no indication what-
soever of the roll and rumble of real thunder.
It has been recognized widely that the irregularity of the

lightning stroke accounts for features of the thunder signa-
ture. Thus the nearest parts are heard first and the furthest
last, since the signals propagate with approximately the speed
of sound. This in itself can account for the long duration
of some thunder records, considering the length of the light-
ning channel, but no one appears to have attempted to model
the process in any detail to see whether such a process can
yield realistic thunder signatures. This has been the object-
ive in the work reported herein, an outcome of several years'
intermittent effort. Specifically, the aim was to reconstruct
the thunder heard at a specified location, within a certain
approximation, given the shape of the lightning stroke.
The rationale for and details of the model of thunder
generation are developed in the main text. Here we merely
outline the process. Starting with a digitized photograph of
a lightning stroke, the computer models the physics of the
sound generation process in terms of a sequence of overlapping
"sonic booms.11 The result of the calculation is a time
history of the sound pressure at the observer location. These
digital data are converted into electrical signals that vary
with time. The varying signal is fed into an anrplifier and
loudspeaker to produce the sound of thunder.
This project is, in a sense, the inverse of one carried
out by A, A. Few et al.5 They reconstructed the shape of the
lightning channel (in three dimensions'.) from measurements
of the thunder recorded by an array of microphones. (More
specifically, phase coherence in the array dictated the vector
direction of sound arrival as a function of time; tracing the
vectors back to time zero located the source points delinea-
ting the lightning channel.) Both approaches help clarify
the acoustics of thunder, but they illuminate different facets.
The Few approach tends to be more of a diagnostic tool for
lightning (particularly cloud-to-cloud lightning, largely
hidden), whereas the present approach is oriented more toward
MODEL OF LIGHTNING THUNDER PROCESS 79
explaining the origin of detailed features of thunder.
Model for Thunder Generation
A lightning stroke involves a virtually instantaneous mas-

sive heat release along a long, tortuous channel. Thus it is
very much like an explosion. From the acoustic point of view,
we model the stroke as an irregular line distribution of point
explosions. The theory of -point explosions, amply supported
by experiment, tells us what to expect. At large distances,
the impulsive wave from the point of the explosion has evolved
by nonlinear distortion into a characteristic shape; the
pressure vs distance (or time) signature resembles a letter

"N" - a sonic boomN-wave.
Our simplified model then postulates that impulsive spheri-

cal waves are emitted simultaneously from each point of the
lightning stroke and evolve independently. At large distances
they all have the character of N-waves. The sound heard by
an observer at a specified location is built up as indicated
schematically in Fig. 1. The impulsive waves, eventually
N-waves, from each point of the irregular channel travel down
a fan of rays converging on the observer. The N-waves arrive
at different times, with much overlapping (not shown in Fig.
l), because of the varying lengths of the rays that they have
to travel.
Fig . 1 Model for the lightning thunder

process (refraction ignored) .
Fig. 2 Cutoff due to refraction. Atmospheric

temperature gradients curve the sound rays so
that observer hears only those "between "a" and
"b"; a ray like "c" will pass over his head.
Thus the lower part of the lightning stroke
cannot be heard at the observer's location
(except for some rumble owing to diffraction).
In reality the sound rays are not really straight the

wave fronts are not really spherical owing to refraction
by the temperature gradients (principally) and winds in the
atmosphere. The rays are curved, normally concave upward.
This has the additional effect, shown in Fig. 2, of producing
a "cutoff" due to refraction. The sound from the lower part
of the lightning stroke cannot reach the observer.
Our model is deliberately simplified in replacing the

curved rays by straight ones. It is further simplified, as a
pragmatic requirement, by replacing the actual three-dimen-
sional stroke by its two-dimensional projection on a plane,
namely its photograph. The net effect of both simplifications
is equivalent to replacing the actual lightning stroke by a
distorted one. The distorted channel, however, will have a
qualitatively similar random irregular character; hence the
thunder signature predicted for it, although not the actual
thunder traceable to the real channel, nevertheless should
be qualitatively similar to real thunder.
A typical cloud-to-ground lightning stroke requires about
100 usec, but it has been modeled herein as instantaneous.
This assumption is equivalent acoustically to a relative
positional error between the top and bottom of the stroke of
only several centimeters vastly less than the accuracy with
which the channel can be defined from a photograph. A stroke

frequently has branches, but this is a complication that we
have decided to omit; our examples are limited to a single
branch-free cloud-to-ground channel. Finally, the lightning
discharge is typically a multiple event, the discharge being
repeated several times along the same channel at roughly 30-
msec intervals. A double repetition is incorporated in our
computer model at times 28 and 60 msec after the initial
stroke.
For the purposes of computerizing the thunder model of Fig.
1, the lightning photograph must be digitized. With the
photographs available the resolution permits digitizing to

about 100 segments, each of the order of 60 m in length; that
is, the lightning channel is in effect represented as a zig-
zag chain of straight segments (Fig. 3)* We can argue (details
later) that these "straight" segments must have a fine struc-
ture that the photograph cannot show (inset B, Fig. 3) In-
corporating this, in a fashion to be explained, there are
ultimately some 2000 straight-line segments in our model of a
lightning channel, averaging around 3 m in length.
The N-wave pressure signals emanating from each point along

such a lightning segment can be integrated analytically. This
was done by Wright and Medendorp7 for application to the sound
of an electric spark. The resulting composite wave form
radiated by the entire segment (spark) is shown on the left-
hand side of Fig. k. For comparison,the measured wave form
from the sparkT is shown on the right-hand side of Fig. h.
The difference in shape is the result of nonlinear steepening
in the latter case, as suggested by Whitham.8 in the context
of sonic boom. The wave forms from each line segment adopted
for the computer simulation are shown in Fig. 5. The compres-
sion (positive) pulse is assumed always to have steepened to
a shock wave, but the rarefaction (negative) pulse is assumed
unchanged from the theoretical shape; the areas are adjusted
to be equal.
The duration of the basic "building block" N-wave was taken

as 5 msec at 1 km from the source, guided by the empirical
fact4 that the spectrum of thunder typically has a dominant
peak around 200 Hz. From one point of view, the spectrum of
a single such N-wave peaks around 200 Hz. From another point
of view, the zero crossings of close-packed nonoverlapping
N-waves will be around 200 Hz, roughly resembling the situation
in a thunder signature. Following the nonlinear theory, the
basic N-waves are taken to stretch as they propagate along the
rays according to the law A ~ \/j5n r/r0. The pressure ampli-
tude decays with distance according to Ap ~ 1/r ^J?n r/r0. The
Fig. 3 Digitization of lightning

stroke (two methods of segment
subdivision shown).
Theory Experiment
Fig. 4 Azimuthal variation of pressure

signal from a line source.
Oblique Angles
Fig. 5 Wave steepening assumed in the
computer generation of thunder.
1/r spherical spreading decay is augmented "by the pulse

stretching effect. The reference length ro is taken as 10 m
for convenience.
We return now to the question of the structure of the light-
ning channel that cannot "be resolved in a photograph. Each of
the 100 or so apparently straight-line segments into -which we
can subdivide the stroke presumably is made up of somewhat
shorter zig-zag subsegments, a fine structure that we cannot
make out. If this were not so, the zero-crossing frequency
of a 10-sec thunder signature would approximate 100/10 = 10 Hz,
instead of the roughly 200 Hz observed. To build up the zero-
crossings to an order of 200 Hz, therefore, we postulate that

each of our 100 measured straight segments of the lightning
stroke is subdivided into 20 randomly selected subsegments.
If the nonlinear steepening of the pressure signatures of
individual subsegments (Fig. 3> inset A) were not included,
the subdivision would have no effect. The signatures from the
subsegments would simply integrate into the relatively smooth
signature of a full segment. In particular, the 19 pulses at
the junctures of the subsegments would cancel out. Thus this
particular scheme of subdivision is but an artifice to in-
crease the frequency 20-fold, and is not very defensible on
physical grounds. A more realistic approach would have postu-
lated the subsegments to have irregular inclinations one to
the other a sort of "random walk" path (Fig. 3 5 inset B) .
Some exploratory calculations have been made with random walk
configurations, but the resulting thunder signatures were quite
unrealistic. However, the deterioration probably was more the
result of a reduction from 20 to 2 or 3 subsegments than of
the random walk feature.
Computational Requirements
The thunder signature is calculated at discrete times. The

interval between "samplings" determines the upper (Nyquist)
cutoff frequency for the calculated thunder. For the present
program this upper cutoff was selected at 1000 Hz, which cor-
responds to a sampling interval of 0.5 msec.
Based on the foregoing sampling rate, a thunder duration

on the order of 10 sec, and 2000 rays (rather than the three
of Fig. 11) to be scanned sequentially, on the order of 107 to
108 pressure evaluations, would be required. Even with large
high-speed computers this order of computations is unaccept-
able. The solution to this difficulty is a limited selective
search for those segments of the lightning channel whose sig-
nals are reaching the observer (Fig. l) at a given instant.
Only a limited length of channel is searched at a time, greatly

reducing the amount of computation. Since, in general, the
base of a lightning stroke is the closest to a ground-based
observer? the onset of the thunder signature (neglecting re-
fraction) is expected to originate from this region and at
later times to originate further up the lightning channel; thus
the region of search proceeds upward from the base of the
stroke. The repeated stroke calculation is via a shift regist-
er to minimize the increase in computer time. Even with this
optimization, 1-2 min of IBM-370 computer time are required
for each second of thunder calculated. Before listening tests
were implemented, 5 sec of this computer-generated thunder was
thought to be adequate to assess its characteristics, and on

this basis most of the records generated were terminated at
5 sec.
Digital to Analog
The pressure-time thunder signature calculated in the
previous section is in digital form and, as such, may be plot-
ted and Fourier analyzed but is not available for a subjective
listening test. The acquisition of a minicomputer (Hewlett-
Packard 2100A) with high-speed digital-to-analog capability
made the conversion of the digital pressure-time signature to
an analog voltage feasible. The mass of calculations was
performed on an IBM-370 and stored on digital magnetic tape.
This data tape was then read by the minicomputer and converted
to a real-time analog signal. The program for the minicomputer
scales the output data for the maximum possible accuracy. The
analog signal produced in the digital-to-analog process is a
series of steps in voltage vs time. The carrier represented
by this staircase is not part of the simulation and was thus
removed by appropriate low-pass filtering. The signal thus
produced was normally recorded on a Nagra tape recorder for
later playback through an amplifier-loudspeaker system. The
first "synthetic thunder", however, did not involve a tape
recorder as intermediary.
Results
As has been described, computer programs have been developed

that enable the calculation of a theoretical thunder signature
from a particular lightning channel and its conversion to a
real-time analog voltage signal. The computation has been
carried out for a particular case, among others, in which the
observer is 0.8 km from the base of a 6-km lightning stroke
(Fig. 3). The pressure-time history for this calculated thun-
der is shown in Fig, 6 together with a thunder signature (from
a different lightning stroke) recorded by Few.^ There is a
| 100ms _>)
Pressures in
2 <*-*^ Arbitrary Units
-1 H-
Fig. 6 Computer-generated "thunder" (upper
trace) and pressure-time signature (lower
trace4) of an example (not the same) of
real thunder (hand traced).
qualitative similarity between the two signatures even though

the two originate from two independent and presumably quite
different lightning channels. On playback through a loud-
speaker system, the computer-generated signal sounded surpris-
ingly like real thunder, exhibiting the characteristic rumble
and roll. The similarity was sufficient to fool an unsuspect-
ing colleague out of sight of the apparatus, and in view of an
outside door; he called out in dismay "Is it starting to rain?"
For some cases in which the calculated record length was arti-
ficially terminated after 5 sec, two independent thunder cal-
culations were spliced to yield a more reasonable listening
length. This is the case for the demonstration tape.
From the calculated results, two tentative correlations

between the lightning channel shape and the resulting thunder
signature may be made. First, the roll of the thunder is
thought to correspond to amplitude modulation dictated by the
large-scale shape variations of the lightning stroke (the
amplitude is a well-defined function of local inclination?).
The "carrier" signal may be attributed to the small-scale ir-
regularities in the lightning stroke which are not visible in
the photograph, (in cases where the "pseudo" fine structure
obtained by subdividing each segment was not employed, the
signal became a series of discrete "pops" separated by quiet
intervals.)
Conclusions
The model of the generation and propagation of thunder em-

bodied in the program described has been found to yield real-
istic-sounding thunder signatures. However, the program
involves a number of simplifying assumptions that bear exam-
ination. Replacement of the three-dimensional lightning
channel by its two-dimensional projection in a photograph is
not thought to impair the realism, nor is the neglect of cur-
vature of the sound rays owing to atmospheric refraction.
(However, neglect of diffraction of sound from the "cutoff
zone" will falsely suppress some low-frequency rumble.) More
questionable is the assumed shape of the pressure signal radia-

ted from a li*ne segment (Fig. 4). This is tied in with the
assumption that the individual source points along the light-
ning stroke radiate independently. The independence principle
is certainly an oversimplification, since it is partly invali-
dated by nonlinear effects; however, again it is thought that
realism will not be much impaired. Perhaps the most question-
able aspect concerns the fine structure of the lightning
channel that is missing from the photograph. The assumptions
underlying the invention of a substitute fine structure will
have to be re-examined.
References
Malan, D. J., Physics of Lightning, English Universities

Press, London, 1963, 176 pp.
2
Remillard, W. J., The Acoustics of Thunder, T.M. 44, Acoustics
Research Lab., Harvard University, Cambridge, Mass., 1960.
Bhartendu, Acoustics of Thunder, Ph.D. Thesis, Physics Dept.,

University of Saskatchewan, Saskatoon, Saskatchewan, 1964.
14
Few, Dessler, Latham, and Brook, "A Dominant 200 Hz Peak in
the Acoustic Spectrum of Thunder,'" Journal of Geophysical Re-
search, Vol. 72, No. 2k, Dec. 15, 1967, pp. 6149-6154.
Few, A. A., "Lightning Channel Reconstruction from Thunder
Measurements," Journal of Geophysical Research, Vol. 75, No.
36, Dec. 20, 1970, pp. 7517-7523.
Plooster, M. N., "Numerical Model of the Return Stroke of the

Lightning Discharge," The Physics of Fluids, Vol. l4, No. 10,
Oct. 1971, pp. 2124-2133.
Wright, W. M. and Medendorp, N. ., "Acoustic Radiation from

a Finite Line Source with N-Wave Excitation,11 The Journal of
the Acoustical Society of America, Vol. ^5, No. 5, May 1968,
pp. 966-971.
o
Whitham, G. B., "The Flow Pattern of a Supersonic Projectile,"
Communications on Pure and Applied Mathematics, Vol. V, 1952,
pp. 301-3^

PANEL DISCUSSION: ACOUSTIC WAVE PROPAGATION
THROUGH THE ATMOSPHERE
M. Friedman, Columbia University: The papers presented on

Acoustic Wave Propagation through the Atmosphere were con-
cerned with the question of what happens to the noise on its
way through the atmosphere from the source to the observer.

The general problem of wave propagation is a well-established
one with a long history. From the aeroacoustician's point of
view, however, there are two new elements that need to be con-
sidered; they are not really new, but they are new in the
sense that they now may have to be taken into account. u Both
received some consideration in the papers in this session.
The two elements are:
1) We are dealing with wave propagation over long dis-
tances, ranging from 700 to 7000 m, and this implies the need
for much greater care in the application of analytical tech-
niques. The approximate methods that have been customary for
a long time in wave propagation theory are no longer adequate
because of severe cumulative effects over such long distances.
2) The second element which I found interesting has to do
with taking into account that aspect of real gases known as
relaxation effects. Aerodynamicists have known for a long time
that relaxation effects can be quite important, particularly
with strong shocks and high temperatures. What is surprising
is that they can be important in aeroacoustics even at the
levels of weak shocks. This fact became apparent in a contro-
versy over what causes the observed shock thickening. It first
had been contended that shock thickening could be attributed
to cumulative effects of turbulence over long distances. Then
it was pointed out that perhaps the same kind of shock thicken-
ing could be attributed to relaxation effects. One of the
papers presented some experimental evidence in an attempt to
identify the dominant mechanism. As the author himself pointed
out, however, his experiments were probably too limited to be
decisive. Nevertheless, I think it is an interesting question
that deserves some further research. Obviously, relaxation
effects are important, and cumulative turbulence effects are
also important. Which one of these effects plays a dominant role
and at what stage in the wave propagation? That is the question
to be settled.
89



DEVELOPMENT OF A NEW COMPUTER SYSTEM FOR
AIRCRAFT NOISE PREDICTION
John P. Raney*
NASA Langley Research Center, Hampton, Va.

Abstract
The paper presents an overview of the activities of NASA's

Aircraft Noise Prediction Office (ANOPO). The principal goal
of ANOPO is to develop a comprehensive, user-oriented, Aircraft
Noise Prediction Program (ANOPP). ANOPO's activities in sup-
port of ANOPP development are discussed briefly. They include
acquisition, implementation, and evaluation of an in-house,
interim collection of programs and implementation of a plan for
acquiring, in the form of Key Technology Documents, state-of-
the-art methodology for aircraft noise prediction. The paper
is devoted primarily to a presentation of the general archi-
tecture and functional capability planned for ANOPP, together
with the rationale supporting major design decisions.
Introduction
In July 1973, NASA established an Aircraft Noise Predic-

tion Office (ANOPO) at the Langley Research Center. The
purpose of creating this new organization was to provide both
a focal point for NASA's aircraft noise prediction activities
and an appropriate interface with other agencies and industry.
In addition, the ANOPO charter directs the timely creation of
a new, integrated, user-oriented Aircraft Noise Prediction
Program (ANOPP). ANOPP will be specifically tailored to meet
NASA's requirements for aircraft noise prediction and will be
used extensively by NASA to evaluate and quantify the benefits

*Head, Aircraft Noise Prediction Office.
93
94 J. P. RANEY
[AIRCRAFT MEASURED
MODELING FLYOVER
DATA DATA
AIRCRAFT
MANEUVER
DATA
/ INPUT
X TO"B"
WYLE
DATA
BASE
B
NEF
w
X Y NEF
PROGRAM
N L PROGRAM
E
NOISE
GRID NOISE
GRID
PRINT
GENERAL L NEF
CONTOUR R CONTOURS
PROGRAM C
Fig. 1 Interim aircraft noise

prediction program.
AIRCRAFT NOISE PREDICTION PROGRAM 95
expected from proposed noise reduction projects and research

activities.
Current Noise Prediction Capability
In support of the preparation of a viable ANOPP develop-

ment plan, ANOPO has acquired and installed at Langley an
interim system for aircraft noise prediction consisting of a
family of contemporary, independently developed capability as
follows (Fig. 1):
A. 1) An aircraft source noise modeling program written
by the Boeing Company for the NASA Ames Research Center. >^
2) An aircraft engine noise synthesizer developed by the Noise

Effects Branch of the NASA Langley Research Center.
B. A Noise Exposure Forecast (NEF) contour program

written by Bolt Beranek and Newman (BBN) for the United States
Air Force.^
C. An NEF contour program written by Wyle Laboratories

for the Department of Transportation.
D. An extensive data base of noise data for the civil

fleet prepared for the FAA.^>5
The interim programs presently are being utilized by

ANOPO, by Langley project offices, and occasionally by other
Government agencies. Some of the elements of the interim
programs may be selected for incorporation in ANOPP; however,
the interim programs themselves eventually will be discarded in
favor of the more comprehensive and flexible ANOPP system.
ANOPP Development
Prediction Methodology
The nature and state of the art of the methodology to be

implemented in an applications program such as ANOPP are
extremely important. For example, empirical methods rely more
heavily on measured data (such as aircraft flyover data in the
form of noise, thrust, altitude, or NTA curves) for problem
input, whereas analytical methods may rely on analytical models
to generate values of output variables. The size and nature
of the input and output data bases also are obviously dif-
ferent. Sophisticated analytical methods may involve the
solution of large eigenvalue problems and require large quan-
tities of core and central processor time. It is obviously
important, therefore, to possess an accurate assessment of the
96 J. P. RANEY
state of the art of noise prediction methodology at present

and equally important to know the rate of change advance
of this methodology. Production of the Key Technology Docu-
ments discussed below is providing ANOPO with this information.
Based on participation in author conferences, editorial

committees, and discussions with other reviewers and interest-
ed participants in preparation of the Key Technology Documents,
it is clear that noise prediction methodology is in its
infancy or, at best, childhood. Accepted methods range from
highly to semi-empirical. Analytical modeling of noise gener-
ating mechanisms has only very recently attracted serious
attention; however, any aircraft noise prediction program that

is to enjoy a reasonable half-life should include provision
for analytical methods. There exists a clear consensus that
prediction methodology will experience a rapid advance during
the next decade while moving to less empirical, more analyt-
ical modeling techniques. It therefore is concluded that,
because of the rapidly changing state of the art, ANOPP should
provide to its users maximum flexibility for updating, modi-
fying, or replacing the noise prediction methodology portions
of the program.
Key Technology Documents
To assure that state-of-the-art technology is implemented

in ANOPO, a series of noise prediction technology documents
are being generated by ANOPO with the cooperation of other
NASA centers, other Government agencies, and industry. These
documents bear approximately a one-to-one correspondence to
the functional (or computational) modules planned for ANOPP.
The areas covered by individual documents include (Fig. 2):
Human Factors, Propagation, Jet Noise, Airframe Noise, Blown
Flaps, Fan/Compressor Noise, Core Noise, Duct Treatment, Rotor/
Propeller Noise, and Shielding.
The continually updated Key Technology Documents combined

with inputs from periodic ANOPO/User Seminars will constitute
che mechanism for assuring the implementation of current
prediction technology in the ANOPP system.
Potential Users
Contacts with NASA Headquarters and other agencies, to-

gether with the Key Technology Document activity, have helped
ANOPO identify potential users of ANOPP. Several classes of
users have emerged whose primary interests are indicated in
table 1.
AMES HUMAN FACTORS
PROPAGATION
JET NOISE
FLIGHT AIRFRAME NOISE
BLOWN FLAPS
FAN/COMPRESSOR NOISE
LANGLEY CORE NOISE

DUCT TREATMENT-
ROTOR/PROPELLER NOISE
LEWIS SHIELDING- ANOPO

Fig. 2 Key technology documents.
Table 1 Potential Users of ANOPP
Source Single Multiple

USER noise event event
modeling exposure cumulative
design exposure
NASA
FAA
DOD
Engine
manufacturers
Airframe
manufacturers
EPA
Airport managers
Consultants
98 J. P. RANEY
The needs of ANOPP users range from sophisticated analyt-

ical source modeling and design to empirical computation of
cumulative noise exposure. Some users will wish to use ANOPP
only in connection with making community exposure estimates
and the sensitivity studies related thereto. An engine design
group will be interested in analysis and evaluation of possible
engine configurations based on analytical models of noise gen-
erating mechanisms. Analog or one-third octave data will be
required by the latter and noise levels vs distance by the
former. Without compromising NASA's requirements or incurring
additional costs, every reasonable effort will be made to
assure ANOPP compatibility with the needs of a broad user
community. Therefore, consistent with NASA's requirements,

ANOPP should be structured to support computations ranging from
detailed analysis of engine noise source to cumulative exposure
contours based on measured flyover data bases. Analytical
source modeling and single event received noise computations
must be provided. Data bases of measured (or computed) flyover
data should be maintained to support computation of cumulative
exposure indices.
ANOPP Logical Levels
As shown in Fig. 3, ANOPP will provide four logical

levels of computational sophistication intended to satisfy
(SINGIE EVENT)
AF-BBN
DOT-WYLE
Fig. 3 ANOPP logical levels.

AIRCRAFT DATA BASE: EFFECTIVE NOISE LEVEL (EPNL, ETC.) VS SLANT RANGE
STRAIGHT LINE SEGMENTS DESCRIBE AIRCRAFT OPERATIONS
SUITABLE FOR ROUGH ESTIMATES OF COMMUNITY NOISE ENVIRONMENT

MINIMAL KNOWLEDGE OF AIRCRAFT NOISE REQUIRED FOR USE
Fig. 4 ANOPP I.
AIRCRAFT DATA BASE: NOISE LEVEL (dBA, PNdB, ETC.) VS RANGE AND DIRECTIVITY
SMOOTH, CURVING AIRCRAFT TRAJECTORY
ESTIMATES TIME HISTORY OF AIRCRAFT NOISE LEVEL AND PROVIDES REFINED
ESTIMATE OF COMMUNITY NOISE ENVIRONMENT
GENERAL KNOWLEDGE OF ACOUSTICS AND AIRCRAFT NOISE REQUIRED FOR USE
Fig. 5 ANOPP level II.
the needs of various user groups and to provide a self-
contained systematic means for validating and improving the
state of the art of aircraft noise prediction.
Level I (Fig. 4 ) , the simplest operational mode of the

program, is intended to serve civil engineers and community
planners who have minimal knowledge of the complex technology
of aircraft noise prediction. This level of ANOPP is charac-
terized by the user of time-integrated flyover (Noise, Thrust,
Altitude) data, in such units as EPNdB, to compute measures
of community noise environment such as NEF contours. The AF-
BBN and DOT-Wyle programs operate at this level.
Level II (Fig. 5) of ANOPP is intended to serve aero-

nautical engineers in making systems studies involving general
100 J. P. RANEY
aircraft types as well as the noise control engineer who re-

quires greater knowledge of the community aircraft noise
exposure than the time-integrated estimates provide. Level II
is based on the use of computed or measured values of noise
levels, such as PNdB's and d B A f s , which vary during the
aircraft flyover. There are no existing programs which use
this prediction methodology.
Levels III (Fig. 6) and IV (Fig. 7) of ANOPP will be used

to predict the time-dependent noise spectrum for an analysis
of the aircraft component noise sources and an analysis of the
V SQURCE NOISE = I AIRCRAFT
^^_ + JET + FAN +. . U J
DIRECTED 1/3 OCTAVE BAND LEVELS
PROPAGATION
COMPONENT DATA OR EMPIRICAL FORMULAS

PREDICTS REAL-TIME SPECTRUM AT RECEIVER
SUITABLE FOR AIRCRAFT/ENGINE SYSTEMS |_
STUDIES 1/3
IN-DEPTH KNOWLEDGE OF AIRCRAFT NOISE
BAND NUMBER
REQUIRED FOR USE
REAL-TIME SPECTRUM RECEIVED

Fig. 6 ANOPP level III.
SOURCE NOISE = I AIRFRAME + J E T + F A N +. ..
DIRECTED CONTINUOUS, DISCRETE SPECTRA
PROPAGATION
ANALYTICAL MODELS OF NOISE SOURCE COMPONENT

PREDICTS REAL-TIME RECEIVED SPECTRUM INTENSITY
SUITABLE FOR DETAILED, ADVANCED DESIGN STUDIES
FOR EXPERTS ONLY
FREQUENCY
Fig. 7 ANOPP level IV.
aircraft flight. These levels are intended for the use of
engineering and research specialists in aircraft noise. Level
III is based primarily on empirical formulas for the noise of
different aircraft source components. This level will be
suitable for making detailed systems studies of aircraft/
engine configuration. The NASA-Boeing program operates at
this level. Level IV will be the repository for the most
advanced acoustical technology and may be used in an experimen-
tal sense for technology validation and improvement or for
detailed designs of advanced low-noise components for aircraft.
ANOPP Availability/Accessibility
Making ANOPP available or easily accessible to all users

is a challenge. Some users will own computing equipment and
others will not. Among those who own computers, a wide variety
of hardware and supporting software will be represented.
It is obvious from the "Prediction Methodology11 and "Po-
tential Users" sections that ANOPP will not be a member of the
class of FORTRAN programs that compiles at run time. As a
matter of fact, it will be a relatively large program, perhaps
one-third the size of the NASTRAN program for structural
analysis. Therefore, a third design decision is obvious as a
result of the previous two: ANOPP will require its own
executive system and data base manager.
Other factors influencing accessibility involve the fact

that it will cost substantially more and take longer to build
and maintain a machine-independent ANOPP than to support ANOPP
on only one type of computer. Present estimates suggest that
the cost of developing ANOPP to run on both CDC and IBM ma-
chines is 1.7 to 3.0 times that of either machine alone and
that maintenance costs would be about 1.5 times greater for
two machines than for one. Therefore, because ANOPP will be
a large program and because of the additional costs associated
with support of ANOPP across several machines, ANOPP will be
designed to operate and will be optimized to CDC 6600/7600
machines under the SCOPE and KRONOS operating systems. Also,
because of the previous decision and because commercial com-
puter networks are equally accessible to all ANOPP users, it
is anticipated that ANOPP will be made available through a
commercial computer network.
Summary
ANOPP should provide the capability to model noise sources

analytically, to predict accurately single event noise, and to
compute community noise exposure resulting from airport
102 J. P. RANEY
operations. Because of the rapidly changing state of the art

of prediction technology, ANOPP should provide maximum flexibi-
lity for updating or replacing the noise prediction methodology
portions of the program. To support both empirical and analyt-
ical prediction methods, comprehensive data base management
capability is required. All of the above militate in favor of
an executive manager for ANOPP. Because of the heterogeneous
nature of the potential ANOPP user community and because of the
additional costs and time required to support ANOPP on more
than one type of computer, it is proposed that ANOPP be de-
signed to operate only on CDC equipment and that it be made
available to users through a commercial computer network.
During the ANOPP development period, the ANOPO interim

programs will be maintained and upgraded as required to satisfy
the requirements of NASA noise prediction activities. Work on
the Key Technology Documents will continue, and the method
recommended by each document will be coded and tested by ANOPO
prior to installation in ANOPP. Completion of ANOPP executive
system and data base manager is estimated to be in the fall of
1976 with implementation of a complete package of prediction
modules about 1 yr later.
References
^Dunn, D. G. and Peart, N. A., "Aircraft Noise Source and

Contour Estimation," NASA CR-114649, July 1973.
2
Crowley, K. C., Jaeger, M. A., and Meldrum, D. F., "Aircraft
Noise Source and Contour Computer Programs User's Guide," NASA
CR-114650, July 1973.
^Reddingius, N. H., "Community Noise Exposure Resulting from

Aircraft Operations: Computer Program Operator's Manual," AMRL
TR-73-108, July 1973.
^Goodman, J. S., et al.," Aircraft Noise Definition, Phase I.

Analysis of Existing Data for the DC-8, DC-9, and DC-10
Aircraft," Kept. FAA-EQ-73-5, August 1973.
^Williams, B. G. and Yates, R., "Aircraft Noise Definition,

(1) Summary (2) Model 707 (3) Model 727 (4) Model 737 and (5)
Model 747," Kept. FAA-EQ-73-7, December 1973.
AIRCRAFT FLYOVER NOISE MEASUREMENTS
E. L. Zwieback*
Douglas Aircraft Company, McDonnell Douglas Corporation,

Long Beach, CA.
Abstract
The evolution of the noise measurement techniques reflect
the varied reasons for the measurements and also reflect the
advancements in electronic instrumentation. Current techniques
are described by a number of standard procedures, not com-
pletely definitive. The results of measurements by various
organizations may be, in some cases, noise levels which are not
comparable. Described are basic typical procedures utilized in
making comprehensive measurements. Differences in procedures
are indicated. General limitations include variations in
the f l i g h t effects of noise sources, in the atmospheric propa-
gation of noise, in the recorded noise signal near the ground,
and in data processing methods.
Introduction
Aircraft flyover noise measurements have been conducted
for approximately 30 years; the measurement objectives, equip-
ment, and procedures have grown increasingly complex to satisfy
a wide variety of requirements. Current measurement objectives
are*detailed acoustic source levels for research and develop-
ment; flyover noise levels to demonstrate compliance with
aircraft noise certification and operator specifications,
levels for development of noise-abatement operations, levels
over a wide range of aircraft operations for community noise
exposure forecasting, and monitored levels for airport oper-

Conference, Hampton, Va., March 24-26, 1975,
^Manager, Acoustics and Vibration Testing.
103
104 E. L. ZWIEBACK
ational information. Detailed acoustic source levels (acoustic

pressures as a function of emission frequency and emission
angle) are useful for a variety of f l i g h t research purposes and
also, of course, for preliminary and product development of
jet-powered aircraft. Compliance demonstrations include stand-
ardized measurements and standardized f l i g h t procedures to
arrive at definitive single-valued levels for specific regu-
latory or contractual purposes. The development of noise-
abatement operations usually requires comparative noise
measurements, p r i m a r i l y in subjective-criteria terms. Another
measurement activity is for the purpose of community noise
exposure forecasting. Specifically included are noise levels
at long distances from the aircraft, these low noise levels

being significant because of the integrated noise exposure
forecasting systems being developed. And f i n a l l y , the airport
noise monitoring systems, functioning in many parts of the
world, result in large numbers of noise measurements being
acquired and cataloged in simple form for operational noise
impact information and, in some cases, for airport noise
compliance purposes.
For an understanding both of the current measurement
procedures and the existing base of flyover noise data, it is
useful to review briefly the history of aircraft flyover noise
measurements. The evolution of the noise-measurement equip-
ment and procedures can be traced from the I940fs, when the
f l i g h t noise of a DC-3-type airplane was recorded on f i l m and
the octave-band sound pressure levels determined up to a
frequency of about 1400 Hz.1 Magnetic tape-recording techni-
ques and condenser microphones for f i e l d use became a v a i l a b l e
in the early I950fs, and their use alleviated the difficulties
associated with determining the noise at high frequencies.
With the advent of turbojet- and turbofan-engined transport
aircraft in the late I950fs and early I960?s the impact of
flyover noise on airport communities increased, and measurement
and abatement efforts increased accordingly. Reference 2
describes recommended measurement practices developed in the
early I960!s. Two significant flyover noise-measurement test
programs, involving a number of test organizations, were con-
ducted in 1968. The programs3'1* presented opportunities for
development and/or implementation of standard techniques and
comparison of the simultaneous measurements obtained by differ-
ent organizations, although these efforts were not the prime
objectives of either program. Federal aircraft-noise-certifi-
cation regulations were established in 1969 and incorporated
more complex subjective noise criteria than those used in the
I960fs. The required complex measurement tasks became feasible
through the development of one-third octave-band parallel
filter systems, or real-time analyzers, together with a variety
AIRCRAFT FLYOVER NOISE 105
of f l e x i b l e d i g i t a l data systems for aircraft testing.

Measurement procedures have been developing in the same
sequence, as a result of the increasing accuracy, repeatability,
and resolution requirements for test programs, and the a v a i l -
a b i l i t y of the sophisticated instrumentation needed to make
the measurements. Wide-range instrumentation, in particular,
has been developed as the characteristics of flyover noise can
vary substantially in terms of frequency and amplitude distri-
bution. To illustrate the frequency distribution characteris-
tics and variations, Fig. I depicts acoustic spectra for
typical flyover noise measurements for JT3D-powered aircraft
(during landing approach) and JT8D-powered aircraft (during
takeoff climbout). Significant differences in frequency

distribution are indicated.
It is the purpose of this paper to describe, in general
terms, current equipment and techniques utilized to make com-
plete precision (comprehensive) measurements of the flyover
noise of jet transport aircraft. Indicated are differences in
measurement equipment and techniques, depending on the ob-
jectives and type of airplane involved. Discussed also are
general measurement limitations as the result of a variety of
technical uncertainties and practical problems. Although
directly relevant, accuracy values of flyover noise measure-
ments are discussed only indirectly. This is because of a
variety of interactions (both known and unknown) between the
characteristics of flyover noise and the measurement procedures.
F i n a l l y , as a result of the general measurement l i m i -
tations, some research efforts to reduce the impact of the
ONE-THIRD
OCTAVE-BAND
SPL (dB)
125 250 500 IK 2.5K 5.0K 10K

GEOMETRIC MEAN FREQUENCY (Hz)
Fig. I Flyover noise spectral characteristics.
106 E. L. ZWIEBACK
measurement limitations are suggested. Noise-measurement

procedures and techniques for types of aircraft other than
conventional jet-powered aircraft, such as propeller-driven
aircraft, powered-Iift-type aircraft, and helicopters, are not
included in this paper.
Current Procedures
To accomodate the variety of measurement objectives, a
number of standards and procedures have been developed by both
industry and government agencies. These standards include
those developed for: I) research and development,5 2) airport
noise monitoring within a state, 3) U.S. government noise

certification,6 4) m i l i t a r y aircraft noise definition, 7 and
5) noise definition and certification for use by various
countries.8 These and related existing standards are listed
i n Tab Ie I.
As a result of different objectives and different
effect}vity dates, some measurement differences have developed
between the various standard procedures listed in Table I.
Differences include height of the microphone above the ground,
s i d e l i n e microphone location, subjective measures of noise
(ranging from A-weighted sound levels to effective perceived
noise levels), and procedures to calculate and adjust measured
noise levels. The height of the microphone above the ground
surface ranges from 4 ft (for certification) to 20 ft (for
monitoring). S i d e l i n e microphone locations for noise certifi-
cation can be either at 1500 or 2100 ft lateral distance from
the runway centerline. Calculation of perceived and effective
perceived noise levels includes differences in minimum
perceived noisiness values (basic subjective conversion quan-
tities), tone or discrete-frequency effects, and noise level
minimums for duration effects. Specifically, the minimum
perceived noisiness is either I or 0.I Noy. Tone, or discrete-
frequency, effects can be calculated using 10-step or 7-step
procedures, depending on the document, with resulting diff-
erences. Duration effects are considered only above a 90-PNdB
floor, or the floor is neglected, depending on the definition
document. Several of the measurement procedures and standards
currently are being re-examined for appropriate modifications.
Significant efforts by both industry and government represent-
atives recently have been devoted to proposed modifications of
Ref. 5 and 8 which may eventually result in standards that are
more definitive.
Current procedures and standards for aircraft noise
measures do have many detailed descriptions for the acquisi-
tion, processing, and normalization of the test data. But,
Table 1 Standards for the measurement of aircraft flyover noise
Meas. Calc.
Number Date Title proc. Equip, proc.
ARP 796 1965 SAE measurement of aircraft exterior noise in the field. X X
FAR part 36 1969 Noise standards: aircraft type certification. X X X
R507 1970 ISO recommendation. Procedure for describing aircraft X X X
noise around an airport.
R1761 1970 ISO recommendation. Monitoring aircraft noise around X
an airport. 3>
11
PUC title 4, 1970 California aeronautics noise standards X X X
0
subchapter 6 5
^
ICAO annex 16 1971 International standards and recommended practices: X X X n
-H
aircraft noise.
IEC 179 1965 IEC recommendation, precision sound level meters X ~n
r~
IEC 225 1966 IEC recommendation, octave, half-octave and third-octave X 0
band filters intended for analysis of sounds and vibrations. <
n
SI. 11-1966 1966 American national standard specification for octave, half- X 73
octave, and third-octave band filter sets. o
ARP 1080 1969 SAE frequency weighting network for approximation of X 11
CO
perceived noise level for aircraft noise. m
SI. 4-1971 1971 American national standard specification for sound level X
meters.
ARP 866 1964 SAE standard values of absorption as a function of tempera- X
ture and humidity for use in evaluating aircraft flyover
noise.
ARP 865A 1969 SAE definitions and procedures for computing the perceived X
noise level of aircraft noise.
ARP 1071 1972 SAE definitions and procedures for computing the effective X
perceived noise level for flyover aircraft noise.
<-i
108 E. L. ZWIEBACK
because of r e l a t i v e newness of some of the techniques and the

complexity of the procedures, a number of items are not defined
completely, and, as a result, the variation can result in
differences in normalized data produced by various organi-
zations. Items vaguely defined include nature of the ground
surface below the microphone, ranging from "perfectly reflect-
ing" to !f not-excessive I y absorptive"; noise path lengths,
acoustical or optical; data time synchronization for noise
and source position; symmetry procedures for pairs of s i d e l i n e
recordings; adjustment for acoustical and electrical ambient
noise influence; accounting for spectra that are incomplete
due to r e l a t i v e l y low noise l e v e l s in certain frequency bands;
noise s i g n a l processing dynamic response and data averaging

requirements; deletion of calculated tone effects due to
ground surface reflection; and selection of engine and airplane
parameters (such as thrust, fan speed, jet velocity, airspeed,
gross weight) for noise source adjustments.
Measuring Equipment
Comprehensive measurements of aircraft flyover noise

require several types of recording equipment: noise levels,
aircraft operations, aircraft space-position, sound-path
weather, and; compatible data processing. The resulting
instrumentation system is diagrammed in Fig. 2. Figure 3
indicates various types of recording equipment deployed in the
f i e l d . The equipment and procedures currently u t i l i z e d to
obtain comprehensive aircraft flyover noise measurements vary
w i t h different test organizations. Techniques for recording
aircraft operations vary from manual reading of a v a i l a b l e
cockpit instruments to high-sample-rate d i g i t a l recording of
special engine instrumentation. Weather recording varies
from airport tower and hand-held instrument readings at a
central location to automatic recording of weather conditions
throughout the noise propagation path, including the use of
atmospheric instruments on the test aircraft. Various types
of recording equipment used for aircraft space-position include
I) a combined system of ground-mounted cameras and an aircraft
altimeter; 2) ground-mounted phototheodolite systems; 3) ground
radar-type (both electromagnetic and o p t i c a l ) automatic track-
ing systems; and 4) aircraft-mounted downward-looking cameras,
together w i t h ground target arrays. A variety of noise record-
ing equipments are a v a i l a b l e and can include high-frequency
and/or high-sensitivity microphones, single-channel magnetic
tape recorders w i t h large dynamic range, and/or multichannel
FM magnetic tape recorders with high-frequency pre-emphasis
networks. F i n a l l y , noise data processing equipment varies from
one-third octave-band s e r i a l l y stepped f i l t e r (with a chart
NOISE DATA ACQUISITION NOISE DATA

REPRODUCER AND
MONITORING AND 1/3 OCTAVE-BAND
RECORDING SYSTEM SPECTRUM ANALYZER
AIRCRAFT OPERATIONS
DATA REPRODUCER
DATA ACQUISITION AND FLYOVER
AND PROCESSOR
RECORDING SYSTEM NOISE
DATA
AIRCRAFT SPACE COMPUTING

DATA REPRODUCER
cvQTFM
POSITION TRACKING AND
AND PROCESSOR
RECORDING SYSTEM
WEATHER DATA DATA REPRODUCER

ACQUISITION AND
AND PROCESSOR
RECORDING SYSTEM
Fig. 2 Comprehensive measurement system for aircraft flyover

noise.
WIND WEATHER
SOUNDING SOUNDING
BALLOON AIRPLANE
RETROREFLECTOR
PORTABLE OPERATIONS
CALIBRATION FACILITY
AND CABLE ,, 1 ^ WEATHER STATION
NOISE MICROPHONE VEHICLE ^ V,4 V (TYPICAL)

(TYPICAL)-.
-4*._JjL_
Q--^_ ^
.
NOISE TEST CENTRAL
RADIO COMMUNICATIONS
Fig. 3 F i e l d development of flyover noise measurement equip-

ment.
recorder) to automatic digital processing systems incorporating

a one-third octave-band p a r a l l e l f i l t e r system (real-time
analyzer). Most organizations involved in aircraft noise
measurements are replacing the s e r i a l l y stepped f i l t e r w i t h
the real-time analyzer for quantity data processing.
The advantages of integrating the diverse equipment and

procedures for coordinated and e f f i c i e n t measurement programs
appear obvious; two current types of integrated measurement
capabilities are described in Refs. 9 and 10. Currently, an
110 E. L. ZWIEBACK
evaluation of the most complete and comprehensive noise measure-

ment systems is being conducted for the U.S. government.11
Measuring Techniques
Special airplane operating procedures have been developed
for noise measurements to acquire the variety of separate types
of test data required and also, in part, to m i n i m i z e high-cost
aircraft f l i g h t operations. To appreciate the magnitude of
aircraft noise test costs, the costs sometimes are indicated
in terms of the f l i g h t time required for testing. Compre-
hensive noise testing of a large transport aircraft requires

about one f l i g h t hour for five flyover test runs, with the
flight testing restricted by stringent test weather conditions.
The specific procedures for aircraft operations involve precise
and constant power settings, precise f l i g h t path control, and
close coordination with ground test units. In current noise
recording procedures, it is recognized that the flyover noise
recordings must be as free as possible from the effects of
both the local microphone station location and the specific
instrumentation system. Therefore, three types of associated
are necessary: I) reference sound pressure level; 2) total
ambient noise level (at the station location); and 3) reference
frequency-response signal for the instrumentation system.
Special data procedures have been developed to integrate
four separate types of test data on a time-synchronized basis
for the resulting normalized noise levels. The four types of
data are I) airplane operational parameters describing the
noise source; 2) airplane (noise source) space position; 3)
atmospheric parameters describing the sound-path propagation
conditions; and 4) airplane-emitted sound pressure levels (SPL)
recorded near the ground. These procedures transform the
measured SPL data into normalized SPL data, with the effects
of test technique and test condition variations removed within
the capabilities of available procedures. Alternatively,
integrated noise levels may be computed for the test-day
conditions. Figure 4 illustrates some of the digital data
processing steps required to obtain the test-day (measured)
integrated noise levels.
Extensive and complex equipment, together with the several
special test procedures, may be justified and utilized only for
comprehensive noise measurements of large transport aircraft.
Data acquisition and data processing equipment and procedures,
reduced in scope from the extensive types described, probably
are sufficient for limited types of noise measurement and
certain types of aircraft. Limited types of noise-measurement
INPUT CORRECTION INPUT NOISE LEVELS-

AND INPUT AMBIENT
1/3-OCTBANDAT
ADJUSTMENT LEVELS NOISE LEVELS 1/2-SEC INTERVAL
AVERAGING ON
"PRESSURE" BASIS FOR
N INTEG PERIODS
IE
ADJUSTMENT OF INPUT
NOISE FOR
CALCULATION OF AMBIENT NOISE
CORRECTION TO _L
INCLUDE SYSTEM CONVERSION OF INPUT
FREQUENCY RESPONSE, NOISE TO CALIB SOUND
MIC PRESS. RESPONSE PRESS. LEVELS
ANDSPL REFERENCE JL
LEVEL CALCULATION OF
PNL VALUES
JL
CALCULATION OF TONE
CORRECTIONS AND
PNLT VALUES
CALCULATION OF
DURATION CORRECTION
AND EPNL VALUE
Fig. 4 Basic d i g i t a l data processing flow for measured

flyover noise levels.
apply to I) noise-abatement operations (both for development
and for definition), 2) preliminary aircraft-noise-definition
testing, and 3) airport noise monitoring. Candidate types of
aircraft for reduced-scope noise-measurement efforts include
1) aircraft with noise levels well below a p p l i c a b l e limits.
2) aircraft with minor configuration changes, or 3) special-
purpose aircraft, few in number.
General Limitations of Flyover Noise Measurements
As discussed in the previous section, a number of factors
result in noise test data that only approach, but never achieve,
the status of being "absolutely correct" data. These factors
include the variety of noise-measurement objectives, the
evolution of the noise-measurement technology, the variations
and options in noise-measurement standards, and the actual
utilization and limitations of the test measurement systems.
"Absolutely correct" flyover noise measurements cannot exist
because measurement definitions and adjustment procedures are
incomplete and w i l l , to some extent, interact with the type of
noise being measured. Of course, as indicated earlier,
significant efforts to improve existing definitions and pro-
cedures for completeness currently are being made. The general
limitations of flyover noise measurements are discussed below.
To some extent, these limitations are integral to any set of
noise data. Both the accuracy and the extended application of
112 E. L. ZWIEBACK
noise data sets are affected by the limitations discussed;

however, the significance of the limitations probably varies
for any given data set, as determined by the particular con-
ditions of the data set,
Noise Sources
The flyover noise sources of a jet transport aircraft are
numerous and consist of several different types. Flight
effects can affect the i n d i v i d u a l noise source levels s i g n i f i -
cantly. Jet noise, turbomachinery noise, and aircraft non-
propulsive noise a l l are subject to the effects of flight
conditions. Therefore, adjustments to measured data for large

variations in f l i g h t conditions should be validated experi-
mentally before applying such adjustments. Where shielding
or reflection effects are possible, because of aircraft wing
configurations, caution should be exercised in determining
noise source levels with wing (flap) configuration changes.
In addition, the engine noise sources on large aircraft are
distributed on the aircraft with separation distances as much
as 100 ft. Therefore, any type of far-field data analysis
that assumes a single-point noise source must be limited to
propagation distances that are large compared with the appro-
priate separation distances of the noise sources of the
ai rcraft.
Noise Propagation
The propagation characteristics of aircraft noise through
the atmosphere have been the subject of a large number of
analytical and experimental efforts. References 12 and 13
describe two of these recent efforts. To some extent, measured
noise data are affected by a number of atmospheric conditions
such as temperature and moisture levels and their gradients,
wind levels and their gradients, and wind and temperature
fluctuations (turbulence). Specifically, for molecular
absorption due to atmospheric moisture levels, certain ranges
of moisture content in th.e atmosphere result in relatively
large sound absorption for certain frequencies. Significantly,
there are ranges of moisture content where the sound absorp-
tion changes greatly for small changes in moisture content.
The typical characteristics of sound absorption (molecular)
as a function of moisture content are shown in Fig. 5. As an
example, in the one-third octave band centered at 4kHz, atmos-
pheric moisture content below about 4 g/m^ result in large
gradients of sound absorption. With the uncertainties present
in determining atmospheric moisture content, calculated values
of sound absorption may have errors in these moisture content
ranges.
MOLECULAR ABSORPTION OF SOUND

REFERENCE: SAE ARP 866
FREQUENCY OF 4000 Hz, TEMPERATURE OF 70F
MOLECULAR 20
ABSORPTION, aMOL
(dB/1000 FT)
4 6 8 10 12 14
ABSOLUTE HUMIDITY (gm/m3)
Fig. 5 Atmospheric absorption characteristics.
Air-to-ground propagation losses for the subjectively

important spectral components of most types of aircraft noise
generally are dominated by atmospheric absorption. Near-
horizontal noise propagation is also influenced in addition by
atmospheric refraction effects due to gradients in atmospheric
propagation conditions. F i n a l l y , noise propagation at long
distances, in addition to atmospheric absorption and refraction
effect, may be influenced by scattering due to atmospheric
fluctuations. The current standard data-adjustment procedures
account only for atmospheric absorption (based on surface
conditions) and do not account for the effects of nonstandard
weather conditions throughout the sound propagation path.
Without the proper considerations taken into account for each
set of noise data, significant data limitations exist,
especially when adjusting data to greatly different propagation
paths or atmospheric conditions.
Noise Recording
Recording noise signals accurately near the ground surface
requires considerable care if the test program objective is to
obtain noise data free of ground surface effects. On the
contrary, most measurement programs do not include this object-
ive, and the results are aircraft noise data combined with
particular ground surface effects. For hard reflecting
surfaces, the surface effects are quite large but fairly
straightforward. These surface effects are a function of sound
wavelength and multipath interactions. Straightforward adjust-
ments are possible assuming that the angles of incidence are
114 E. L. ZWIEBACK
ONE-THIRD 20
OCTAVE-BAND
LEVEL,
10
^^H
REFLECTION
^
DIFFERENCE -i r
(dB) 0
(DIFFERENCE
BETWEEN NOISEE SOURCE HEIGHT: 400 FT
-10
TOTAL AND
MICR OPHONE HEIGHT: 4 FT
DIRECT SPL's)
NOISlE SOURCE ANGLE: 65 DEC
-20 HARC) REFLECTING SURFACE
WHIT E NOISE SPECTRA
CALC ULATED PER SAE AIR 1327 (PROPOSED)

50 125 250 500 IK 2.5K 5.0K 10K
Fig. 6 Acoustical effects produced by a reflecting plane.
known, and micrometeorological effects near the surface are

insignificant. Figure 6 illustrates the spectral magnitudes
of the effects of a hard reflecting surface for a typical
flyover condition, using the calculation methods of Ref. 14.
For natural ground surfaces with finite but complex (and
probably nonuniform) acoustical impedances, the effects are
smaller but more complicated, because the impedance also may
vary with the incidence angle and the frequency. In any case,
varying ground surface effects usually are found in all exist-
ing measurements, perhaps exaggerated for s i d e l i n e measurements.
In general, ground surface effects tend to increase aircraft
flyover noise levels by 3 dB relative to free-field (direct-
only) levels.
Other recording limitations include the evaluation of
significant noise signals during the flyover, that is, the
influence of existing ambient noise on the recorded flyover
noise signals. This problem was encountered frequently in
the past; however, techniques have been developed in recent
years to account for the influence of steady-state ambient
noise. The i n a b i l i t y to discriminate between the ambient
noise and low levels of flyover noise results in the 'Moss"
flyover noise measurement. This situation, the loss of par-
ticular frequency-band levels, is a common occurrence for
many current types of flyover noise measurements. Figure 7
illustrates this point, showing typical total ambient (combined
acoustical and electrical) noise levels, together with the
measured maximum noise levels of a large transport aircraft
during takeoff (with power cutback) at a distance of about
5300 ft. Flyover noise levels in the frequency bands above
frequencies of 2000 Hz are at or below the total ambient noise
A-WTD
LEVEL
75
70 s* "~- ^. f s ____ AIRCRAFT TAKEOFF AIRCRAFT
s
/
s
V (WITH CUTBACK) MEASUF ED
X A T A LONG DISTANCE
60
^x >
^\ \
rv
ONE-THIRD 50 ^
OCTAVE-BAND \ AMBIENT
\ 1 <
SPL
40 \
(dB RE 2 x 10-5 \
k:^
N/M2)
30 TOTAL AMB IENT NO SE LEVE LS^y. ^ L/2-IN.-D A
FOR RECOR DING
MICROPh ONE
20 SYSTEM -\
*- h-J
10 /
l-IN.-DIA MICROPHONE ' ^^^^ ^-.-^^
5
50 125 250 500 IK 2.5K 5.0K 10K
Fig. 7 Typical low-amplitude noise levels for aircraft
flyovers.
levels. Future problems are anticipated with; persistently

increasing levels of ambient noise at test sites, uncertainties
as to the time-stability of ambient noise levels during flyover
noise testing, and lower levels of recorded aircraft noise.
Data Processing
Data processing of the analog recordings of the transient
acoustic signals into integrated noise measures for reference
aircraft and weather conditions requires a number of different
types of data operations. One of the fundamental operations
involves the frequency filtering and conversion into root-mean-
square (rms) values of the analog acoustic signals. Filtering
may be done with analog or digital techniques. Significantly,
the specific shape of the one-third octave filters, always
nonideal, w i l l vary within a standard envelope. The nonideal
filter shapes can interact with the noise spectra character-
istics involving discrete frequencies, and the interaction can
be compounded by doppler-shift effects to produce tone levels
influenced by filter-band edge crossings. The narrow-band
spectral time history of flyover noise signals (during a land-
ing approach) shown in Fig. 8 illustrates the concept of the
doppler-shift/fiIter-band interaction. Standard band edge
frequencies of some of the one-third octave filter bands are
indicated in the figure.
116 E. L. ZWIEBACK
1/3-OCT
SOUND
PRESSURE LEVEL
(dB)
3 4 5 6 7 8 9 10 NOTE: FAN ROTOR TONE

FREQUENCY (kHz) DISPLAY ENHANCED
Fig. 8 Narrow-band spectral time history of landing approach

flyover noise.
The conversion into rms values involves signal averaging

or smoothing. The signal averaging is especially significant
for the usual case of the transient random signals of aircraft
flyover noise. Several types of signal averaging are feasible,
such as summation (integration) or exponential (digital
equivalent of Resistor-Capacitor (RC) smoothing), so long as
(in the standard case) the output data exhibit dynamic response
characteristics s i m i l a r to the response of a "slow" sound level
meter. The results of the standard dynamic response test^8
for a summation-type digital system indicate that integration,
or averaging, times between I and 1-1/2 sec w i l l meet the
majority of current standard response requirements. The effect
of two different averaging times was evaluated for a large
number of flyover noise recordings over a range of flyover
transient conditions (represented by the aircraft angular
velocity relative to the ground microphone).15 The results are
shown in Fig. 9 and indicate that the maximum tone-corrected
perceived noise level (PNLTM) decreases with increasing averag-
ing time. The opposite trend occurs with the duration correct-
ion, with the net result being approximately no change for the
effective perceived noise level (EPNL). For low-altitude or
high-speed flyover (high angular rate relative to the micro-
phone), significant extended-decay characteristics have been
observed for time-series noise data, because of the use of
standard averaging times. F i n a l l y , decreased angular reso-
lution of the data can occur easily for high-angular-rate
2
1/3 OCTAVE-BAND SIGNAL DATA PROCESSING
NOISE LEVELS
CHANGE WITH
AVERAGING TIME
CHANGE (dB/SEC)
0 10 20 30 40 50
FLYOVER ANGULAR RATE AT OVERHEAD (DEG/SEC)
Fig. 9 Change of flyover noise levels with averaging

time changes.
FLYOVER DIRECTIVITY
ANGULAR RATE CHANGE
AT OVERHEAD DURING DATA
(DEG/SEC) SAMPLE PERIOD

(RELATIVE TO (DEGREES)
GROUND 30 STD-AVG OF
MICROPHONE 1-1/4 SEC
20
LOCATION)
10
200 400 600 1000 20003000

AIRCRAFT FLYOVER HEIGHT (FEET)
Fig. 10 Aircraft flyover angular rates and directivity
changes.
flight testing combined with "standard" data averaging. The

range of aircraft angular rates and also aircraft directivity
maximum changes during a spectrum sample period (for "standard"
data averaging) are shown on Fig. 10.
Adjustments for the influence of ambient noise throughout

the flyover are based on the assumption that the ambient noise
remains reasonably unchanged from the particular pre- or post-
flyover ambient-record ing time. To m i n i m i z e errors in identi-
fying and evaluating only flyover noise signals, high-confi-
dence tolerance intervals for the ambient noise levels are
selected. However, the net results, in some cases, are a
number of spectra containing frequency-band noise levels
118 E. L. ZWIEBACK
identified as being ambient-noise dominated. These particular

band noise levels are not used throughout subsequent data
operations. Therefore, in many cases, flyover noise measure-
ments w i l l provide incomplete spectral information relative to
the complete 24-band audio frequency range.
Data adjustments to reference aircraft and reference
propagation conditions include only a specific number of
factors. As noted earlier, these limited factors do not
include nonstandard atmosphere of effects on atmospheric
propagation which might involve nonuniform absorption, ray-
bending or refraction, and fluctuations or scattering phenomena,
F i n a l l y , computation of the relatively complex noise measures,

such as effective perceived noise level (EPNL!s), involves
many steps. These data-processing steps result in data limited
by some variations in the minimum perceived noisiness values,
methods to replace ambient-dominated SPL!s for both perceived
noise level and tone correction calculations, and the deter-
mination of duration-correction intervals.
General V a l i dity
Because of the high cost of obtaining accurate and repeat-
able noise measurements, existing sets of aircraft flyover
noise data are being utilized in a variety of ways, some of
which never were anticipated during the original data-acquisi-
tion programs. The v a l i d i t y of existing data for other a p p l i -
cations is of real concern. As an example, airport community
noise levels in the form of noise contours often are used in
noise control procedures and regulations regarding aircraft
operations. Supportive of these activities, aircraft noise
data bases using existing sets of noise data are being created
by a number of interested organizations. Such efforts are
indicated in Ref. 16 and 17. In developing data bases,
particularly when extrapolating flyover noise levels from
existing sets of data, the limitations' discussed previously,
together with long-distance propagation effects s t i l l not w e l l
understood, can combine to render that portion of the data
base misleading, if not erroneous. A detailed discussion
concerned with the v a l i d i t y of centerlirve data extrapolated
to s i d e l i n e conditions is presented in Ref. 18. It concludes
that the accuracy of noise impact areas calculated currently
is extremely dubious.
The statistical v a l i d i t y of comprehensive noise measure-

ments, in terms of the repeatability of data scatter, has been
determined many times for large transport aircraft during
specific FAR-36 noise certification tests. In terms of the
90% confidence l i m i t s for the mean value expressed in effective

perceived noise levels, takeoff noise measurements usually w i l l
exhibit confidence l i m i t s of less than I EPNdB, whereas confi-
dence l i m i t s for landing approach noise measurements u s u a l l y
w i l l be less than 1/2 EPNdB.
On the other hand, absolute accuracies of flyover noise
measurements are not w e l l established. A considerable amount
of research remains to be done on the significance of the
variety of measurement limitations discussed, especially for
the aircraft noise data bases being developed. Significant
efforts should be directed to understand more completely I)
the aircraft f l i g h t effects on noise source variations, 2)

the effects of atmospheric propagation of aircraft noise and
corresponding data adjustment procedures, 3) the need and
methods for m i n i m i z i n g ground surface effects on recorded
noise, and 4) data processing variations of transient noise
signals. Until that research is accomplished, the accuracy
of flyover noise measurements and derivative data bases cannot
be defined completely in quantitative terms.
Summary and Recommendations
The techniques and equipment necessary for making compre-
hensive measurements of aircraft flyover noise have progressed
significantly with the increasing emphasis on the subject.
Procedures to describe flyover noise levels have become
increasingly complicated. In addition, the utilization and
application of the noise measurements have become greater than
o r i g i n a l l y intended.
The demonstrated repeatability of comprehensive noise
measurements is considered adequate for current specific
purposes. For a number of reasons, noise-measurement
accuracies cannot be determined quantitatively for a general
case. General limitations of aircraft flyover noise measure-
ments exist which, with sufficient research, may be m i n i m i z e d
to enhance the accuracy and application of the measurements.
The characteristics of noise source variations with f l i g h t ,
noise propagation effects through the atmosphere, ground
surface effects, and data processing effects of the transient
noise signals a l l are subjects recommended for further
research.
References
1
Wiener, F. M. and Marquis, R. J., "Noise Levels Due to an
Airplane Passing Overhead,ft Journal of the Acoustical Society
of America, Vol. 18, No. 2, Oct. 1946, pp 450-452.
120 E. L. ZWIEBACK
2
McPike, A. L., "Recommended Practices for Use in the Measure-
ment and Evaluation of Aircraft Neighborhood Noise Levels,"
Paper 650216, 1965, Society of Automotive Engineers, Warrendale,
Pa.
^Committee A-21, "Comparison of Ground-Runup and Flyover Noise
Levels," AIR 1216, A p r i l 1972, Society of Automotive Engineers,
Warrendale, Pa.
^"Noise Certification Procedure Demonstration," Acoustic Rep.

160, 1968, British Aircraft Corp., Ltd., Weybridge, England.
5
Committee A-2I, "Measurements of Aircraft Exterior Noise in
the Field," ARP 796, 1965, Society of Automotive Engineers,
Warrendale, Pa.
6
"Part 36 Noise Standards: Aircraft Type Certification,"
Federal Aviation Regulations, 1969, Federal Aviation Admin-
istration, Department of Transportation, Washington, D.C.
7
Bishop, D. E. and Galloway, W. J., "Community Noise Exposure
Resulting from Aircraft Operations: Acquisition and Analysis
of Aircraft Noise Performance Data," Rep. AMRL-TR-73-l07,
Sept. 1974, Aerospace Medical Research Laboratory, Wright-
Patterson Air Force Base, Ohio.
8
"Aircraft Noise," International Standards and Recommended
Practices, Annex 16 to the Convention on International C i v i l
Aviation, 1971, International C i v i l Aviation Organization,
Montreal, Canada.
^Zwieback, E. L., "Flyover Noise Testing of Commercial Jet
Airplanes," Journal of Aircraft, Vol. 10, No. 9, Sept. 1973,
pp 538-545.
10
Auzolle, S. and Hay, J., "Method of Measurement and Analysis
of Noise of an Aircraft in Flight," Tech. Transl. TTF-14, 058,
Dec. 1971, NASA, Washington, D.C.
11
"Aircraft Noise Measurement Systems," Request for Proposal
WA5R-4-0095, Dec. 1973, Federal Aviation Administration,
Washington, D.C.
12
M i l l e r , R. L. and Oncley, P. B., "The Experimental Determin-
ation of Atmospheric Absorption from Aircraft Acoustic F l i g h t
Tests," Contractor Rep. CR-I89I, Nov. 1971, NASA, Washington,
D.C.
13
Smith, C. M., "Atmospheric Attenuation of Aircraft Noise, M
Rep. HSA-HAD-R-GEN-214, Sept. 1973, Hawker Siddley Aviation,
Ltd., Hatfield Herts, England.
14
Committee A-21, "Acoustic Effects Produced by a Reflecting
Plane," Proposed AIR 1327, A p r i l 1973, Society of Automotive
Engineers, Warrendale, Pa.
15
Stouder, D. J., "Use of Digital Averaging Techniques for
the Analysis of Aircraft Flyover Noise," Proceedings of
Inter-Noise 74, 1974, Washington, D.C., pp 137-140.
16
"Aircraft Sound Description System," Federal Aviation
Administration Order, Aug. 1973, Federal Aviation Adminis-
tration, Department of Transportation, Washington, D.C.
17
"Noise Assessment for Land-Use Planning," Circular
II60AN/86, 1974, International C i v i l Aviation Organization,
Montreal, Canada.
18
Seykra, C. A., Storey, W. C., and Yates, R., " V a l i d i t y of
Aircraft Noise Data," Journal of the Acoustical Society of
America (Abstract only), Vol. 55, No. X6, Suppl., A p r i l 1974,
pp S48.

REVIEW OF THEORY AND METHODS FOR THE PREDICTION OF
GROUND EFFECTS ON AIRCRAFT NOISE PROPAGATION
*
Paul B. Oncley
MAN-Acoustics and Noise, Inc., Seattle, Wash.
Abstract
Sound transmitted or received near a surface is modified
by an interference pattern that is determined both by the geom-
etry and by the acoustic constants (complex impedance and prop-
agation constant) of the surface. The classifical treatment,
assuming a point source and single frequency, has been expanded
by Howes and by Hoch and Thomas to include extended sources and
various types of filters including one-third octave. It is not
adequate, however, to deal with the frequent cases of propaga-
tion at near-grazing incidence. The analyses of Rudnick and of
Ingard are examined and linked with the Hoch-Thomas equations.
Several more recent approaches are discussed briefly, but for
experimental verification better information on the soil con-
stants is required. The available data are discussed, and
methods for improving the data are suggested. Advantages and
questions associated with microphones on the surface are given,
and it is shown why the foregoing analyses have been more suc-
cessful in extrapolation and noise prediction than in correct-
ing experimental data. Experimental and theoretical results
are compared.
I. Introduction: Classic Theory
The analysis of sound propagation close to Earth's surface
is of great practical importance and presents many difficulties
because of the irregularities and inhomogeneities both in
Earth's surface and in the air immediately above the interface.

* Senior Research Scientist.
123
124 P, B. ONCLEY
To approach the problem, we must make simplifying assumptions,

which commonly include a surface that is flat and smooth, spec-
ularly reflecting, and with an acoustic impedance that is ei-
ther uniform in all directions or definable in terms of the im-
pedance perpendicular to the surface ("locally reactive"). The
air is also considered as quiet and homogeneous. The geometry
of the problem is given in Fig. 1.
A source at height h emits sound that reaches the receiv-
er at height h by a direct path
= [D2 + (hs-hr)2]%
and a reflected path

= [D
The difference1 between the two paths is Ar = r2-r.. The clas-

sical analysis gives the velocity potential $ ai the receiver
as
$r = A1[(exp(-1kr1)/r1) + Rp(exp(-ikr2)/r2)]exp(iu>t) ^
l/^) + Rp(exp(-ikr2)/r2)]exp(io)t) (1)
SOURCE
RECEIVER
Fig. 1 Test geometry.

PREDICTION OF GROUND EFFECTS 125
The reflection coefficient R p is defined as
zsin* - [1 - (k/k9)2cos2i^
R _ __________^ /2)
p
zsin* + [1 - (k/k2)2cos2^
Here z is the complex impedance ratio of the surface,

ZP/PQC, and k^ is the complex propagation constant in the sur-
face medium. The propagation constant in air is assumed real
and equal to w/c. The grazinq angle is the complement of the
incidence angle. Equation (1) can be rewritten as
$r = (A 1 /r 1 )exp(-ikr 2 +io)t)[l + (r 1 /r 2 )R exp(ikAr)] (la)
The ratio (r-i/r 2 ) for the case of propagation near the

ground is approximately 1 and can be dropped. We shall consid-
er initially the case in which z 1 (rigid boundary) and $ is
large enough that
0 9 1
zsirty [1 - (k/k 2 ) cos ^
so that R is nearly +1. Then the magnitude of $ is dependent

on the phase angle kAr, ranging from near zero for kAr = TT,
Sir, etc., and giving pressure doubling for kAr = v\2i\. For val-
ues of DAr, the first minimum fg is given by
f0 - Dc/4hshr (3)
The resulting interference pattern, such as the one shown by

the solid curve in Fig. 2, is superposed on the spectrum of the
source under test, making interpretation difficult.
II. Effect of Finite-Width Filters
Since most airplane noise analysis is based on octave or

one-third-octave filter data, the interference pattern usually
is seen in a more diluted form, but bands containing minima
still show diminished sound pressure, and low-frequency bands
and those with peaks still give readings 4 to 6 dB above free-
field values. An analysis by Howes^ provides a means of calcu-
lating the correction for each filter band, for several types ~
of filter. His approach has been continued by Hoch and Thomas
in France. They express the result in terms of the ratio of
126 P. B. ONCLEY
X- IU 1 1 i 1 1 1 i ] 1 i 1 1 1 1 i 1 1
' ' i 111
3
'A A
""^N/^^\
/ \ /
I: o / \ V
1/

00
0-
/ \i i
5 ? -10 - \/\ / i
S r^
*~ SJ-20 -
30
T 2
!
0 100 10 00 i0,000
Frequency (Hz)
Fig. 2 Calculated pressure level deviation from field
conditions; r=200 ft, hs=hr=7.07 ft,is surface
equivalent to concrete, is surface equivalent to
Quietone.
the mean square pressure at the receiver to that which would

have existed under free-field conditions:
p _ 2 / 9
(4)
where C(T) is the autocorrelation coefficient of the noise sig-

nal of spectral density w(F) across the filter band, (^^5)-
This has been shown to be
rfh
Jf Dw df
C(T) = (5)
VF) df
The time lag T is the sound transit time across Ar, or
T = Ar/C. For noise that is relatively uniform across the
band, w(F) may be replaced by a constant WQ so that Eq. (5)
becomes
sin [>T(f-f )]
C(T) COSTTT (fk+O
D a (6)
Expressing the ratio of mean square pressures of Eq. (4)

in decibels and substituting the ratio S for n/r2, we obtain
for an infinitely narrow band (f = f = fb), as in Eq. (la),
AN = 10 log 1Q [1 + S2 + 2S cos(2^rf)] (7)
When the filters are of constant percentage bandwidth,

Eq. (6) becomes
AN = 10 log1Q {1 + S2 +
+ 2S[sin(aAr/Xi)/(aAr/Xi)] cos (gAr/A.,)} (8)
where a = 7r(Af/f..), and 3 = ZirVl + ^(Af/f.. ) 2
Values of Af/f-j are 0.707 for octave bands and 0.2316 for
one-third octaves, giving a = 0.725 and 3 = 6.325 for one-third
octave measurements or, in terms of k.Ar (k. = 27rf./C),
AN = 10 log 1Q {1 + S2 +
+ 2S[sin(0.1158k 1 Ar)/(0.1158k i Ar)]cos(1.00668k i Ar)} (8a)
III. Effect of an Extended Source
The interference pattern, somewhat diluted by one-third

octave filtering, is weakened even further by the finite size
of the source, which in the foregoing discussion has been taken
as a point. Particularly in the case of large jet engines,
sound arises from an infinite number of points distributed over
a large volume. The geometry of each source with respect to
the receiver gives rise to a different value of Ar for each.
This can be simulated by taking a number of points, say 5,
evenly spaced[ throughout the source volume, and computing the
power ratio pT /Pep for each case, and then averaging the
ratios. Since eacn Ar gives a different fg, the sharp dip is
replaced by a broad trough.
IV. Finite Impedance Case
In the more general case of reflection from a porous medi-

um such as soil, sod, crushed rock, etc. ,it is no longer true
128 P. B. ONCLEY
that zl and that Rn^l. Moreover, experimental data invaria-

bly show that the interference pattern is shifted downward in
frequency like the dashed curve in Fig. 2. The significance of
this is evident from Eq. (la). The value of Rp is complex and
may be written as |Rp e1'6, so that the expression in parenthe-
sis becomes
[1 + (r1/r2)|Rp|exp(ikAr + 16)]
Obviously, the frequency of the first minimum fQl is now the

frequency at which (kAr + 6) = IT. The same concept is ex-
pressed in the Hoch-Thomas formulation [they give (-6)]:
AN = 10 log 1(J {1 + S 2 | Q | 2 +
+ 2S|Q|cos[(27rArA) + 6]} (9)
i r^
Here | Q | e is equated to the reflection coefficient
1>6
Rp|e . When constant proportion bandwidth filters (such as
one-third octave) are used, Eq. (8) is modified similarly:
N = 10 log 1Q {1 + S
+ 2S|Qi|[sin(aAr/X.)/(aAr/Xi)]cos[e(Ar/X.) + 6-]} (10)
The subscripts i indicate that |Q-j|, x-j, and 6-j are evalu-
ated at the geometric center frequency of the respective filter
bands.
V. Propagation at Near-Grazing Incidence

It long has been recognized that the classical analysis
fails for propagation parallel to the surface. In Eq. (2),
when ip->0, Rn becomes -1 for all values of z and (k/k2), and,
since under these conditions r2 = r-j and Ar = 0, the parenthe-
sized term in Eq. (la) goes to zero. Thus, propagation should
be impossible at grazing incidence. Experiments show that, not
only is there no minimum as ^0, but there is an increase in
signal strength at low angles.
The first satisfactory resolution of this classic dilemma
was given by Rudnick,^ who in 1947 applied to the acoustic case
the methods used by Sommerfeld,5 Van der Pol,6 and Norton7 in
the analogous problem of electromagnetic wave propagation over
an absorbing Earth. Rudnick showed that Eq. (1) is valid only
for a nonexistent plane wave, and that use of spherical wave

propagation functions led to the replacement of R by
R p + F(w)(1-R p ) (11
where
F(w) = 1 + i2Vw"e"w / exp(-u^)du (12

-i\/w~
and
r> o
12kr [l -
9
(1-Rp) cos > z
F(w) can be expanded to give
F(wH-2e-w(w + ^Tr+^yr+ 773T+...)+ie"Wv^" (14)
or, for |w|>10, by
F(w) = - [-1
2w (2wr (2w)
Although Rudnick's complicated expressions were formidable
in 1947, they are handled readily by modern digital computers.
Moreover, it readily can be seen that the coefficient Q in e-
quations such as (9) and (10) actually describes the entire
reflected signal, and for the spherical wave analysis we can
define
Q = |Q|ei6 = Rp + F(w)(l - R p ) (16
Thus the previous equations can be used directly for the analy-
sis of one-third octave band data and where extended source is
considered.
It is useful to consider the behavior of the "numerical
distance11 w and the "boundary loss factor" F(w). For example,
w is directly proportional both to propagation distance r2 and,
through k = 2-rrf/c, to frequency. For near-grazing cases,
cos2\|; approaches 1 and (1 - RD)2 approaches 4, and for most
130 P. B. ONCLEY
cases k2k, so that [1 - (k/k2)2cos2*] is not much different

from 1. For the values of z usually associated with grassy
soil or plowed fields, |w| exceeds 10 for rxf products>25,000.
Since Eq. (15) converges rapidly, at larger values of |w|, F(w)
is essentially proportional to l/|w|, and so at distances above
a few hundred feet and frequencies above a few hundred Hertz
the signal strength [proportional to F(w) by expression (11),
since Rp -* -1] falls off at 6 dB per double distance in addi-
tion to the inverse square loss, or a total of 12 dB per double
distance, and it also drops at a 6 dB per octave rate with fre-
quency. Over2 acoustically hard surfaces like concrete or wa-
ter, where z may be of the order of 10", the 12 dB per double
distance loss will not occur except at very long distances and
very high frequencies. This is in accord with the observed

fact that sound propagates much better over water than over
Earth. It also has important implications in calculating the
propagation of airplane sideline noise around airports and of
highway traffic noise.
A graphic solution of Eq. (12) is given in Figs. 3 and 4,
in the form given by Ingard. This clearly shows the inverse
proportionality of F(w) and |w| for w>10. The phase of F(w) is
of great importance in determining the frequencies of maxima
and minima, since it 1is6 an important component of iQle1'6, just
as the phase of Rle ' is in the application of Eq. (9),
Arg w 6
15
30
45
60
75
90
0.004
M
Fig. 3 Graph of the reflection function F(w) as
a function of the numerical distance w. Magnitude
of F(w) vs magnitude of w; phase of w as a para-
meter.
Fig. 4 Graph of the reflection function

F(w) as a function of the numerical distance
w. Phase of F(w) vs magnitude of w; phase
of w as a parameter.
using Q = Rp at higher grazing angles. Note that, as the graz-

ing angle increases, Rp goes positive, reducing the importance
of F(w) in expression ul). At the same time, through the re-
duction of (1 - Rp)2 and of cos^ in Eq. (13), w rapidly in-
creases, and F(w) becomes very small. Thus, for grazing angles
above 15 or 20, the Rudnick solution approaches the classic
one.
In Rudnick's analysis, the surface was assumed to be a
homogeneous porous medium. The case of a locally reactive med-
ium was treated subsequently by IngardS and by Lawhead and
Rudnick,^ who came up with essentially the same equations.
More recently, the problem has been studied by Delaney and
BazleyJO who based their analysis on the work of WiseJT
Other important contributions have been made by Piercy and
EmbletonJ2 Pao and EvansJ3 Aylor,'4 and Wenzel.'^ Delaney
and Bazley, however, arrive at precisely the same equation as
the classical analysis [Eq. (1)J, and their curve for Rp shows
the same value of -1 for grazing incidence, so that it is not
132 P. B. ONCLEY
clear that they have contributed to the solution of the low-in-

cidence-angle dilemma. Piercy and Embleton follow basically
the Rudnick-Ingard approach, as does Aylor (who is primarily
concerned with vegetation effects). Pao and Evans extend the
analysis to an intermediate transition layer: a useful step,
since in reality the Earth-air interface is not well defined.
Wenzel's analysis is quite different, although it leads to sim-
ilar asymptotic conclusions in limiting cases. Both Wenzel,
and Piercy and Embleton point out the analogy of the third term
in the propagation equation, such as F(w)(l - Rp) in expression
(11), to the ground wave in radio wave propagation. Experimen-
tal verification of the various analyses probably will depend
on getting better values for the acoustic constants of the
surface.
VI. Evaluation of Constants

The application of the preceding analyses has been ham-
pered by the inadequacy of experimental values for the acoustic
impedance and propagation constant of soils. The most useful
data on impedance are from studies by Walker and Doak^6 and
Dickinson and DoakJ? The latter reference points out the
problems of applying conventional impedance tube techniques to
soil in situ, and the variability of impedance with moisture
content, insolation, porosity, vegetation, and other factors as
measured with an open-field method at normal incidence. The
data from a number of measurements in North Wales are summa-
rized in Fig. 5.
There is a good reason to suspect that soil impedance may
vary with incidence angle, but no study of such variability has
been reported. It should be possible to make such measurements
using Eqs. (1) and (2) at incidence angles above 15. Exten-
sion to near-grazing angles using the Rudnick formulation would
be more difficult but possible.
Data on the complex propagation constant are even less
satisfactory. Rudnick used values for insulating building ma-
terials with data calculated theoretically based on ScottIS and
BeranekJ9 A classic theoretical reference 2is that of Zwikker
and Kosten. More recent work is by Burns, ^ Smith and Green-
korn,22 and Attenborough and Walker.23 The complex propagation
constant can be computed if complex impedance and complex den-
sity are known. Virtually no experimental data on transmission
in soil are available. Fortunately, Eqs. (2) and (13) are not
critically dependent on k2, since |k2l is usually appreciably
greater than 1, and the term (l-k/kzcos^) is omitted frequently
I l T i l l
2 Re(z)
0
o
T- O
+> ~<-
<a
T3
0)
Q_
-6
-8
IM(z)
-10 Grassy surfaces

Sharp sand
-12 Moisture 0 to 10%
Freshly raked soil
i___i i i i
200 500 1000
Frequency (Hz)
Fig. 5 Typical measurements of ground impedance.17
A comment is necessary on the sign of z. From Eq. (2),

representing the radical by M and z slnip by a + IB,
9 9
a + b -M 2iab
a + b2 + 2a + M
2
a 2
+ b + 2a + M
M is usually close to 1 + iO, and a is positive, and so,

to get a value of Rn in the first or second quadrant, b also
must be positive. Conventionally, the compliant reactance of
porous materials is qiven a negative sign, but to get the first
minimum (kAr + 6 = TT) below the hardwall minimum (as it is in-
134 P. B. ONCLEY
variably) the reverse convention must be accepted, and the re-

actance (typically -10 to -4 by Dickinson's data) must be
treated as positive. The same reversal is necessary in the
spherical wave equations.
VII. Applications in Data Correction and Extrapolation
There are two main ways in which the preceding equations
can be used. First, there is a need to correct out the effects
of interference in data experimentally collected, and second,
the expressions can be used, once equivalent free-field spectra
have been obtained, to extrapolate those spectra out to any de-
sired distance and height.
The first has been only partially successful. If the max-

imum and minimum frequencies in the correction do not match ex-
actly the frequencies experimentally obtained, the result can
be an uncorrected dip in one frequency band plus an undesired
peak in an adjacent one. Experimental data with two micro-
phones symmetrically disposed around the source may show quite
different interference patterns, and the minima at a fixed po-
sition often may shift into an adjacent band in a matter of
minutes during a test run. Since the speed of sound is strong-
ly dependent on temperature, a change in temperature can pro-
duce a shift in the interference pattern. Not infrequently al-
so, as on a cold sunny day, the temperature near a surface like
a concrete slab may be many degrees warmer than that of the up-
per air. Wind gradients, turbulence, and wind currents flowing
from a jet engine under static test also can distort the trans-
mission paths so that the actual value of kAr may be quite dif-
ferent from the calculated value.
Data correction holds more promise for measurements over
porous ground than over concrete, since in the former case var-
iations in kAr are less significant. Fig. 6 shows the differ-
ence between calculated corrections (AN) and measured values.
The measured values are the differences in one-third-octave
band pressure levels between data measured at a 215-ft distance
over a crushed rock surface and those obtained simultaneously
from a microphone supported by a balloon, 215-ft above the jet
engine source. Microphone and source were 13-ft above the sur-
face in the first case. A microphone above the source gives a
smooth pattern, generally accepted as the free-field value.
The measurements were made by General Electric at Cincinnati.
Calculated values are our own, using Eq. (10), including ex-
tended source averaging, and calculating [Q-jleMî) by Eqs.
(11, 12, and 16). The impedance values based on Dickinson's
data were (5 + i 15) P c at 100 Hz and (4 + i 4.0)Pc at 1000 Hz,
+5
QQ
s o
-5
50 TOO 1000 5000

Frequency in Hertz
Fig. 6 Computed and measured deviation from free field
(data from General Electric Company).
extrapolated linearly with log frequency to other points. k2/k
was chosen arbitrarily at 6 + 16. The agreement is excellent:
within a decibel in most frequency bands.
Better results over hard surfaces are obtained by locating
the microphone as near to the surface as possible. For scale-
model testing where high frequencies are particularly impor-
tant, the microphone should be recessed into the ground plane,
with the diaphragm exactly flush. Thus, by Eq. (3) the first
null frequency can be made high enough to avoid the entire test
band. Note that data so recorded are 6 dB above free-field
values. Care also must be taken to avoid thermal and wind gra-
dients that could cause "shadowing" and high-frequency loss due
to ray bending.
Surface microphones give good results for overflight mea-
surements at high elevation angles, with the microphone re-
cessed in a 4 x 4 sheet of 1/2-in. or 3/4-in. plywood faced
with aluminum. At lower angles (below 24, for example), the
effect of ground absorption is again important. However, good
results can be obtained with a microphone directly on a porous
surface using corrections as in the foregoing, provided that
the impedance is known. The correction curve for a surface
microphone is much smoother than at higher microphone positions
and can be used without concern for kAr errors. Where a micro-
phone height of 4 ft is specified (as in noise certification
specification FAR-36), more repeatable results could be ob-
tained by using a surface microphone and extrapolating the data
to 4 feet using a standard ground impedance value. This is not
now legal for certification purposes, however.
136 P. B. ONCLEY
Either the plane wave or spherical wave analyses can be

used successfully for noise prediction and extrapolation, al-
though the Rudnick expressions should be used for grazing an-
gles below 15. In prediction, the average or most probable
value is desired, and so it is not necessary to match the con-
ditions at a particular time and position as it is in data cor-
rection. Noise prediction accuracy will, however, still depend
on accurate values for z and k2 under representative condi-
tions. Tests with known z and k2 values under controlled con-
ditions and at various angles $ also are needed to validate the
equations and to permit a choice between the Rudnick analysis
and that of Delaney and Bazley. It is a profitable field for
future study.
References
, J. W. (Lord Rayliegh), The Theory of Sound, Dover,
New York 1945, Vol. 2, p. 78.
2
Howes, W. L., "Ground Reflection of Jet Noise," TN-4260,
April 1968, NACA.
3
Hoch, R. and Thomas, P., "Influence des reflexions sur les
spectres de pression acoustics des jets," 1st Collogue
d'Acoustique, 1968, Toulousse.
4
Rudnick, I., Journal of the Acoustical Society of America, 19,
348 (1947). ~~~
Sommerfeld, A., Annalen der Physik, 28, 665 (1909); also 81,
1135 (1926). ~ ~
6
Van der Pol, B., Physica 2_, 843 (1935).
Norton, K. A., Proceedings of the Institute of Radio Engi-
neers, 24^ 1367 (1936); also 25, 1203 (1937).
o
Ingard, U., Journal of the Acoustical Society of America, 23,
329 (1951). ~
g
Lawhead, R. B. and Rudnick, I., Journal of the Acoustical
Society of America 23_, 541; also Z3, 546 (1951).
10
Delaney, M. E. and Bazley, E. N., Journal of Sound and Vibra-
tion 13_, 269 (1970).
]1
Wise, H., The Bell System Technical Journal 8, 662 (1929).
12
Piercy, J. E. and Embleton, T. F. W . , "Effect of Ground on
Near-Horizontal Sound Propagation," Automotive Engineering Con-
gress , 1974, Paper 740211, Society of Automotive Engineers.
13 Pao, S. P. and Evans, L. B., Journal of the Acoustical Socie-
ty of America 49_, 1069 (1971).
Aylor, D., Journal of the Acoustical Society of America 51,
197; also 5j_, 414 (1972).
Wenzel, A. R., Journal of the Acoustical Society of America
55_, 956 (1974); also Oncley, P. B., Journal of Sound and Vibra-
tion J3_ (1), 27 (1970).
16
Walker E. J. and Doak, P. E., "Effects of Ground Reflection
on the Shape of Sonic Bangs," 5th Internation Congress on
Acoustics, 1965, Paper L-55, Liege.
Dickinson, P. J. and Doak, P. E., Journal of Sound and Vibra-

tion U_ (3), 309 (1970); also Dickinson, P. J., "Free Field
Measurement of the Normal Acoustic Impedance of Ground Sur-
faces," Doctoral Thesis, 1969, University of Southampton.
18
Scott, R. A., Proceedings of the Physical Society, 58, 165,
Pt. 2 (1946).
IQ
Beranek, L. L., Journal of the Acoustical Society of America
1^, 248 (1942).
20
Zwikker, C. and Kosten, C. W., Sound Absorbing Materials,
Elsevier, New York, 1949.
21
Burns, S. H., Journal of the Acoustical Society of America
49^ 1 (1970).
op
Smith, P. G. and Greenkorn, R. A., Journal of the Acoustical
Society of America 52^ 247 (1971).
23Attenborough, K. and Walker, L. A., Journal of the Acoustical
Society of America 49_, 1331 (1970).

METHODS FOR THE PREDICTION OF
AIRFRAME AERODYNAMIC NOISE
James D. Revell* and Gerald J. Healyt

Lockheed-California Company, Burbank, Calif,
and
John S. Gibson*
Lockheed-Georgia Company, Marietta, Ga.
Abstract
Two of the possible methodologies for the prediction of

airframe (nonengine) aerodynamic noise are described as: 1)
the whole aircraft method; and, 2) the aircraft drag element
method. A brief review of background data is presented, and
the two prediction methods are covered in detail. The whole
aircraft method is based entirely on measured airframe noise,
whereas the aircraft drag element method is formulated from
knowledge of the drag coefficients of the major airframe noise
contributors with a number of constants evaluated from measured
airframe data.
Background
Airframe noise is defined as the nonpropulsive aerodynamic

noise of an aircraft moving through the air. This problem
first became of interest from the standpoint of military air-
craft detectability and led to early tests of small gliding
aircraft.1~3 These results imply an airframe noise floor for
large aircraft at landing approach^>5 which appears to be con-
firmed by a few tests on large aircraft^"^ and by more recent
analytical studies, > 1 0 which estimate the noise floor to be
FAR 36 minus 10 EPNdB. The above results pertain to large-
aspect-ratio wings. However, some recent tests on low-aspect-

* Manager, Acoustics R&D Staff
'Research Scientist
Âcoustics Group Engineer
139
140 J. D. REVELL, G. J. HEALY AND J. S. GIBSON
ratio aircraft (an F106B delta wing,^ and the Jetstar-^) seem
to indicate lower noise levels than earlier predictions would
imply.3>13 ê present paper describes recent developments
in prediction methods which appear to be useful. Reference 10
should be consulted for other methods.
Two possible approaches to the prediction of airframe

aerodynamic noise can be described as 1) the whole aircraft
method; and, 2) the aircraft drag element method. These two
methods have produced useful results which are described
herein. The present paper is a much shortened version of
Ref. 14 which should be consulted for more details.
Whole Aircraft Method
This is an empirical method where the OASPL for a clean

aircraft is defined in terms of simple aircraft design para-
meters and speed. A single nondimensional frequency spectrum
is used in conjunction with a characteristic ("Strouhal") fre-
quency. The various parameters are fitted statistically to
flyover test data. Two well known methods are those of Refs.
3 and 10; the method of Ref. 3 will be shown in its original
form (for recent refinements, see Ref. 13 ). The method
is fast, simple, and suitable for a slide rule or desk calcu-
lator. The chief weakness is its lack of generality for
various "dirty" configurations and low aspect ratio wings. The
original airframe aerodynamic noise equation developed by
Lockheed for the actual measured overall sound pressure level
(OASPL), is
OASPL = 10 log [(1/h)2 (V4W/C ) (C/b)l + 11.2 (1)

IU L L J
where h is the altitude in feet, V is velocity in knots, W is

weight in pounds, C and b are the wing chord and span, respec-
tively, in feet, and C-^ is the lift coefficient. The ratio
C/b is the reciprocal of aspect ratio. The OASPL is actually
related to the sixth power of velocity, since weight is approx-
imately equal to lift, which, in turn, is related to the second
power of velocity. A more detailed and refined analysis of
these same data^ resulted in the following equations:
OASPL = 10 log1Q [(sinc)>/r)2 (M6S/AR4)] + 179.9 (2a)
OASPL = 10 log [(sinc)>/r)2 (V6S/AR4)] + 28.0 (2b)
where sine)) accounts for the lift dipole directivity (<(> = 90

for overhead position), r is the observer-to-aircraft range in
PREDICTION OF AIRFRAME NOISE 141
0.125 0.25 0.5 1 2 4 8 16 32

NONDIMENSIONAL FREQUENCY, f/fmax
Fig. 1 Nondimensional aerodyanmic noise spectrum.
meters, V is the velocity in meters per second (M is the flight
Mach number), S is the wing area in square meters, and AR is
the aspect ratio. Equations (2a) and (2b) are recommended for
calculation of the far-field airframe noise OASPL from aero-
dynamically "clean" configured transport aircraft. These
equations have shown good results except for low-aspect-ratio
wings11>i^ ; however, an improved correlation is shown in a
later section by the drag element method.
A frequency spectrum, in terms of one-third-octave band

levels relative to the OASPL vs nondimensional one-third-octave
band center frequency, was developed from data acquired in
Ref. 3, and is presented in Fig. 1. The frequency defining
the peak of the spectrum level is given by
max (1.3) (V/t) (3)
where 1.3 is the Strouhal number determined from the data, V

is velocity, and t is the mean wing thickness, at the location
of the mean wing chord (C). The mean wing chord is defined as
C = S/b where S and b are wing area and span, respectively.
"Clean aircraft" results generalize roughly to a "dirty" land-
ing approach configuration by adding 9 to 12 dB. Distinctions
between "clean" and "dirty" aerodynamic configurations require
a more refined method called the "drag element method."
The Drag Element Method
This method, first presented in Ref. 9, assumes that air-

frame aerodynamic noise is a small fraction of the mechanical
Table 1 Noise prediction parameters3
j drag component
Kj, D
tr
dB ft/
1) Wing p r o f i l e drag 78 Sw 3 2.6 0.1

2) Wing induced drag 63 Sw 3 0 1/TT
3) Fuselage drag 76 7rdl f /2 3 2.6 0.05
4) Nacelles 76 Tidl U S /2 3 2.6 0.05
5) Horizontal tail 78 S HT 3 2.6 0.1
6) Landing gear 85 C S ) 1 2.6 0.2
7) Leading edge slat 78 S

LES 3 2.6 0.1
Common reference values for all components are

C
DR = -0106> SR = 100 ft-2vR = 192 knots,
hR = 500 ft., (VTE/V)R = 0.974
kplanform areas.
energy dissipated by the various components of drag of the air-

plane, as did Lighthill^ for jet noise. The various com-
ponents each have a unique characteristic Strouhal frequency;
therefore, the total acoustic energy covers a wide frequency
range, requiring a spectral energy summation to combine the
radiated noise of multiple drag components. Preliminary
results^ show good correlation with earlier data in Refs. 3
and 7. The extension of Ref. 9 is the basis of this section.
The bulk of flyover noise data discussed in Refs. 1-14

display a velocity to the sixth power dependence which charac-
terizes dipole noise. " Therefore, the present method assumes
that the primary sources are random lift dipoles distributed
along the wing trailing edge. These dipole strengths are
determined by lift force fluctuations, caused by pressure
fluctuations, with correlation lengths related to boundary layer
thickness. These trailing edge pressure fluctuations are
governed by Reynolds stresses due to velocity fluctuations in
the boundary layer and the near wake .-^ According to
Ref. 18, the wake effect on wing pressure fluctuations
has an extra component due to the turbulent swirling shear
layer surrounding the trailing vortices, implying that there
is an effect related to induced drag, estimated herein by
empirical data of Ref. 7. The trailing edge dipole model
adopted is like that of Ref. 19, except that each component
source strength is related to a function of the component
drag coefficient. The types of drag related noise discussed

are 1) profile drag type noise caused by the nearly two dimen-
sional boundary layer flow past the trailing edge of a wing,
trailing edge flap, leading edge slat or similar device,
2) landing gear of bluffbody drag noise; and 3) induced drag
noise. The radiated noise due to the axisymmetric boundary-
layer flow past the closure of the fuselage also is treated
as a special form of profile drag noise.
The OASPL for each drag component j is expressed as

OASPL, = + 10 log in F. where
-i I f -i 11 rP l f u ? l f \ 7 l / \
^ ^-ll^-H-il I /-^1/r^-l (4)
Recommended values of the various parameters are given in

Table 1. The Strouhal frequency (peak of the radiation
spectrum) is defined by
f
St " <Str)VTE(J)/deq(J) (5)
deq(j) = CD(j) c(j) (6)
where d e g is the equivalent cylinder diameter defined as the

product of element profile drag coefficient times chord for
a streamlined element. The equivalent cylinder concept for
streamlined bodies is a generalization of the case of circular
cylinder where C-^ = 1 and d = d.
An empirical one-third-octave band spectrum is derived from

flight data of Ref. 3 in a nondimensional form (Fig. 1) as
x = f^/^St' where f^ is the center frequency of a given one-
third-octave band. The chosen spectrum fits data for the DC-3
and Convair 240^ for low lift coefficient conditions where pro-
file drag noise dominates. This same spectrum shape has given
useful results and is employed for all components analyzed in
this paper; therefore, the spectrum is changed only via the
Strouhal frequency. The component spectrum equation and the
normalized one-third-octave band energy spectrum are given in
Eqs. (7) and (8), respectively.
(SPL)1/3. = OASPL. + 10 log |S(x)|. (7)
J/z
S(x) = SQn(x) exp - as lx-1 (8)
s /?
SQ(x) = x D / K; x < 1
S Q ( x ) = K/\/x~; 1 < x < 2
S Q ( x ) = K\/2/x; 2 < x < 4
S 0 ( x ) = K(N/2/4) ( 4 / x ) 2 ; x > 4 (9)
where a s = 0.002 and K = 0.15. The individual spectra so

obtained then are summed over the multiple components to
obtain the composite spectrum.
The theory discussed in Ref. 14 shows that clean wing

profile drag noise level depends on the cube of the profile
drag coefficient c^p . Two factors of c^ arise from the
assumption that trailing edge pressure fluctuation is propor-
tional to cjp by analogy to flat plate data20, while the third
arises because of the spanwise correlation length which is
assumed proportional to trailing-edge boundary-layer thick-
ness, and thereby to c^p . Also in Ref. 14, a factor of VTE/V
to the exponent 2.4 arises from the use of boundary-layer
theory relations between local trailing-edge velocity to free-
stream velocity ratio, VT/V and drag coefficient, plus the
assumption that the noise relates to the sixth power of the
trailing edge velocity. The assumptions yield two partially
offsetting functions of VTE/V. The trailing-edge velocity
ratio can be determined from surface pressure data for a wing,
or from boundary-layer theory calculations (Ref. 14 gives a
rough empirical formula). The theory in Ref. 14 (Eq. 40) gives
a result like Eq. (4) with the drag exponent n = 3 and m = 2.4.
In predicting the profile drag noise of wings with

deflected flaps, a modified drag function is recommended,
because a continuation of drag coefficient to the third power
tends to overpredict the incremental flap noise of the C-5A.
Increased flap profile drag decreases the Strouhal frequency
compared to a clean wing in Eq. (5). The equation for
deflected flap profile drag noise is
3 2
OASPLW ~ 10 logi n "AC

10 C C
\ DR /
+ nondrag terms in Eq. (4) (10)
Leading-edge slats, horizontal tail, fuselage, and

nacelles are all characterized by streamwise boundary-layer
growth and trailing-edge dipole forces similar to those
associated with wing profile drag noise and are governed by

the same prediction method. For leading-edge slats on thick
wings at high lift coefficients, the trailing-edge velocities
can be quite high, with VTE/V up to 1.5. Leading-edge slats
have a small chord (about 10% of the wing chord) and a low drag
coefficient which lead to high Strouhal frequencies that affect
the high-frequency spectrum.
In Ref. 9 , the landing gear noise was evaluated by includ-

ing its drag with the fuselage. Further analysis of the data
of Refs. 7, 8, 11, and 12 has raised the assessment of the land-
ing gear noise in relation to the profile drag noise of wings
with deflected flaps. This relative change is implied in the

parameters of Table 1. Also, further analysis of landing gear
effects for the C-5A^ and JetStar^ has led to a new approach.
A typical bluffbody drag coefficient C^^ = 1 with an exponent
of 1 is used to define the "proper" effective landing gear
cross-sectional area Ajj-g which is derived from the airplane
landing gear drag coefficient, referred to wing planform area.
It is also convenient to define the span of the landing gear
as b-n-g which permits the evaluation of an equivalent cylinder
diameter (deq)p- (or bluffbody frontal width). The equations
defining A^g and (deq)g are:
A =AC
Trg D V C DTT =AC
DgSw
d = A ,b = Ac S /b (12)
TTg TTg/ TTg Dg W TTg
Wing induced drag/trailing vortex dipole noise predictions

involve the fundamentals of turbulent fluid dynamics 1*7
which show that turbulent wakes will induce an increment
to the wing trailing edge pressure. In Ref. 18 this idea is
applied to the swirling turbulent shear layers about the
trailing vortices emanating from the side edges of deflected
flaps on wings typical of landing approach conditions. This
analysis employs empirical turbulence level estimates and a
vortex growth law derived from turbulent eddy viscosity con-
cepts. The resulting noise equation shows nearly an induced
drag cubed law at typical landing approach lift coefficients
and flap settings. In Ref. 145a similarity argument is given
wherein the characteristic velocity is an average vortex swirl
80
00
o
a.
C/)
70
DQ
LU WING PROFILE DRAG
60 .FREQUENCY 222 Hz
PROFILE DRAG
O
O CONTROLS SPECTRUM
I I I I
Q
FLTTEST (HEALY, REF. 3)
SE 50
^THEORY
LU
40
50 100 200 500 1k 2k 5k 10k
FREQUENCY, Hz
Fig. 2 Theory vs gliding flyover test data for "clean" wing

Convair 240.
velocity in the near wake for a flapped wing segment with a

section lift coefficient c-^ given by:
v t /V (2TiARf(dvtxob)) (13)
The characteristic length is assumed to be an average

vortex core radius rc = bf(dvtxob)/2, where bf is the flap
segment span, dvtxob is an average ratio of vortex core dia-
meter to flap segment span, and ARf is the flap segment
aspect ratio. The associated Strouhal frequency becomes
fs = v /(2irr ). These results combine to give the following
relation for OASPL due to trailing vortices
(QASPL)vtx = 6 3 + 1 0 log
Qvtx = < I-
(14)
The empirical constant 63 dB was derived from C-5A test data

(40 flap setting at a lift coefficient C L = 1.48).
Drag Element Method Airframe Noise Predictions
Comparisons will first be made for large-aspect-ratio

clean wings. Figures 2 and 3, taken from Ref. 9, show free-
70
OQ
o
Q_
C/D
Q 60 S
1
LU WING
< 50 PROFILE DRAG-
FREQUENCY O
120 Hz 0
Q i I
CL THEORY I
1
5 40 A FLIGHT TEST 11
LU
0 FLIGHT TEST 13
O
30
o
50 100 200 500 1k 2k 5k 10k
FREQUENCY, Hz
Fig. 3 Theory vs gliding flyover test data for "clean 11 wing
DC-3.
field calculations, compared with flyover test data from

Ref. 3, respectively, for the minimum weight Convair 240 at
192 knots and for the DC-3 at maximum weight for two speeds.
Figure 4 shows predictions for the clean C-5A (flight test
A-l)' with and without ground reflections. The interpretation
of the results of Figs. 2 and 3, however, now is slightly
different from that presented earlier.^ Currently,the clean
wing profile drag radiation dominates the entire frequency
spectrum for all cases shown in Figs. 2 and 3. Previously,9
it was estimated that induced drag controlled the 108-knot,
minimum speed DC-3 case (CL = 0.66) at high frequencies. This
effect now is masked by the profile drag noise, because of the
new parameters shown in Table 1. The results in Figs. 2 and 3
did not consider atmospheric absorption, which explains the
apparent overprediction of noise levels above 5000 Hz. Recent
calculations for the DC-3 and Convair 240 produce OASPL values
falling between those values derived,in Ref. 3, from the com-
posite spectrum and the lower OASPL values derived from the
smoothed spectra. Calculations for the clean C-5A shown in
Fig. 4, include ground reflections which tend to be smoothed
by time-averaging in the flight data. It is believed, there-
fore, that Figs. 2-4 provide good substantiation for the pre-
sent predictions of clean wing profile drag, which is the basis
for the predictions of all of the other components of airframe
noise, except for the landing gear and the induced drag effect.
CD 100
"O <MEAUSRED (RUN A-1)
-FAN FUNDAMENTAL
Q.
CO
90
o Qn
80
p
6 PREDICTED
DC (GROUND PREDICTED
/
f 70 I REFLECTED) (FREE-FIELD)
\ / \
V ESTIMATED AERO NOISE^\

. I i i I i il__________I i , I i
60
20 40 70 100 200 400 700 1k 2k 4k 7k
FREQUENCY, Hz
Fig. 4 C-5A Galaxy flyover test data vs theory for "clean"
configuration.
Comparisons will next be made for large-aspect-ratio wings
with landing gear plus flap extension. Figures 5 and 6 present
results of freefield calculations which compare favorably with
C-5A test data7 for flap angles of 16 and 40, respectively,
with landing gear down. Results of recent calculations with
the current method differ negligibly from Ref. 9, from which
Figs. *5 and 6 were taken. The current method of evaluating
landing gear effects, compared to that of Ref. 9, has resulted
in a reassessment of the spectral composition of the low-
frequency part of the spectra shown in Figs. 5 and 6. Pre-
viously, this portion of the spectrum was attributed entirely
to the profile drag of the wing with deflected flaps, whereas
the present analysis indicates that the landing gear produces
equal noise. These predictions also indicate that, at high
frequency, the one-third-octave SPL values are dominated by
the induced drag noise (above 125 Hz at the 40 flaps angle,
CL = 1.5, and a>ove 200 Hz at the 16 flap angle, CL = 0.96).
Comparisons of landing gear drag noise with flaps

retracted are shown in Figs. 7 and 8. Figure 7 shows the
results of the current prediction method with those from C-5A
flight test A-2 (landing gear down, flaps retracted).7 The
calculations are shown with and without ground reflections for
a 5 ft microphone elevation. Figure 8 shows predictions of
the effect of landing gear deployment for the NASA Edwards FRC
JetStar (flight test 18).12 The analysis assumes that the gear
drag was distributed equally among the fuselage, the wing, and
the gear as a separate element since this aircraft has gear
elements both on the wing and the fuselage.
100 I
CD LANDING GEAR - INDUCED DRAG
T3
CONTROLS SPECTRUM CONTROLS SPECTRUM-
35 90
GQ
LU
80
<
O 70
O
Q EXTRAPOLATED-
- PRESENT THEORY
DC
I O dvtxob = 0.006 AERO NOISE
h- 60 (REF.7)
LJLJ A dvtxob = 0.007 1 I I
FLIGHT TEST RUN A-3 (GIBSON, REF. 7)
50
20 40 70 100
200 400 700 1k 2k 4k 7k
FREQUENCY, Hz
Fig. 5 C-5A Galaxy flyover test data vs theory for 16 (40%)
flaps, landing gear extended.
\ r^
en luu
/^
~fa.
"0
-^^
^
^^^
_i
kT^Xj i v \ZH
Sssj A/ N
90
S LA NDING
^H
' 0
<
- irMDUCEC
GE AR
CO ^_^D RAG '^V^
^ r-r n
LU 80 \ji\j NTROLS
> SP ECTRUM
"* c ONTRO LS
-NVSo
< O PECTRLJM~
NLv
vv <J

6
70
PRESENT THEORY
FAN FUNIDAIV ENTAL
^[ V<3
>^^
EXTF APC iLATEC
f*~
QC
AERC)NO ISE-^" 0
V dvtxob = 0.008 \V
E 60 (REF 7) V
O dvtxob = 0.007 \
LU s
%
FLIGHT TEST RUN A- 5 ( G H3sorSI, REF. 7)
o
20 40 70 100
200 400 700 1k 2k 4k 7k 10k
FREQUENCY, Hz
Fig. 6 C-5A Galaxy flyover test data vs theory for 40 (100%)
flaps, landing gear extended.
Predictions of noise for low-aspect-ratio clean aircraft

have been a problem until development of the present method.
It has been noted that early prediction methods have over-
estimated considerably the noise of low-aspect-ratio wings,
03 110
PREDICTED MEASURED
(FREE-FIELD) (RUN A-2)
Q_
CO A
100 FAN FUNDAMENTAL
CO
90
o N
*. ^N
P 80 PREDICTED/ <V NN
G (GROUND REFLECTED) ESTIMATED
OC
I
LU
70
20 40 70 100 200 400 700 1k 2k 4k 7k
D
FREQUENCY, Hz
Fig. 7 C-5A Galaxy flyover test data vs theory for "clean"
wing, landing gear extended.
CO
CL
CO
CD
LU
<
CJ
O
Q
OC
z
o
500 1k 2k 5k 10k
FREQUENCY, Hz
Fig. 8 Effects of landing gear deployment on JetStar with
flaps retracted: theory vs flight data.
especially the delta wing F106Bn(600 leading edge sweep angle,

aspect ratio 2.08). Figures 8 and 9 show comparisons of analy-
tical vs flight test results for the JetStar,12 and Fig. 10
shows the same comparison for the F106B.11 Figures 8-10 all
show good agreement between the present method and the test
data.
CD
T3
Q.
CO
<
00
0
0
Q
CC
- ENGINE NOISE
- ONESECOND AVERAGE (SEE TEXT)
50 100 200 500 1k 2k 5k 10k

FREQUENCY,Hz
Fig. 9 Effects of velocity on "clean" wing JetStar: theory
vs flight data.
LJJ
FLIGHT DATA
70
G CO
Eo
IZ
60
40 63 100 160 250 400 630 1k 1.6k 2.5k 4k 6.3k

FREQUENCY, Hz
Fig. 10 Theory vs flyover test data for "clean" F-106B.
The dash-dot curve in Fig. 9 (JetStar flight 15 at 365

knots) represents an average of the acoustic energy over a
time interval of 1 sec, centered about the visual overhead
time, and is estimated to give a value 3 dB below the calcu-
lated spectrum at source emission time. As can be seen, the
measured flyover noise spectrum falls between this curve and
that predicted for the instantaneous noise signature.
Finally, the F106B data are considered. Figure 10 shows

a comparison between the current prediction method and the
data of Ref. 11, adjusted to freefield conditions. As can be
seen, the predictions are in good agreement with flight test
data, although somewhat downward shifted in frequency. In this
case, the Strouhal number for large-aspect-ratio aircraft was
used and may not be applicable to delta wings. The nacelles
contribute about 1 dB in the frequency range between 250 and
1600 Hz, whereas the fuselage makes a significant contribution
below 200 Hz, actually dominating the predicted spectrum below
125 Hz, where the agreement between the extrapolated flight
data and the prediction is very good, indicating that the
selections of the fuselage noise level constant and Strouhal

number are reasonable.
In summary, the present method affords good predictions

of the F106B and JetStar which had been overpredicted seriously
by earlier methods; therefore, it is believed that some pro-
gress has been made in this respect as exemplified by the good
comparisons with flight test data shown in Figs. 2-10.
Concluding Remarks
Two possible approaches to the prediction of airframe

aerodynamic (nonpropulsive) noise have been discussed which
entail different levels of efforts. Prediction procedures
and results are presented for both the whole aircraft method
and the drag element method. These methods, in the opinion
of the authors, reflect the state-of-the-art in each of these
categories. The choice of methods depends largely on the
effort appropriate to the study in question. The whole air-
craft method is presently limited to clean configured large-
aspect-ratio wings, whereas the drag element method is capable
of handling a wide range of aircraft types and flight con-
figurations. It is believed that this method represents a
significant advance, but that much work needs to be done,
especially in the assessment of landing gear effects. The
method is considered useful for preliminary design although
some knowledge of aerodynamic drag and velocity field esti-
mation is required; however, fairly rough estimates have
provided good results.
The airframe aerodynamic noise has been referred to as

"the ultimate noise barrier," since it represents the minimum
noise an aircraft in motion will generate in the absence of
propulsion system noise. This establishes a lower bound for
practical engine noise reduction. The present method confirms
the airframe noise floor level estimated at FAR 36 minus
10 EPNdB at landing approach as given in Ref. 9. This con-

clusion implies a definite limitation on the near term achieve-
ment of large noise reductions for aircraft similar to the
current subsonic aircraft. Beyond this point, further vehicle
noise reduction would, of necessity, require an aircraft
designed to be aerodynamically quieter.
References
^-"Far-Field Aerodynamic Noise Measurement Program ," Rept .

LR 23640, June 1970, Lockheed-California Company.
2
Smith, D. L., Paxson, R. P., Talmadge, R. D., and Hortz,
E. R. , "Measurements of the Radiated Noise from Sailplanes,"
TM-70-3-FDDA, July 1970, Air Force Flight Dynamic Laboratory.
, G. J., "Measurement and Analysis of Aircraft Far-Field

Aerodynamic Noise," CR-2377, Dec. 1974, NASA.
, J. S. , "The Ultimate Noise Barrier - Far Field

Radiated Aerodynamics Noise," Inter-Noise f 72; International
Conference on Noise Control Engineering, Oct. 4-6, 1972,
Washington, D.C., pp. 332-337.
5
Blumenthal, V. L., Streckenbach , J . M. , and Tate, R. B. ,
"Aircraft Environmental Problems," Journal of Aircraft,
Vol. 10, No. 9, Sept. 1973, pp. 529-537.
Gibson, J. S., "Non-Engine Aerodynamic Noise: The Limit

to Aircraft Noise Reduction," Proceedings of the Institute
on Noise Control Engineering (Inter-Noise '73), Aug. 22-24 ,
1973, Copenhagen, Denmark, pp. 445-454.
'Gibson, J. S., "Non-Engine Aerodynamic Noise Investigation

of a Large Aircraft," CR-2378, Oct. 1974, NASA.
o
Gibson, J. S. , "Recent Developments at the Ultimate Noise
Barrier," Paper 74-59, Aug,. 1974, International Council of
the Aeronautical Sciences.
^Revell, J. D., "The Calculation of Aerodynamic Noise

Generated by Large Aircraft at Landing Approach," 87th
Meeting, Acoustical Society of America, April 1974, Paper JJ9.
10
Hardin, J. C., Fratello, D. J., Hayden, R. E., and
Kadman, Y. , "Prediction of Airframe Noise," TN-D7821, NASA.
Burley, R. R., "Preliminary Measurement of the Airframe Noise

From an F-106B Delta Wing Aircraft at Low Flyover Speeds,"
TMX-71527, March 1974, NASA.
I O
^Lasagna, P. L. and Putnam, T. W., "Preliminary Measurements
of Aircraft Aerodynamic Noise," AIAA Paper 74-572, June 1974.
13
Healy, G. J., "Aircraft Far-Field Aerodynamic Noise - Its
Measurement and Prediction," AIAA Paper 75-486, March 1975,
(published elsewhere in this volume).
14
Revell, J. D., Healy, G. J., and Gibson, J. S., "Methods

For the Prediction of Airframe Aerodynamic Noise," AIAA Paper
75-539, March 1975.
Lighthill, M. J., "On Sound Generated Aerodynamically,

I: General Theory," Proceedings of the Royal Society,
Ser. A, Vol. 211, 1952, pp. 564-587.
16
Curle, N., "The Influence of Solid Boundaries Upon Aero-
dynamic Sound," Proceedings of the Royal Society (London),
Ser. A , Vol. 231, p p 504-514, 1 9 5 5 . " "
Townsend, A. A., page 27, The Structure of Turbulent Shear Flow,

Cambridge University Press, Cambridge, England 1956.
18
Revell, J. D., "Induced Drag Effects on Airframe Noise,"
AIAA Paper 75-487, March 1975 (published elsewhere in this
volume).
1
Hayden, R. E., "Noise from Interaction of Flow with Rigid
Surfaces: A Review of Current Status of Prediction
Techniques," CR-2126, 1972, NASA.
20
Schloemer, H. H., "Effects of Pressure Gradients on
Turbulent Boundary Layer Wall Pressure Fluctuations," Journal
of the Acoustic Society of America, Vol. 42, No. 1, pp 93-113,
July 1967.
REVIEW OF THEORY AND METHODS FOR TURBINE
NOISE PREDICTION
D. C. Mathews,* R. T. Nagel,+ and J. D. Kester*
Pratt & Whitney Aircraft, East Hartford, Conn.
Abstract
The state-of-the-art of turbine noise prediction is reviewed. A literature sur-

vey was conducted, and current methods for turbine noise prediction are discussed
iand compared. The technology required for advanced prediction methods is
discussed, including generation of noise by turbine blade and vane elements, pro-
pagation and attenuation of the noise through blade rows, and noise radiation
through the turbulent jet to the far field. Research areas are suggested which are
critical to the future improvement of turbine noise prediction methods.
I. Introduction
The noise signature from a modern aircraft engine includes contributions

from several engine components. As technology has progressed, methods have
been developed to reduce noise from the major sources. Although the net effect
on the community has been to reduce noise exposure, the problems that face the
aircraft engine designer have become more complex. The contribution to overall
engine noise of components that once could be ignored now may be dominant
factors. Turbine noise is an example of one such component. As the impor-
tance of turbine noise has increased, greater emphasis has been placed on the de-
velopment of methods to predict this noise. This paper presents the current
state-of-the-art in turbine noise prediction methods and discusses possible direc-
tions for the development of improved technology in this area.
To put turbine noise in its proper perspective, it must be considered along

with all other engine noise sources. These sources are illustrated for the JT8D-
109 turbofan engine in the cross-sectional drawing of Fig. 1. In the early turbo-
jet engines, jet exhaust noise was the dominant factor. With the advent in the
Presented as Paper 75-540 at the AIAA 2nd Aero-Acoustics Conference,

Hampton, Va., March 24-26, 1975.
* Assistant Project Engineer.
+
Analytical Engineer,
* Chief, Acoustics Engineering,
155
156 D. C. MATHEWS, R. T. NAGEL AND J. D. KESTER
FAN NOISE TURBINE NOISE JET NOISE

COMPRESSOR CORE
NOISE ENGINE
NOISE
Fig. 1 Turbofan engine component noise sources (JT8D-109

engine).
TOTAL ENGINE NOISE
PEAK
NOISE
PNdB
TOTAL ENGINE NOISE

TURBINE NOISE (UNTREATED]
. , 1 1 . 1 , 1
1960 1965 1970 1975 1980
DATE OF ENGINE INTRODUCTION
Fig. 2 Importance of turbine noise in turbofan engines.
early 1960's of the more efficient turbofan cycle engines, jet exhaust noise was
reduced markedly but at the expense of increased whine from the fans. Second-
generation high-bypass-ratio turbofans, such as the JT9D, CF-6, and RB.211,
provided even lower levels of jet exhaust noise and introduced new features to
provide significant reductions in fan noise. As a result of the fan and jet noise
improvements provided by these high-bypass turbofan engines, it was revealed
that noise from the turbine made a significant contribution to the engines' far-
field noise signature.
The significance of turbine noise is illustrated by the trend curve of Fig. 2

at aircraft approach power. Significant contributions of turbine noise from high-
TURBINE NOISE PREDICTION 157
bypass engines such as the JT9D required that acoustical linings be installed in
the tailpipes of these engines to improve the noise signature. Based on the trends
illustrated by Fig. 2, it is obvious that turbine noise will continue to be a signifi-
cant factor in the noise signature of future aircraft engines. It is this trend that
motivates the need for accurate and proven methods to predict turbine noise.
II. Current Status of Turbine Noise Predictions
Methods to predict noise from engine components fall into two basic cate-
gories: empirical and analytical. Analytical methods are possible only when the
aerodynamic and aeroacoustic characteristics of the source generation, propaga-
tion, and radiation processes are recognized and understood to the extent that
they can be described by reliable mathematical models. Although the develop-
ment of an analytical prediction method is the ultimate goal for any prediction
system development effort, this goal seldom is realized fully in practice. A
variety of elements are required to model analytically the propagation of fan
noise, including the modal propagation concepts introduced by Tyler and
Sofrin, the functions developed by Sears^ to calculate the response of airfoils
to gusts, and the models introduced by KajP' ^ and Mani^ to allow the calcula-
tion of noise propagation through upstream and downstream vane rows. Even
though relatively massive technical efforts have been undertaken to develop
analytical prediction methods for fan and jet noise, the methods in current use
are not purely analytical but are a blend of analytical and empirical techniques.
The importance of the turbine as a noise source was recognized relatively
recently, and, therefore, turbine noise prediction methods have not received as
much attention as fan and jet noise sources. Our survey of the literature on tur-
bine noise prediction methods has disclosed that few analytical models of tur-
bine noise problem elements are reported, and that most prediction methods are
purely empirical. Rather than focusing on details of the unsteady aerodynamic
processes that are at the heart of the physics of turbine noise generation, these
methods tend to focus on the thermodynamic parameters that are used to de-
scribe the mean properties of flow through the turbine. Typical steady-state
parameters that often are utilized include mass flow, temperature, pressure ratio,
turbine speed, and, in some cases, overall work extraction.
Because noise generation, which results from unsteady aerodynamic pro-

cesses, is being correlated with steady-state turbine operating parameters, it is
obvious that the relationships involve several implicit assumptions. These rela-
tionships are discussed in the following sections for several current prediction
methods. It should be recognized that all of the reported methods have com-
mon inherent limitations because of the similar nature of the correlation para-
meters used:
1) The magnitude-of-time unsteady processes are related to mean flow pro-
cesses. Although this assumption is valid marginally for a given turbine design,
it is obviously unwarranted for any new designs except those that are merely
aerodynamically "scaled" models. In cases where noise predictions are required
for studies of "paper engines" or to develop parametric trend studies, only mean
flow information is available. Consequently, there will be a continuing require-
ment for prediction systems that embody this concept. However, it should be
recognized that the accuracy of these prediction systems is limited by this very
basic assumption.
2) Similarity of design. The rationale that underlies empirical turbine noise

procedures that extrapolate data from one design to predict noise from another
is that the designs are "similar." By similarity, it is meant that the aerodynamic
features of the turbine are nearly identical or that the range of aerodynamic
parameters open to the turbine designer is restricted by turbine design criteria.

Typical factors established by design criteria include pressure ratio per stage,
blade and vane aspect ratios, gap-chord ratios, and airfoil thickness and camber
distributions. Turbine design criteria are developed by each engine manufacturer
based largely on past experience, and these criteria may differ substantially be-
tween manufacturers. Consequently, it may well be the case that an empirical
procedure that relates turbine noise to turbine mean flow properties may provide
acceptably accurate estimates for designs of a given "family" but fail when ap-
plied to a design that incorporates different criteria.
3) Data base orientated. It is common that the most complete and accur-
ate turbine acoustic data available to a manufacturer for use in an empirical pre-
dection system are those measured from its own products. Because noise is a
factor in the competitive sale of engines, it is common that these data are not re-
leased publicly. Consequently, it is difficult to establish the extent to which a
prediction procedure developed by one manufacturer applies to predict accurate
levels for designs made by others. With this background, some of the current
procedures will be discussed in the following section to illustrate the similarities
and differences encountered in state-of-the-art turbine noise predictions.
III. Description of Current Turbine Noise

Prediction Methods
Emphasis in this section is placed on the prediction of turbine noise levels

rather than on a detailed description of the predicted spectral and directivity
characteristics. Later sections, which compare the various methods, will provide
illustrations of the predicted similarities and differences in these characteristics.
Rolls-Royce Method
Smith and Bushell" presented an early empirical prediction procedure that

was based on data obtained from full-scale Rolls-Royce engines and turbine rigs.
The turbine noise data were separated into discrete tone and broadband compo-
nents. Following procedures developed previously for fans and compressors, the
discrete tones were assumed to be generated by the interaction of rotor blade
wakes with downstream stator vanes and interactions of stator vane wakes with
downstream rotor blades. The broadband noise was assumed to be due to ran-
dom unsteady lift fluctuations on the airfoils induced by freestream turbulence.
The noise characteristics are related primarily to the blade inlet relative velocity.
Other parameters, such as density, turbulence intensity, blade incidence, solidity,
interstage spacing, and number of stages, were assumed merely to modify these
characteristics. Smith and Bushell acknowledge that an accurate correlation of
data which accounts for effects of all the preceding parameters is extremely dif-
ficult to develop, and therefore offer correlations based on only a few of these
parameters. The correlation for the final-stage fundamental tone SPL (dB) at
the peak far-field angle (100-ft sideline distance) is given by the following rela-
tion:
peak tone SPL - 1 0 log m - 20 log (S/C) + 30 log ( 1 1 1 6/CL) +

f
l (Vrel) (1)
where
m = core engine mass flow, Ib/sec
CL - speed of sound at the turbine exit, fps
S/C = rotor/stator spacing normalized by the upstream blade
chord
V re j = blade inlet relative velocity
The turbine tone levels from full-scale engines did not increase with speed
as rapidly as those observed from tests on model turbine rigs, with f j (Vre],)
from model data following roughly a 30 log (Vrej) behavior and full-scale engine
data following approximately 10 log (Vrep. Similarly, the broadband correla-
tion from Ref. 6 is given by
peak - - oct SPL = 10 log m + 30 log (1 1 16/CL) +

h (VreP (2)
In both cases, the lack of collapse of data using the suggested empirical func-
tions f j (Vrej) and $2 (Vrep *s significant- The correlated broadband SPL levels
depend on the number of stages of the model or engine, with increasing values
of the function i^ ^reP corresponding to increasing numbers of stages. The
authors maintain that each upstream stage imparts more turbulence into the
flow and thus increases the broadband noise generated. According to Smith and
Bushell, a saturation point eventually is reached in the turbulence level of the
flow beyond which further increases in noise are not apparent.
Directivity curves also are presented in Ref. 6. Although directivity is
shown to be a function of turbine speed, it generally peaks at about 120 from
<X
^^
^VD
^*^sO
MAX SPL AT
PEAK ANGLE QJT9D ^\33
DJT8D ^x
NORMALIZED BY
_ AJT3D Xn
SIZE,
NO. OF STAGES,
SPACING, VLl
D\\
TEMPERATURE,
STAGE WORK. T Q\
5dB X
~ ^ cSv
i i i
2.6 2.7 2.8 2.9 3.0
LOG1Q (V T |p LAST STAGE)
Fig. 3 Pratt & Whitney Aircraft turbine noise correlation.
oJT9D
QJT8D
AJT3D Oo
w
ou^
MAX SPL AT HIGHLY LOADED TURBINE
PEAK ANGLE
NORMALIZED BY SIZE,
NO. OF STAGES, 5dB A
SPACING,
DAD D
TEMPERATURE 8 ?
LIGHTLY LOADED TURBINES
2.6 2.7 2.8 2.9 3.0

LOG10 (V T(P LAST STAGE)
Fig. 4 Turbine noise data correlation without stage work
parameter.
the inlet axis for both turbine rigs and engines. The spectrum is assumed to be
the same at all far-field angles, but a distinction is made between the spectrum
from full-scale engines and that from a turbine rig.
Pratt & Whitney Method
The turbine noise procedure developed at Pratt & Whitney Aircraft (P&WA)
predicts peak angle one-third-octave band levels, polar directivity patterns, and
the one-third-octave band spectral characteristics of turbine noise. The proce-
dure was based on a correlation of data from JT9D, JT8D, and JT3D engines,
corrected for turbine mass flow and number of stages. The observed effects of
blade/vane spacing,7 together with turbine inlet temperature and turbine stage
work, also are included to obtain a good data collapse when plotted against the
turbine wheel speed as shown in Fig. 3. The dominant parameters in this proce-
dure are the turbine stage work and turbine speed. The need for a stage work
term in the correlation became apparent during development of the procedure.
Specifically, it was found that turbine noise from the highly loaded JT9D tur-
bine was approximately 10 dB higher than that from the more lightly loaded
JT3D and JT8D engine turbines, even after normalizing by the parameters sug-
gested in Ref. 6. This lack of "collapse" is shown in Fig. 4. The same data,
when normalized by an additional stage work term, correlated much better as
shown in Fig. 3. It is difficult to assess the influence of the number of stages on
turbine noise levels. One approach is to assume that turbine noise levels are
directly proportional to the number of stages in the low-pressure turbine. This
was assumed in the P&WA method and is included in the correlation of Fig. 3.
Spectral correlations in the prediction procedure include contributions

from both turbine discrete tones and broadband sources. The spectral peak is
centered on the average of the stage blade passage frequencies, and the spectral
characteristics are assumed to be the same at all angles. The far-field directivity
pattern is predicted to peak at about 115 from the inlet axis.
NASA Method
A turbine noise prediction method for interim use in the NASA Aircraft
Noise Prediction Program (ANOPP) was selected in early 1974 by Krejsa and
Valerino? The method basically is that derived by Smith and Bushell,6 but the
levels were adjusted to provide agreement with the turbine noise data from addi-
tional turbofan engines that were available. The empirical technique predicts
far-field turbine noise levels, directivity, and one-third-octave band spectra. The
procedure considers the broadband and discrete tone components of turbine
noise separately, and assumes that both components are related to the inlet rela-
tive tip velocity of the turbine last stage, the primary mass flow, and the local
speed of sound at the turbine exit. The effects of stator/rotor spacing on the
discrete tone level are also considered.
The correlation equation of Ref. 8 for turbine tone levels was noted by
NASA to overpredict results from the P&WA JT8D engine by approximately 10
dB. It was assumed in Ref. 8 that this difference was attributable to the com-
mon flow exhaust geometry of this engine. It therefore was recommended in
Ref. 8 that, for turbofan engines with the primary nozzle exit plane upstream of
the fan nozzle exit, a correction factor of -10 dB be applied to the tone level pre-
diction. It also was assumed in Ref. 8 that both tone and broadband compon-
ents have spectra shapes that are centered at the fundamental blade passage fre-
quency calculated for the last stage. The spectra shape assumed for turbine
broadband noise was based on fan broadband noise characteristics.
General Electric Methods
The General Electric Company (G. E.), under contract from the FAA,9"11
recently developed both an analytical prediction technique for discrete frequency
turbine noise generation and an empirical approach. These prediction techniques

are discussed in the following paragraphs.
Analytical Prediction. The approach used in the analytical fomulation is

quite similar to that developed previously for fans and compressors. It was rea-
soned that viscous wakes impinge on downstream blade rows and generate un-
steady circulation about the blades. This produces periodic fluctuations in blade
lift which excite acoustic spinning modes in the duct. Since the wakes from
highly cambered, thick-trailing-edge turbine blades in a favorable pressure gra-
dient are different from those from typical fan blades, a description of the wake
thickness and velocity defect was obtained by G.E. from cascade tests. When
this description was coupled with the remainder of the theoretical noise-genera-
tion model and the predictions were compared with in-duct measurements from
a G.E. single-stage turbine rig, good agreement was obtained, particularly with
regard to rotor/stator spacing effects. The predicted trends, however, are inade-
quate for describing the far-field noise from multistage turbines, since proce-
dures were not provided to account for the effects of attenuation through down-
stream blade rows, propagation from nozzles, or radiation effects through ex-
haust shear layers. Therefore, it was necessary to resort to empirical methods
for predicting turbine noise from multistage full-scale engines.
G. E. Empirical Prediction. Relationships are presented in Ref. 11 which
relate turbine noise levels to various steady-state turbine operating parameters.
One such expression was produced by correlation of data from several G. E. en-
gines and is given by
peak SPL = 40 log (AT/T)turbine - 20 log Vtip + 10 log A + 165 (3)
This expression predicts the maximum one-third-octave band SPL of tones plus
broadband at a sideline distance of 200 ft and at an angle of 120 from the inlet
axis, the assumed angle of maximum noise radiation. The correlation parameter
of turbine ideal work extraction normalized by the inlet enthalpy AT/T also can
be expressed in terms of the turbine pressure ratio by
7-1/7
AT/T=l-(l/Prr n (4)
where
Pr = turbine pressure ratio
V^p = blade tip speed of the dominant stage, fps
A = core nozzle exit area, ft
7 = ratio of specific heats
Although the effects of blade/vane spacing were observed to be important

in establishing the level of tones, a spacing term was not included in the empiri-
cal expression of Eq. (3) which applies for the combination of tones and broad-
band noise. Good agreement with G.E. engine data was reported when using
this expression.
Suggested directivity patterns and spectra are presented in Ref. 11 which

are different for approach or takeoff conditions, but the peak level always occurs
at 120 from the inlet. The spectral shape is assumed identical for each far-field
angular position at a given engine power setting. However, the spectra fall off
faster at high frequencies for the takeoff conditions than for the approach con-
ditions.
IV. Comparison of Current Prediction Methods

The empirical procedures discussed in the previous section are similar in

several respects but differ significantly in other important areas. None are based
on detailed "models" of the physics of noise generation, propagation, and radia-
tion mechanisms. All are based on empirical correlations of steady-state operat-
ing parameters with measured turbine noise data from a limited number of en-
gines. Therefore, the applicability of a given method to a new turbine design is
questionable. Some of the similarities and differences between these procedures
are presented in this section, along with a quantitative comparison of predicted
levels with measured turbine noise data from an engine that was not included in
the data base for any of the procedures.
Comparison of Empirical Correlating Parameters
A summary of empirical correlation parameters contained in the previously

discussed prediction methods is shown in Table 1. It was assumed for all pro-
cedures that a turbine size parameter is required. All but the General Electric
method use primary mass flow, whereas the G. E. procedure selected the core
engine nozzle exit area. All procedures assume that turbine noise levels are
directly proportional to the engine size parameter, as in Eqs. (1) and (3).
Turbine temperature also is a common parameter in the various prediction

systems, but it sometimes is expressed in terms of sonic velocity. An expression
for dipole noise, developed from Lighthill's theory of aerodynamic sound, is the
basis for the temperature dependence assumed in most procedures. The temper-
ture in the G. E. procedure1 * is used only to normalize the turbine work term,
as seen in Eq. (3).
Another parameter that is common to all procedures is turbine speed. The

exponent that relates noise increases to turbine speed varies, however, from
large negative values to moderate positive values, depending on which procedure
is used. Several methods incorporate the influence of rotor/stator spacing on
the prediction of turbine tone noise levels, as shown in Table 1. The P&WA and
NASA procedures make use of a -10 log (spacing) term. This dependence is in
cr>
Table 1 Comparison of empirical turbine P

noise correlation parameters o
Turbine noise correlation parameters Jj

Engine Rotor Number ^
Turbine noise size Turbine Turbine stator Turbine of stages Fan duct L?
Prediction method component m A temp. speed spacing work (N) length ^
Tone x x X X ;H
Rolls-Royce6 Broadband x x X

D>
0
Tone and m
Pratt&Whitney 12 broadband x x x x x x x rt*
Tone x x X X X
C_,
8
NASA Broadband x x x
General Tone and m

Electric broadband x x X X 3
good agreement with the experimental findings in Ref. 7. The -20 log (spacing)
dependence used by Rolls-Royce" in Eq. (1) was based on earlier fan noise ex-
perience. Since the time of development of the earliest turbine noise prediction
method by Smith and Bushell, it has been suggested by several investigators
(see Table 1) that a turbine work parameter would be required to "collapse"
turbine noise data adequately from engines with significant differences in load-
ing characteristics. Some procedures, however, do not incorporate a work term.
A fairly recent development in turbine noise technology was the recog-
nition of the effects of fan duct length on observed far-field turbine noise
characteristics. After the tones have propagated through the primary engine
tailpipe, they must pass through the turbulence in the fan shear layer. In this
process, some of the acoustic energy in the tones is scattered to other angles and
frequencies. These phenomena were discussed in more detail by Mathews and
Peracchio12 and are discussed in Sec. V of this paper. NASA and P&WA in-
clude this effect by using an empirical fan duct length correction. The latter
method makes use of data obtained on a turbofan engine tested with three dif-
ferent length fan ducts. Kazin et al. also have observed this phenomenon and
present an empirical method for predicting the decrease in tone energy due to
scattering. This method, however, is not included explicitly in the G. E. turbine
noise prediction procedure.
Application of Predictions to JT8D-109 Engine
Having identified the key aspects of a variety of current procedures, the

relative "accuracies" of these procedures will be illustrated in this section by
showing comparisons of predicted turbine noise levels with levels measured from
tests of a JT8D-109 engine. The JT8D-109 engine, illustrated in Fig. 1, was
derived by "refanning" a Pratt & Whitney JT8D-9 engine under a NASA con-
tract. The objectives of these modifications were to demonstrate a significant
reduction in jet exhaust noise by increasing the engine bypass ratio from about
1 to 2, and to reduce fan noise by the extensive use of acoustic linings in the
fan duct work. With noise from other sources significantly suppressed by the
modifications mentioned, turbine noise from the three stages of the low-pressure
turbine easily could be identified in the engine noise spectra as illustrated in
Fig. 5.
Performance and geometric characteristics of the JT8D-109 engine turbine

required as input to the various prediction systems (Sec. Ill) were obtained and
the turbine noise characteristics predicted for the JT8D-109 turbine operating at
several engine operating conditions. Data from this engine were not used in the
data base of any of the prediction systems discussed in this paper.
Predicted Turbine Noise Levels. Predicted noise levels, in terms of one-

third-octave band sound pressure level (SPL), are compared in Fig. 6 with mea-
APPROACH POWER
120 FROM INLET AXIS
TURBINE NOISE
3 4 5 6 7 8 9 10
FREQUENCY KHz
Fig. 5 Typical JT8D-109 engine noise spectra.
MEASURED DATA PREDICTED LEVELS
-JT8D-109 ENGINE D GENERAL ELECTRIC
A NASA
PRATT & WHITNEY
ROLLS ROYCE
TYPICAL APPROACH POWER
4000 4500 5000 5500 6000 6500 7000 7500

CORRECTED TURBINE SPEED-RPM
Fig. 6 Comparison of measured and predicted turbine

noise levels.
sured values over a range of engine speeds. The levels shown in Fig. 6 represent
the sum of the tone and the broadband SPL at the peak radiation angle. It is
seen that both measured and predicted turbine noise levels increase gradually
with rpm and generally tend to level off at high engine speeds. The previously
described prediction systems envelope the measured data with a band approxi-
mately 7 dB wide.
It is difficult to evaluate, from the single comparison in Fig. 6, all of the
shortcomings or strong points of the various procedures examined, nor is this
the intent of the current paper; Fig. 6 does reveal, however, the general range of
uncertainty that potential users of any procedures should expect. Since the pro-
cedures are empirical in nature, each is subject to the inherent shortcomings dis-
cussed in Sec. II, and, therefore, the 7-dB variation shown in Fig. 6 is not sur-
MEASURED DATA
- JT8D-109 ENGINE
PREDICTED DIRECTIVITY
0 GENERAL ELECTRIC
6 NASA
PRATT & WHITNEY
__. ROLLS ROYCE
90 100 110 120 130 140 150
POLAR DIRECTIVITY ANGLE FROM INLET AXIS- DEGREES
Fig. 7 Comparison of measured and predicted turbine

noise directivity.
(JT8D-109 ENGINE, APPROACH POWER, 120)

1/3 OCTAVE BAND CENTER FREQUENCY
1000 2000 5000 10,000
2000 4000 6000 8000 10,000

FREQUENCY Hz (NARROW BAND)
Fig. 8 Comparison of narrow-band and one-third-octave

band spectra (JT8D-109 engine, approach power,
120).
prising. If data from an engine with a highly loaded turbine had been used in the
comparisons, the scatter in the predictions probably would have exceeded the
7-dB band shown, since two procedures (see Table 1) do not include a turbine
work term.
Predicted Turbine Noise Directivity. Figure 7 illustrates a comparison of

predicted polar directivity patterns (tone plus broadband) with JT8D-109 data
at a typical aircraft approach power setting. The measured peak level of turbine
MEASURED DATA
- JT8D-109 ENGINE
PREDICTED SPECTRA
-D-GENERAL ELECTRIC
-A-NASA
PRATT & WHITNEY
APPROACH POWER,
120 FROM INLET AXIS
2 2.5 3.15 4 5 6.3 8 10
ONE-THIRD OCTAVE BAND FREQUENCY

(KHz)
Fig. 9 Comparison of measure and predicted turbine

noise spectra.
noise occurs at an angle of 115 from the inlet axis. All predicted peak angle
locations are seen to be within 5 of the data. A fairly large variation exists,
however, in levels predicted at forward angles, with the predictions lying within
approximately 5 dB of the data at an angle of 90. With the exception of the
method of Ref. 6, the agreement is better at the farther aft angles. Although
Smith and Bushell" did not suggest a specific directivity, the Rolls-Royce curve
shown is the average of several patterns presented in Ref. 6.
Predicted Turbine Noise Spectra. Most prediction methods express results

in terms of one-third-octave levels. However,before comparing a measured JT8D-
109 one-third-octave band spectra with the various predictions, several problems
associated with the use of one-third-octave spectra for turbine noise should be
noted. Figure 8 depicts, for the JT8D-109 engine, both a one-third-octave spectra
and a narrow-band spectra of the same signal. Figure 8 shows that the entire tur-
bine noise spectra between 5000 and 10,000 Hz is defined by only four points
when using a one-third-octave representation, making it impossible to determine
the relative contributions of the tones and the broadband noise. In addition,
slight changes in engine speed can alter the spectra shape significantly if it is
accompanied by a shift of one of the tones into a different one-third-octave band.
It becomes apparent that, even though the required output of a prediction
method is expressed in terms of one-third-octave levels, detailed analysis of
narrow-band spectra must be included in the development of these methods.
Figure 9 illustrates a comparison, at a 120 angle, of several predicted spec-

tra (tone plus broadband) with the measured turbine noise portion of a JT8D-
109 spectrum at approach power. Two of the three procedures included in this
figure predict the correct peak frequency of turbine noise, but both the General
Electric and NASA procedures overestimate the width of the spectra for this
particular engine.
V. Requirements for Advanced Prediction Methods
The comparisons presented in the previous section clearly indicate that

current turbine noise prediction methods must be improved to achieve a rea-
sonable level of accuracy. Improvements in empirical prediction systems can be
accomplished both by increasing the range and accuracy of the data base used in
development of the procedure and by the identification and use of correlation
parameters that more nearly reflect the physics of the actual noise-generation,
propagation, and radiation process. Although more data always are useful, the
more immediate requirement appears to be the identification of more appro-
priate correlation parameters. In order to accomplish this, models must be de-
veloped for the complex mechanisms of noise generation by turbine rotor and
stator rows, propagation through downstream stages and exhaust nozzles, and
radiation through the turbulent exhaust flow to the far field. Development of
these models will involve both the formulation of analytical expressions to de-
scribe turbine noise generation, propagation, and radiation phenomena and the
research tests necessary to verify the accuracy of the models. In the remainder
of this section, the elements required for each of these models are described,
together with a discussion of the status of available models.
Turbine Noise Generation
Narrow-band spectral analysis (see Fig. 8) reveals that tones are the domi-
nant feature of turbine noise. It therefore is clear that a model to predict the
generation of tones at the fundamental and harmonics of turbine blade passing
frequency must be given highest priority. Because of the similarities between the
turbine and fan/compressor design features, it is possible to postulate that blade/
vane interaction noise-generation processes are similar in each. Therefore, some
of the findings from fan/compressor noise research may be applied to model
turbine tone generation.
In fans and compressors, it has been found that a major source of noise
generation, called "interaction" noise, is attributable to the surface pressure
fluctuations on rotor airfoils as they "chop" through the wakes from upstream
vanes, and on downstream vanes as wakes from upstream blades pass by. These
sources both produce noise at blade passing frequency and its harmonics. Such
processes clearly are present in turbines, but with some notable differences due
to basic differences between the aerodynamics of turbines and compressors.
It is not uncommon for the pressure drop across one turbine stage to be as
large as the pressure rise across six to eight compressor stages. This is possible
because the boundary layers on turbine airfoils are subjected to favorable pres-
sure gradients and are therefore much less likely to separate than in the case of a
compressor. Because of these favorable gradients, a much higher degree of aero-
dynamic tolerance is associated with turbine stages. Therefore, the blade and
vane configurations are much different from those of compressors. Turbine
blades typically are highly cambered to achieve the required stage work. This
leads to much steeper axial pressure gradients along the airfoils and to wake
characteristics that are different from those for compressor stages. It also is pro-
bable that the surface pressure response to perturbations in the incoming flow is
much different in turbines than for compressors. Research tests are required to
quantify these aerodynamic differences between compressors and turbines.
Analytical models of compressor noise generation have used the Silverstein

model to calculate the wake velocity deficit characteristics behind an airfoil,
and the method of Kemp and Sears^' ^ to calculate the airfoil surface pressure
response to the "gust" associated with a passing wake. Although these models
are sufficiently accurate for application to relatively lightly loaded compressor

blades having modest camber,they must be revised substantially, as in Ref. 10,
to account for the highly loaded conditions associated with turbine blades.
In addition to turbine tones, narrow-band analysis (Fig. 8) also reveals the

presence of a broadband "haystacking" in the far field. Generation of this noise
has been attributed to several possible mechanisms, including the following: 1)
interactions of turbine blade and vane airfoil surface pressures with fine-scale
turbulence"; 2) unsteadiness in the strengths and locations of blade wakes that
provide random modulation of interaction tones; and 3) spectral scattering of
tones by turbulence in the exhaust flow. Although both mechanisms 1 and
2 may be present in turbines, current evidence suggests that mechanism 3,
spectral scattering of tones by exhaust turbulence, is the most important source
of the observed turbine broadband noise. Because this mechanism is associated
with radiation rather than generation, it is discussed in the "Turbine Noise Radia-
tion" portion of this section.
In summary, blade-vane interaction noise is the most significant factor that

must be modeled in an analytical turbine noise prediction system. Although
work done to date to model this source mechanism in compressors can give
valuable insight into the type of model required, the model must be modified
substantially to account for the aerodynamic and mechanical differences be-
tween turbines and compressor stages and then verified by experimental tests.
Turbine Noise Propagation
Many of the concepts that have been developed to predict noise propaga-
tion in compressors also apply to turbines, but some new factors are introduced
because of the higher levels of aerodynamic loading associated with turbines.
The most significant factor in noise propagation in both turbines and compres-
sors is the concept of "acoustic spinning modes." Experience with fan noise has
shown that the pressure field associated with an isolated subsonic rotor does not
propagate. However, wheen stationary vanes are introduced either upstream or
downstream, noise at blade passing frequency may propagate. Propagation oc-
curs when acoustic interaction patterns are generated which "spin" at super-
sonic speeds even when the rotor speed may be well below sonic. * Depending
on the number of blades and vanes used in a turbine row, interaction patterns
are generated which may or may not propagate. Patterns that do not propagate
are said to be "cutoff." Cutoff patterns attenuate rapidly near the rotor and do
not carry significant amounts of acoustic energy to the far field. Cutoff typically
is achieved by use of a large number of stationary vanes upstream and down-
stream of a rotor stage. Fans typically require about twice as many vanes as
blades to achieve cutoff. In turbines that operate at elevated temperatures and
lower tip Mach numbers, cutoff may be achieved with fewer vanes. The concept
of cutoff has been used widely to reduce noise levels from fans,but it has not
been applied extensively to turbines. Although there is no fundamental reason

why this technique cannot be applied, it is probable that structural and/or weight
penalties would be encountered.
Noise propagation through vane rows in compressors has been modeled by

Mani and others. > ' Similarly, the attenuation of turbine noise through
downstream stages must be considered, but the process is more complicated in
turbines than in compressors. The complications associated with turbines arise
from the larger velocity, pressure, and temperature gradients as well as the higher
blade camber that exist in turbines. These factors must be included in a turbine
noise-propagation model and verified by experiments.
The internal contour of the tailpipe is expected to influence the propagation

of turbine noise. Fan noise experience with contoured inlets has shown that, as
the inlet diameter is reduced forward of the fan, the wall Mach number of the
spinning mode acoustic patterns is reduced. A pattern that would propagate in
a cylindrical duct can be caused to decay in a contoured duct with a suitable
diameter reduction. Models to predict this behavior in inlets have been pub-
lished^' and could be extended to predict the propagation of turbine tone
acoustic spinning modes in contoured exhaust ducts.
From the preceding discussion, it is obvious that the propagation of tur-
bine tones involves several complicated processes. A detailed knowledge of the
turbine flowfield will be required as input to the turbine noise-propagation mod-
el, and several simplifying assumptions must be made. However, results of pre-
vious work in modeling fan noise propagation phenomenon could serve as a con-
venient starting place to develop the required model for turbine noise propaga-
tion.
Turbine Noise Radiation
Models have been developed to predict far-field noise radiation from

flanged1** and unflanged 1 ^ duct terminations. These models have been used
fairly successfully to predict fan noise radiation from inlets. However, the radia-
tion of turbine noise presents a more complicated problem, as illustrated in Fig.
SCATTERING BY
EXHAUST TURBULENCE
REFRACTION THROUGH
VELOCITY AND TEMPERATURE
GRADIENTS
EFFECTS OF FLOW ON NOZZLE

TERMINATION ACOUSTIC IMPEDANCE
-EFFECTS OF FAN DUCT
EXIT LOCATION
Fig. 10 Elements required for turbine noise radiation

model.
SPL-dB
3 4 5
FREQUENCY-KHz
Fig. 11 Effects of fan duct configuration on
turbine noise.
10. A turbine noise radiation model must address the effects of jet flow on the
nozzle termination acoustic impedance and be developed to apply to a wide
range of exhaust geometries. For example, common flow nozzles exist on many
bypass engine installations in which the primary exhaust nozzle exit plane is lo-
cated well upstream of the final jet nozzle that passes both fan and primary flow.
The calculation of nozzle termination impedance for this type of geometry
would be extremely difficult and has not been attained to date.
In addition to the effects of coaxial nozzle exit impedance on noise radia-

tion, the noise reaching the far field from the primary nozzle exit plane must
pass through two turbulent mixing regions: one at the shear interface between
the primary and fan streams which involves mixing of different temperature
streams, and one at the fan-ambient shear interface. The turbulent eddies con-
vected in these regions provide "discontinuities" in the flow from which turbine
tone scattering has been observed.10-12 This scattering is responsible for chang-
es in both spectral and directivity characteristics of the radiated noise. Whereas
the bulk of the acoustic energy at the primary nozzle exit is contained in dis-
crete tones, significant spectral "haystacks" of broadband noise are observed in
the far field in a frequency range about each tone, as shown in Fig. 8. The scat-
tering mechanism is discussed in the following paragraphs.
As a tone passes through the turbulent region of the exhaust shear layers,
variations in the speed of propagation of the sound waves are induced. These
changes may be produced by temperature variations that cause the sound speed
to fluctuate, or by random refraction and convection of the sound by the tur-
bulent eddy. Both effects may cause random phase variations along a wavefront
and give rise to both spatial and frequency scattering of the tone. -22
The turbulence intensity and thickness of a shear layer strongly influence

the degree of tone scattering which occurs. Since the fan flow shear layer char-
acteristics are dependent on the axial distance from the fan nozzle, the shear
layer from a "short fan duct" configuration that has the fan nozzle well up-
stream of the primary nozzle would be developed more fully than that from a
longer fan duct at the point where turbine tones would pass through. Figure 11
shows the effect of fan duct geometry on the far-field turbine noise spectra from
a turbofan engine tested with both short and long coplanar fan ducts. For the
short fan duct, it is seen that more of the turbine tone energy is scattered to ad-
jacent frequencies by the turbulence in this thicker shear layer than by the very
thin shear layer associated with the long fan ducts.
Careful analysis of far-field turbine noise levels, integrated over frequency

and far-field angles, reveals that no new source of acoustic energy is introduced
by the scattering phenomenon. The energy in the "haystack" surrounding each
tone merely is transferred from that originally contained at the tone frequency.
Advanced turbine noise prediction procedures will require models for the
radiation phenomena described previously. Further experimental research is
needed, however, to verify the relationships that exist between eddy size, con-
vection velocity, and tone frequency, so that accurate models can be obtained.
VI. Concluding Remarks
Several procedures have been developed by government agencies and indus-

try which present methods to predict turbine noise using empirical procedures.
These procedures are similar in some respects but incorporate some notable dif-
ferences in the correlation parameters used. Also, each procedure employs dif-
ferent empirical data to reflect the measured data base that uniquely is available
to the particular agency or company. Among the more significant differences
noted among the procedures was the inclusion or lack of a term to account for
effects on noise of turbine work extraction. Evidence was presented to suggest
that such a term is required.
A measure of the "accuracy" of the published procedures was illustrated by

comparisons of predicted noise characteristics with turbine noise measurements
from a relatively new experimental engine, the JT8D-109 refan engine. Noise
measurements from this engine, which was tested under NASA contract, were
used because extensive turbine noise data were available, and because data from
this engine were not used in formation of any of the empirical predictions. Re-
sults of the comparisons of measured vs predicted noise levels, directivity, and
spectra show that none of the procedures achieved "perfect" agreement. In
general, these comparisons suggest that there is considerable room for improve-
ment in most of the procedures before they can be considered reliable. The sug-
gested direction for improvement of these procedures is not to acquire more tur-
bine noise data but rather to conduct the experimental and analytical research
necessary to develop correlation parameters that are related more closely to the
physical processes of noise generation, propagation, and radiation.
The ultimate goal of turbine acoustic research and development efforts is to

develop and verify reliable analytical models of the turbine noise-generation, pro-
pagation, and radiation processes. Because of similarities between turbine tone
generation and fan noise, some of the analytical formulations developed to pre-
dict fan noise can be adapted for use in turbine noise models. Significant differ-
ences exist between fan and turbine aerodynamics, however, and these differ-
ences must be addressed by both experimental and analytical programs in order
to extend fan noise models to predict turbine noise.
From the results of our survey, it appears that much more research is re-
quired before reliable analytical models of turbine noise are available. Work
conducted toward the formulation of analytical models should provide better
insight into the physical processes involved. This insight should lead to the selec-
tion of more fundamental correlation parameters for improved empirical pre-
diction systems. Because of the importance of turbine noise in the design of
future engines, the development of improved turbine acoustic technology must
be given a high priority in future acoustic research and development programs.
References
r, J. M. and Sofrin, T. G., "Axial Flow Compressor Noise Studies,"

SAE Transactions, reprint, Vol. 70, 1962.
Sears, W. R., "Some Aspects of Non-Stationary Airfoil Theory and It's

Practical Application," Journal of the Aeronautical Sciences, Vol. 8, No. 3,
Jan. 1941, pp. 104-188.
^Kaji, S. and Okazaki, T., "Propagation of Sound Waves through a Blade

Row, I. Analysis Based on the Semi-Actuator Disk Theory," Journal of Sound
and Vibration, Vol. 11, No. 3, 1970, pp. 339-353.
Kaji, S. and Okazaki, T., "Propagation of Sound Waves through a Blade

Row, II. Analysis Based on the Acceleration Potential Method," Journal of
Sound and Vibration, Vol. 11, No. 3, 1970, pp. 355-375.

5
Mani, R. and Horvay, G., "Sound Transmission Through Blade Rows,"
Journal of Sound and Vibration, Vol. 12, 1970, pp. 59-83.
6
Smith, M. J. T. and Bushell, K. W., "Turbine Noise - It's Significance in the
Civil Aircraft Noise Problem," Publ. 69-WA/GT-12, Nov. 1969, American
Society of Mechanical Engineers.
7
Fisk, W. S., et. al., "Supersonic Transport Noise Reduction Technology
Summary - Phase I," Rept. FAA-SS-72-43, Dec. 1972, Department of
Transportation, Federal Aviation Administration.
^ Krejsa, E. A. and Valerino, M. F., "Interim Prediction Method for Turbine

Noise," TM, July 1974, NASA.
^Kazin, S. B., et. al., "Core Engine Noise Control Program, Vol. I - Identifi-
cation of Component Noise Sources," Rept. DOT-FA72WA-3023, May 1974,
Federal Aviation Administration.
10
Kazin, S. B., et. al., "Core Engine Noise Control Program, Vol. II - Iden-
tification of Noise Generation and Suppression Mechanisms^' Rept. DOT-FA72
WA-3023, May 1974, Federal Aviation Administration.
1
1 Kazin, S. B., et. al., "Core Engine Noise Control Program, Vol. Ill - Pre-
diction Methods," Rept. DOT-FA72WA-3023, June 1974, Federal Aviation
Administration.
Mathews, D. C. and Peracchio, A. A., "Progress in Core Engine and Turbine

Noise Technology," AIAA Paper 74-948, Aug. 1974.
13
Silverstein, A., Katzoff, S., and Bullivant, W. K., "Downwash and Wake
Behind Plain and Flapped Airfoils," Rept. 651, 1939, National Advisory Com-
mittee for Aeronautics.
Kemp, N. H. and Sears, W. R., "The Unsteady Forces Due to Viscous

Wakes in Turbomachines," Journal of the Aeronautical Sciences, Vol. 22, 1953,
pp. 478-483.
^Kemp, N. H. and Sears, W. R., "Aerodynamic Interference Between Moving

Blade Rows," Journal of the Aeronautical Sciences, Vol. 20, No. 9, Sept. 1953,
pp. 585-597.
1 Amiet, R., "Transmission and Reflection of Sound by Two Blade Rows,"

Journal of Sound and Vibration, Vol. 34, No. 3, 1974, pp. 399-412.
^ Mathews, D. C. and Nagel, R. T., "Inlet Geometry and Axial Mach Number
Effects on Fan Noise Propagation," AIAA Paper 73-1022, Oct. 1973.
18
Morse, P. M. and Ingard, K. U., Theoretical Acoustics, McGraw-Hill, New
York, 1968, p. 471.
Levine, H. and Schwinger, J., "On the Radiation of Sound from an Un-
flanged Circular Pipe," Physics Review, Vol. 73, 1948.
20 Howe, M. S., "Multiple Scattering of Sound by Turbulence and Other In-

homogeneities," Journal of Sound and Vibration, Vol. 27, No. 4, 1973, pp. 455-
476.
21
Kraichnan, R. H., "The Scattering of Sound in a Turbulent Medium," Jour-
nal of the Acoustical Society of America, Vol. 25, No. 6, Nov. 1953, p. 1096.
22
Rudd, M. J., "A Note on the Scattering of Sound in Jets and the Wind,"
Journal of Sound and Vibration, Vol. 26, No. 4, 1973, pp. 551-560.
REVIEW OF THEORY AND METHODS FOR COMBUSTION
NOISE PREDICTION
* +
R.E. Motsinger and J.J. Emmerling
General Electric Company, Cincinnati, Ohio
Abstract
The state-of-the-art theory of engine combustion noise is

reviewed. The results of a literature survey are summarized,
and formulas for the prediction of combustion noise are pre-
sented. A compilation of necessary empirical data is given,
and the accuracy and range of validity of these data are indi-
cated. These formulas and data include effects of noise
generation by the combustion process, scaling laws for spectral
content, acoustic power, and thermoacoustic efficiency. Methods
of predicting engine combustion noise generation are recom-
mended. Research that is critical to the improvement of com-
bustion noise prediction is identified.
Nomenclature
B = acoustic transmission coefficient, dimensionless

D = diameter, ft (m)
F = ratio of fuel to airflow rates
K = constant, depending on engine
L = length, ft (m)
m = mass flow rate, Ibm/sec (kg/sec)
N = number of turbine stages -
OAPWL = overall sound power level, re 10
Presented as Paper 75-541 at the AIAA 2nd Aeroacoustics

The recent information included in this paper from Ref. 4
developed under Contract DOT-FA72WA-3023.
R.E. Hotsinger, Manager ATD&PM, Acoustic Design Technology
J.J. Emmerling, Engineer, Acoustic Design Technology
177
178 R. E. MOTSINGER AND J. J. EMMERLING
2
OASPL = overall sound pressure level, re 20 yN/m
P = total pressure, psf (N/m2)
SPL = sound pressure level for one-third-octave band center
frequency, re 20 yN/m2
T = total temperature, R (K)
V = velocity, fps (m/sec)
6 = ratio of total pressure atmosphere or reference
pressure, P/PQ
Subscripts
a = air
b = flame speed
c = combustor
d = directivity factor
e = equivalent
f = frequency
L - laminar
N = nozzle
R = reference
T = turbine
0 = atmospheric or freestream
3 = combustor inlet station
4 = combustor exit station
5 = turbine exit station
I. Introduction
The most significant characteristic of combustor noise

detected in the far field is its location in the spectrum. It
occurs in the low-frequency region, generally peaking in level
in the 500-Hz octave band such that it is confused easily with
jet noise. The spectral distribution correlates closely with
that for jet noise from conical jet nozzles. The frequency of
peak noise, however, is unrelated to the conical jet nozzle
Strouhal number. This is illustrated in Fig. 1, a plot of the
spectral distribution from four turboshaft engines with the SAE
AIR 876 jet noise spectrum superimposed. The four engines
range in size from 250 to 22,800 hp, and the data include not
only aircraft engines (T58, T63, and T64) but also a land-
based gas turbine for electrical power generation (DM 625
model N). Jet noise is not a significant contributor for any
of these engines; nevertheless, the jet noise spectrum, with
the peak-noise frequency arbitrarily centered to coincide with
that of the four engines, represents a reasonable approxi-
mation to the engines spectra. The frequency band of peak
noise was the same for all four engines, and the levels of the
COMBUSTION NOISE PREDICTION 179
10
9
e o
CQ -10
n3
-20
o T63 (250 HP)
9 n T58 (1500 HP)
O T64 (2800 HP)

Q
1-30 _ A DM625 MODEL N (28,200 HP)
CQ SAE AIR 876 JET NOISE SPECTRUM
-40
o 100 500 1000 5000 10000
FREQUENCY, Hz
Fig. 1 Combustor noise spectra for turboshaft engines.
QUIET ENGINE "A" FULLY SUPPRESSED
150 Ft. (45-7m)

140(2.44 rad) FROM INLET
70%
100 200 500 1000 2000 5000 10000

-H FREQUENCY, Hz
Fig. 2 Turbofan engine noise spectra showing combustor noise.
spectra in Fig. 1 are normalized with respect to the level of

this band.
A low-frequency noise peak also occurs for highly sup-

pressed turbofan engines, as shown in Fig. 2.1 There are two
humps, one peaking at 125-160 Hz and the other at 630 Hz. The
lower-frequency hump stems from distortion of jet noise by the
first ground-reflection reinforcement. The higher-frequency
hump is attributed to combustor noise with a varying level of
fan and jet noise contribution, depending on the power setting.
It is not possible to compare the full spectral distribution
from this turbofan with that from the turboshaft engines be-
cause of the contribution of other noise sources from this
engine. Note, however, that the peak frequency (630-Hz
one-third-octave band) occurs in the same octave band (500-Hz)
as the distribution shown in Fig. 1.
The data of Fig. 2 contain so many combined effects that

it is difficult to establish the actual amount of noise con-
tributed by the combustor. There are not only jet noise and
ground reinforcement/cancellation effects, but also some evi-
dence of fan noise contribution to the 500-Hz octave band
(even for the highly suppressed fan).
These data illustrate the difficulties in identifying

combustor noise contributions to the far field for turbofan
engines. The data in Fig. 1 show no such difficulty for
turboshaft engines. Surprisingly, for turbojet engines at low
power settings, combustor noise also is relatively easy to
identify. As discussed in the engine data correlation section
of this paper, turboshaft engines produce in the order of 8 dB
more noise than turbofans; whereas, turbojets produce in the
order of 16 dB more. Thus, it is important in many appli-
cations to be able to predict engine-radiated combustor noise.
II. State-of-the-Art Review
The attack on the problem of predicting combustor noise is

based on both theoretical and experimental approaches, with the
latter including both combustor component tests and direct
measurement of engine data. Engine data show that the theo-
retical and combustor component test approaches, which consider
the combustor noise-generation process per se, must be extended
to include the effects of nonuniformities associated with the
combustion process as they propagate or are transmitted through
the turbine and exhaust nozzle. Much work is currently under-
way on both the basic combustion process and the extension of
theoretical predictions and component data to include factors
associated with the engine, and it is premature at this time to
attempt to summarize the latest progress being made. However,
in the past few years, significantly improved understanding of
the character of the problem has been achieved, and enough
progress has been made to provide quantitative results in a
form that will allow an interim basis for predicting combustor

noise levels, spectra, and far-field radiation patterns.
Literature Review
Essentially two mechanisms have been considered analyti-

cally in the literature as the cause of comb ustor-associated
noise: 1) the combustion process itself, which gives rise to
acoustic volumetric and pressure fluctuations; and 2) the con-
vection of turbulence and temperature fluctuations through
regions having a mean pressure gradient, such as turbine stages
and exhaust nozzle, which gives rise to noise generation. Much
more attention has been devoted to the first item, both ana-
lytically and in combustion component testing, and this mecha-
nism, therefore, forms the basis of the correlations described
in this paper. On the other hand, the results from consider-
ation of the second item quantitatively predict the location
of the frequency of peak noise generation (which already has
been noted experimentally in Fig. 1) and are reported qualita-
tively to check reasonably well with "measured (inferred)"
core engine sound power; consequently, this approach is one
requiring further research.
Consideration of the first item has provided a basis for

developing correlating parameters to use for evaluation of
measured engine combustor-radiated noise. Reference 3 provides
a summary of the results of this approach, including a summary
of pertinent published information available as of mid-1974.
Additional information since has become available in Ref. 4,
providing quantitative data correlations for predicting com-
bustor noise from aircraft turboengines.
Tables 1 and 2, adapted from Ref. 3, summarize applicable

information in the published literature. Table 1 lists the
kind of acoustic information discussed in each reference and
the basis of the information, whether theoretical or empirical
data. Table 2 compiles the parameters used to compute low-
frequency core noise power level or, alternately, sound pres-
sure level.
Sound Level. The form of the equation in which the para-

meters of Table 2 are used is
OAPWL (or OASPL, if appropriate) =
const. + 10 log-in [product of the parameters]

Table 1. Literature search survey3
Acoustic Sound
power pressure
Reference level level Spectrum Directivity
5
Bushell .. D D
Marshall . D D D
Ho and Tedrick D D
0
Motsinger D ... D D
j 10
Grande T ... ...
Arnold T
12
Gerend ... D D D
14
Dunn ... D ...
1S
Strahle T ... D D
16
Swan ... ... D .
17
Abdelhamid T D D
1Q
18
Plett . . T D D
19
Chiu T . ...
20
Chiu T,D D ...
T = theoretical analysis, D = empirical data.
The first two equations in the table, from Ref. 7, were

developed from empirical methods combined with dimensional
analysis using combustors from small gas turbine engines,
first in component tests and then installed in the engines.
The first equation was derived from the component tests,
whereas the second was based on the engine tests. It is note-
worthy that, in the operating range of the combustors, the
level of noise from the engine was less than that from the
component rig.
The third equation, from Ref. 8, was developed empiri-

cally using experimental correlations from Ref. 9 in combi-
nation with measured combustor pressure fluctuations and
noise data from a turboshaft engine. The same parameters were
used in Ref. 4 to correlate data from a number of other
engines, as discussed later, with the result that it was
ell ..
... MIN.r-ld....~,.~ Purchased rrom American Institute of Aeronautics and Astronautics
Table 2. Compilation of parameters used to compute low-frequency core noise

Mass flow Fuel
Ref. Constant Compute Temperature Pressure r-ate mixture Velocity Dimension
7 81 OAPWL
~4-1;)~J [~J U+F T~ [V4 ] [De]
7 23 OAPWL
~14-l;t{~ [~2] U+ FJ4 [V:J [ De1 n
0
3:
c:o
K~-l;)T~J~ [F~/Par~ . .. ... c:

8 56.5 OAPWL
[rha.J
(/)
-1
.......
0
:z
10 a OAPWL
[C~~T~~~ ~vPoJ [m~~~/~J U+F] [V: V3] [DD:~ :z
0
.......
(/)
[T:J rYRJ3 Un~~;/~ ...

fT1
12 b OASPL
d
. .. -0
;;0
fT1
0
.......
e
~ARJ2 ~J3 [m~~;/~ .. . . .. n
14 c OASPL -1
.......
0
:z
15 OAPWL .. . ... ~LrH
1+F
~ 2.68 1.3~
\;j 'til [D~84J
a205.5 + 10 log [BNB T]
b[-15 for annular, -6 for can or can-annular] 5 dB
c76 + correction for combustor type, engine directivity
d200-ft sideline, 110 0 from inlet
eat I-meter radius 00
w
necessary to adjust the value of the constant to be used,

depending upon the type of engine, whether turbojet, turbo-
shaft, or turbofan. This, of course, indicates that another
variable in addition to those considered must be taken into
account.
The fourth equation, from Ref. 10, is somewhat more

complicated than is listed in the table because the "constant"
includes the 10 logln [B B ] term. These two terms are defined
as
>T E nozzle transmission coefficient
N
T /T
o N
B E turbine transmission coefficient
These terms, which do not appear in any of the other equations,

impose a reduction in the computed noise level reaching the far
field depending upon the inverse of the nozzle temperature
ratio and terms involving the inverse of the turbine tempera-
ture ratio. This is, as more energy is extracted by the tur-
bine and as more potential energy is converted to kinetic
energy by the exhaust nozzle, the lower will be the combustor
noise at the far-field receiver. The author claims agreement
of his correlation with high- and low-bypass-ratio turbofan
engines, but checks against other types of engines (e.g.,
turbo jets or turboshafts) were not made.
The fifth and sixth equations, from Refs, 12 and 14, are
very similar in terms of the functions and parameters used, and
they are based upon correlation of engine data. The details
included in the original references are more extensive in Ref.
14; for example, the effect of source motion on far-field
directivity pattern (source convection factor in flight) is
included. Again, both references differentiate among engine
types .
The last equation per Strahle is based on the results of

theoretical analyses and correlation of component combustor
tests. In a later paper, Strahle suggests the following form
based on theoretical considerations:
OAWPL = const. + 10 Iog1() [(1/T^/T^) (P4) (Fa)V2D2]
where
a = 0, fuel lean primary zone
a = 1, fuel rich primary zone
In this latest form, terms involving the combustor environment

in the engine are included.
Considering all of these correlations, there appears to be
general agreement that there are three basic parameters as a
minimum: turbine inlet temperature (in terms of the tempera-
ture itself, or a temperature rise, or fuel/air ratio), pres-
sure level of the combustion zone, and a size parameter in
terms of either the airflow rate or a dimension. There also is
theoretical and experimental basis for including velocity as a
parameter. It is of interest to note that those authors who
include the velocity term have used a combination of analysis
and atmospheric pressure combustor component tests to develop

their basic correlation. In the aircraft engine application,
the choice of fuel and the airflow Mach numbers possible for
acceptable designs are limited to a very narrow range; thus, it
is not surprising that those correlations derived primarily
from engine test data&>12,14 do not include either the flame
speed or airflow velocity. No doubt, to make the extrapo-
lation from combustor component tests to compute engine-
radiated noise, this parameter must be included. In the
interim, it is recommended that engine-derived correlations
be used.
The size parameter appears, in the form of a dimension,

in those correlations that have been developed from analytical
and component tests. Those derived from engine data use a
mass flow parameter; this could, of course, have been expres-
sed alternately as the product of a dimension and a velocity
term, but the size parameters and velocity parameter are
different effects and should be kept separate. Of those
correlations using the mass flow parameter, one set uses the
mass flow rate per se, and the other uses "corrected" mass flow
rate in the form of a flow function based on the temperatures
and pressures existing in the combustor within an engine.
When considered in conjunction with the pressure parameter,
the grouping from Ref. 10 causes the pressure level dependence
to be eliminated, whereas that from Refs. 12 and 14 reduces to
essentially the same pressure dependence given in Ref. 8.
The pressure ratio parameter is a critical one in the

correlation, since significant variations exist from one
engine design to another, and since the pressure ratio of the
engine varies significantly over the operating range from idle
to full power. The engine-derived correlations include terms
that reduce to the combustor pressure level squared.
186 R. E. MOTSINGER AND 0. J. EMMERLING
The temperature parameter also is an important term but

one in which the literature shows least agreement. It should
be noted that Strahle implicitly includes a temperature param-
eter equivalent to the combustor temperature rise by his use of
the fuel/air ratio F; in his recent paper, he concludes:
"The acoustic power is independent of fuel/air ratio at low
fuel/air ratios and makes a transition to a linear dependence
on fuel/air ratio above a fuel flow at which the primary
stabilization region becomes fuel rich.11 Reference 7 states
that the acoustic power is proportional to the combustor tem-
perature rise raised to the second power for combustor compo-
nent data but raised to the fourth power when the same combus-
tor was tested in the engine. (Note that the temperature term
in the denominator and the pressure term combine to form a

term proportional to the air density at the combustor
discharge.) References 8 and 10 state that the acoustic power
is proportional to the combustor temperature rise squared.
References 12 and 14 state that it is simply the combustor
discharge temperature squared. It should be noted that the
use of the combustor temperature rise squared is consistent
with the observation that the acoustic conversion efficiency
is directly proportional to the fuel/air ratio, as reported in
Refs. 23, and 9; the correlation of engine data from Ref. 4,
discussed in the following, used this form.
Spectrum and Directivity. As already pointed out, com-

bustor noise is difficult to identify from turbofan engine
data even at the frequency of peak noise. Accordingly, Ref. 3
adopted the recommendation of Ref. 14 to use the in-flight SAE
spectrum-1-3 with minor modifications; also, the prediction for
the frequency of peak noise from the same reference was
adopted, neglecting the Doppler shift factor,2^ with the
reservation that the validity of the equation was open to
question at that time. More recent information presented in
the next section confirms the SAE spectrum shape as a
reasonable envelope of test data but shows that the frequency
of peak noise is the same (within about 1/3-octave band) for
a wide variety of engine types, sizes, or power settings.
Likewise, Ref, 3 adopted the directivity recommendation of
Ref. 14. Additional data presented in the next section indi-
cate that the angle of maximum noise is 120, as originally
recommended, but that the levels at angles forward and aft of
the peak decrease more rapidly.
Recent Results
Data from seven engines have been reported recently in

Ref. 4. These engines include three turbojet engines with
J85 300o# THRUST)

170 D
J65 8000# THRUST ? TURBOJET
GE4 6oooo# THRUST)
O GE12 15OO SHP
A T64 3600 SHP
160 l\ TF34 9275# THRUST)
OQEP "A
PQ 150
I
J
140
co
130
O
o
120
100
Fig. 3 Power level correlation.

nominal thrusts from 3000 to 60,000 Ib, two turboshaft engines
of 1500 and 2850 shaft horsepower, and two turbofan engines of
9275- and 22,000-lb thrust. These data include variations, in-
cluding engine-to-engine and power setting variations, of core
engine weight flow from 3 to 600 Ib/sec, pressure ratio from
about 2 to 18, and combustor temperature rise from about 800
to more than 1400F.
Power Level* Based on the power level parameters deter-

mined from the T64 engine,^ the combustor OAPWL data for these
engines correlate as shown in Fig. 3. Clearly, there is a
difference among the three basic engine types with regard to
the noise level radiated to the far field. The level for the
turbofans is the lowest, the turboshafts intermediate, and the
turbojet the highest0 The approximately mean lines through the
data for each of the three types of engine are separated by
8 dB, so that the correlation indicates turbojets to radiate
16 dB more noise than turbofans, all other factors being equalo
A TF34 HIGH BYPASS RATIO TURBOFAN (9OOO# THRUST)

- JT3D LOW BYPASS RATIO TURBOFAN* (l8000# THRUST)
X T64 TURBOSHAFT (3OOO SHP)
D J85 TURBOJET (4000# THRUST)
O GE4 TURBOJET ( 60000# THRUST)
V GARRET COMBUSTOR COMPONENT TEST **
* REFERENCE NO. 24
** REFERENCE NO. 23
SAE 8?6 FLIGHT

JET NOISE
SPECTRUM
100 200 400 800 1600

1/3 OCTAVE BAND FREQUENCY, Hz
Fig. 4 Combustor noise spectra for turbofan

and turbojet engines.
At the present time, the cause for this unexpected behav-

ior must be the subject of further research. It is of
interest to note, in this regard, that those engines with the
lowest steady-state temperature drop across the turbine pro-
duce the greatest noise level in the far field. Generally,
those turbines with the lowest steady-state temperature drop
also consist of the fewest number of turbine stages. Thus, a
turbine attenuation mechanism is suggested by these data.
Spectrum. The combustor noise spectral distributions for

those engines in which the spectral shape could be identified
at the far-field angle of maximum noise are shown in Fig. 4; in
addition, the spectrum for the JT3D reported in Ref. 24 is
shown.. Data from a combustor component test reported in Ref.
23 are included. These data are supplemented by those already
presented in Fig. 1. For comparison, the SAE jet noise
spectrum is shown.
10
_ ]EXPECTED VALUE
h+^ n % *
v x<
1
>
\
S *'''" V
v N
O \
-5 JLpr
<>
o J85 - 4c0 Hz
n TF34 - 4oo Hz
-10 \ A GE4 - 400 Hz
<> F1O1 COMPONENT - 400 Hz
0 CF6 COMPONENT - 400 Hz
X ,T64 - 400 Hz
-15
60 80 100 120 140 160 180
ANGLE FROM INLET, (DEGREES)
Fig. 5 Combustor noise directivity.
These data reveal a remarkable constancy in the one-third-

octave band frequency of peak noise: it occurs at 400 Hz one
1/3-octave band. This is independent of the type or size of
engine and, in fact, occurs even in the component rig data.
Also, it can be seen that the SAE jet noise spectrum represents
an upper envelope of the data, and this shape is recommended
for predicting the suppression required in designing a com-
bustor noise suppressor. Note, however, that in determining
the power level data presented in Fig. 3 the spectral distri-
bution for the T64 engine was used; if the SAE jet noise
spectrum is to be used in prediction procedures based on the
data of Fig. 3 to determine the peak one-third-octave band
level, then the appropriate adjustment must be made for the
difference between OAPWL and maximum one-third-octave band PWL.
Directivity. The data in Fig. 5 include the data from

Ref. 4 for the directivity of combustor noise using those
engine data for which the combustor noise peak one-third-
octave band level could be identified clearly. In addition,
data from two combustor component test rigs (described in Ref.
4) are included. The peak noise angle is 120 from the inlet,
and the levels decrease more rapidly both fore and aft than the
correlation proposed in Ref. 3. The expected-value line shown
on the figure is recommended. Margin of error is indicated by
the data scatter.
III. Recommended Prediction Method
As a result of the new information now available, the

following correlations are recommended as an interim method for
predicting combustor sound power level, spectrum, and direc-
tivity.
Sound Power Level
The correlation from Fig. 3, distinguishing among engine

types, is recommended as follows:
OAPWL = 10 Iog10 ma [(T4 - T3> (VW^] 2 + K... (1)
where
K = 4 8 for turbofans
= 56 for turboshafts
= 64 for turbojets
m = core engine airflow rate, Ibm/sec
a
and where the temperature difference is in degrees Rankine.

The empirical constant K for each engine type is typical for
the engine designs used in the data base, which are, with only
one exception, from one engine manufacturer. The value of the
constant K depends upon the spectral distribution assumed for
the core noise. The values listed were based upon the use of
the T64 spectrum, which results in the OAPWL being 6.8 dB above
the peak level one-third-octave band. Thus, in using Eq. (1)
to determine the peak one-third-octave band, use
PWL of the peak one-third-octave band = OAPWL - 6.8 (2)
This procedure was followed in reverse order in determining

the OAPWL in Fig. 5 to provide a consistent basis for all
engine types, including those for which the spectral distri-
bution could not be discerned because of other contaminating
noise sources (e.g., fan or jet).
Spectrum
The SAE jet noise spectrum is recommended as an upper

envelope of the spectral distribution, with the one-third-
octave band frequency of peak noise being placed at 400 Hz
( one 1/3 octave band). This distribution should be applied
relative to the PWL of the peak one-third-octave band calcu-
lated by Eqs. (1) and (2).
Directivity
The directivity index as given in Fig. 5 is proposed for

consideration as an alternate to that recommended in Ref. 3

unless some conservatism or margin-of-safety is desired in the
predicted far-field pattern0 If so, the recommendation of
Ref. 3 should provide assurance of overpredicting rather than
underpredicting the far-field pattern relative to the maximum
angle, which is 120 from the inlet in either case. Note that
the directivity in Fig. 1 of Ref. 3 should be converted to a
directivity index to be consistent with the recommendations
for the OAPWL and spectrum given in the foregoing.
IV. Suggested Research Areas
The following areas require further research to improve

upon the recommended method presented in this paper:
1) Investigate whether the low-frequency noise, called

combustor noise in this paper, is generated by the combustor
itself or by nonuniformities associated with the combustion
process propagating through regions having a mean pressure
gradient.
2) Investigate the possible causes for the differences

in radiated power level from atmospheric combustor-component
tests and from the same combustor installed in engines, and
establish a correlation so that component test data can be
used for predicting combustor noise installed in engines.
3) Investigate the possible causes for the different

noise levels radiated from different types of engines, in-
cluding analysis and tests of turbine and nozzle low-frequency
noise attenuation.
4) The effect of combustor pressure level on combustor

noise should be determined.
5) The proper form of the temperature parameter, or

fuel/air ratio, should be determined.
6) The observation that the frequency of peak noise

level is essentially constant should be investigated further
by analysis and by additional data. Further data are needed
to establish the spectral distribution with less data scatter.
7) Controlled tests of engines with minimum contami-

nation from other noise sources and from ground effects should
be conducted to improve the resolution of the far-field
directivity pattern.
8) Methods for reducing combustion noise should be in-

vestigated. This should include various combustor geometries
and various current and advanced methods of fuel injection and
liner mixing geometries.
9) Methods of suppressing combustor noise should be

investigated.
References
?
Kazin, S.B., and Paas, J.E. , "NASA/GE Quiet Engine A?
Acoustic Test Results, 11 CR-12175, Oct. 1973, NASA.
2
Pickett, G.F., "Turbine Noise Due to Turbulence and Tem-
perature Fluctuations," Eighth International Conference on
Acoustics, July 23-31, 1974, London.
Huff, R.G. Clark, B.J., and Dorsch, R.G., "Interim Pre-

diction Method for Low Frequency Core Engine Noise,"
TM X-71627, Nov. 1974, NASA.
4
"Core Engine Noise Control Program, Final Report," Rept.
FAA-RD-74-125, Aug. 1974, U.S. Department of Transportation,
Federal Aviation Administration System Research Development
Service, Washington, D.C.
Bushell, K.W., "A Survey of Low Velocity and Coaxial Jet

Noise with Application in Prediction," Journal of Sound and
Vibration, Vol. 17, No. 2, July 1971, pp. 271-282.
Marshall, D.A.A., "Sources of Noise in Aero-Engines," 1st

International Symposium of Air Breathing Engines, June 19-23,
1974, Marseille*
Ho, P.Y. and Tedrick, R.N., "Combustion Noise Prediction

Techniques for Small Gas Turbine Engines," Inter-Noise 72; In-
ternational Conference on Noise Control Engineering, Pro-
ceedings, Oct. 4-6, 1972, Washington, D.C.
o
Motsinger, R., "Prediction of Engine Combustor Noise and
Correlation with T64 Engine Low Frequency Noise," R72AEG313,
1972, General Electric Company.
9
Knott, P.R., "Noise Generated by Turbulent Non-Premixed
Flames," AIAA Paper 71-732, June 1971.
10
Grande, E., "Core Engine Noise," AIAA Paper 73-1026,
Oct. 1973.
Arnold, J.S., "Generation of Combustion Noise," Journal of

the Acoustic Society of America, Vol. 52, No. 1 (Pt. 1)
January 1972, pp. 5-12.
12
Gerend, R.P., Kumasaka, H.A., and Roundhill, J.P., "Core
Engine Noise," AIAA Paper 73-1027, Oct. 1973.
;also, AIAA Progress in Astronautics and Aeronautics-Aero-
acoustics; Jet and Combustion Noise; Duct Acoustics, Vol. 37,
edited by H. T. Nagamatsu, J. V. O'Keefe, and I. R. Schwartz,
1975, pp. 305-326.
13"Jet Noise Prediction," Aerospace Infor. Rept. 876, July

1965, Society of Automotive Engineers,
Dunn, D.G. and Peart, N.A., "Aircraft Noise and Contour

Estimation," CR-114649, 1973, NASA.
Strahle, W.C., "A Review of Combustion Generated Noise,"

AIAA Paper 73-1023, Oct. 1973;also, AIAA Progress in Astronautics
and Aeronautics-Aeroacoustics; Jet and Combustion Noise; Duct
Acoustics, Vol. 37, edited by H. T. Nagamatsu, J. V. O'Keefe,
and I. R. Schwartz, 1975, pp. 229-248.
Swan, W.C. and Simcox, C.D., "A Status Report of Jet Noise
Suppression as Seen by an Aircraft Manufacturer," First Inter-
national Symposium on Air Breathing Engines, June 18-22, 1972,
Marseille.
194 R. E. MOTSIN6ER AND J. J. EMMERLING
Abdelhamid, A.N., Harrje, D.T., Plett, E.G., and SummerfieId,

M., "Noise Characteristics of Combustion Augmented High-Speed
Jets," AIAA Paper 73-189, Jan. 1973. ;also, AIAA Journal, Vol.
12, March 1974, pp. 336-342.
Plett, E.G., Chiu, H.H., and Summerfield, M., "Research on

Noise Generated by Ducted Air-Fuel Combustion Systems," 1973,
Princeton University.
19 Chiu, H.H. and Summerfield, M., "Theory of Combustion
Noise," Fourth International Colloquium on Gas Dynamics of
Explosions and Reactive Systems, July 10-13, 1973, San Diego,

Calif.
20
Chiu, H.H., Plett, E.G., and Summerfield, M., "Noise Generated
by Ducted Combustion Systems," AIAA Paper 73-1024, Oct. 1973.
;also, AIAA Progress in Astronautics and Aeronautics-Aero-
acoustics: Jet and Combustion Noise; Duct Acoustics, Vol. 37,
edited by H. T. Nagamatsu, J. V. O'Keefe, and I. R. Schwartz,
1975, pp. 249-276.
Von Glahn, U. and Goodykoontz, J., "Forward Velocity Effects

on Jet Noise with Dominant Internal Noise Source," 86th Meeting
of the Acoustical Society of America, Oct. 30-Nov. 3, 1973, Los
Angeles, Calif.
22
Strahle, W.C. and Shivashankara, B.N., "Combustion Generated
Noise in Gas Turbine Combustors," Paper 75-GT-27, Dec. 1974,
American Society of Mechanical Engineers.
23
Ho, P.Y. and Tedrick, R.N., "Combustion Noise Prediction
Techniques for Small Gas Turbine Engines," Rept. SD-8006,
May 10, 1972, Garrett AiResearch Manufacturing Company of
Arizona.
24
Mathews, D.C. and Peracchio, A.A., "Progress in Core Engine
and Turbine Noise Technology," AIAA Paper 74-948, Aug. 1974.
PANEL DISCUSSION: AIRCRAFT NOISE PREDICTION
J. P. Raney, NASA Lanc|ley Research Center: The purpose of

the aircraft noise prediction session was to focus upon the
requirements for development of a generally accepted computer
program for noise prediction. We are interested in establish-
ing a mechanism that will tell us when we have a satisfactory
consensus for a particular technology area. There is, among

other things, a great need for advances in analytical modeling,
which at the present time is in its infancy. The greatest
overall need is to develop and to verify experimentally analyti-
cal methods of modeling noise generating mechanisms.
Flow field. The first step, then in improving the ability
to predict noise, is to build a consistent and complete aero-
thermodynamic model for the aircraft engine. This model re-
quires a general thermodynamic balance of the different stages
of the engine cycle, a description of the turbulent atmosphere
being drawn into the engine inlet, the flow and flow gradients
in the engine ducts and blade rows, and the wall boundary
layers and blade wakes. When this information is developed,
the work on the acoustics problem of the engine may begin.
Fan noise. The details of the flow field discussed above
are especially important for fan noise prediction. Large-scale
atmospheric turbulence is drawn into the engine inlet, causing
a nonuniform axial flow into the fan blades. As the blades
rotate, this nonuniform flow causes unsteady loads on the
blades due to the varying angle-of-attack. These unsteady loads
radiate dipole noise in harmonics of the blade passage fre-
quency. They also generate broadband noise due to the random
fluctuations of the blade load amplitudes and phases. This
noise caused by inlet flow distortion has been identified as
a key technology area for which research is needed. The under-
standing of both the fluid mechanics and the acoustics of the
inlet flow distortion problem is necessary for advancing the
state of the art of fan noise prediction.
Combustion noise. The unsteady combustion process in the
engine generates a low-frequency noise which sometimes has been
confused with low-frequency jet noise. The available predic-
tion theory for this noise is empirical in nature and does not
195
196 DISCUSSION: AIRCRAFT NOISE PREDICTION
account for the fact that this noise must be carried through
the turbine and exhaust nozzle before it is radiated to the
far field. In order to understand this phenomenon better, a
good understanding of the flow through the turbine and exhaust
nozzle and the effect of this flow on the combustion noise
transmission are required. Again, basic thermodynamics and
fluid mechanics are an inseparable part of the acoustic pre-
diction problem.
Turbine noise. Like combustion noise, turbine noise
presently is predicted by empirical formulas which account for
only the gross variables of the problem. The few analytical
models which have been attempted use concepts similar to fan
noise in which the blades are replaced by concentrated dipoles

which represent the unsteady blade loads. However, such models
may be completely inappropriate in a turbine with high solidity
stages of highly cambered airfoils. The presence of many
stages in the turbine greatly attenuates the sound of all but
the last stage so that the sound generation and transmission
process in the turbine is quite complicated. A fundamental
approach based on realistic models of the turbine flow is
needed for turbine noise prediction. Turbine noise radiation
is influenced also by the unsteady flow field of the jet. Tones
generated by the turbine are transformed into broadband noise
as they radiate through the unsteady turbulent jet flow. This
process has been called "haystacking" because of the charac-
teristic shape of the broadband noise which results from this
process. In turbine noise, an understanding of this effect of
unsteady turbulent flow on sound propagation is required for
improvement of our predictive ability.
Duct acoustics. Noise from sources inside of the CTOL
engine may be attenuated by the addition of sound-absorbing
material inside the nacelle. Very precise complex analytical
models of duct transmission have been developed; however, these
analyses are for idealized duct and flow models. In an actual
engine, the duct wall boundary layer significantly affects the
attenuation of the sound, especially in the inlet. Thus, a
realistic description of the flow is necessary before a pre-
diction can be made. Also, these precise analytical models of
duct transmission are based on a linear boundary condition, the
duct wall impedance. It is known that the acoustic materials
used in engine nacelles are nonlinear at the sound intensities
which occur in these engines and that the flow over the
materials has a major influence on this property. Therefore,
a primary area where work is needed in duct acoustics is the
modeling of duct problems with realistic nonlinear boundary
conditions.
DISCUSSION: AIRCRAFT NOISE PREDICTION 197
Much of the work in duct acoustics in the past 10 years

has been developed using the modal theory of sound transmission.
Unfortunately, researchers have carried idealized transmission
analyses to extremes, making predictions of attenuation based
on a single-mode assumption. Attempts also have been made to
generate pure modes in the laboratory in order to verify their
properties. Real engine noise sources, however, are always
represented by a large number of modes interacting in a complex
manner, and this must be accounted for in any realistic pre-
diction attempt. The modeling of real sources as well as real
boundary conditions is necessary for improving the state of
the art in duct acoustics. Also, modal theory is not essential
to the duct propagation phenomenon; it is only a tool. Other

tools are now being considered. One which shows promise is
the finite element method which has reached a high level of
development in the field of structural analysis. Just as duct
acoustics modal theory evolved from electrical transmission
line theory, the finite element techniques of structural analy-
sis may be developed into an acoustic transmission line theory
which will be competitive with modal theory in the prediction
of duct acoustic effects. The development and comparison of
both of these methods is a fertile area for further research.
Jet noise. Jet noise is one of the oldest subject areas
of concern in the overall CTOL noise problem. In spite of
this, our predictive capability for real-world jet noise prob-
lems is not well developed. Presently, an empirical formula
is being used by NASA for jet noise predictions. The diffi-
culty with empirical formulas is that each is derived to
represent only a certain set of data. The SAE A-21 committee,
for example, has an empirical jet noise prediction formula
which no doubt represents their data, but NASA and SAE pre-
dictions are different. They are different because they are
based on different data. To eliminate these differences, it is
necessary to develop a unified data base for jet noise. The
data which are entered into this base should be required to
meet certain standards e âbl^shed by the peer group of experi-
mentalists in this field. These experimental standards will
rule out certain carelessly conducted experiments and define
the subset of jet noise data which will be included in the jet
noise data base. The gathering of this information also will
define additional experiments to be carried out. Then, if an
empirical correlation is made, only one formula may be con-
sidered "best." This is the formula with the least variance
of the estimate.
A unified data base also will serve to define the direc-
tion that analytical work in jet noise should take. If
198 DISCUSSION: AIRCRAFT NOISE PREDICTION
empirical formulas are inadequate, analytic models based on

Lighthill's, Phillip's, or Lilley's equation may be used. In
these partial differential equations, the source terms must be
modeled by some assumed turbulent flow. Here again the basic
fluid mechanics of turbulent flow enter the picture. It is
necessary to compare a sequence of models for the source terms
in both Lighthill's, Lilley's, and other jet noise formulations
to see which provides the least variance of the estimate
against a unified data base. When comparing the solutions to
partial differential equations, however, the accuracy of the
prediction is not the only criterion which may be cited to
determine which of several methods may be best. The cost of
prediction, as judged by computation time, for example, is

another factor which must be considered. Perhaps the most
important consideration of all is, does the predictive equation
provide a realistic method for achieving noise reduction? All
of these factors must be considered in arriving at a "best" jet
noise prediction method.
Airframe noise. Besides the engine, the various components
of the airframe may radiate significant amounts of noise during
the landing approach of a CTOL aircraft. Here again we are at
the empirical formula level in our state-of-the-art prediction
capability. Presently, we use a formula developed for aircraft
in the "clean" configuration from a limited but well-defined
data base. It is recognized that the extension of flaps and
landing gear will increase the airframe noise by 10 dB or even
more, so that the present prediction method is an interim
device used for order-of-magnitude estimations. A promising
empirical approach which accounts for the effects of flap ex-
tension is the drag element noise theory. In this theory, each
airfoil is assumed to produce a noise in proportion to the cube
of its drag coefficient. This theory is related to the analytic
theory of edge noise which is probably the dominant component
of airframe noise. Edge noise theory, however, depends on the
turbulent flow conditions at the trailing edge of an airfoil,
so we see a fundamental dependence of the acoustics of airframe
noise on the fluid dynamics of the airfoil. Research in this
field must proceed along a consistent path using valid models
of the turbulent boundary layers in comparably valid acoustic
theories. Experiments in edge noise must simultaneously study
the fluid dynamics of the turbulent flow and the noise radia-
tion. Precise flight tests also are required to validate
empirical theories such as the drag noise theory.
Noise propagation. CTOL noise must propagate over large
distances before it reaches the community. The character of
the noise is modified during this propagation process due to
DISCUSSION: AIRCRAFT NOISE PREDICTION 199
the dependence of attenuation on such factors as frequency,

temperature, and humidity. Fortunately, available prediction
methods account for the more important absorption processes,
classical absorption and molecular absorption, if the ambient
atmospheric conditions are known along the ray from the source
to the observer. There remains some controversy about the
effects of atmospheric turbulence on propagation which must be
resolved by careful experimental work. A more important re-
search area relates to the effects of ground absorption on the
propagation of sound. There is a strong theoretical base for
prediction of ground absorption, but these prediction methods
depend on the impedance of the earth surface which is seldom,
if ever, known. Thus, careful studies are required to develop

a data base of ground impedance data for the various types of
terrain which are involved in the aircraft noise propagation
problem. The development of these data by careful experiments
will greatly improve the accuracy of our noise prediction
methods.



THE USE OF HARTMANN GENERATORS AS SOURCES OF HIGH-
INTENSITY SOUND IN A LARGE ABSORPTION FLOW-DUCT
FACILITY
D. L. Mart lew*
National Gas Turbine Establishment, Pyestock,

Farnborough, Hants, United Kingdom
Abstract
Air driven sources of high-intensity sound, comprising

groups of Hartmann generators, were developed for the large
absorption facility at NGTE. A single, 1 in. generator
demonstrated an acoustic power of 2 kw, a tuning range of 1
octave, a power range of 20 dB, and a peak efficiency of 5%.
An "in-duct" source was developed utilizing 32 Hartmann
generators and incorporating an air-collection feature. It
produced tones at multiples of 1 kHz and gave 160 dB OASPL in
the test duct. Precautions in measurement and analysis,
necessitated by the tonal spectrum, were examined and a band-
width of 50 Hz was shown to assure correct interpretation of
data. Finally, a simpler source was developed for the same
duty, comprising groups of 4 in. Hartmann generators in the
reverberation chambers. This produced intense sound at
frequencies down to 400 Hz, and could be adjusted to give a
smoother, broadband spectrum.
Introduction
The steady broadening of the program of aero-engine noise

research in the United Kingdom has required a corresponding
extension in experimental facilities. The first large-scale
special-purpose anechoic facility was provided in 1966 at the
Rolls Royce plant at Ansty, near Coventry, for the noise
Presented as Paper 75-529 at the AIAA

2nd Aero-Acoustics Conference, Hampton, Va.,
March 24-26, 1975
^Principal Scientific Officer
203
204 D. L. MARTLEW
testing of engine fans and compressors. More recently, the

extensive Noise Test Facility has been built at the National
Gas Turbine Establishment at Pyestock, near Farnborough. This
comprises two units, the absorber facility, completed in 1972,
and the anechoic facility, which came into operation in 1974.
The absorber facility consists of a flow duct, in which a

Mach number up to 0.7 may be achieved with a working cross-
section of 4 ft2 (0.37 m2). The latter is situated between
two reverberation chambers, each about 20 ft (6.1 m) wide,
16 ft (4.9 m) high, and 14 ft (4.3 m) long and some 4500 ft3
(130 m3) in volume. The duct flow is provided by a large
centrifugal fan, and splitter silencers are used to control the

noise emitted from the intake and exhaust of the facility and
to isolate the upstream reverberation chamber from the noise
of the fan. A line diagram of the layout is given in Fig. 1.
This paper concentrates on the absorber facility, and its

main purpose is to present and discuss the experience accum-
ulated over nearly three years in the use of multiple Hartmann
generators as very high-powered sources of artificial sound for
testing absorptive linings in the flow duct. The development
of an "in-duct" noise source and its operational characteristics
are described. This produced the strong discrete tones that
are characteristic of the spectrum of Hartmann generators, and
the effect of these tones on the measurements is considered.
Similarly, the presence of discrete frequencies means that
certain precautions have to be taken to ensure that the
DOWNSTREAM
REVERBERATION
CHAMBER
IN-DUCT NOISE GENERATOR

POSITIONS
Fig. 1 Diagram of absorber facility

THE USE OF HARTMANN GENERATORS 205
measurements are interpreted correctly during analysis.

Finally, an account is given of the development of a simpler
noise generator system, again based on multiple Hartmann
generators, having improved operational flexibility and a more
nearly broadband spectrum.
General Characteristics
of Hartmann Generators
In its conventional form, the Hartmann generator consists

of a convergent nozzle delivering a supersonic jet of air

which impinges directly onto the open end of a resonance
tube,2-4 as shown in Fig. 2. If the distance between nozzle
and tube, or cup, is chosen correctly, the shock wave that
stands between them becomes unstable in position and responds
vigorously to the acoustic pressure wave in the resonance tube.
The resulting shock motion, coupled with the alternate filling
and emptying of the tube, causes the emission of high-intensity
sound with a spectrum consisting of a fundamental tone at a
frequency related to the depth of the resonance tube or cup,
accompanied by an extensive family of harmonics and an
underlying broadband component (Fig. 3). Clearly, provided
that suitable air supplies are available, the Hartmann
generator provides a simple and inexpensive means of
generating high-intensity sound, and as such it has been
investigated in some detail at NGTE, where it has been used
successfully in a number of experimental situations. Before
dealing with the development of the noise source systems for
the absorber facility, it is pertinent to consider some of the
NOZZLE-- RESONATOR CUP PWL-dB(re10~ 1 2 W)

150
130
GAP CUP DEPTH
Fig. 2 The Hartmann acoustic

generator
110
Fig. 3 Typical spectrum of 2 4 6 8 10

Hartmann generator FREQUENCY-KHz
206 D. L. MARTLEW
111
Fig. 4 One-inch Hartmann generator
characteristics of a single Hartmann generator, measured

principally on the simple 1 in. diam unit shown in Fig. 4.
Effect of Gap between Nozzle and Cup
This distance primarily affects the acoustic output of the

generator through the stability of the shock wave. Best
results are obtained when the cup is situated in the first
shock cell of the jet, i.e., of the order of one nozzle
diameter downstream of the exit with pressure ratios of about
3/1, although oscillation can be obtained at other settings.
As a rule, a gap of one nozzle diameter has been adopted as
standard, and this has generally yielded peak acoustic
efficiencies of 5 to 6%. Since the shock cell length reduces
with pressure ratio, a smaller gap would be required to obtain
optimum output at pressure ratios down towards 2/1. However,
in the applications under discussion, reduction in blowing
pressure was used as a means of reducing acoustic output, and
the achievement of peak output at all pressures was not of

importance.
Effect of Cup Dimensions
Although tests have shown that cups of greater or lesser

diameter than the nozzle can operate effectively, no signif-
icant advantage was found, and -the configuration adopted as
standard uses cups equal to the nozzle diameter.
The principal variable in the cup is its depth, which

controls the frequency of oscillation. It is possible to use a
long resonance tube if a low frequency is required, and, on the

other hand, an oscillation can be obtained with zero cup depth.
Efficiency falls off at these extremes, and the optimum
output has been found to correspond with a cup depth equal to
the nozzle diameter. Variation of cup depth in the range 0.7
to 1.4 diameters can normally be used without excessive loss
of output and allows adjustment of the fundamental frequency
over about one octave. The relation connecting frequency with
cup depth, treating the latter as a one-quarter wave tube, may
be written in the form:
c/fD = 4K [(d/D) + kj
where f is the fundamental frequency, D is the cup diameter,
c is the velocity of sound, d is the cup depth, K is a constant,
and k is an end correction for the cup. This relationship is
fitted to measured frequencies of a variety of Hartmann
generators in Fig. 5, and it is seen that the cup behaves
approximately as a one-quarter wave tube with an end correction
of 0.77 diam . The value of the velocity of sound at the
stagnation temperature was assumed to be appropriate.
Effect of a Central Needle
In its optimum form, with nozzle-to-cup spacing, cup

diameter, and cup depth all equal to nozzle diameter, the
15
10
CUP DIAMETERS f Y If
PRESSURE RATIOS 2-2-3.4
Fig. 5 Relation between A. 3

frequency and cup depth D
208 D. L. MARTLEW
Hartmann generator, although capable of good acoustic

efficiency, had a marked tendency to behave erratically. The
oscillation would often appear in bursts and occasionally be
reluctant to start at all. The addition of an axial rod or
needle extending from the nozzle exit to the base of the cup
had a markedly beneficial effect in making the oscillation
more regular and starting completely certain. 4" >^ It was
convenient to use screw-threaded rod of one-eighth of the
nozzle diameter supported in the base of the cup, although
plain rod was equally effective. An additional benefit was
that a useful acoustic output was obtained at subsonic blowing
pressures, although obviously the mechanism of shock
oscillation was no longer valid. It is seen from the power

curves in Fig. 6 how the addition of a needle results in a
smooth reduction in output as pressure ratio is reduced to as
little as 1.2, conferring a controllable 20 dB, i.e., 100/1,
power range on the Hartmann generator.
Effect of Nozzle Pressure Ratio
In terms of acoustic power, the effect of varying nozzle

pressure is covered by the remarks of the previous section and
by Fig. 6. Maximum output was obtained, for the various sizes
and designs of Hartmann generator that were investigated, at
pressure ratios in the range 2.8/1 to 3,1/1, a definite fall
being apparent at 3.5/1. For the 1 in. unit the peak output
was measured in an anechoic chamber, and confirmed in a small
reverberation chamber, as 153 dB re 10~12w, i.e., about 2 kw,
corresponding to an efficiency of 5% based on isentropic
expansion from the nozzle supply pressure to ambient pressure.
The variation of the acoustic output by reducing pressure to
subsonic levels is accompanied by spectral changes. At
supersonic conditions an extensive family of harmonics is
present, whereas at subsonic conditions the higher harmonics
die out. Pressure has little effect on the frequency of the
oscillation except that there is a slight tendency for it to
rise at values above 3.2/1. A slight rise is also observed
as pressure falls through the subsonic range.
^RELATIVE POWER
WITH NEEDLE
~j-^ Fig. 6 E f f e c t of needle on

PRESSURE RATIO power output
These exploratory tests on the Hartmann generator thus

showed that for optimum output the cup diameter, cup depth,and
spacing between cup and nozzle should all equal the nozzle
diameter. A needle of diameter equal to one eighth of the
nozzle diameter greatly improves the steadiness and range of
operation. Peak output occurs at a pressure ratio of about 3/1,
at which condition the generator has a substantially uniform
directivity.
Development of the "In-Duct"

Noise Generator
The specification of the absorber facility called for a

test duct of 4 ft2 (0.37 m) cross section, a maximum Mach
number of 0.7, and5 an overall sound pressure level of at least
160 dB re 2 x Kf NnT2, with the sound transmitted either with
or against the flow. Measurements were called for at
frequencies in the range 500 Hz to 20 kHz and this, together
with considerations of the mean air velocity, led to the choice
of rectangular reverberation chambers of some 4500 ft3 (130 m3)
volume with proportions approximating those giving the most
uniform distribution of modes. In principle, the absorption
of a given acoustic treatment in the duct is obtained by
comparing the differences between the average sound pressure
levels in the upstream and downstream reverberation chambers
when the duct is untreated and when it is treated. For this
purpose, a suitable source of sound is placed in either the
upstream or downstream reverberation chamber, according to the
direction of propagation it is required to study.
In view of the considerable size of the reverberation
chambers and the presence of large openings in them containing
absorptive splitters, an exceedingly high power sound source
appeared to be necessary. One estimate, made at the design
stage of the facility, indicated that well over 100 kw of
acoustic power would have to be generated in either reverber-
ation chamber if a SPL of 160 dB were to be produced in the
working section. It was hoped that to some extent the losses
from the chambers into the splitter banks could be minimized
by providing some sort of perforated but acoustically
reflective barrier with sufficient area to pass the airflow,
but the conflicting requirements for minimum flow pressure loss
but maximum acoustic reflection made it difficult to rely on a
lower source power being acceptable. The experience already
gained at NGTE and recounted in the foregoing section had
ranked the Hartmann generator as the preferred noise source.
Easily within the capabilities of the existing air supplies,
its simplicity and ruggedness weighed in its favor in
210 D. L. MARTLEW
comparison with available sirens and electrodynamic or electro-

pneumatic sources. It offered a useful degree of control and
tuning, and it was possible to argue that the tonal spectrum,
which is very similar in nature to that of typical aero-engine
fans, would confer a degree of realism to the testing of
liners for application to fan ducts, although, in common with
most other sources, it was not possible to complete this
realism by controlling the modal distribution of the generated
sound.
In order to get around the problem of generating appar-
ently impractical quantities of sound in the reverberation
chambers,an "in-duct" noise source was conceived and, after
certain confirmatory tests successfully demonstrating its

principles, was built for the absorber facility. The sound
was produced by a number of Hartmann generators, situated
around an annular recess formed in the wall of the ducting near
to the working section where the inside diameter was 5 ft 6 in
(1.7 m). The diagram in Fig. 7 shows the arrangement, which
employed 24 generators of 1-in. diam and 8 of 2-in. diam, all
of fitted with needles. The total airflow passing through all
these nozzles amounted to almost 50 Ib/sec (23 kg/sec), and
this was prevented from interfering with the main airflow
passing to or from the test duct by the novel expedient of
fitting each individual Hartmann generator with a cylindrical
sleeve, of about three times the nozzle diameter, placed around
the cup. These collected each jet and delivered it to a mani-
fold from which it was ducted away from the facility to a
silenced exhaust under its own momentum. A set of exploratory
tests was performed to reveal any degradation of performance
caused by operating Hartmann generators in close proximity to
one another and by partially enclosing them. The principle of
air collection also was explored experimentally and shown to be
entirely viable; full capture was obtained against a
significant back pressure with the mouth of the collector tube
in the plane of the cup lip, and it was clearly possible to
trade the quantity collected against the position of the
CUP-^ NOZZLE/COLLECTOR TUBE /
NOZZLE AIR SUPPLY DISCHARGE

Fig. 7 Arrangement of "in-duct" noise generator
Fig. 8 Demonstrator section of

"in-duct" generator
collector and the back pressure in the actual installation. A

small section, containing one 2-in.-diam and two l-in.~diam
units and fully representing the actual design that evolved as
a result of these tests, was constructed (Fig. 8), and its
performance was measured. A peak power of 5 to 6 kw was
observed, from which it was estimated that the full-size
generator would deliver 40 to 50 kw. On the simple assumption
that this energy would divide itself between the 4-ft2 test
duct and the 24-ft2 duct communicating with the adjacent
reverberation chamber in proportion to these areas it was
predicted that the desired 160-dB SPL would be realized in the
test section with reasonable certainty.
The in-duct noise generator was installed in the completed

facility, and the cup depths were set to cause the 2-in. units
to resonate at 1 kHz and the 1-in. units at 2 kHz. Figure 9
shows the spectra of the sound at entry to the test section at
the pressure ratios required to give 160-,150~,and 140-dB
160 SPL-dB BANDWIDTH 50Hz
PRESSURE RATIO 3:1

140 OASPL 160dB
120
100
A 6 10
FREQUENCY - KHz
Fig. 9 Spectra of "in-duct"
noise generator
212 D. L. MARTLEW
OASPL. It was found possible, during the commissioning of the

facility, to cutback the collector tubes to well downstream of
the cup lips and still to obtain full capture of the jets and
thus to minimize the loss of sound into the collector system,
It was also found necessary to add a little throttling to
increase the back pressure in the collectors until they were
drawing through ejector action only a slight excess above the
nozzle flow. In this way, all interference by the noise
generator air with the distribution of the main rig flow was
eliminated, and velocity traverses in the test section were
unable to detect any change when the nozzle air was turned on
or off. The in-duct noise generator operated in this form
with complete reliability for the first two years of activity

in the absorber facility.
Measurement and Analysis Considerations

Using the "In -Duct" Generator
A number of assumptions, observations and manipulations

have to be made in order to deduce from a flow duct experiment
the absorption due to the application of acoustic treatment to
the duct. These processes are examined now in order to
demonstrate that the siting of the noise source in the duct
rather than in one of the reverberation chambers does not make
them untenable. Similarly, the spectrum of the sound produced
by the Hartmann generators contains a number of strong tones,
and the possible effects of these on the apparent performance
of an acoustic lining are discussed.
Partition of Energy Upstream and Downstream

The absorption due to a given acoustic treatment is deduced
from the difference between the hard-walled and the soft-walled
transmission losses. Both transmission losses are measured
under chosen environmental conditions, e.g., overall sound
pressure level and flow Mach number, and are defined by an
equation of the form
TL = Li - Ls + 10 logic K
where LI and L2 are the average sound pressure levels in dB in

the source and receiving chambers. K is a constant dependent
on the acoustic losses in the system. The attenuation due to a
given liner is then given by
10 Io
8io KSoft/Khard
It is usually assumed that the constant K does not change
between the hard-walled and the soft-walled situation and the
last term in the equation above disappears, making a knowledge

of the value of K unnecessary.
With the noise source positioned in the flow duct,the

sound level in the adjoining reverberation chamber is used as
a measure of the acoustic power entering the test section. It
can be shown that transmission loss is now given by
TL = Li ~ L2 10 loga /(I - a) + 10 log K
where the terma/(l -a) represents the split of acoustic

power between the test duct on one side of the source and the
reverberation chamber on the other. Thus, an additional
assumption has to be made when using the change in the
difference between reverberation chamber levels as a measure
of attenuation, namely that the energy split does not change
between the hard and soft walled situations. Obviously this is
a difficult assumption to prove,but confidence in its
correctness has built up with experience and no discrepancy
has been encountered which has cast doubt upon it. More
recently, after cross checks using sources in the reverberation
chambers, the very close agreement in measurements made with
the quite different sound sources has provided strong
circumstantial evidence in favor of the measurement technique
used with the in-duct source.
Averaging SPL in the Reverberation Chambers
A number of linear traverses were made in each reverber-

ation chamber during the commissioning of the absorber facility
in order to determine the diffuseness of the sound field and
to aid the choice of the number and positions of microphones
which would give a stable average value of sound pressure
level in any frequency band of interest. Fig. 10 illustrates
a typical result for a narrow frequency band containing the
principal tone from the in-duct generator and which shows the
___________TRAVERSING MICROPHONE
TRAVERSING MICROPHONE
10dB
._!_
FIXED MICROPHONE
FIXED MICROPHONE
10SEC-H MKHz 6/oBANDWIDTH

Fig. 10 Reverberation chamber
traverses
214 D. L. MARTLEW
greatest variation in level during the traverses. The fixed

microphones show that local fluctuations in level of around
2 dB occur in periods of the order of a second. If they were
due to variations in source power, they were slow enough to be
observed simultaneously at the other fixed microphones. No
such correspondence was found,and it therefore seems probable
that the fluctuations are principally due to a constant
shifting of the standing wave pattern in the chamber as a
result of the unsteady phasing of the oscillations of
individual generators. The traces from the moving microphones
show signs of a spatial pattern in sound pressure level, but
the most frequent variations are the same as seen at the fixed
microphones. This shifting of the sound field has a smoothing

effect which simplifies the determination of mean sound
pressure level from a small number of microphones, provided
that a reasonable time average of each is taken. Experience
has shown that stable and repeatable values of mean SPL can be
obtained using only two fixed microphones in each chamber,
although three are used where possible.
Effect of Duct Flow on Spectrum
When the sound from the noise generator propagates

through the duct in the presence of airflow, some scattering
of acoustic energy from the peaks of the tones has been
observed. This takes the form of a broadening of the tones,
accompanied by a slight reduction in the height of the peaks,
and is due to unsteady propagation through the turbulent flow.
Fig. 11 illustrates how the phenomenon becomes increasingly
evident as speed increases. Checks have shown that no
significant loss of energy takes place, and the main effect
upon the processing of data is that, when upstream and down-
stream sound pressure levels are subtracted in order to obtain
transmission loss, a pair of small spurious kinks appear in
SPl-dB
TEST DUCT ENTRY
100
DOWNSTREAM REV. CHAMBER
4 6 10
FREQUENCY - KHz
Fig. 11 Effect of flow on
spectrum shape
the spectrum at the position of each tone. The selection of

larger bandwidths for analysis reduces the effect, but the
necessity for restricting bandwidth in order to minimize the
risk of incorrectly evaluating attenuation peaks, as discussed
in the following section, means that this solution cannot be
adopted without care. To a considerable degree, the spurious
irregularity disappears in the final attenuation spectrum,
since the subtraction of two transmission loss spectra, each
with similar spurious variations, is involved. At worst, it
means that a slight visual smoothing of the finished result has
to be made.
Effect of Bandwidth of Analysis
The attenuation of a sound-absorbing system is the ratio

of the sound power entering the system to that leaving it at a
given frequency. A correct value for attenuation cannot, in
general, be obtained by comparing sound powers entering and
leaving the system when they are measured over a band of
frequencies centered upon the frequency of interest, unless the
distribution of source power and of the attenuation across the
band are almost uniform. In practice, sound is always analyzed
into bands of considerable width; for example,the use of one-
third octave bandwidth is very common and some of the early
results from the absorber facility were handled in this way
before a rapid means of narrow-band analysis became available.
It soon became evident, however, that some of the sharper
attenuation peaks were being underestimated or even missed,
and 50-Hz constant bandwidth analysis was adopted as standard
as soon as possible in order to avoid these problems. Some
simple calculations showed the extent to which experimental
results could be distorted by using 25% (approximately one-
third octave) bandwidth. Fig. 12 shows two examples; in one
case a sharp attenuation peak is underestimated by over 6 dB,
even though the source spectrum is flat^whereas in the other
the presence of a narrow tone near the true attenuation peak
results in apparent attenuations which, if deduced from 25%
bandwidth levels, would be even more misleading. Not only
could the value of peak attenuation be underestimated by some
8 dB, but the frequency at which it occurred would appear to be
higher than is actually the case. These calculations showed
that, in general, the sharpness of the tone did not have a very
strong effect but that, if possible misinterpretation was to be
kept comparable with other experimental errors, then 5% band-
widths were to be recommended". Accordingly, all test data from
the absorber facility are processed automatically at 50-Hz
216 D. L. MARTLEW
RELATIVE FREQUENCY
0-6 0-8 1-0 12 1-4 1-6
or
dB
-20 L APPARENT ATTENUATIONS

D FROM25/,BANDWIDTHS
X FROM 5*/. BANDWIDTHS
ÂCTUAL ATTENUATION
-10
dB
-20
Fig. 12 Effect of bandwidth

on apparent attenuation
constant bandwidth through a Hewlett-Packard digital Fourier

analyzer.
Development of a Reverberation
Chamber Noise Source
Although the in-duct noise generator was used regularly

during the first two years of operation of the absorber
facility and gave reliable results, two shortcomings became
apparent. The first of these was simply that the generator
was a large and cumbersome piece of equipment which was moved
from one end of the test ducting to the other when changes in
the direction of sound propagation were required. The second
and potentially scientifically serious shortcoming was that
very little acoustic power was produced below the fundamental
frequency of 1 kHz of the 2-in. Hartmann generators. With
requirements appearing for testing forms of duct treatment
having peak attenuation at around or even below 1 kHz, the
possibility was that the worsening signal-to-flow noise ratio
at the low frequencies would impose an unacceptable restriction
on the range of the facility. Also, it had been determined
OASPL-dB
160
RATIO
Fig. 13 Output of a four-inch

Hartmann generator
that the reverberation chamber losses were actually much less

than had been assumed during the design of the facility,
consequently, an investigation was commenced, in parallel with
the main test program, into the sound levels that could be
produced by a 4-in.-diam Hartmann generator situated within
one of the reverberation chambers. Figure 13 shows the
overall sound pressure level of such a unit fitted with a
needle but without air collection. A range of pressure ratios
and cup depths was covered, and the results were consistent
with the earlier general conclusions given in the first part
of the paper. A clear maximum output is obtained at 2.8/1
pressure ratio, with a cup depth of 4 in., giving a fundamental
frequency of 500 Hz. The other cup depths tuned the generator
to 400 and 600 Hz, and the resultant overall sound pressure
levels were reduced by only 1 and 2 dB,respectively. It was
thus indicated that, with the single 4-in. unit producing up to
161 dB OASPL in the reverberation chamber, corresponding to
about 158 dB in the test duct, a group of three such generators
could ensure the requisite 160 dB in the duct.
It was considered preferable, while setting up the system

of multiple 4-in. generators to replace the in-duct device, to
tune each of the three units to different fundamental
frequencies so as to produce a spectrum in which the tones were
distributed more uniformly. This would obviously reduce the
significance of the effects of flow on spectrum shape, and of
bandwidth of analysis, which were discussed in the previous
section in relation to the in-duct generator. An additional
argument for a more uniform distribution of tones may be
advanced as follows. In the parametric type of experimental
study which has formed a large part of the test program for the
absorber facility, the OASPL at entry to the test duct is very
often treated as an independent variable as a matter of
218 D. L. MARTLEW
, SPL-dB
PRESSURE RATIO 2-8:1
BANDWIDTH 10 Hz
a. SINGLE 4" GENERATOR

FREQUENCY - KHz
Fig. 14 Mixed tuning of

multiple Hartmann generators
practical convenience. From a more fundamental viewpoint, some

at present uncertain function of acoustic velocity through the
facing of the lining and of the flow velocity near the surface
would probably be more correct and there is consequently a
possible risk of misinterpretation in comparing, for example,
the deduced resistance of two similar linings at the same
OASPL where one is driven at its tuned frequency by an
intense discrete frequency while the other is off tune. It is
not thought that such misinterpretation has occurred to a
significant degree in results obtained with the tonal spectrum
of the in-duct noise generator, but the risk must certainly be
reduced by the change to a more uniform spectrum.
Figure 14 compares the 10-Hz bandwidth spectrum of a

single 4-in. generator with that of a group of three set to
give fundamentals at about 400, 500, and 600 Hz. An inter-
esting feature is the way that interaction between the
individual units has produced large numbers of sum and
difference frequencies. The sum components serve to make the
spectrum much more evenly filled above the fundamental
frequencies; however, the difference frequencies appear only
at low levels and do not fill out the spectrum below 400 Hz.
The combination produced 164-dB OASPL in the reverberation
chamber, which indicated a loss of around 1 dB compared with
the sum of the powers of the generators operated singly.
Figure 15 shows the arrangement as in current use. An
existing 2-in. generator has been added to the group in order
to fill partially the gap in the spectrum around 700 Hz. The
50-Hz bandwidth spectra of the sound produced by the group in
the test duct are presented in Fig. 16 for three pressure
ratios. Comparison with the spectra for the in-duct source
IFig. 15 Noise generators in reverberation chamber

160 .SPL-dB
BANDWIDTH 50 Hz
PRESSURE RATIO 2-8:1

5PL160dB
120
100
4 6 8 10
FREQUENCY - KHz
Fig, 16 In-duct spectra with

reverberation chamber source
shown in Fig. 9 shows how the distribution of sound energy has

been extended to lower frequencies and spread more uniformly.
There is still a gap at 700 Hz, particularly at low pressure
ratios, but this probably could be filled by replacing the 2-
in. unit by a specially designed generator.
A number of comparisons were made between the attenuation

of several lining configurations determined with the use of the
in-duct source and with the reverberation chamber source.
Very good agreement was found in nearly every case, as in the
example illustrated in Fig. 17. This shows, for a typical
lining configuration and for a duct Mach number of 0.3 and an
OASPL of 150 dB, the attenuation measured with the in-duct
source and that with a 2*in. Hartmann generator in the
220 D. L. MARTLEW
ATTENUATION - d B
30 50 dB OA5PL
MACH No. +0-3
89/. PERFORATE FACING
20 - 30 r
1-5" CELL DEPTH
10- 20 - 30
0 L 10 - 20 - 30
0L 10 - 20
0L 10
1 2 3 4 5 6 7 8 10
FREQUENCY - KHz
@ IN- DUCT NOISE GENERATOR SPECTRUM AS IN FIG. 9
(B) ONE 2" GENERATOR IN REV. CHAMBER. SPECTRUM SIMILAR TO
S ONE 4" GENERATOR IN REV. CHAMBER. SPECTRUM SIMILAR

THREE A" GENERATORS IN REV. CHAMBER. SPECTRUM SIMILAR
TO FIG.14a
TO FIG.Kb
Fig. 17 Attenuations with various noise sources
upstream reverberation chamber, having a spectrum very similar

to the in-duct source; agreement is very close at all
frequencies. Changing to a 4 in. generator in the reverber-
ation chamber and lowering the fundamental frequency from 1 kHz
to 500 Hz again results in close agreement, but with a slight
fall in the peak attenuation. A similar result is obtained
with the broader spectrum of the multiple noise generator.
The agreement obtained in these comparisons has done much to
increase confidence in the general validity of measurements
made in the absorber facility.
The indication is that the results are not sensitive to

the nature and position of the noise source and therefore
truly reflect the behavior of the duct itself. Furthermore,
the change in the source is sufficient to indicate that it is
unlikely that in either case certain duct modes were being
excited preferentially and that the assumption of equal energy
in all possible modes is therefore reasonable. This has
important implications when deducing the acoustic properties
of a lining configuration from the measured duct attenuation.
The process amounts to finding the boundary conditions
representing the wall lining which, when applied to the
solution of the duct wave equations, give the best agreement
with measurement. A separate solution exists for each of the
possible modes of propagation, and the computation of duct

attenuation requires information on the distribution of energy
between the possible modes of propagation. In the absence of
direct data,it is usually assumed that all of the modes are
excited with equal energy.
Conclusions
The experience recounted in the paper has shown that the

Hartmann generator may be used as a source of high-intensity
sound for the accurate measurement of the performance of
absorptive treatment in flow duct rigs. It is simple, robust
and inexpensive, although pressure air supplies must, of

course, be available. The performance and optimum config-
uration of single generators has been explored and the
beneficial effect of an axial needle exploited. It has been
confirmed that it is practical to operate Hartmann generators
in groups in constricted spaces and to collect their jets to
prevent aerodynamic interference with flows in the vicinity.
These principles have been applied successfully to the
provisioning of an "in-duct" noise source for a large flow
duct facility.
Advantages and disadvantages follow from the tonal nature

of the spectrum of the Hartmann generator, and certain
precautions in measurement and analysis have been suggested to
ensure the correct interpretation of flow-duct measurements.
An alternative source, also using multiple Hartmann generators
but having a more uniform distribution of acoustic power in
the spectrum, has been described.
References
^Flintoff, J. L., "The Ansty Noise Facility - Its Design,

Instrumentation and Future Commitments," The Aeronautical
Journal, Vol. 75, No. 726, June 1971, pp. 397-406.
^Hartmann, J. and Trolle, B., "A New Acoustic Generator: The

Air Jet Generator," Journal of Scientific Instruments,
Vol. 4, No. 4, January 1927, pp. 101-111.
o
J
Brun, E. and Boucher, R. M. C., "Research on the Acoustic
Air-Jet Generator: A New Development," Journal of the
Acoustical Society of America, Vol. 29, No. 5, May 1957,
pp. 573-583.
222 D. L. MARTLEW
Savory, L. E., "Experiments with the Hartmann Acoustic

Generator," Engineering, Vol. 170, August 4, 1950, pp. 99-100;
also Vol. 170, August 11, 1950, pp. 136-138.
5
Bracher, E., Maresca, C., and Bourmay, M. H., "Fluid Dynamics
of the Resonance Tube," Journal of Fluid Mechanics, Vol. 43,
Pt. 2, August 1970, pp. 369-384.
OUTDOOR JET NOISE FACILITY: A UNIQUE APPROACH
R. A. Kantola*
General Electric Company, Schenectady, NY
Abstract
Outdoor acoustic facilities are used widely for jet noise
experiments because of their relatively low cost when compared
to anechoic chambers. However, many problems do exist with
outdoor facilities. It is the intent of this paper to de-
scribe a new facility at the General Electric Corporate Re-
search and Development (CRD) Center which minimizes these draw-
backs. The unique combination of capabilities that this out-
door facility offers includes hemispherical microphone cover-
age, permanently installed microphones, acoustically treated
ground plane, and real-time data processing. This facility is
intended primarily for high-temperature jet noise research and,
as such, has a silenced burner capable of operation to 2000R.
These features result in a facility with minimal ground reflec-
tions and a short turnaround time so that testing in short pe-
riods of favorable weather is possible. Most important, these
features do not compromise the capability of the facility to
obtain high quality jet noise measurements on both cold and
heated jets.

Conference, Hampton, Va., March 24-26, 1975. This study was
carried out in part under Department of Transportation Contract
DOT-OS-30034. The author is indebted to Howard W. Avery, who
carried out the mechanical design of this facility, and to John
R. Trudeau, Richard E. Warren, and Ivan H. Edelfelt, who car-
ried out the software programming of the computer. J. R. Tru-
deau, in particular, is to be thanked for his willingness to
suggest and pursue new and innovative software techniques, such
as the real-time isometric plots of the acoustic spectra.
^Project Engineer, Research and Development Center.
223
224 R. A. KANTOLA
Introduction
To define the capabilities of this facility, calibration

data of the acoustic arena are presented as to the free-field
nature of the radiation patterns, effect of the ground plane
and other obstructions, and contamination by combustion and
pipe-borne noise. Data obtained using this facility, for
round convergent jets, are compared with information in the
literature. To provide an example of the use of this facility,
axial noise source location measurements of cold round subsonic
jets using the "hole-in-the-wall" technique also are described.
The "hole-in-the-wall" experiments determine axial jet noise
source distributions by the use of a translating absorbent

chamber and calibrated irises to separate the noise upstream of
the iris (which is absorbed in the chamber) and downstream of
the iris which is measured by the microphone array.
Nomenclature
R = microphone measurement radius

H = height of the jet axis above ground
X = axial distance from jet exit
0 = angle between individual microphones and the jet axis
fy = azimuthal angle between microphone array and the ground
plane
D = jet diameter
dg = aperture diameter
SPL = sound pressure level in one-third-octave bands, re
0.0002 ybar
OASPL = overall average of SPL -.3
PWL = sound power level in one-third-octave bands, re 10" w
OAPWL = overall average of PWL
pj = jet density
PO = ambient density
Aj = jet exit area
Vj = jet exit velocity
Tj = jet exit total temperature
f = frequency
a) = density exponent for PWL
Facility Description
Acoustic suppression between combinations of elemental jet

flows is quite small on a total power basis and requires de-
tailed azimuthal far-field measurements to allow the investiga-
OUTDOOR JET NOISE FACILITY 225
Fig. 1 CR&D hot jet noise facility.

tor to determine the relative importance between different pro-
posed suppression mechanisms. In this facility, a hemispheri-
cal ly swept array of microphones is provided to survey the
far-field directivity patterns of nonaxisymmetric nozzles or
suppressor configurations as shown in Fig. 1. Twelve % in. B&K
model 4133 microphones are attached to a traversing boom that
pivots about the jet axis. These microphones are positioned
every 10, starting at 6 = 20 to the jet axis and ending at
6 = 130. To avoid an obstruction in the jet plume, a large
hoop is used to provide a centerless pivot on the downstream
end of the microphone boom. The boom can be moved to any azi-
muthal angle by the two overhead cables. Since the paths tra-
versed by the microphones are circular arcs centered on the
jet axis, any deviation of the radiation patterns from axisym-
metry can be detected easily.
Outdoor acoustic facilities are extremely dependent on
weather conditions, and, in areas where the weather is very
changeable, as in the Northeastern United States, the ability
to respond quickly to favorable weather conditions is crucial
to the utilization rate of the facility. To avoid long startup
and shutdown times, a hermetically sealed microphone holder
was designed to allow permanent installation. An additional
benefit of this approach is that the electronic noise floor is
measured easily when the microphones are covered. An acousti-
226 R. A. KANTOLA
PERFORATED
HASTELLOY X
LINER
INLET AIR
Fig. 2 Burner-muffler construction.
ANGLE TO
THE JET
AXIS g
100 Hz. 80kHz

j OCTAVE -CENTER FREQUENCY
Vj =1006 fps , Tj =I903R, d = 2 IN.

Fig. 3 Real time isometric display of the acoustic spectra,
cally treated surface is used to minimize the ground reflection

effects. By using large sheets of acoustical foam, a reason-
able reduction of the ground reflection problem can be obtained
with minimal time required to lay down and take up the cover-
ings. To allow testing during the winter months, the 30- x
28-ft. concrete pad is heated electrically to remove ice and
snow.
To provide the heated air for the high-temperature tests,
two heaters are used. A large natural-gas-fired heat exchanger
pre-heats the air to about 400F, and this warm air is fed into
the burner end of the combustor muffler through a 4-in. pipe,
as can be seen on Fig. 1. Two small JP4 combustors are used
to provide the remainder of the heat addition. To prevent com-
bustion noise from contaminating the jet noise downstream of
the burners, acoustically treated baffles are used to prevent

a line-of-sight path. The wall of the plenum is lined with 2
in. of Kaowool and faced with a 1/8-in.-thick perforated sheet
(45% porosity) of Hastelloy X. A layout of this muffled burner
is shown on Fig. 2. Jet exit temperatures of 1990R have been
obtained with no thermal damage.
Data acquisition is controlled by a HP 2100 series mini-
computer that obtains the acoustic signals from a GR 1921 real-
time one-third-octave band analyzer and samples the tempera-
tures and pressure signals. By the use of a scanning multi-
channel amplifier, GR 1566, each microphone signal is analyzed
sequentially, and the signal level of each one-third-octave
band (100 Hz to 80 Hz) is stored on magnetic tape. The effec-

tive frequency range depends on the microphones used also. For
operational monitoring, a three-dimensional plot of the one-
third-octave band analysis of the sound pressure level, SPL,
of each microphone is displayed on an oscilloscope as the mi-
crophone array is sampled. A typical oscilloscope display is
shown in Fig. 3. For backup and when longer averaging times
are necessary, the acoustic signals can be recorded simultan-
eously on a Sangamo Sabre IV tape recorder. After all of the
signals have been accumulated, the computer corrects the data
for nonuniform response of the microphones and can correct for
any known non-free-field effects of the arena. Using these
corrected values of the sound pressure level, the computer
then calculates the overall average sound pressure levels for
each microphone, the one-third-octave band acoustic power
levels, and the overall acoustic power level. The raw and cal-
culated data are then stored on magnetic tape.
While the computer is processing the acoustic data, simul-
taneous measurement and calculation of all pertinent parameters
for determination of the nozzle exit conditions and ambient
conditions also are carried out and recorded on magnetic tape.
As all of the pertinent data exist on one magnetic tape, the
acoustic information is normalized readily by the computer im-
mediately following the test.
Acoustic Purity and Calibration Tests
To assess the non-free-field effects of the arena, several
varieties of tests were conducted. These tests used two types
of loudspeakers, a direct radiation type for low frequencies
and a horn-coupled driver for the midrange of frequencies.
Calibration jet noise tests also were conducted to compare with
previously reported results. The following sections will dis-
cuss these tests and their implications.
228 R. A. KANTOLA
Acoustic Purity-Speaker Tests

The principal aim of these tests was to determine the ef-
fects of the ground plane and support structure on the free-
field quality of the data. In particular, measurements were
made to determine the spherical divergence, ground reflection
patterns, and the reflections from the boom and support struc-
ture.
Spherical Divergence. To determine whether or not the
arena exhibited spherical spreading (that is, if the sound
pressure level varied as an inverse function of the source to
microphone distance), an enclosed direct radiation loudspeaker
with an effective output range from 60 to 5000 Hz was used.

The speaker was pointed directly at a far-field microphone. In
the horizontal position, both the speaker and the microphone
were positioned 6 ft. above the concrete pad. An 18- x 18-ft.
area of the concrete pad was covered with two 4-in. layers of
acoustical foam, the top layer a 90-ppi (pores per inch) rec-
ticulated foam and the bottom layer a 90-ppi non-recticulated
foam. For these tests, the far-field microphone was mounted
on a separate tripod, and the large boom was stowed in the op-
posite quadrant. A control microphone (12 in. from the speaker
exit plane and on the speaker centerline) was used to monitor
the near-field output of the speaker. A random noise generator
with a pink-noise filter was used to excite the loudspeaker.
The measurements were carried out at an angle of 90 to
the jet axis and at three different radii, R = 6, 8, and 10 ft.
To determine the effect of the boom angle, both horizontal and
vertical measurement planes were used. Fig. 4 illustrates the
results. The arena exhibited spherical divergence between the
8- and 10-ft. radii, within the 1 dB experimental repeata-
bility of this particular speaker.
Ground Reflection Corrections. One of the most difficult
problems in outdoor acoustic testing is the proper assessment
of the influences of the ground plane on the acoustic data.
The most consistent surface is a smooth concrete pad. However,
the effects, although consistent with a particular acoustic
source, are very large. The effect is not only dependent on
the acoustic arena and the relationship of source to receiver
geometry, but also on the nature and physical extent of the
source. Without prior knowledge of the source characteristics
(randomness in time and location), the proper ground correction
will be very difficult to predict, particularly for the middle
frequencies. The approach taken here is to reduce the ground
90 f=2500Hz
6 =90
frO
=90
SPL
(dB) MICROPHONE
80
* -MICROPHONE
SPEAKE
6' ÂCOUSTICAL PAD
(8" THICK)
70
SPLa20L06|0(l/R)
_L _L _L
4 6 8 10 12
MICROPHONE RADIUS (FEET), R
Fig. 4 Spherical divergence tests (loudspeaker acoustic source),
J
4
o o
3 n
0
A n D
2 >-
1 _ -, 5hJo ^j^Ljcr 0
D
nnn0
0- o
AdB D AAA2ft AA D
-1 ft A
DA n
-2 - n o
O
-3 A
o HARD PAD 1
D 4" FOAM LINING >
-4 _ TESTS
---POINT SOURCE PREDICTION * * . ^Vl'g'^
-5
WITH H A R D PAD '
D
-6
-7
-8 1
n
IO Z I0 5 10" 10
1/3 O C T A V E FREQUENCY ( H z )
Fig. 5 Ground reflection corrections for different ground
treatments.
230 R. A. KANTOLA
reflections to less than 2 dB in the frequency range above

400 Hz. To do this, the concrete pad was covered with 4- x
6-ft. acoustical foam sheets. Foam sheets, rather than rigid
sheets with wedges, were used to minimize the time required to
install and remove the covering.
To test the effects of this acoustical covering, a direct
radiation loudspeaker was used. The approach adopted here is
to calibrate the speaker during the test by pointing the
speaker vertically and placing the microphone at the desired
radius directly above the speaker. A near-field microphone
(1 ft. from the speaker) was used to insure that the speaker
characteristics did not change during the test. In this man-

ner, the ground reflection corrections for the hard pad itself
and with two different thicknesses of foam were measured.
In Fig. 5, the effects of the surface condition on the
ground reflections can be seen. For the speaker-microphone
distance R of 10 ft. and with both the speaker and microphone
6 ft. above the hard pad, the first cancellation (null) occurs
near 100 Hz and the first reinforcement (peak) near 200 Hz.
When the foam sheets are used, two effects occur simultaneous-
ly; first, the impedance of the surface is altered and, in-
stead of a simple reflection, the surface absorbes some of the
incident sound and causes phase shift in the reflected sound;
and secondly, the geometric relationships are altered. The
first effect results in a smoothing effect on the first two
peaks and nulls, but at the middle frequencies the locations of
the peaks and nulls also change. Acceptable performance is
achieved with the 8-in.-thick foam. The ground reflection cor-
rection is reduced to less than 2 dB above a frequency of
200 Hz.
Effects of Microphone Boom and Support Structure. In or-
der to determine if the microphone boom and the saddle (guide
plates that ride on the hoop) had any effects on the data, a
series of special tests was conducted. Because of the dis-
tances and sizes of the obstructions, the main effect antici-
pated was sound scattering in a range of frequency higher than
the ground reflections. A high frequency horn driver and ex-
ponential horn were used to determine the effects of the boom
and support structure on the SPL spectra. The details of
these particular tests can be found in an earlier paper^.
As a result of these tests, the maximum acceptable micro-
phone radius was found to be 9 ft. This microphone radius was
used for the studies reported throughout the remainder of this
paper.
<<, 0 = 30*;
JET NOISE
Vj = l009fp$
530R
20
1/3 OCTAVE BAND CENTER FREQUENCY (Hz)

Fig. 6 Facility noise comparisons.
Facil ity Noise

The capability of this facility to measure low-level acous-
tic signals is determined by the level of noise contaminants.
These contaminants are the electrical noise floor, the ambient
acoustic background noise, and any pipe-borne acoustic noise
associated with the pumping machinery and throttling valves.
Fig. 6 shows these contaminants in relation to typical jet
noise spectra at a jet velocity of about 1000 fps. For the
electrical background and the pipe-borne or flow noise, the
maximum level and minimum levels for each one-third-octave
band that occurs in any of the 12 microphones are shown. This
is the most critical way of presenting the data. As can be
seen, the electrical noise dominates the background above 8000
Hz.
232 R. A. KANTOLA
Flow noise as shown here is determined by removing the 2-

in. nozzle termination from the 12-in. plenum and passing the
same mass flow through the system as with the jet nozzle in
place. Removing the nozzle changes the termination impedance
of the pipe line and also causes more upstream throttling, and,
as a result, more pipe-borne noise escapes than would with the
2-in. nozzle in place. This overestimate of the flow noise
content of the jet noise measurement is confirmed by the lack
of contamination of the low frequency end of the jet noise
spectrum. By scaling the jet noise by the eighth power of
velocity and the flow noise by the sixth, the facility should
be capable of measuring cold jet noise down to about 400 fps
with a 2-in. nozzle,for frequencies above 400 Hz.

Jet Noise Calibrations of the Facility
Round convergent jets were used to provide comparison data
on both hot and cold jets and assess burner noise contamina-
tion. Two sizes of jets were used: a 1-in. exit diameter jet,
and a 2-in. jet. The tests covered a jet temperature range from
ambient to 1990R and a jet exit velocity range from 600 to
over 1900 fps. The facility is capable of slightly higher tem-
peratures, as after about 1 hr. of running at the highest tem-
perature very little thermal damage was seen. These tests were
conducted for the most part with the 8-in.-thick acoustic lin-
ing on the concrete pad, although some limited testing was
done with a bare pad. The majority of these tests were con-
ducted with the microphone boom in the horizontal position, ^
= 0. Other boom angles also were tested to see if there were
any significant differences in the results. For the most part,
the jet flow was subsonic, although a few slightly underexpanded
jets were tested.
Overall Sound Power Level. The overall sound power is cal-

culated by assuming a non-reflective ground plane and is based
on the signal content in the one-third-octave frequency bands
from 100 Hz to 80 k'Hz. For the microphones used here, however,
the frequency range is limited to 40 kHz. No ground reflection
correction is applied to the individual microphone spectra
that are used in the remainder of this paper.
When the data are normalized by the methods recommended

by Hoch2 et al, the corrected overall power is seen to be in
excellent agreement with Hoch's generalized curve, as seen in
Fig. 7 Very close agreement also is seen between the 1- and
2-in. nozzle data.
COLD, D = 2", H A R D PAD, W I T H W I N D S C R E E N S

180 Q BURNER ONLY ^ D=2"
< P R E H E A T ONLY I LINED PAD
y BURNER + PREHEAT f W I T H
A COLD J WINDSCREENS
O COLD, D = 2", LINED PAD
D COLD, D = l "
NO CORRECTION FOR GROUND
REFLECTION
160 /
HOCH'S (REF. 2)
GENERALIZED
CURVE
COAPWL
(dB)
140
COAPWL = OAPWL-IOLOG, O [AJ(/>J// ) O) W ]
120 _L_
-0.2 -O.I 0 O.I 0.2
LOG,0(Vj/A0)
Fig. 7 Corrected overall acoustic power.
Of particular importance is that the higher-temperature,

low-velocity runs are not contaminated by burner noise. To
verify this, three ways of heating the flow are used. These
included running the preheater only, running the burner only,
and operating with both the preheater and the burner. In all
cases, the scatter of the data due to manner of heating the
air is less than 1 dB, except where the jet flow is under-
expanded. This underexpanded jet flow data point is identified
in Fig. 7 by the small tick mark attached to the symbol.
One-Third Octave Power and Pressure Spectra. To achieve

valid comparisons with other experimental jet noise data, the
spectral data taken on this facility are normalized to remove
the effects of variations in jet nozzle flow area and the var-
iation in the ratio of the jet density to the ambient density.
At this point, it should be stated that the correction for the
variation in density ratio is applied uniformly across the
234 R. A. KANTOLA
170 O OLSEN 4.0" DIA. COLD JET (REF.4 )

O AHUJA 2.84" DIA. COLD JET (REF. 3)
a NGTE 1.78" DIA. COLD JET (REF. 5 )
160
o CR a D 2" DIA. COLD JET
150
Vj-IOOO FT/SEC
140
130
O
o
120
NO -^00-0 *
V j ~ 600 FT/SEC
100
90
-1.2 -0.8 -0.4 0.4 0.8
Fig. 8 Normalized power spectra, cold jets.
whole spectrum. Hoch and his co-workers^ have shown that at

the low and medium velocities heating of the jet not only al-
ters the overall levels but shifts the spectrum to the lower
frequencies. Results confirming the experiments of Hoch et
al^ will be discussed later. For these tests, the density ratio
exponent GO is that as was determined by Hoch^ for normalizing
power levels. Fig. 8 illustrates the results3 obtained on the
CRD facility as compared with other results ^, and 5 fr0m
both anechoic and outdoor jet noise test facilities. The data
obtained at the CRD facility are well within the scatter
exhibited by the other facilities. The variation in the high-
frequency response in the microphones and effect of wind-
screens (if used) have been removed from the CRD data presented
here.
In the low-frequency portion of these data, ground reflec-
tion perturbations are noticeable, although not as large as
determined by the loudspeaker tests. Because these perturba-
tions are in the low end of the spectrum where the one-third-
octave bands are narrow, the location of the peaks and nulls
move from one frequency band to an adjacent band with varia-
tions of nozzle size, jet velocity, and temperature, and also
with the angle to the jet axis. For this facility, the acous-
tic spectra are not corrected for these ground reflections,
Cff^b Vj=l9l9fps
170 QCP ^ V,S94.R
o o
O O fVs
/% Ô
7" I6
o 0
*~o
O
^ n
u ^J*= I5l7fpt Q
w
o
2* 150 0
0
0
/o /%
^ 140
Q_ 0 Tj=l903 p R q^
LU O ^*
0 0^ D = 2"
8 0 Q
^>)0^ T a =486R
ro
LINED PAD
0 ^***b
Vj = 744fps %^ = 0 P
120 0 ^ L= 1990 R
no 1 0 1 1 1 1
-2
Fig. 9 Normalized power spectra , Tj^l900R.
because of the complexity of the corrections and the small mag-

nitude 1 dB (above 400 Hz) of the perturbations.
The normalized power spectra of heated jets are shown in
Fig. 9 for a range of jet velocities. The slight shift of the
spectral peak to lower Strouhal numbers as the velocity is
lowered agrees with the hot jet data of Hoch^. As mentioned
earlier, heating a jet at a constant velocity will cause a
change in the noise spectrum. Hoch2 found that this phenomenon
is more pronounced at very high jet velocities and results in
a lowering of levels for all frequencies. This trend weakens
for the low frequencies as the velocity is lowered, and at low
velocities the low-frequency noise increases with increasing
jet temperature. To clarify the spectral shift, the one-third-
octave spectrum is subtracted from the overall power level and
shown on Fig. 10
236 R. A. KANTOLA
-10
-20
j = I 5 l 7 f p s , Tj = l85lR
-30 D V j = I489fp$, T j = l 5 0 9 R
o D = 2"
o Ta=486R
-40 LINED PAD
-50 -
-60 1 I 1 1 J
- 2 - 1 0 1 2
LOG,0(fD/Vj)
Fig. 10 Effect of jet heating at VjMBOO fps.
Directivity Results. The directivity patterns of the over-

all acoustic pressure obtained from this facility are compared
to Lush's^ anechoic chamber data in Fig. 11 4 The overall pat-
terns agree well, except at the shallowest angles. Both the
1- and 2-in. jets are displayed with quite close agreement,
except at the shallow angles. This slight disagreement could
be due to nonspherical spreading at this angle because of the
extended nature of a jet noise source.
The azimuthal measurement capabilities of this facility

were used to determine the axial symmetry of the sound field
by conducting a series of tests with a 2-in. diam. round noz-
zle. The PWL spectra was found! to be within h dB for boom
azimuthal positions varying from vertical to horizontal, ex-
cept at frequencies less than 200 Hz where it is within 1 dB.
o LUSH, D = l", R=IO' (REF.6)
I U 1 .X
NOTE:CR 8 0 DATA CORRECTED

TO 1" DIAMETER
r|OI4f s
A A A /^ J P AND 10' ARC
o o 8 D / ^ V, = l009fps
100 - / n^Â A
V J =980fps / o
OASPL
(dB) DDo * 00 on o D
90
o o
D %\% 0
80 8
^Vj6H4o0fo
| Vj=605fps
0 Q
Vj=59lfps
70 i i i i i I I 1 1 1 1 1 1
30 60 90 120
ANGLE TO JET AXIS, 9 (DEGREES)
Fig. 11 Directivity patterns for various velocities.
The variation in the low-frequency end is due to the ground

reflection variation with the azimuthal angle, ty.
"Hole-in-the-Wall" Tests
Background
As an illustration of the utility of this facility, some
brief experiments on 11the location of jet noise sources, using
the "hole-in-the-wall or "wall-isolation" technique, are de-
scribed. The "hole-in-the-wall" experiment is a means of de-
termining the axial distribution of noise sources in a jet
plume. There are several versions of this method, but the com-
mon feature is that the jet is caused to flow through an aper-
ture that is made as small as possible without affecting the
jet or producing an additional noise through buffeting or
toroidal edge-tones. The idea is to separate the jet noise
into two parts, upstream of the aperture and downstream of the
aperture.
Of course, the aperture cannot separate completely the
noise, as there is noise leakage through the aperture in both
directions, with preference for contributions from upstream
sources to appear downstream of the "hole-in-the-wall." Potter
and Jones?, in the original version of this method, used a re-
verberant chamber and a movable jet nozzle. They established
238 R. A. KANTOLA
JET
H APERTURE
CHAMBER A CHAMBER B
TOTAL
Fig. 12 Acoustic leakage and non-closure of source distri-

bution.
closure on the noise sources first by determining the distri-

bution by upstream measurements and then by reversing the ex-
periment and determining the distribution by downstream mea-
surements. In Fig. 12, a schematic representation of a "hole-
in-the-wall" experiment is shown using two back-to-back rever-
berant or anechoic chambers (A and B) with an aperture between
them. Now,in this experiment, the acoustic power in both A and
B will be measured simultaneously. The sum of the acoustic
power must be a constant Py if there is no interaction noise.
Suppose that the aperture could separate the noise completely;
then as the jet was withdrawn into chamber A, the power mea-
sured in chamber B, PR, must decrease. However, because of
leakage, the measured value of Pg will be larger than the true
value. The acoustic power measured in A will likewise be de-
creased by an identical amount so that "reversing" the experi-
ment as done by Potter and Jones? does not provide closure,
as it reproduces the same cumulative power distribution in
both cases.
Fig. 13 "Hole-in-the-wall" experimental apparatus.
In a more recent study, McGregor and Simcox^ used a mov-

able jet and a fixed absorbent chamber to absorb the upstream
noise and measured the downstream noise with an outdoor micro-
phone array. This method is more attractive in that it does
not require that the reverberant characteristics of the chamber
be calibrated, and the jet itself is not as subject to possi-
ble excitation due to high acoustic level inside the chamber.
In this method, it is necessary, however, to calibrate the
aperture diameter vs. axial distance to maintain closure on
the total jet power.
Experiments
In this study, a slightly different version of the Mc-
Gregor and Simcox^ method was used with a translating absor-
bent chamber and a fixed jet nozzle and microphone array.
Fig. 13 shows the absorbent chamber mounted on a cart that can
traverse axially on a set of rails. The chamber is constructed
of ^ in. transite and lined with 2 in. of high-temperature
acoustic material, Koawool. Two industrial mufflers are used
to silence the entrainment air inlets. A series of orifice
plates is used to provide the aperture.
To determine the minimum aperture size for a given axial
station, a series of tests without the absorbent chamber was
conducted. The total acoustic power was measured with a ser-
ies of orifice plates placed at various axial positions in
240 R. A. KANTOLA
o 2 JE APERTURE
2" JE C H A M B E R
l"j CHAMBER
Vj - 990 999 fps
Tj = 512'
BARE P,
Fig. 14 Optimum aperture versus axial distance.
120 i-
INTERACTION
NOISE
110
PWL
(dB)
O X / D =13, d 0 / D = 8
D X/D =14, d 0 / D = 8
100 D=l"
Vj =992 f p s
Tj=5llR
B A R E PAD
ABSORBENT CHAMBER
90 I I
10
1/3 OCT. B A N D - F R E Q U E N C Y (Hz)
Fig. 15 Effect of axial location of aperture.

I03 ICT 10s

1/3 OCTAVE BAND C E N T E R F R E Q U E N C Y ( H z )
Fig. 16 Power Spectra as a function of aperture location,

1" jet.
the jet flow, and the minimum aperture sizes were determined.
Results from both the aperture-alone tests and with the absor-
bent chamber in place are shown in Fig. 14. The range of sizes
was very close to the relation (dQ-D)/x = 0.5, where D is jet
diameter, dg is aperture diameter, and x is axial distance from
jet exit. This corresponds to a truncated cone starting at the
jet lip and flaring out with a half-angle of 14.
From these measurements, it is found that it is virtually
impossible not to generate interaction noise, except when the
orifice is located very close to the nozzle exit. Where does
this noise appear in the measured far-field spectrum? When
far downstream, for example when x/D > 6, the interaction spec-
trum peaks are separated well enough to be identified. An exam-
ple of this is shown in Fig. 15, where data obtained for two
axial locations of the same orifice plate are shown. The in-
teraction noise is seen to dominate in the low frequencies.
The decrease in PWL at the high frequencies also can be seen
clearly as the axial distance of the orifice plate from the
nozzle exit is increased. When the orifice plate is closer
to the nozzle, the interaction peak frequency becomes closer
to the jet noise peak frequency, as would be expected from a
Strouhal number consideration. In these cases, the separation
of the jet noise and the interaction noise becomes more diffi-
242 R. A. KANTOLA
140 i
0=
Vj = 9 9 0 - 9 9 9 f p s
Tj = 5llR
120
B A R E PAD
ABSORBENT CHAMBER
IIOl
0 10 15 20 25
X/D
Fig. 17 Decrease of measured acoustic power with aperture
location.
cult. In Fig. 16, data from the mapping of a 1-in. convergent

jet illustrate this trend. The data of McGregor and Simcox8
for a slightly different case using a moving nozzle, a heated
sonic jet, and with a microphone array limited to the forward
arc only show a larger dropoff of PWL with axial distance and
better separation of the interaction noise and the jet noise.
The diameter of the aperture also would have an effect, but
this information is not available in the McGregor and SimcoxS
paper, so that an exact comparison cannot be made.
To calculate the distribution of the overall sound power,

OAPWL, from the measured PWL requires that the interaction
noise be eliminated. For x/D less than 6, the low-frequency
noise is assumed to be that of the undistributed case, x/D =
0. For x/D greater than 6, the low-frequency portions are
assumed to be of similar shape to the x/D = 0 case, but at a
-10 r-
-12
-14
-16
-18
Vj = 990-999 fps
Tj=5llR
-20 BARE PAD
ABSORBENT CHAMBER
-22
-24 I J
0 5 10 15 20 25
X/D
Fig. 18 Axial source distribution (% OAPWL/X/D).
lower level. Computing the OAPWL in this manner yields the

cumulative axial distribution shown in Fig. 17. A simple
numerical differencing scheme for the data shown in Fig. 17 is
used to determine the axial distribution-of overall sound shown
in Fig. 18. The location of the peak of the sound distribu-
tion between 6 - 8 diam. agrees well with McGregor and Simcox8
heated jet data, but the shape of the curve is somewhat dif-
ferent.
To define the location of a source of a given frequency,

a similar procedure is used as for the overall power. For
frequencies above 5 kHz, the method is free of the interaction
noise contamination. In Fig. 19, the peak contributions at
high frequencies are found to be close to the jet exit and
244 R. A. KANTOLA
Or
S -10
0.419
0=1"
= 990-9991 ps 0.838
-20
j = 5llR
BARE PAD
ABSORBENT CHAMBER
-30
10 15 20
X/D
Fig. 19 Axial source distribution at various Strouhal numbers,

1" jet.
generally agree with other "hole-in-the-wall" experiments and

other types of source location methods. The "hole-in-the-wall"
technique is a relatively simple technique and requires little
in the way of specialized equipment, for simple flow geome-
tries. The method is, however, better suited to measurements
of the high-frequency portion of the noise spectrum because of
the interaction noise contamination of the pure jet noise.
References
^Kantola, R. A., "Outdoor Jet Noise Facility, A Unique Ap-
proach," AIAA Paper 75-530, 1975.
2
Hoch, R. G., Duphochel, J. P., Cocking, B. J., and Bryce, W.
D., "Studies of the Influence of Density on Jet Noise," Jour-
nal of Sound and Vibration, Vol. 28, No. 4, June 1973, pp.
649-668.
3
Ahuja, K. K., "Correlation and Prediction of Jet Noise," Jour-
nal of Sound and Vibration, Vol. 29, No. 2, July 1973, pp. 155-
iw.
4
01sen, W. A., Gutierrex, 0. A., and Dorsch, R. G., "The Ef-
fect of Nozzle Inlet Shape, Lip Thickness, and Exit Shape and
Size on Subsonic Jet Noise," TMS-68-1182, 1973, NASA.
^Cocking, B. J., "The Effect of Temperature on Subsonic Jet

Noise," Rept. 331, 1974, National Gas Turbine Establishment.
, P. A., "Measurement of Subsonic Jet Noise and Compari-
son with Theory," Journal of Fluid Mechanics, Vol. 76, No. 3,
1971, pp. 477-500.
^Potter, R. C. and Jones, J. H. , irAn Experiment to Locate the
Acoustic Sources in a High-Speed Jet Exhaust System," 1967,
presentation to Acoustical Society of America.
^McGregor, G. R. and Simcox, C. D., "The Location of1 Acoustic
Sources in Jet Flows by Means of the 'Wall-Isolation Tech-

nique," AIAA Paper 73-1041, 1973.

FACTORS IN THE DESIGN AND PERFORMANCE OF
FREE-JET ACOUSTIC WIND TUNNELS
Y. Kadman* and R.E. Hayden*

Bolt Beranek and Newman Inc., Cambridge, Mass.
Abstract
Free-jet wind tunnels that have good acoustic character-

istics and low background noise are very versatile tools for
studying aerodynamic noise. The physical, acoustic, and aero-
dynamic characteristics of a new, subsonic wind tunnel,
recently completed, are described. The tunnel combines an un-
usually long jet and a 22,000-ft3 test chamber in a compact
package. Anechoic wall treatment and jet collector and dif-
fuser design are solved in a cost-effective manner. The
methods and results of the tunnel calibration tests, both
acoustic and aerodynamic, are given.
Introduction
In the design of machines or components with low noise

generation at the source, the understanding and characteriza-
tion of the underlying noise-generating mechanisms are of ut-
most importance. The definition of noise-generation mechanisms
and source-reduction techniques is currently an integral part
of the early design stage of aircraft components, fans, and
underwater vehicles.
To overcome current deficiencies in noise prediction

methods and the attendant flowfield analysis, noise radiation
and reduction must be tested under simulated operational condi-
tions. As a consequence, the acoustic wind tunnel has become a

Conference, Hampton, Va., March 2^-26, 1975-
^Senior Engineering Scientist.
247
248 Y. KADMAN AND R. E. HAYDEN
principal research tool for the definition of noise-generation

characteristics of a variety of propulsive devices, such as
propellers, rotors, fans, compressors, jets, and lifting sur-
faces and systems. Investigations of flow noise and flow-
surface interaction use the acoustic wind tunnel extensively.
The basic factors in the design of such facilities are

flow quality, physical layout, collectors, and reduction of
background noise. The Bolt Beranek and Newman Inc. (BBN) sub-
sonic acoustic wind tunnel was designed to be a flexible and
versatile research facility with many diverse applications.
This facility, an enlarged version of an earlier BBN free-jet
tunnel, was designed to accommodate a full spectrum of noise-

producing systems and components, and is used to illustrate the
design principles as presently understood.
Design Requirements and Constraints
Physical Constraints. Several approaches to the physical

layout are possible. Conventional approaches use bellmouth
collectors at l-to-3 jet diameters downstream. This makes the
circuit more efficient but increases background noise and
reduces the field of measurement. The BBN tunnel uses a very
long test chamber, with collector far downstream of the nozzle.
The tunnel location and size were dictated by usable space

and the availability of a 600-hp diesel engine, housed in an
existing engine test facility. Because of the anticipated
testing of a variety of large objects which might require
extensions to or modifications of the tunnel nozzle, a long
test chamber was desired. A 60-ft-long space was available
for the tunnel. Other dimensions were chosen to provide a
reasonable ratio of wall dimensions for operations in the re-
verberant mode. A long free jet was chosen to maximize the
flexibility of the facility and to avoid noise and flow-related
problems of collector/nozzle interaction.1'2
Flow Quality Requirements. As in conventional wind tun-
nels, low turbulence, uniform flow, and appropriate Mach and
Reynolds number ranges were required. A reasonable goal for
the turbulence level in the potential core of the free jet was
0.2%3:>If; below this level, the turbulence does not affect the
transition materially from laminar to turbulent flow on a flat
plate. The recommended velocity uniformity to within a 0.25%
across the jet5 was known to be an achievable goal from our
experience with the previous nozzle design.6
In principle, there is no limit to the test section speed.

For the BBN tunnel, we selected the top speed of the primary
FREE-JET ACOUSTIC WIND TUNNELS 249
jet to be 300 fps, which is the upper limit of the range of

takeoff and approach speeds for CTOL aircraft. A 7-ft2 nozzle,
with a maxim-urn velocity of 300 fps, was designed for those
tests that required higher speeds. Available power dictated a
choice of the l6-ft2 basic nozzle area if we were to achieve a
maximum flow velocity of 1^0 fps. This velocity was judged
sufficient for most of the projected investigations,
judged sufficient for most of the projected investigations.
Acoustic Requirements. Room acoustic quality, background

noise, and tunnel flow noise are the acoustic factors to be
considered. Most of the work performed in an acoustic wind

tunnel is concerned with far-field noise and directivity
patterns. Both the geometric and acoustic far-field conditions
must be met for meaningful test results. The desired cutoff
frequency for the anechoic test chamber was determined by the
far-field requirements. Under certain conditions, the square,
k- x ljft, baseline nozzle can accommodate propellers and fans
of up to an approximately 3-ft span, which requires a minimum
measurement distance of 9 ft from the tunnel centerline if one
is to be in the geometric far field.7
The acoustic far field7 of a dipole is established when

the measurement location distance r from the object is much
larger than A/6. Thus, the facility can be used with con-
fidence down to frequencies less than 100 Hz, because the 9-ft
measurement distance satisfies the acoustic far-field cri-
terion, i.e., r ~ 5(A/6).
Low self-noise is a fundamental requirement of any flow

facility in which acoustic measurements are desired. The most
obvious sources of tunnel self-noise are the prime mover and
the tunnel'drive fan. These sources are controlled using
generally accepted noise-control procedures. A more difficult
problem is dealing with the flow in the room itself, since the
character of that noise is similar in level and spectrum to
that which one desires to study in certain instances, such as
the measurement of airfoil and airframe flow noise. Prior1
experience indicated that the dominant source of flow noise is
the impingement of the free jet on the collector cowl and not
the free-jet noise itself. Since the design of the new tunnel
exhaust was to be similar to the existing BBN flow facility,6
model testing was not performed for the final collector design.
A comparison of noise from various facilities having different
collector placements follows.
WINDOW
MICROPHONE\ I!
-Ê
FAN: 140,000 cfm COMPRESSOR: 6000 cf

AT 1 psi HEAD ATlSpsi HEAD
Fig 1. BBN high-speed free-jet acoustic wind tunnel:

elevation.
Tunnel Circuit
General
Figure 1 is a schematic view of the wind tunnel. By means

of inlet and exhaust ports, the tunnel can "be operated either
in a closed- or open-loop configuration. Extremely cold
weather requires closed-loop operation, whereas hot weather or
tests involving hot flow or exhaust emissions necessitate open-
loop operation.
The Nozzle
The "baseline tunnel nozzle has a bellmouth intake leading

to a rectangular (ll- x 12-ft) inlet cross section that con-
tracts to a square (h- x U-ft) exit cross section to produce a
7.5 contraction ratio. At present, two more "add-on" nozzles
are available. The first is a constant-area square-to-round
transition, and the second is a square (h x 1+ ft)-to-rectangular
(28 x Ij-Q in.) transition. This latter nozzle gives a total
contraction ratio of 16.5. The maximum flow velocity of the
primary nozzle is 150 fps, and the maximum flow velocity of the
smaller rectangular nozzle is 295 fps.
The shape of the primary nozzle was determined according

to the method used by Maestrello8 and Hanson.9 They reported a
successful contraction design, which employed a ninth-degree

polynomial. Its coefficients were determined by specifying the
number of derivatives which must vanish on each side of the
nozzle, which make the rate of static pressure drop gradual,
assuring that there is no flow separation along the nozzle.
Flow straightening and turbulence reduction at the inlet

were achieved by a combination of four 2^--mesh stainless-steel
screens with very small wire diameter (0.0075 in.) to avoid
vortex shedding from the wires and one 8-in.-thick aluminum
honeycomb with 0.625-in. hexagonal cells. As can be seen in
Fig. 1, the nozzle is wholly within the test chamber. All of
the external surface of the nozzle was covered with the same
foam used in the acoustical treatment of the test chamber.
Test Chamber and Test Section
As mentioned previously, the wind tunnel was to be used

mainly for acoustic far-field and directivity measurements. In
addition, the test chamber was designed to be used as an
anechoic chamber without any flow. The internal dimensions of
the chamber are 23 x kk x 20 ft mean height (the room cross
section is not square to improve reverberant characteristics
when appropriate), with a total volume of 22,000 ft3. In this
area, fairly large objects can be tested, and the microphone
can be swept at a full circle around the test object.
In order to generate a high-quality free-field condition,

2-in.-thick open cell foam was applied to the lO-in.-deep
channels of the chamber walls that were formed by the basic
structure. In random incidence absorption tests, this foam had
an absorption coefficient of 0.96 at 500 Hz and 0.8 at 250 Hz.
The cell-like walls of the test chamber improved the foam per-
formance, however, and a lower cutoff frequency of 160 Hz was
achieved.
Tunnel Jet Collector
In the smaller BBN tunnel, it had been found that the con-
cept of a stagnation plate collector was successful for achiev-
ing low noise and minimum recirculation at the expense of
reduced tunnel aerodynamic efficiency. In this design, the jet
impinges on a rigid flat plate and is sucked out around the
stagnation plate perimeter, which avoids the noisy edges of a
bellmouth-type collector. The comparison with other concepts
follows .
This design also is compatible with situations where space

for a long, slowly diverging diffuser is lacking. It also
provides excellent sound attenuation from downstream noise
sources, because the ductwork has a number of acoustically
favorable bends. Although the efficiency of the stagnation
plate is lower than a regular diffuser, the long jet assures
that the velocities at the cowl are low and the losses are
relatively small, compared to those encountered in high-speed
flow. The stagnation plate is treated acoustically like the
rest of the test chamber.
The side and top passages of the cowl continue down to the
fan room. However, to eliminate any acoustic "short circuits"

from the fan, the bottom passage, which is closest to the fan,
splits to the sides and turns down after an 8ft extension.
All passages are treated with cavity-backed acoustic material,
i.e., either 2-in. foam or a combination of 1-in. fiberglass
board and fiberglass cloth plus perforated metal. The insides
of the cavities are covered with 1-in.-thick fiberglass mats.
The flow-turning locations, where the flow forces may destroy
the lining, are covered with 23% open perforated metal that is
0.07 in. thick.
Fan Plenum Chamber and Drive System
The four lined exhaust ducts discharge into the fan plenum
chamber. The double-inlet, double-width centrifugal fan is
centered in a cavity-backed, foam-lined room. The lining pre-
vents the buildup of reverberant energy near the fan. The fan
has no inlet boxes and rests on a concrete slab, which is iso-
lated from the rest of the structure. The fan is surrounded
with screens, which evenly distribute the flow being sucked
from the exhaust ducts.
Isolated from the fan itself, the fan exhaust box is lined
with foam or fiberglass. This lining is covered with screens
or perforated metal, at the locations where the flow velocities
are high. As in the rest of the tunnel, cavities back most of
the lining, which greatly enhances its sound absorption capa-
bility.
The fan is connected to the 600-hp diesel engine by a belt

drive. The tunnel speed is controlled by the diesel throttle.
If flow velocities lower than those provided at the engine idle
speed are needed, air is bled into the fan plenum chamber,
thereby reducing the suction from the wind tunnel.
E3BN HIGH-SPEED WIND TUNNEL

1rURBULENCE ON TUNNEL (j.
1 m FROM NOZZLE FACE
\?
)0 fps
r*
/x2J
^
/
-150
"^-,
V
x^S.
'^^
-70
~^>x J
X
X <>^ X^
\
-80
I 100-/
\- \
\
8 \
\
-90
\
\
- inn i i i i 11 i i 1L_ 11_
ONE-THIRD OCTAVE BAND CENTER FREQUENCY ( H z )
Fig. 2 Turbulence spectrum vs flow speed.
Return Passage and Nozzle Inlet Plenum
The treated fan exhaust duct, as it leads to the air

return passage, diverges to a 100-ft2 area, so that flow noise
and energy losses are minimal. The inlet plenum, upstream of
the nozzle, is treated acoustically to absorb any noise that
propagates upstream from the test chamber through the nozzle.
An acoustically absorptive bellmouth surrounding the nozzle
inlet assures clean flow into the nozzle.
Aerodynamic Performance
With the basic l6-ft 2 nozzle, the tunnel exceeded its

design velocity of ihO fps by about 10%. When the "add-on"
constant-area square-to-round transition was employed, the
maximum velocity increased to 160 fps. The reduced jet length-
to-diameter ratio probably accounts for the increased efficien-
cy. However, because of the increased length-to-diameter jet
ratio, the high-speed smaller nozzle has a maximum flow of
only 295 fps, which is slightly lower than the design velocity
of 300 fps. The mean flowfield was surveyed with a continuous
pitot-tube traverse at the exit plane of the nozzle, and the
mean velocity was uniform to within 0.25% for velocities of
90 fps and above, and within 0.6% at 30 fps.
Figure 2 shows the total turbulence spectrum at a 3.3-ft

distance (l m) from the high-speed nozzle face on the tunnel
centerline. The overall turbulence level is 0.23%. This
value is considered adequate for most tests. To improve both
TUNNEL WALL
Fig 3. Free-field chamber

calibration.
ANECHOIC
TREATMENT
DETAIL
(NOT TO SCALE)
1 2 4 8 1011.5
DISTANCE FROM TUNNEL CENTERLINE(ft)
the low-speed velocity uniformity and the turbulence level,

additional upstream screens are planned. The tunnel energy
ratio5 (i.e., the ratio of test section dynamic pressure to
complete tunnel circuit total pressure loss) was estimated to
be about 1.5; the free jet and the collector were responsible
for 75% of the total circuit energy loss. This low figure for
the energy ratio was considered a minor disadvantage compared
to the advantages of a quiet long jet, large test chamber, and
a compact quiet collector.
Acoustic Performance
Anechoic Chamber Calibration
To evaluate the free-field characteristics of the test

chamber, one measures the sound pressure level along radii
emanating from a point source, which is located at the center
of the chamber. The objective is to determine deviations from
attenuation at the rate of 6 dB per doubling of distance. Be-
cause deviations in a closed space are due to wall reflections,
one would expect the greatest deviation from free-field be-
havior to occur along the radius that intersects the chamber
wall, where it is nearest to the source.
The cross-jet noise source10 was used because it is a good

approximation to a broadband point monopole. The one-third-
3BN HIGH-SPEED FREE JET

OA A'IND TUNNEL - 4' * 4 ' NOZZL E
SPL -LOW NOISE 7 f t FROM TUNNEL
'V
^ 10ft FROM NOZZLE FACE
l\
/-~^ \ 40 fps
\K ^
V^ r-57 N
III
r90
XN
X
\
S
\
^^.XX
' 1
/
X \ s^^^
^
1 -^
oy ^^: ^ -^
S^ X
^^^
NO FL
..^ "Xx
-\
6 3 5 63 1 5 2 50 500 1000 2000 40OO 8OOO 16,000 31
so
3BN HIGH SPEED FREE JET
tfIND TUNNEL -28" x 40" NOZZL
80 \ "LOW NOISE 7 f t FROM TUNNEL C

X 10ft FROM NOZZLE FACE
\a
o 70
\
\
\
N, IN ^\
K
K
.0
-240 ps
^X
:t
<
\
O
60
\ ^^ s
0
v -< 175
~^
\ r^ 5 * ^ - ~^^^
^
^
S 50
N
^^ ^
40 s r 82
\
-^^
30
\
31.5 63 125 250 500 1000 2000 4000 8OOO 16,000 31,500
ONE-THIRD OCTAVE BAND CENTER FREQUENCY ( H z )
Fig. k Background noise levels in

BBN wind tunnel.
octave-band sound pressure levels were measured along various

radii by traversing the path with a 1/2-in. condenser micro-
phone. As expected, the deviations were most noticeable along
the horizontal path normal to the tunnel centerline from the
source to the wall. The results of this "worst-case" condition
are illustrated in Fig. 3. Free-field conditions are accept-
able for most experiments down to 160 Hz. Free-field behavior
is obtained down to 100 Hz along radii that intersect the wall
at greater distances.
Background and Flow Noise
Figure h shows the wind-tunnel background and flow noise

spectra for the various velocities for the two nozzles. The
microphone was located 10 ft from the exit plane of the nozzles,
8 ft to the side of the tunnel centerline. The low-frequency
noise (below 125 Hz) is attributed to the jet buffeting the
structure. However, typical model-scale considerations show
that this low-frequency range is rarely of interest. For a
10:1 scale-model test, the lowest frequency of interest is
B8N LARGE TUNNEL; 28x40in.NOZZLE, X/VA" =10

BBN LARGE TUNNEL; 4x4ft NOZZLE; XA/A~ = 8.5
NASA AMES 7xlOft TUNNEL(TREATED);XvT=2.15
UARL TUNNEL; Xv/A~ = 1.95
- UARL TUNNEL; X,/A~=0.90
MIT AVLTUNNEL; X-/^--0.&7___________
2 -60
Fig. 5 Tunnel noise vs col-
lector placement. A IN ft 2
R IN ft
U IN fps
SPLINdB(re0.002/ibor)
DIMENSIONLESS FREQUENCY,^
200 Hz. It was established that the fan and diesel noise are
not a factor in the flow noise. The background noise compari-
son with other tunnels is shown in Fig. 5-
Lip noise and the impingement of flow on the collector

structure are responsible for the noise above 125 Hz. As
expected, this part of the spectra can be normalized on the
basis of the velocity to the sixth power at constant Strouhal
number, which indicates that the noise sources are dipole in
nature. As reported elsewhere,1'9 the free-jet noise is well
below the tunnel background noise levels for the low, subsonic
velocity range involved.
High lift devices may deflect the tunnel jet and increase
the background noise levels.1 By extending the nozzle with a
tilted section, we measured the effects of the main-jet flow
deflection by angles up to 10. We did not detect any changes
in either the acoustic or aerodynamic performance of the tunnel,
a fact that bore out the validity of the concept of large sepa-
ration between the nozzle and the collector.
As previously mentioned, the collector placement is a

major factor in the tunnel background noise, as well as in the
general utility of the tunnel as a free-field facility. The
effect of collector location and shape is just now being studied
at BBN. Some preliminary comparisons among acoustic tunnels are
given in Fig. 5> where the sound pressure level at approximately
a ^5 observation angle to the jet axis at 1-2 diameters down-
stream is given. Levels are seen to normalize on a U6 basis and
jet exit area. Short X/D placements generally produce more
high-frequency noise in the test section than large X/D loca-
tions. Further work on this aspect of acoustic tunnel design
is needed.
Summary
Some major aspects of the processes used to design and

test an acoustic wind tunnel, have been given. The BBN tunnel
used as an example possesses two unique features: a long,
open-jet test section housed in a large test chamber, and a
stagnation plate collector design. These features, which are
the result of a careful noise/performance tradeoff, produced a
highly versatile cost-effective solution to the problem of
designing a high-performance wind tunnel. Experiments that
demonstrate the tunnelTs ability to accommodate a variety of
tests and simultaneous measurements of acoustics and aerodynamic

performance were described. The characteristics of this tunnel
compare favorably with other open-jet facilities.
References
âterson, R.W. , Vogt,, P.G. , and Foley, W.M. , "Design and

Development of the United Aircraft Research Laboratories
Acoustic Research Tunnel,11 Journal of Aircraft, Vol. 10, No. 7,
July 1973, pp. ^427-^33.
2
Batchelor, G.K., "Sound in Wind Tunnels," Rept. ACA-18, 19^5,
Australian Council for Aeronautics, Melbourne, Australia.
3
Parker, J.D., Boggs, J.H., and Blick, E.F., Introduction to
Fluid Mechanics and Heat Transfer, Addison-Wesley, Reading,
Mass. , 1970.
^Schlicting, H., Boundary Layer Theory, McGraw-Hill, New York,

1968.
5
Pope, A. and Harper, J.J., Low-Speed Wind Tunnel Testing,
Wiley, New York, 1966.
6
Hersh, A.S. and Hayden, R.E., "Aerodynamic Sound Radiation
from Lifting Surfaces With and Without Leading-Edge Serrations,"
CR-llii370, 1971, NASA.
7
Beranek, L.L., Acoustics, McGraw-Hill, New York,
8
Maestrello, L., "UTIAS Air Duct Facility for Investigation of
Vibration Noise Induced by Turbulent Flow Past a Panel," TN 20,
1958, University of Toronto, Institute of Aerospace Studies.
9
Hanson, C . E . , M The Design and Construction of a Low-Noise,
Low-Turbulence Wind Tunnel," TR79611-1, 1969, Acoustics and
Vibration Laboratory, Massachusetts Institute of Technology,
Cambridge, Mass.
10
Ver, I.L., "Acoustical Modeling of the Test Section of the
NASA Langley Research Center Full-Scale Wind Tunnel," Rept.
2280, 1971, Bolt Beranek and Newman Inc.
CORRECTION OF OPEN-JET WIND-TUNNEL MEASUREMENTS
FOR SHEAR LAYER REFRACTION
Roy K. Amiet*
United Technologies Research Center, East Hartford,, Connecticut

Abstract
The problem of sound refraction by a plane, zero-thickness

shear layer is treated by combining a previous solution of Rib-
ner and Miles "with geometrical acoustics. Analytical expres-
sions are given "which allow one to correct far-field measure-
ment angle and acoustic amplitude for the effects of shear
layer refraction. The correction is independent of source type
and the results represent the sound field one "would expect to
measure in a flow that has a freestream extending to infinity.
Preliminary experimental results are in basic agreement, but
further tests are necessary to establish the theory definitely.
Nomenclature
h = distance of source from shear layer

j 5 ^5 5 = unit vectors in the x,y,z directions, respectively
M = tunnel Mach number
P = pressure
r = source to microphone distance

Conference, Hampton, Va,, March 2^-26, 1975. The author is
grateful for the many helpful discussions with personnel at
Bratt & Whitney Aircraft and United Technologies Research Cen-
ter. In particular, discussions with A. A. Peracchio and
A. B. Packman were helpful in formulating the experimental
work, and the assistance of H. P. Day with the experimental
program was appreciated.
^Senior Research Engineer, Aeroacoustics Group.
259
260 R. K. AMIET
XQ = axial distance from source to point at -which sound

ray crosses shear layer
y-, = microphone distance above sound source
z-j_ = separation between layers of double shear layer
a = angle bet-ween shear layer and wavefronts just .
below shear layer
= [(1-M cos0)2 - cos2e] -1/2

0 = angle between shear layer and ray propagation
direction above shear layer
0f
= angle corrected for shear layer effect

0o = value of 0 which begins zone of silence
X = wavelength
5 = parameter defined by Eq. (7)
= phase_______
a = J 2 2 . 2 2.
Vx +p (y +z )
a) = circular frequency
Subscripts
A,B,C = points defined in Fig. 1

i = incident
m = measured
r = reflected
Introduction
In studying the effect of flight speed on a sound source

such as a compressor or a jet exhaust, it is necessary to ob-
tain accurate experimental data under controlled conditions.
An open-jet anechoic wind tunnel,such as that located at United
Technologies Research Center,, can be used to generate such
data. Sound reflection from the walls, which is a problem with
closed section tunnels, is eliminated by the use of an anechoic
chamber5 and the problem of extraneous noise due to flow inter-
action with microphones is avoided, since the microphones are
outside the stream. However, the open-jet tunnel does have the
disadvantage that the sound produced by the device being tested
must pass through the jet shear layer before being sensed by a
microphone outside the flow. In crossing the shear layer, the
sound is refracted, an effect that becomes more important as
SHEAR LAYER CORRECTION 261
the Mach number is raised. Also,, scattering from the turbu-

lence in the shear layer may occur, but this is neglected in
the present paper.
Several previous studies, Refs. 1-7, for example, have

been conducted on this problem. These, however, generally
have considered a specific type of sound source near a shear
layer and have calculated a corresponding directivity curve
rather than addressing the general problem of correcting any
sound source such as a combination of monopoles, dipoles, etc.
Reference 8 proceeds along the lines of geometrical acoustics
as done here. However, Ref. 8 uses the technique for calcu-

lating the directivity pattern produced by a given source in
the presence of a shear layer rather than arriving at a method
for correcting acoustic tunnel measurements independent of
source type. Because the present study corrects both the sound
amplitude and the measurement angle, it can be applied to a
general source.
Theoretical Development
The modeling of the problem is shown in Fig. 1. The ob-

server is at a distance y^-h above the shear layer, and the
WAVEFRONT
Fig. 1 Acoustic source beneath plane zero-thickness
shear layer.
262 R. K. AMIET
sound source is a distance h below the plane, zero-thickness

shear layer. There is no restriction on the size of h. The
observer is assumed to be in both the geometrical and acoustic
far-field of the source; i.e., the source-observer distance is
significantly greater than both the source dimensions and the
acoustic wavelength. Both the source and observer are assumed
to be in a plane normal to the shear layer and parallel to the
flow. The line connecting the source and the observer makes an
angle Qm with the shear layer. The measured angle ^ goes to
zero as the observer moves downstream and to rr as the observer
moves upstream. The actual path of a sound ray is represented
by the line SCO, which, below the shear layer, makes an angle
Q T with the shear layer, and an angle 9 above the shear layer.
The change from Q T to 9 as the sound passes through the shear
layer is a result of refraction by the shear layer. The fluid
densities above and below the shear layer are assumed to be the
same. (This assumption can readily be eliminated.) There is
little change in density across the shear layer of the UTRC
Acoustic Research Tunnel. The Mach number M is assumed uniform
below the shear layer and zero above it.
If the shear layer had not been present so that the uni-
form Mach number M continued out to infinity, the sound on
reaching the former position of the shear layer would continue
to propagate rectilinearly, following the dashed line in Fig. 1
rather than the solid line. Thus, the sound heard at position
0 in the presence of the shear layer would be heard at position
A or B in the absence of the shear layer. If one wishes to
correct the data to an equal sideline distance, point A is
used, whereas if one wanted to correct the data to an equal
radial distance from the source, point B is used.
The method of the derivation is to use geometrical acous-

tics together with the solution of Ribner and Miles for the
transmission and reflection of sound by a plane, zero-thickness
shear layer. The sound measured at the observer point 0 is
traced back by geometrical acoustics to point G+ just above the
shear layer. Knowing the amplitude at point C+, Ribnerfs re-
sults are used to cross the shear layer, giving the amplitude
at point C- just below the shear layer. The amplitude at point
A or B which would exist in the absence of the shear layer then
can be obtained from the sound level at G- by noting that sound
pressure decays inversely as the distance from the source.
Thus5 the pressure that would exist at point A would be the

pressure at point C- times the ratio of distances of the source
from points C and A.
It should be pointed out that it is not necessary for

point C to be in the far field of the source. When the sound
measured at point 0 is used to calculate the sound at point C,
only the far-field component can be calculated, since point 0
is assumed to be in the far field. Thus, an actual measurement
of the sound at point G might not agree with the value given
here unless one were able to separate out the near- and far-
field parts of the measurement.
The details of the derivation are given in Appendix A.

The resulting correction equations are given below:
tan 0' = /(/3 2 cos 0 + M) (la)
y, cot 0m - h cot 0' + (y, -h) cot 6 (ib)
r 2 2 ~l
~l 11/2
where - [d ~M COS 9} -COS Q\
P
A f h f . Q6+(
, , V| ,N/-l l / 2 r 3fl , , i n/. 31 1/ 2
-J [sm0 + ( T -l)C J
PB/PA = sin 0m/sm0' (3)

The first two equations give Q T in terms of 0m. The angle 0
could be eliminated, giving a single equation relating g f to
g.^5 but for simplicity of expression 9 is left as a parameter
here. Equation (2) gives the corrected pressure P^ at an equal
sideline distance, whereas Eq. (3) gives the corrected pressure
at an equal radial distance from the source. Again, 9 appears
as a parameter in these equations and is related to g by Eqs.
When the observer is far from the shear layer so that

y-,h, Eq. (ib) gives Q = 9, and also the terms within the
braces {...} in Eqs. (2) and (3) become unity.
264 R. K. AMIET
It is interesting to compare these results with the re-

stilts obtained by Gottlieb^ for the directivity of a monopole
placed in a stream and extended by Amiet to the case of dipole
sources. As was done here, Gottlieb assumed a zero-thickness
plane shear layer. The distance y-^ was assumed much greater
than h. If the directivity pattern given by Gottlieb is cor-
rected in angle and amplitude using Eqs. (1) and (2) with
h/y-j_ = 0, one finds that the resulting directivity curve is
identical to that produced by a source in a uniform stream with
no shear layer. (See the calculation in Appendix B.) In other
words? Eqs. (1-3) have contained in them the Gottlieb results.
The present results are more general, however, in that they do

not assume a specific type of sound source but rather derive a
correction valid for any types or combinations of sources.
GottliebTs solution amounts to combining the known directivity
of a specific kind of source in a stream together with the
correction in angle and in amplitude presented here to give a
resulting directivity curve for a monopole in a stream in the
presence of a shear layer. Gottlieb did not present the re-
sults as a correction in angle and amplitude, however, and it
is not possible to use his results for correction of the sound
from a general sound source.
An interesting sidelight of this correction is that, if

the observer point 0 is on the y axis directly above the source
and y-j_h, the sound measured by the observer will be just that
which would be measured by an observer at the same point and
with the same source strength but with no flow. In other
words, the sound level produced by the source with the tunnel
on should remain unchanged at this particular observer location
if the acoustic tunnel is turned off, provided that the source
strength can be kept fixed during the process. This is not
obvious, but it can be shown from Eqs. (1) and (2) with h/y., =
0. In essence, it results from the fact that the convective
amplification of sound is zero for an observer at 90 to the
direction of motion of a source.
In order to apply the results given by Eqs. (1-3), it is

not necessary for the acoustic wavelength \ or the body size 4
to be small compared to h. It is only necessary that \ and &
be small compared to the source-observer distance OS so that
the observer is in the acoustic and geometric far field of the
source. For the particular case of y.h, it will be noted
that Eqs. (1-3) are completely independent of h. This was a

point noted by Gottlieb to be true so long as the observer is
not in the so-called "zone of silence."
The zone of silence is the angular region 9 < 9 where 9

is that particular value of 9 for which 9' = 0. From Eq. (1),
this gives = 0 or
sec 00= I + M (ij.)

Angles 9 within the zone of silence will not concern us here,

since they do not correspond to a real value of 9 T . As noted
by Gottlieb, acoustic waves propagating to the far field at
angles less than 9 cannot be matched with acoustic propagating
waves beneath the shear layer but rather are matched with waves
that decay exponentially with the distance h.
It should be noted that the corrected angle 9' does not

cover the entire range 0 to rr when 9m goes, through this range.
Rather, 9* ranges from 0 at the zone of silence to a value 9 !
smaller than TT "when Q = 9 = TT> where, from Eq. (la),
tan &'I = VzM + M2/ (-1 -I- M -h M2) (5)

\s /
Thus, outside the jet stream it is not possible to measure the

sound for values of 9' greater than 9^'. Also, as mentioned
below, measurements made near the value 9* = 9^' probably are
inaccurate because of reflection from the lower shear layer.
This precludes making measurements of the sound radiated for-
ward from the source at small angles to the axis. The sound
is reflected from the shear layer and thus is trapped within
the jet.
Discussion
Typical results of these equations are plotted in Figs.

2 - U. Figures 2 show results obtained for y.,h. The in-
dependent variable in these plots is the measured angle 9 .
Figure 2a gives the corrected angle 9' for several Mach numbers
and Figs. 2b and 2c give the decibel correction to be added to
the measured sound level for equal sideline and equal radius
measurements, respectively. Figures 3 and k show similar
curves for h/y^ = 0.2 and h/r = 0.15, respectively.
R. K. AMIET
10
IE 2
8?
o 0
20 40 60

100 120 140 160 180
o -2 MEASURED A N G L E S
CM
-4
-6
20 40 60 80 100 120 140 160 180

MEASURED ANGLE fl
10 -
8-
E
t 2
20 40 60-^80 100 120 140 160 180 JP^g . 2
MEASURED A N G L E Bm ~ Corrections for h/y., = 0;

-4 (a) Angle correction,
(b) Amplitude correction
-6
to equal radius position,
-8 (c) Amplitude correction
-10
to equal sideline distance.
c
180
160
140
uj 120
_i
O
< 100
o
HI
o 80
20 40 60 80 100 120 140 160 180

MEASURED ANGLE 6m
Fig. 3 Corrections to equal sideline distance with h/y. = 0.2;
(a) Angle correction, (b) Amplitude correction.
^60-^80 100 120 140 160 180

MEASURED ANGLE, 6
20 40 60 80 100 120 140 160 180
MEASURED ANGLE, 6m
Fig. k Corrections to equal radius position with h/r = 0.15;

(a) Angle correction,, (b) Amplitude correction.
One of the major simplifying assumptions used in this deriva-

tion is that there is only a single shear layer, when in fact
the open jet has an upper and a lower shear layer, with the
sound source in the middle. For the present results to be
applicable, the reflection from the lower shear layer should be
negligible. The ratio of reflected pressure to incident pres-
sure for the case of a plane wave incident on a plane, zero-
thickness shear layer was given by Ribner as
r
r .
(6)
This is plotted in Fig. 5, which shows that, except for angles

near the zone of silence and angles near l80? the amplitude of
the reflected wave is small. Thus, if the observer is not near
one of these two limits, the lower shear layer would be ex-
pected to have little effect.
One additional important assumption was that the thickness

of the shear layer could be ignored. Graham and Grahaor made a
calculation of the sound transmission through two plane, zero-
thickness shear layers a distance z, apart. This example
should give some idea of the effect of finite thickness on
shear layer transmission. Figure 6a shows the problem, which
consists of two shear layers separated by a distance z^. The
ro
CT)
00
-f
2
CO
JN ii
\ >
X
I
00
ro
ii
CO
Q o> 2
o

Q
The greatest deviation from unity of this ratio | T(Z-, )/T(0) |

occurs for those values of z./x such that i- = (2n+l) TT/2=^n.
Setting sin 5n = 1 (this of course requires a different value
of z-j/x for each value of Q T ) gives the results shown in Fig.
6b for a Mach number M = 0.5 and M-j_ = 0.25. There is little
effect of finite thickness except at extreme angles.
Experimental Results
Devising an experiment to measure the refraction effect is

complicated by the difficulty of obtaining a source whose di-
rectivity in a uniform stream is known. Measuring the direc-

tivity of a source in stationary air produces no difficulty,
but this directivity would be expected to change in an unknown
manner when the source is placed within a stream.
One of the simplest sources is the compact dipole. The

directivity of a compact dipole in a stream is 'known, and so
the directivity corrected for the presence of the shear layer
can be obtained easily . This is shown in Fig. 7 for a Mach
number of 0.27, along with the directivity of a dipole in
stationary flow for comparison. The observer is assumed to be
far from the shear layer, so that h/y-,l. Curves for three
values of h/x also are shown, and it will be noted that chang-
ing h/x affects only the sound in the zone of silence.
To obtain an experimental check on these theoretical re-

sults, a 1/16 in. diam. cylindrical rod was placed in the po-
O 300 fps EXPERIMENT f - 11 KHz
M = 0.27 |
- M = 0
180
Fig. 7 Directivity of dipole near

shear layer; theory vs experiment.
270 R. K. AMIET
tential core of a 2 in. diam. free jet. Because of vortex

shedding from the rod, a fluctuating dipole with a Strouhal
frequency of about 0.2 (based on rod diameter) was produced.
Acoustic measurements were taken on a circular arc at a dis-
tance of 33 in. from the rod and are denoted by the circles in
Fig. ? Measurements were taken only in the downstream quad-
rant because the upstream sound could be shielded partially by
the jet nozzle, and because very little difference is expected
upstream between directivity of a dipole in stationary air and
in a stream,as is evident from the theoretical curve shown in
Fig. 7.
Since the directivity of a dipole varies rather slowly

with measurement angle, the only substantial difference in
sound amplitudes between a dipole in stationary air and a di-
pole in a jet is near the zone of silence, where a rapid drop-
off in the sound level is predicted. This dropoff did occur
in the measured sound levels at about the proper angle. The
sound level was normalized to 1 at 9m = 90.
Some deviations from the predicted directivity are ex-

pected near the zone of silence, as mentioned previously. Near
the zone of silence, the reflection coefficient of the shear
layer becomes important, so that the validity of approximating
the jet as a single shear layer begins to break down. This may
explain the tendency for the measured directivity curve to ap-
pear somewhat rippled just before the zone of silence is en-
tered. This tendency was predicted analytically in an unpub-
lished study by Lansing and Brown for the case of a source on
the centerline of an axisymmetric jet.
There appears to be some -uncertainty, e.g., Howe^, as to

whether the procedure used here and in Refs. 3 and U gives
correct results in the zone of silence because of shear layer
instabilities. Also, the acoustic prediction for the zone of
silence depends strongly on the ratio h/\. In actuality, the
shear layer has a finite thickness, so that h cannot be defined
accurately. Because of these points, it should not be surpris-
ing that agreement between experiment and theory is unfavorable
in the zone of silence. In this region, however, both theoret-
ical and experimental results have values significantly lower
than those of a dipole in stationary air. Any lack of ability
to predict the sound level within the zone of silence is not
important for purposes of correcting wind-tunnel data, since

measurements made here do not correspond to measurements that
could be made in a freestream without a shear layer. That is,
for 0 in the range 0 < Q < 9 , tan0T as given by Eq. (la) is
imaginary.
Thus5 these experimental results give some verification of

the theory in that the zone of silence has its onset at about
the angle predicted by theory. A better verification perhaps
could be obtained by using a more directional source. Some
effort has been made along these lines, but significant prob-
lems are involved. When a sound source is placed in a stream,
one cannot expect that the radiation pattern of the source will
remain unchanged even if one has no shear layer. Thus, to have
an idea of the amount of shear layer refraction, the source
directivity must be measured inside the shear layer and com-
pared to that outside the shear layer.
The measurement inside the shear layer is made more diffi-

cult by the fact that the inflow microphone must be in the
acoustical and geometrical far field of the source. This dif-
ficulty became more obvious when a small 1/2-in. jet operating
supersonically in the screech regime was used as a source.
This source produced a narrow-band signal that had a very sharp
directivity. However, there was disagreement between the di-
rectivities of the inflow and the far-field microphones, even
when the acoustic tunnel was not operating (i.e., no shear
layer), indicating that the inflow microphone was not in the
far field of the source.
An alternative procedure described below makes use of a

source that need not be directional, but that can give verifi-
cation of the angle correction. A pure tone noise source was
placed in the center of the acoustic tunnel. For this purpose,
a high-frequency (25-kHz) dog whistle was used. It was placed
behind a 2-in. pipe concentric with the tunnel, the pipe tend-
ing to shield the whistle from the effects of the flow. The
idea is then to determine the angle of the wavefronts after the
sound has passed through the shear layer. This can be done by
cross-correlating the output of two microphones (see Fig. 8a).
By comparing the cross-correlation with flow to that obtained
with no tunnel flow, the shift in phase between the cross-
correlation of two adjacent microphones determines the angle of
272 R. K. AMIET
propagation of the far-field wavefront. Given this angle, the

point x at -which the sound emerged from the shear layer can be
calculated. The measured value of Q 1 then is given as tan"
(h/x ). For this particular test, the far-field microphones
were placed at a distance of r = 10 ft. In retrospect, it
might have been preferable to make the measurements nearer to
the shear layer, allowing a better calculation of Q T . The re-
sults of the measurements shown in Fig. 8b are in reasonable
agreement with theory, the theory generally underestimating
the experimentally measured angle correction.
The effects of turbulent scattering have been ignored in

the present analysis. Some preliminary work by Paterson in-
dicate that for frequencies in excess of 20 kHz this may af-
fect the sound level of a tone by several dB, but further ex-
perimental testing is needed.
Conclusions
The dipole produced by a rod in the flow shows the onset

of the zone of silence at the angle predicted by theory. The
amplitude of the sound outside the zone of silence showed good
agreement with theory, but this gives only a weak verification
of the theory because of the rather small variation of ampli-
tude with angle; i.e., there were no sharply defined directiv-
ity peaks with which one could check the angle and amplitude
corrections independently.
90
80 O EXPERIMENT
THEORY
70
Q>
LU 60
< 50
Q
UJ
040
30
20
10
0 10 20 30 40 50 60 70 80 90
MEASURED ANGLE, # m
Fig. 8 Experimental verification of angle correction;

(a) Experimental setup, (b) Theory vs experiment.
As discussed in the text, the use of a source with a

sharply defined directivity presented certain difficulties.
An alternate procedure of cross-correlating the outputs from
nearby microphones to define the wavefront angle gave results
for the angle correction which were in reasonable agreement
with theory. Thus, the tests performed gave results that
agreed with theory. The author feels, however, that there is
f\irther room for experimental verification of various aspects
of the theory.
Appendix A: Derivation of the Correction Equations
As was mentioned in the text, geometrical acoustics along

with the solution of Miles or Ribner will be used to derive the
correction relation. The relations between the various angles
in Fig. 1 can be derived as follows. (Note that all angles
marked in this figure are in the z = 0 plane.) The loci of
points of equal phase (the wavefronts) of the sound wave are
the circles drawn in the figure. These circles propagate out-
ward at the speed of sound and driFt downstream at the fluid
velocity5 the phase of the wavefronts being given by
= t + (Mx-a-)/Co/32 (Al)
to the wavefront intersecting the point (XQ, h), and, by cal-

culating dy/dx: from Eq. (Al), a is found to satisfy the equation
cos a/(sin a + M) = h/Xo = tan 0' (A2)
The relation between a and 6 is found by equating the x com-

ponents of phase velocity across the shear layer. As for the
case of a plane wave incident on the shear layer discussed by
Ribner1,
sec 9 = esc a + M (A3)

The last two equations allow the relation between 0 and Q! to
be written as
1
= /(/32cos#+M)= h/xn
274 R. K. AMIET
or
cos$'
costf = -M] (A5)
Equation (A^J-) is Eg., (la) in the text. Equation (Ib) relating

^ to 9 and q f is derived by noting that
y, cot #m= x -h) cot# (A6)
Using Eq. (A2) gives Eg. (lb).

In order to calculate the corrected amplitude, we first

must calculate the amplitude at point C- just below the shear
layer. To do this, the spreading rate of a ray traveling along
CO must be determined.
The rate of spreading in the xy plane is determined easily

by calculating dxQ/d0 from Eq. (A^)5 giving
dx0/d = (A7)
Figure 9 shows the cross section in the xy plane of a ray tube.

The ratio of the lengths d^/^l is found to be
[(y r h)csc0]dQ
= i +(
c 3
(A8)
sin
The spreading of the ray in the perpendicular plane (the
plane formed by the line OC and the z axis) also must be de-
termined. Ib do this, the amount of refraction by the shear
spreading in xy
layer of a ray propagating out of the xy plane must be deter-

mined.
For a plane wave

P: = 6 (A9)
incident on the shear layer from below, a transmitted wave

Pt = Ae i M -t> (AID)
is produced, where A is some transmission coefficient (possibly

complex) to be described later. The phase expressions Q. and
0, must be of a form such that P, and P^ satisfy the wave
equation and the convected wave equation, respectively, and the
phases of P^ and PJ- must match at the shear layer. The appro-
priate expressions are
i - T^- x sincjb cos\// + z sin< sin\// + y /(i-M sind) cosi//)2-sin2c

(A11)
w [xr sine/) cost// + z sine/) sin\// + y cosc/>|1
t = Q
These equations were expressed in terms of the angles 0 and |,

since these angles are the polar angles of the unit normal to
the transmitted wavefronts, as shown in Fig* 10.
SOUND RAY, n t
Fig. 10 Description of the polar

angles 0, ^, and |f
276 R. K. AMIET
By calculating v@/|v|? the unit normals to the incident and

transmitted waves are
A * sinc cosv// H-t^/d-M Sin + k

n =
i
- M sinqb
r cos\l/
r
(A12)
nt = i sin<^> cosv// + ) cose/) + K sirujb sinv//
In the actual case, the wavefronts are circular rather

than plane and are given by Eq- (Al). By calculating the gra-
dient of Eq. (Al), the unit normal to the actual wavefront is
y@ _
(A13)
O--MX
By comparison -with Eq. (A12), the angle of propagation of the

transmitted wave can be related to the position (x 5h,z) at
which the ray crosses the shear layer. Thus,
sincft
l-Msin<cos\// o~ - M x 0
sin</> sin\// /3 2 z
I- M sine/) cos \l/ cr-Mxo
For the z = 0 plane, ^ = 0, 0 = ^ - 0 , and Eq. (Alâ) becomes

equivalent to Eq. (la). Equation (Al^b) is the equation of
interest at present. By taking the derivative of z with re-
spect to ^, for small ^ we find
z /.2
0/
"
where Eq. (A^) was used to evaluate x^h.
Rather than using the polar angle ^ as measured in the

x-z plane, we wish to use the angle ^ f , which is measured in
the plane perpendicular to the xy plane and along the ray OC,
as shown in Fig. 10.
The relation between ^ and ^f is

r -i-l/2
sini//1 = sin\// cos \// + cot < (Al6)
which becomes for small ^
i//' i//sin< (A17)
Thus, Eq. (A15) becomes
(dz/d^'j , = h/ (Al8)
Figure 11 shows the ray spreading in the plane produced by the

z axis and the sound ray.
The ratio of the two lengths dz., and dz2 is

dz2? dz, + (y,-h)csc0d\// / y, \
- - 1 + -L-, csc0 (A19)
dz, dz, v .. /
The product of the two length ratios given by Eqs. (A8)
and (A19) gives the ratio of ray tube cross-sectional area for
a point in the far field to that for a point just above the
shear layer. In order to conserve acoustical energy in the
ray tube, the acoustical pressure should behave inversely with
the square root of the ray tube cross-sectional area. Thus,
Pc+ /dzp dip

TT = /i L (A 20)
Pm 7dz, dh ^ }
Fig. 11 Ray divergence in a plane

perpendicular to the xy plane and
along ray path.
278 R. K. AMIET
For the case of a plane "wave incident on the shear layer,

the ratio of the transmitted press-ore to the incident press-ore
was found by RLbner to be
p
c_/Pc+ = (l/2
^ [ +sin#(l-Mcos0)2] (A2l)
Using this expression along -with Eqs. (A8) and(A19-A20) then

allows Pn to be calculated in terms of the measured pressure
"~
p
m
Since the pressure decays as r"^ in the far field, kno-wing
_ allows us to calculate the pressure P in Fig. 1. Thus,
C- = (h/r) csc0' = (h/r) [MZ(I- M
Combining Eqs. (A20) and (A22) then gives Eq. (3) for P .
B
For correction to equal sideline positions rather than
equal radius, the ratio of P to P is
A C-
P
A/PC-= h/y,
Combining this with Eqs. (A20) and (A21) gives Eq. (2).
Appendix B: Relation to the Solution of Gottlieb
Gottlieb has found the directivity to be expected from a

monopole sound source placed beneath the shear layer to be
a;
This relation assumes that the observer is in the far-field at

a constant radial distance from the source. The angle Q is
restricted to lie outside the zone of silence (although Gott-
lieb also gave results for the zone of silence). Let us apply
the appropriate correction equations [Eqs. (1) and (3), assum-
ing y.,h] to determine if the correct directivity for a mono-
pole in a stream in the absence of the shear layer can be cal-
culated from Gottlieb1 s solution.
Equation (3), -when combined with Eq. (Bl), gives the cor-
rected press-ore PB in terms of the measured angle ^ as
PB OC (!-McOS#j[M 2 (l-McoS0 m ) 2 + l - M 2 C O S 2 # m ] l / 2 (B2)
From Eq. (1), one can show that

/32cos#m + M
cos 9' = - - y, h
giving
P oc (I/COS0') (/32cosm+M)(l-Mcosm)
Finally, using Eq. (A5) gives the result (since 9 = 9 here)
PB oc (i//32o-J) (l-
where
x' = r cos#' y' = r sin$'
Now Eq. (B5) [except for constant factors such as monopole

strength which were omitted from Eq. (Bl)] is the far-field
solution for a monopole in a stream [see Eq. (1.33) of Ref. 9>
for example]. Thus, the solution procedure used by Gottlieb
has inherent in it the same assumptions, such as geometrical
acoustics, used here. The main difference is that Goftlieb's
procedure gives directivity predictions for each of the source
types (monopole, dipole, etc.), whereas the present procedure
gives a method for correcting the data independent of the
source type.
References
Ribner, H. S., "Reflection, Transmission and Amplification of

Sound by a Moving Medium,n Journal of the Acoustical Society of
America, Vol. 29, No. U, April 1957, pp.
280 R. K. AMIET
O
Miles, J. ., "On the Reflection of Sound at an Interface of
Relative Motion, " Journal of the Acoustical Society of America,
Vol. 29, No. 2, February 1957, pp. 226-228.
^Gottlieb, P., "Sound Source Near a Velocity Discontinuity,"

Journal of the Acoustical Society of America, Vol. 32, No. 9>
September 1960, pp. 1117-1122.
Amiet, R. K., "Propagation of Sound Through a Two-Dimensional

Shear Layer with Application to Measurements in the Acoustic
Research Tunnel," Rept. UAR-lAO, 1972, United Technologies Re-
search Center.
Graham, E. . and Graham, B. B., "Effect of a Shear Layer on

Plane Waves of Sound in Fluid," Journal of the Acoustical Soci-
ety of America, Vol. ^-6, No. 1, July 1969, pp. 169-175.
Lansing, D. L. and Brown, T. J., "Refraction of Sound from a

Source in a Jet," Memo, for H. H. Hubbard, December 1970, NASA
Langley Research Center.
7
'Howe, M. S., "Transmission of an Acoustic Pulse Through a
Plane Vortex Sheet," Journal of Fluid Mechanics, Vol. ^-3,
Part 2, August 1970, pp. 353-367-
o
Csanady, G. T., "The Effect of Mean Velocity Variations on
Jet Noise," Journal of Fluid Mechanics, Vol. 26, Part 1, Sep-
tember 1966, pp. 183-197.
Âmiet, R. K., "Aerodynamic Sound Production and the Method of

Matched Asymptotic Expansitions," Ph.D. Thesis, 1969? Cornell
University; Also Rept. UAR-H223, 1969, United Technologies
Research Center.
10
Paterson, R. W.5 "Passage of High Frequency Pore Tones
Through a Turbulent Round Jet," Rept. UAE-M112, 1973, United
Technologies Research Center.
USE OF A LASER SHADOWGRAPH FOR JET NOISE DIAGNOSIS
Michael J. Rudd*
Bolt Beranek and Newman Inc., Cambridge, Mass.

Abstract
The shadowgraph was first used for quantitative measure-

ments in 1970. With a laser as a light source, the shadowgraph
is proposed as a means of predicting the noise radiated, per
unit volume, from a jet. One of the properties of the shadow-
graph is that its spectrum is related to the fourth spatial
derivative of the density fluctuation, whereas the noise pro-
duction can be shown to be related to the fourth time deriva-
tive. The spatial and time derivatives in a narrow band are
related by the convection velocity of the turbulence. By
utilizing a noise generation model, which is cast in terms of
density fluctuations, the noise radiated in the far field of a
jet can be predicted from measurements made inside the jet.
Introduction
Despite the achievements of modern technology, we still

do not understand fully the basic noise-generating mechanism of
the jet aircraft. Many previous investigations have dealt
unsuccessfully with the location of the noise in the jet.
These studies found that a microphone cannot be placed in the
flow because the nonacoustic pressure fluctuations are much
greater than the acoustic pressures. Some external means of
locating the noise source is required.
To evaluate and improve the performance of jet noise

silencers, we must identify the regions of noise production in
a jet. Several techniques have been proposed for this purpose.

^Senior Scientist.
281
282 M. J. RUDD
W.T. Chu1 suggested that an elliptical dish with a microphone

at its focus be used as a directional antenna, but the resolu-
tion of this system is only one or two wavelengths of sound.
Siddon2 and Scharton and White3 proposed placing a microphone
with a windscreen in the flow and cross-correlating its output
with a farfield microphone. However, cross flow in the
microphone inflow introduces errors into this system, and the
dynamic pressure may contaminate the static pressure measure-
ment. Furthermore, the microphone probe may interfere with the
flow and radiate noise itself.
Accordingly, noninvasive optical (or infrared) techniques

were proposed to locate jet noise sources. The first of these

techniques employs the Laser Doppler Velocimeter. To predict
noise with this instrument, however, the three velocity compon-
ents must be measured simultaneously to form the six independ-
ent components of the second time derivative of the stress
tensor. The Laser Velocimeter is not yet capable of measuring
the stress tensor. Meecham and Ford1* and Ribner5 demonstrated
that the Reynolds stress tensor used in the original Lighthill
theory could be replaced by the pressure, which is more easily
measured. Hence, remote optical techniques employing this
parameter have been proposed.
Rudd6 suggested a modification of the shadowgraph tech-

nique originally proposed by Roe7. In this modification, a
large-diameter laser beam passes through the flow and produces
shadow bands at some distance on the other side. These shadow
bands are then analyzed by a mask, and the transmitted light is
collected by a photodetector. In this paper we discuss the use
of the laser shadowgraph technique in identifying the sources
of noise in a jet.
Theory of Noise Generation
In turbulent flow, the primary stresses arise from the

inertia of the fluid. Since the viscous stresses, molecular
heat conduction, and diffusion are negligible, we can consider
the flow to be inviscid and isentropic. The equations of con-
tinuity and momentum, as used by Lighthill, are
(3p/3t) + (3/3x) (pv) - 0 (1)
and
(3/3t) (pvi) + c2(3p/3x) = -(3T /3x.j) (2)

USE OF A LASER SHADOWGRAPH 283
where the instantaneous stress at any point is T.. = pv.v. +

(p-c2p)6.., c is the average speed of sound, and 6.. is the
Kronecker delta. However, if the flow is isentropic, which is
an appropriate assumption for flows of moderate Mach number
which are not too hot, then p-c2p = OCpc^M4), where M is Mach
number of the flow. Therefore, T.. ~ pv.v.. By differentia-
tion and subtraction, we can derive a third equation, i.e.,
32p/3t2 - c 2 V 2 p = 32T.,/3x.3xJ = 32pv.v./3x.3x. (3)

!J ! J ! J 1 J
Equation (3) is equivalent to the wave equation on the

left-hand side and a source term from the fluid stresses on
the right-hand side. Ribner5 separated the pressure fluctua-
(o)
tions p-p0 into an incompressible part p and a compressible
part p . Ribner called the former part "pseudo-sound pres-
sure," i.e., p = p + p . Using Eqs. (1) and (2), we can
obtain
[32(pvv )/3x3x ] - 32p/3t2 = - V2p
Thus, we can set
V2P(0) = - 82(pv.v.)/8xi9xj (4)
and
V 2 p (1) = 32p/3t2 (5)
Using the isentropic relation p = c 2 p, we can write Eq. (3) as
1/c2 (32p/3t2) - V 2 p - 32pviv./3xi3x. (6)
By subtracting Eq. (4) from Eq. (6), we obtain
1/c2 [32P(l)/9t2] - V2p(l) - - 1/c2 [32p(c)/9t2] (7)

which is Ribnerfs so-called "Dilatation equation." Equation
(7) relates the sound-pressure fluctuations p'1' to a generat-
ing pseudo-sound pressure p'0'. The solution of Eq. (7) is
where s is a position outside the flow, q is inside the flow,

and ( } denotes an evaluation at a retarded time, t = t -
284 M. J. RUDD
|s-q|/c. The integration is taken over the volume of the flow.

If we are concerned just with the acoustic far field, the
equation becomes
p(l)(I,t) = - l/4uc2s / [32p(0)/3t2]d3q
Now, since the acoustic pressure p is much less than the

pseudo-sound pressure p , we can substitute pc2 ~ p in-
side the flow.
p d . t ) = ~ 1/4TTS / ( 3 2 p / 3 t 2 ) d 3 q
when p(s,t) is the sound pressure outside the flow, and p is

the density inside.
The quantity p(s,t) is time-varying with a mean value of

zero. Therefore, we prefer to measure a time-independent
quantity, such as the autocorrelation of p(s,t),
2 2 2 2 2
P (,t) * P (I,t) = + i/(4u) J dV J 0 P/3t * 3 p/at )dV
where * denotes the correlation operation, and q f and q" are
pairs of points in the fluid.
Let us now consider the sound radiated in a frequency

band centered at 0) and also by unit volume of turbulence. We
take the Fourier transform in time of the above autocorrelation
function and drop the first volume integral. In addition,
since we are considering a single frequency, the time deriva-
tive can be replaced by U). Then,
P 2 (I,U)) l/167T 2 s 2 / [o) l| p 2 (a))x] dq
where p2(o)) and p2(oa) represent the spectral components of the

far-field pressure and the density fluctuations. Now, the fre-
quency on the right is evaluated in retarded time, and we wish
to measure it in a stationary frame of reference. Therefore,
we multiply it by a factor (1-M cosS)"1, where M is the con-
vection Mach number of the turbulence and 9 is the angle be-
tween the direction of flow and s. Thus,
P 2 (~s,0)) = 03Vl67F 2 s 2 (l-M c cosB)k / p 2 [o)/(l-M c cos6) ] dq
Now, the integral on the right represents the spectral

component at a frequency u) of the autocorrelation function of
the density fluctuations. Let us Integrate this in the three
directions, i.e., x, y, and z, where x is parallel to s, and

y and z are normal to it. Then,
2
r (I,co)
P = co"L L L /16TT2s2(l-M cosG)4 p2[o3/(l-M cos6)] (8)
x y z c c
where L , L , and L are the correlation lengths, and

x y z
2
p [u)/(l-M cos0)] is the spectral component of the density
fluctuations measured at a point. We shall use the laser
shadowgraph to measure this expression for the far-field sound
radiation. Note that the Doppler shift is included in the

factor (1-M cosS)""1 in the spectral term and the convective
amplification in the term (1-M cosG)"4 outside the spectral
term.
Laser Shadowgraph
Description of the Instrument
Roe7 first proposed the shadowgraph as a quantitative

instrument for the measurement of turbulence, and Rudd6 later
suggested the addition of the laser as a light source. Al-
though the shadowgraph has measured flow visualization for many
years, it has been used only as a qualitative instrument to
visualize shock waves or a turbulent wake.
The instrument consists of an arc lamp, or a spark light

source, whose light is collimated to produce a parallel beam of
light that is passed through the flow region and falls on a
screen, or photographic plate, on the other side. A silhouette
of any solid object in the flow is produced. In addition, the
light will be refracted by density variations in the flow to
produce light and dark bands. Astronomers have found that
these shadow bands exist when starlight is refracted by atmos-
pheric turbulence. As the light and dark bands are convected
by the passing wind, the brightness of the star appears to
fluctuate. This is the cause of the twinkling, or scintilla-
tion, of stars at night. Furthermore, these shadow bands are
responsible for the "poor seeing" of stars when they are seen
to quiver in a telescope.
Until 1965, the shadowgraph was a purely qualitative

instrument with an ill-defined relationship between the inten-
sity of the shadow bands and the turbulence which produced
them. Then Townsend8 derived a theory for the production of
shadow bands. He demonstrated that only about 0.2C tempera-
286 M. J. RUDD
ture fluctuations were needed to produce the twinkling of the

stars. The main result is
I 2 ( , m ) = 8TTN 2 z sin2 [ ( 2 +m 2 )R/2N] | $ ( , m ) | 2 (9)
where I2(,m) is the power spectrum of the shadow bands with

wave numbers and m. N is the wave number of the light, z is
the thickness of the turbulence, |$(,m)|2 is the power spec-
trum of the refractive index variations in the turbulence, and
R is the distance from the turbulence to the observer. The
power spectrum is the Fourier transform of the autocorrelation
function of the flow parameter.
In the original derivation, the turbulence was assumed to

be thin; hence, the sensitivity is proportional to the thick-
ness of the turbulence. However, if we are dealing with a jet,
the turbulence is not correlated over the whole light path.
Accordingly, we can replace the thickness z of the turbulence
by the correlation length L z which is defined as the area under
the normalized spatial correlation function in the z direction.
The power spectrum of the shadow bands then becomes
I 2 ( , m ) = 8iTN 2 L sin 2 [ ( 2 -hn 2 )R/N] )<!>(, r a ) | 2 (10)
z
^
Let us consider only wave numbers much less than (R/2N)2.
Then, Eq. (10) becomes
I2(,m) = 2TfR2Lz (2-Hn2)2 |$(,m)|2 (11)
Since the wavelength of light does not appear, this result also
can be obtained by geometrical optics. Similarly, let us inte-
grate over the shadowgraph along the y direction. Then the
spatial spectrum |$(,m)|2 becomes |$(,0)|2, since the inte-
gration acts as a low-pass filter. In turn, this quantity is
equal to the area under the normalized spatial correlation
function in the y direction multiplied by the power spectrum in
the x direction, i.e., L |$()|2. Then, the power spectrum of
the shadow bands in the x direction is
I 2 ( ) = 2^LzL 4|<K)|2 (12)
Now, if we have turbulence with a wave number propagating at

a speed U , then the frequency seen in a stationary frame of
reference will be 0) = U . Therefore,
I 2 (co) - 27TR 2 L z L (u>Vu) |*(o>)f 2 (13)

provided that the convection velocity does not vary too rapidly
with frequency; i.e., U varies slightly over a narrow fre-
quency band. Now, the term $(00) is related to the density
spectrum by a constant, i.e.,
P2(u)) = 1.94 x 107 |$(o>) 2 SI units
Therefore, we can now rewrite Eq. (8) for the generation of

sound as follows:
P 2 (I,U)) = 1.94 x IQ 7 (U^L x /327T 3 R 2 s 2 ) I 2 [co/(l-M c cosO)] (14)

Thus, we can predict the intensity of the far-field sound from

three measurements of U , L , and I2(o)).
Use of the Shadowgraph for Jet Noise Diagnosis
For jet noise diagnosis, we have suggested that a

parallel laser beam be projected through the jet to produce a
shadowgraph at a distance R behind the jet as shown in Fig. 1.
A grating of wave number is placed over the shadowgraph
image at a distance R from the flow. The grating is oriented
normal to the direction of sound propagation of interest. The
light passing through the grating is collected on a photodiode,
and the electrical signal is fed to a spectrum analyzer. The
expected spectrum will peak at a frequency f, have a bandwidth
of Af, and a total power of <>2 (a)). Then the convection veloc-
ity and correlation length are given by U = 2irf/ and
L = U /Af = 27Tf/Af. Then we have
x c
p2[s,27Tf(l-M cosG)] = 1.94 x icf1* (2-rrf /A)5/R2s2Af $2 (CD)
Thus, from these measurements, we can measure the sound radi-
ated in a direction 9 at a particular frequency. To determine
the radiated sound at other frequencies, gratings of different
BEAM
EXPANDER-
-^x
PHOTO-
DIODE
0-
A
LASER
SLIT
NOZZLE
Fig. 1. Laser shadowgraph system.
288 M. J. RUDD
spacings must be employed. Refraction effects cannot be in-

cluded since they occur outside the region where the measure-
ments are made; thus, they can provide sources of error in the
measurement.
The procedure for diagnosing jet noise is as follows:
1) Prepare a range of gratings over the spatial fre-

quency range of interest.
2) Place a grating over the shadowgraph at an angle 6,

and measure signal power center frequency and bandwidth.
3) Calculate radiated sound power.
4) Repeat for different angle 6.
5) Repeat for different grating spacings.
6) Repeat for different positions in the flow.
In principle, the laser shadowgraph could be an absolute

instrument since its sensitivity can be calculated from first
principles. However, as a check, the system should be cali-
brated by a sound wave whose pressure fluctuations can be mea-
sured with a microphone. A sound source, such as a conven-
tional driver producing levels of 120 dB re 0.0002 ybar or
greater, would be suitable.
Broadband Measurements
It might be thought that the use of a slit over the

shadowgraph, instead of a grating, would permit us to make
simultaneous measurements at all spatial frequencies. However,
we would have to measure the correlation length and convection
velocity separately. Since the correlation length and convec-
tion velocity vary with frequency, the labor involved would be
as great or greater than the narrow-band technique proposed
here.
Cross-Correlation Measurements
Some authors have suggested correlating the shadowgraph

output (or, indeed, any other output) in the flow with the far-
field sound. This approach is invadid for two reasons: First,
a Doppler shift exists between frequencies in the flow and the
sound outside. If we try to correlate two phenomena of differ-
ent frequencies, then the result is zero, even if the two are
causally related. Correlation does not imply cause, and, in-

deed, cause does not always imply mathematical correlation.
Second, point measurement cannot be used since sound is gener-
ated from a volume integral.
Conclusions
By three easy measurements, the laser shadowgraph can

diagnose the sound radiated from a jet as a function of angle
and frequency. The noise is analyzed in frequency bands, and
the power, frequency, and bandwidth of each band are measured.
However, we must restrict the quantitative use of the instru-

ment, so that the density fluctuations are small compared with
the Mach number and the phase fluctuations produced in the
laser beam are small compared with IT/2. Then, the shadowgraph
can give meaningful quantitative data and predict the noise
radiation as a function of frequency, angle, and position along
the jet axis. The shadowgraph is inexpensive, it will operate
in hot combustion flows, and it is relatively insensitive to
noise and vibration. However, the theory of sound generation
in hot flows at high Mach numbers must be developed further
before quantitative data can be obtained.
References
, W.T., "Turbulence Measurements Relevant to Jet Noise,"

1966, UTIAS Rept. 119, University of Toronto.
2
Siddon, T.E., "New Correlation Method for the Study of Flow
Noise," Proceedings of 7th International Congress of Acoustics,
Budapest, Vol. 4, August 1971, pp. 533-536.
3
Scharton, R.D. and White, P.H., "The Simple Pressure Source
Model of Jet Noise," The Journal of the Acoustical Society of
America, Vol. 52, July 1972 (Pt. 2), pp. 399-412.
^Meecham, W.C. and Ford, G.W., "Acoustic Radiation from Iso-

tropic Turbulence," The Journal of the Acoustical Society of
America, Vol. 30, April 1958, pp. 318-322.
5
Ribner, H.S., "Aerodynamic Sound from Fluid Dilations A
Theory of the Sound from Jets and Other Flows," UTIA Rept. 86,
1962, University of Toronto, Institute of Aerophysics.
6
Rudd, M.J., "The Velocity Measurement of Phase Objects," Pro-
ceedings of Electro-Optic Systeg^, IP ^J.y jteagurement, pre-
sented at University of "Southampton7" September 25-26, 1972.
290 M. J. RUDD
7
Roe, G.E., "An Optical Study of Turbulence,11 Journal of Fluid
Mechanics, Vol. 43, September 1970, pp. 607-635.
8
Townsend, A.A., "The Interpretation of Stellar Shadow-Bands as
a Consequence of Turbulent Mixing," Quarterly Journal of the
Royal Meteorological Society, Vol. 91, January 1965, pp. 1-6.
EFFECTS OF TRANSDUCER FLUSHNESS ON FLUCTUATING
SURFACE PRESSURE MEASUREMENTS
Richard D. Hanly*
NASA Ames Research Center, Moffett Field, Calif.

Abstract
The installation flushness of pressure transducers has

been found to influence strongly the quality of fluctuating
surface pressure data in supersonic attached turbulent
boundary-layer flow. A systematic investigation of the
effects of flushness on surface pressure fluctuations beneath
a 4-in. (10.16-cm) thick boundary layer was conducted in the
NASA-Ames 9- by 7-ft siupersonic wind tfunnel at Mach numbers of
1.68, 2.0, and 2.5. Flushness effects on transducer signal
amplitude, power spectral density, coherence, and narrow-band
convection velocity were evaluated. Two identical transducers
were installed 0.5 in. (1.27 cm) apart in the wind-tunnel wall
and aligned in the direction of the airstream. Results show
that a very slightly protruding transducer produces adverse
effects on the desired pressure fluctuation measurements,
whereas a submerged transducer produces small effects. Because
of the great difficulties encountered when attempting to
install precisely flush a number of transducers and maintain
that flushness throughout a wind tunnel or flight test, indi-
cations are that a better method for enhancing the uniformity
and spatial correlation of fluctuating surface pressure data
is to submerge transducers deliberately beneath a surface ori-
fice about 0.010 in. (0.0254 cm) deep. It also was found that
as Mach number increases the undesirable effects of nonflush
transducers are alleviated.
Nomenclature
Cn = fluctuating pressure coefficient
"rms
f = frequency, Hz

^Research Scientist.
291
292 R. D. HANLY
G = power spectral density function

M^ = freestream Mach number
q^ = freestream dynamic pressure
TI = upstream transducer, movable
T2 = downstream transducer, flush
U^ = freestream velocity, fps
Uc = convection velocity, fps
y = transducer protrusion or submergence, in. (T^ only)
y = coherence function
6 = boundary-layer thickness, in.
Introduction
Many investigations of dynamic loads on aerospace vehicles

require accurate and reliable measurement of fluctuating sur-
face pressures using transducers mounted in the surfaces of
flight vehicles or wind-tunnel models. It has been found
necessary to investigate factors that affect the quality and
reliability of fluctuating surface pressure measurements
because these measurementsûnder attached turbulent boundary
layers,have revealed inconsistencies that are difficult to
explain. It became apparent from these investigations that
slight surface imperfections in the installation of transducers
are more critical in affecting local surface flow and the
desired measurements than originally was expected.
A systematic investigation of this phenomenon was con-

ducted using two 0.125-in. (0.318-cm) fluctuating pressure
transducers mounted on the wall of the Ames 9- by 7-Foot Wind
Tunnel under a 4-in.-thick boundary layer at Mach numbers of
1.68, 2.0, and 2.5. The effects of transducer flushness are
illustrated by comparing overall root-mean-square fluctuating
pressure coefficients, power spectral densities, coherence
functions, and narrow-band convection velocities for a series
of transducer flushness settings for Mach number 1.68. Data
and discussion for Mach numbers 2.0 and 2.5 are included in
AIAA Paper 75-534.* These show quite clearly that as
Mach number increases, the undesirable effects of nonflush
transducers decrease.
Apparatus and Instrumentation
The test apparatus consisted primarily of a specially

made bolt, Fig. 1, the head of which was flush with the inside
surface of a dummy window in the 9- x 7-ft. tunnel. Two trans-
ducers were mounted in the head of the bolt: one movable out-
ward and inward from the inside tunnel wall surface, and one,
0.500 in. (1.27 cm) downstream, fixed as flush with the surface
EFFECTS OF TRANSDUCER FLUSHNESS 293
O-RING
SEAL
WIND
TUNNEL
WALL
Fig. 1 Installation of movable and flush transducers,
as possible. The movable upstream transducer was mounted on

the end of a shaft that was threaded with a 1/4-40 screw thread
to produce a calibrated axial movement of 0.025 in. (0.064 cm)
for each full turn. A pointer and graduated disk combination
gave good resolution to 0.001 in. of transducer protrusion or
retraction. Special care was exercised in installing the two
transducers as nearly flush as possible; final positioning of
the movable transducer for each data point was accomplished by
always rotating the shaft clockwise to eliminate any possible
hysteresis effects.
The transducer signals were amplified with wide-band d.c.

instrumentation amplifiers and recorded by an FM 14-track mag-
netic tape recorder having a flat frequency response from d.c.
to 20 kHz. The power spectral density, coherence, and convec-
tion velocity measurements, covering a frequency range from
10 Hz to 20 kHz, were obtained using a hybrid analog-digital
computing process with 106 discrete frequency points in six
ranges with varying bandwidths.2
Fluctuating Pressure Coefficients
Fluctuating pressure coefficient C is plotted in

Prms
Fig. 2 for both the movable upstream transducer and the
294 R. D. HANLY
VARIABLE HEIGHT
TRANSDUCER, T1
FLUSH TRANSDUCER, T2
l \ . . /WIND TUNNEL
f:( WALL
BOLT
SHAFT
-.012 .008 .012
Fig. 2 Effects of transducer protrusion and submergence on

overall fluctuating pressure coefficients.
flush downstream transducer ^2 as functions of TI protru-
sion or submergence y, normalized by the boundary layer thick-
ness 6. The most noticeable effect of transducer protrusion,
as y/6 increases from 0 to maximum, is that the signal from
the protruding transducer increases by a factor of 6, while
that from the downstream transducer increases by only a factor
of 2. The airflow separates off the leading edge of the pro-
truding transducer causing a substantial rise in Cp with
increasing protrusion; however, the disturbance is not con-
vected strongly downstream. In contrast, the amplitude of the
Cp for Tj was unchanged over the greater portion of the
submerged range; only between y/6=0 and -0.002 was there a
slight increase in amplitude. This undoubtedly was caused by
the flow reattaching on the slightly submerged transducer
diaphragm after separating over the upstream corner of the
transducer hole. This small airflow disturbance dissipated
very quickly with downstream distance and had no noticeable
effect on the downstream transducer, the Cp amplitude for
T remaining unchanged over the entire submerged range of T^.
A somewhat surprising result is that the minimum Cp

for T]_ did not occur at flush position but rather at
y/6 = +0.0005 for all three Mach numbers. This phenomenon
could possibly have been caused by a very slight misalignment

of the transducer diaphragm with the plane of the tunnel wall,
resulting in a small amount of tilt in different directions as
the transducer was rotated for axial placement. Thus it may
have been that at y/6 = 0.0005 there occurred the most favor-
able angle of diaphragm tilt for minimum response, rather than
at y/6 = 0. Please note, however, the extreme sensitivity of
the Cp to slight variations in y/6 from absolute flush-
ness. From y/6 = +0.0005 to -0.0005 a small discontinuity
even when compared to the thickness of the laminar sublayer of
the boundary layer the C increased from 0.0028 to
Prms
0.0042, an increase of 50%.
Power Spectral Densities
Power spectral density functions (PSDs) for the fluctuat-

ing pressures, as measured by Tj and T2, are shown in
Figs. 3-5. In all cases, nondimensionalized spectral density
function GU00/q006 is plotted as a function of nondimensional-
ized f6/Uoo. Figure 3 depicts the PSDs of the pressure fluc-
tuations at TI at various values of y/6 from flush to full
protrusion and Fig. 4 shows the PSDs at various values of y/6
from flush to full submergence. The overall levels of the PSDs
10i-3r y/6 = 0.01129
y/6 = 0.00451
y/6 = 0.00135
10,-4
n-5
O
FLUSH
y/6 = 0.00090
y/6 = 0.00045
10,-6
= 1.68
= 4.43 in.
10 -7
10 -3 10-2 10,-1 1.0 10
f/U 0
Fig. 3 Effects of transducer protrusion on power spectral
density.
296 R. D. HANLY
10~J r
= 1.68
= 4.43 in.
/y/5 = -0.00045
10 -4
y/5 = -0.00090
10 -5
10 -6
10"
10 -3 10 -2 1.0 10
f5/uO
Fig. 4 Effects of transducer submergence on power spectral
density.
10~3 r
y/5 = 0.0113
10,-4 v/5 = 0.00677
y/5 = 0.00452
/\ V /5 =
0-00226
y/5 = -0.0113 thru

y/5 = +0.00090
1 fl-
IVU = 1.68
5 = 4.43 in.
"
io-33 10~ 1.0 10
Fig. 5 Effects of upstream transducer protrusion and submer-

gence on the flush downstream transducer PSDs.
are consistent with the variation of Ti1 Cn shown in

Prms
Fig. 2. The PSDs show no discernable peaks, indicating that
the random character of the fluctuations beneath the supersonic
turbulent boundary layer was not altered by the T^ misalign-
ment with the surface.
The extremely small protrusions of Tj from the flush

position caused primarily a loss of amplitude of the frequency
components in the lower end of the spectrum, and accounts for
the dip in the Cp curve in Fig. 2. There is no corres-
ponding abrupt change in PSD shape for a slightly submerged T]_
(Fig. 4). When the PSDs for protruding II (Fig. 3) and those
for submerged TI (Fig. 4) are compared, it is evident that the
curves for negative y/6 show much less spread than do those
for positive y/6, as is to be expected from knowledge of the
Cp curves. The submerged PSD shapes, however, when com-
pared to that of the flush position, show a higher degree of
uniformity than do the protruding PSD shapes, indicating that
all frequency components of the signal are affected to nearly
the same extent in the submerged case but not in the protruding
case.
The effects of varying the height of T]_ on the PSDs of

the pressure fluctuations at the downstream transducer are
shown in Fig. 5. A single curve represents all of the PSDs
from T for the entire T]_ y/6 range from maximum submerg-
ence to +0.0009. Only when y/6 was above this value did the
protruding transducer begin to affect the response of the
downstream transducer, and then by raising the T2 PSD ampli-
tude slightly, not by changing the shape of the spectrum.
Coherence
Most current investigations of surface pressure fluctua-

tions beneath turbulent boundary layers include the study of the
the spatial correlation of the pressures. The coherence func-
tion, which is a measure of the extent to which two transducers
sense the same pressure disturbances, commonly is employed for
this purpose. Figure 6 shows the coherence y as a function
of nondimensionalized frequency f6/U00 for the two transducers
T1 and T2 -at various protruding and submerged positions of Tj.
In the case of a protruding Tj, even the smallest incre-

ment of +y/6 caused a significant loss of coherence between
the fluctuating pressures measured by Tj and T2 (Fig. 6).
Because Tj protruded slightly above the surface, a local dis-
turbance was generated by its leading edge. The disturbance,
298 R. D. HAMLY
/FLUSH = 1.68
,y/5 = 0.00045 = 4.43 in.
,y/5 = 0.00090
/y/6 = 0.00226
1.0 r
FLUSH
y/5 = -0.00045
10-3
Fig. 6 Effects of transducer protrusion and submergence on the

coherence function.
sensed strongly by TI but not large enough to propagate to

T2, thereby caused a loss of coherence between the signals from
TI and T2. On the other hand, submerged TI caused only a
gradual degradation of coherence over the entire range of sub-
mergence, thus indicating that both transducers continued to
sense primarily the same fluctuating pressures, namely those of
the attached turbulent boundary layer.
Since it can be assumed that two transducers that are

absolutely flush will result in the highest degree of coherence
measurement at any particular separation distance, the coher-
ence function should reveal whether transducer TI was closer
to the true flush position at y/6 = 0 than it was at the
first increment of protruding y/6 where the lowest Cprms
was measured. Figure 6 reveals that the highest coherence was
obtained, in fact, at y/6 = 0, which then can be considered
to be the closer of the two y/6fs to the true flush position,
regardless of whatever mechanism was responsible for the mini-

mum
"'Prms occurring at y/<5 = +0.0005.
Narrow-Band Convection Velocity
Narrow-band convection velocity, which is the mean speed

at which each frequency component of the pressure fluctua-
tions is convected downstream, like the coherence function, is
frequently of interest in the study of the turbulent boundary-
layer flow field. Convection velocity measurements are signif-
icant only when there is coherence between the pressures sensed
by the transducers. In the lower frequency regions, however,

convection velocity is indeterminate because phase angle is
very near zero regardless of the amount of coherence. When
significant phase angles can be determined from the co- and
quad-spectral densities, convection velocities can be obtained.
Figure 7 shows narrow-band convection velocities (normal-

- 1.68
= 4.43 in.
1.0 r
6 - y/5 = +0.00045 / /
f6/Uo
1.0 p
.8 -
y/5 = -0.00090
.6 -
.4 - y/6 = -0.00226
I _J
10- 10- 1.0 10
IU
I ^00
Fig. 7 Effects of transducer protrusion and submergence on the

narrow-band convection velocity.
300 R. D. HANLY
ized by the freestream velocity UOQ) obtained from Tj and T2

as T]_ was protruded and submerged. Transducer misalignment
with the surface caused the narrow-band convection velocities
to decrease below the flush-case convection velocity wherever
the coherence function was affected strongly (see Fig. 6). The
reduced convection velocities are indicative of the influence
from lower velocity disturbances being introduced to the
boundary-layer turbulence near the surface due to the discon-
tinuity caused by Tj. As with the other functions shown, a
protruding transducer caused much more deviation from the flush
case "normal" convection velocity than did the submerged trans-
ducer.
Summary and Conclusions
The results of this investigation show that it is

extremely difficult to install pressure transducers flush
enough to avoid disturbing the flow field. Then too, the
majority of so-called flush transducer installations on wind
tunnel models and flight vehicles are inherently nonflush
because they are on contoured surfaces, the degree of flush-
ness depending on the radius of curvature of the surface and
the diameter of the transducer. The adverse effects may be
greater on wind tunnel data than on flight data because the
smaller-scale wind-tunnel models have thinner boundary layers
and greater surface curvature for the same size transducers.
The problem of installing transducers absolutely flush to

the surface is greatly compounded when large numbers of trans-
ducers are required for a particular investigation. A recent
test (at Ames) required the installation of 240 fluctuating
pressure transducers in a single model. This type of test
makes it highly desirable that a method of installation be
employed that provides ease of installation yet causes negli-
gible effects on the fluctuating surface pressure data.
The data presented in this paper show that the signals

from a submerged transducer, for a selected range of submer-
gence, were less affected by slight variations of depth (i.e.,
y = -0.010 0.002 in.) than were the signals from a nearly
flush transducer (i.e., y = 0 0.002 in.). Therefore, an
appropriate transducer installation is one in which the trans-
ducers are deliberately submerged with relatively wide toler-
ances to a nominal depth beneath the surface. This technique
allows latitude in installation accuracy while ensuring more
consistent and uniform fluctuating surface pressure data than
from so-called flush transducer installations.
A standard transducer mounting has been adopted at Ames

Research Center that submerges the transducers nominally
0.010 in. (0.0254 cm) beneath a surface orifice that has the
same diameter, 0.047 in. (0.1194 cm), as the active diaphragm
of the transducer model currently used. The surface orifice
is especially useful when curved surfaces are involved because
the orifice can be considerably smaller than the body of the
transducer and the degree of surface curvature has almost no
effect. This technique has an added advantage in that the
smaller diameter hole tends to prevent airflow reattachment on
the diaphragm of a submerged transducer; thus, the slight
amplitude rise observed in the data when the transducer was

submerged just below the surface is eliminated.
This paper has illustrated important effects of transducer

flushness on the measurement of fluctuating surface pressures
for the case of a relatively small transducer installed beneath
a thick supersonic attached turbulent boundary layer. It could
be expected that somewhat different, but still significant,
results would be obtained for transducers having different
sizes relative to the boundary-layer thickness. It has been
found that fluctuating pressure measurements beneath separated
flow have not been affected by transducer flushness to the same
extent as have those beneath attached flow. This would be
expected because the velocity of the separated flow near the
surface is very low and the fluctuating pressure levels are
high relative to any disturbances induced by a lack of trans-
ducer flushness.
The conclusions of this investigation may be summarized as

follows:
1. Protrusion of a transducer above the surrounding sur-

face by as little as 0.00056 can cause large effects on overall
rms level, spectral content, and coherence of surface pressure
fluctuations beneath supersonic attached boundary layers.
2. Submergence of a transducer below the surrounding

surface produces relatively small effects on overall rms level,
spectral content, and coherence.
3. Because of the difficulty of installing and keeping a

transducer precisely flush in a flat or curved surface, instal-
lations should be submerged deliberately, preferably between
-0.0026 and -0.0056.
4. The effects of transducer protrusion and submergence

decrease with increasing supersonic Mach number.1
302 R. D. HANLY
References
R. D., "Effects of Transducer Flushness on Fluc-

tuating Surface Pressure Measurements," AIAA Paper 75-534,
1975, Hampton, Virginia.
2
Lim, R. S., and Cameron, W. D., "Power and Cross-Power
Spectrum Analysis by Hybrid Computer," TM X-1324, 1966, NASA.
PANEL DISCUSSION: AEROACOUSTIC INSTRUMENTATION
P. F. Massier, Jet Propulsion Laboratory: In the field of

experimental aeroacoustics, a large variety of instruments is
being utilized to characterize the inside flow and the noise
radiated outside. These studies are oriented toward advancing
the understanding of noise generating mechanisms and have the

general goal of trying to relate what is going on inside to
the effects outside. The problem of characterizing the noise
sources inside the flow has led us along essentially three
paths: First is the utilization of instruments that are physi-
cally placed within the jet stream to obtain measurements of
the fluctuating quantities and of the mean flow variables.
Second, recognizing that the introduction of physical bodies
into the flow can generate additional noise sources, it per-
haps would be more desirable to make use of noninvasive tech-
niques. Thus some investigators have concentrated on this
type of instrumentation, still with the intent of directly
evaluating the fluctuating quantities. The third is instru-
ments used outside the flow which measure the effect, that is,
the noise, but which are arranged in such a manner that appar-
ent noise source distributions can be inferred from these
measurements.
During this Conference, we were exposed to all of these
methods. It did seem to me, however, that investigators are
now at a point where more discussion can be devoted to results
rather than to the development of the instrumentation that
they used for the purpose of obtaining the results. Perhaps
this is one indication that we have made some progress. I
probably will not mention all of the instruments that were
discussed; however, examples of invasive instrumentation, the
first category mentioned, include a remote-sensing microphone
probe used in hot jets. Also, a 3-dimensional drag sensing
probe, which is based on the use of differential and total
magnetic flux to separate three orthogonal planes of movement
of a sphere, has been developed. This enables one to track
the fluctuating momentum flux, the Reynolds stress, and the
rotational velocity fluctuations as well as their directions.
There are problems of magnetic interference with this kind
of probe. The instruments used outside the flow to determine
apparent noise sources (that is, the third category) include
303
304 DISCUSSION: AEROACOUSTIC INSTRUMENTATION
pairs of microphones with correlations of their signals. Also

the reflecting directional microphone has been used noninvasively.
Then in the other category, which includes noninvasive
methods to obtain fluctuating quantities, first the sodium line
reversal method; here the spectrum containing the D-line is
channeled to one photomultiplier and that of neighboring lines
to a second photomultiplier. The outputs contain both ac and
dc signals, and from these outputs by certain manipulation
both the average temperature and the fluctuating temperature
supposedly can be determined. This was a proposed technique.
There is also a description of the crossed beam laser schliren
method, and experimental results using this method were shown

for a high-temperature supersonic jet. By using the informa-
tion of the density fluctuations obtained with the beams,
together with the experimentally evaluated mean shear, the
outer correlation of the radiated noise was calculated by a
theory that is suitable for Mach wave emission. This theory
is a modification of that developed by Ffowes-Williams and
Maidanik. The calculated noise field agreed well with that
obtained by using microphones outside the jet. A controversial
point arose, not whether the laser schlieren should be used,
but rather whether account must be taken regarding polarization.
Then another author showed analytically that the laser shadow-
graph could be used to obtain quantitative results of the ra-
diated sound in a particular direction and at a particular
frequency. This is not a new concept, but nevertheless it is
refreshing to come to grips with it. The laser velocimeter
also has proven to be a useful technique for obtaining impor-
tant gas-flow data; however, there are complex problems re-
garding its use for various applications. It was perhaps
agreed that information regarding the noise source developed
by an inverse approach, that is, from far-field measurements,
gives useful information; however, it was argued by Ffowcs-
Williams that, if one wants to know something about the source,
it is necessary to get directly to the source.
The following remarks are offered as a summary of some of
the more interesting comments that were made about test tech-
niques. First, the hole-in-the-wall method of determining the
axial distribution of the noise sources in a jet was discussed
again. In this technique, the jet is caused to flow through
an aperture separating the jet noise into two regions, up-
stream and downstream of the aperture; then the noise is
measured in the two regions. Either the jet or the aperture
can be moved axially. There was also a description of the
acoustic wind tunnel which could be used, for example, to
simulate forward speed effects and to study the generation of
DISCUSSION: AEROACOUSTIC INSTRUMENTATION 305
sound for flow interaction with solid bodies. Then in another

paper a highintensity sound source was discussed. This has
application for evaluating performance of absorbing materials
used to line duct walls. The Hartmann generator was suggested
as a method of providing such a source. This device consists
merely of a convergent nozzle delivering a supersonic jet of
air which impinges directly onto the open end of a resonance
tube. Then if the distance between the nozzle and the tube is
chosen correctly, the shock wave which stands between them
becomes unstable in position and responds vigorously to the
acoustic pressure in the resonance tube. There is a resulting
high-intensity sound with a spectrum consisting of a fundamen-
tal tone accompanied by a family of harmonics and an underlying

broadband component. A rod can be inserted between the nozzle
exit and the base of the tube to make the oscillation more
regular.

Index to
Contributors to Volume 46
Abrahamson, A.L. Wyle Leung, K.A., University of'Toron-

Laboratories. . . . . . . . . . . . . . . . 3 t o . . . . . . . . . . . . . . . . . . . . . . .77
Amiet, Roy K., United Tech- Liu, C.H., NASA Langley Re-
nologies Research Center . . . . 259 search Center . . . . . . . . . . . . . . 51
Chow, P.L., Wayne State Univer- Maestrello, L., NASA Langley
sity . . . . . . . . . . . . . . . . . . . . . . 51 Research Center . . . . . . . . . . . . 51
Ellis, N.D., University of Toronto Martlew, D.L., National Gas Tur-
. . . . . . . . . . . . . . . . . . . . . . . . . 77 bine Establishment, United
Emmerling, J.J., General Electric Kingdom. . . . . . . . . . . . . . . . . 203
Company . . . . . . . . . . . . . . . . 177 Mathews, D.C., Pratt & Whitney
Gibson, John S., Lockheed- Aircraft. . . . . . . . . . . . . . . . . . 155
Motsinger, R.E., General Electric
Georgia Company. . . . . . . . . . 139

Hanly, Richard D., NASA Ames company. . . . . . . . . . . . . . . . . 177
Research Center . . . . . . . . . . . 291 Nagel, R.T., Pratt & Whitney Air-
Hay den, R.E., Bolt Beranek and craft . . . . . . . . . . . . . . . . . . . . 155
Newman Inc . . . . . . . . . . . . . . 247 Oncley, Paul B., MAN-Acoustics
Healy, Gerald J., Lockheed- and Noise, Inc. . . . . . . . . . . . . 123
California Company........ 139 Raney, John P., NASA Langley
Huang, Ming N a n , Wyle Research Center . . . . . . . . . . . . 93
Laboratories. . . . . . . . . . . . . . . 35 Revell, James D., Lockheed-
K a dm an. Y., Bolt Beranek and California Company........ 139
Newman Inc . . . . . . . . . . . . . . 247 Ribner, H.S., University of'Toron-
Kantola, R.A., General Electric to . . . . . . . . . . . . . . . . . . . . . . . 77
Company . . . . . . . . . . . . . . . . 223 Rudd, Michael J., Bolt Beranek
Kester, J.D., Pratt & Whitney Air- and Newman Inc. . . . . . . . . . . 281
craft . . . . . . . . . . . . . . . . . . . . 155 Tubb, P.E., University of Toronto
Kurts, D., University of Toronto . . . . . . . . . . . . . . . . . . . . . . . . . 17
. . . . . . . . . . . . . . . . . . . . . . . . . 77 Wenzel, Alan R., NASA Ames
Lam, F., University of Toronto Research Center . . . . . . . . . . . . 67
. . . . . . . . . . . . . . . . . . . . . . . . . 77 Zwieback, E.L., Douglas Aircraft
Company . . . . . . . . . . . . . . . . 103

Aeroacoustics Acoustic Wave Propagation - Aircraft Noise Prediction - Aeroacoustic Instrumentation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aeroacoustics Acoustic Wave Propagation - Aircraft Noise Prediction - Aeroacoustic Instrumentation

Uploaded by

Copyright:

Available Formats

Downloaded by Boeing Company on July 15, 2013 | http://arc.aiaa.org | DOI: 10.2514/4.

Martin Summer-field, Series Editor-in-Chief

Technical papers from AIAA 2nd Aero-Acoustics Conference,

Published by the American Institute of Aeronautics and

American Institute of Aeronautics and Astronautics

The MIT Press

Library of Congress Cataloging in Publication Data

AIAA Aero-acoustics Specialists Conference, 2d,

(Progress in astronautics and aeronautics; v. 4346)

All rights reserved. No part of this book may be reproduced in any

Conference and Technical Program Committees xxxiv

Chapter I Acoustic Wave Propagation through the Atmosphere 1

Propagation of Aircraft Noise over Long Distances

Measured Effects of Turbulence on the Rise

Sound Scattering from Atmospheric Turbulence 35

Scattering of Coherent Sound Waves by

Saturation Effects Associated with Sound

Computer Model of the Lightnings-Thunder

Panel Discussion: Acoustic Wave Propagation

Chapter II Aircraft Noise Prediction 91

Development of a New Computer System for

Aircraft Flyover Noise Measurements 103

Review of Theory and Methods for the Prediction

Methods for the Prediction of Airframe

Review of Theory and Methods for

Panel Discussion: Aircraft Noise Prediction 195

Chapter III Aeroacoustic Instrumentation 201

The Use of Hartmann Generators as Sources of

Outdoor Jet Noise Facility: A Unique Approach 223

Factors in the Design and Performance of Free-Jet

Acoustic Wind Tunnels 247

Correction of Open-Jet Wind-Tunnel Measurements

Use of a Laser Shadowgraph for Jet

Effects of Transducer Flushness on Fluctuating

Panel Discussion: Aeroacoustic Instrumentation 303

Index to Contributors to Volume 46 307

Conference and Technical Program Committees xxxiv

Chapter I Jet Noise 1

Tests of a Theoretical Model of Subsonic Jet Noise 3

New Evidence of Subsonic Jet Noise Mechanisms 27

Bernoulli Enthalpy: A Fundamental Concept

A Potential Flow Model for Calculation of Jet Noise 91

Jet Noise Source Location by Cross-Correlation

An Experimental Study of the Structure and Acoustic

Measurement and Prediction of Jet Noise in Flight 137

Jet Noise Modeling: Theoretical Predictions and

Sound Pressures of a Choked Jet Oscillating

An Experimental Study of Shock-Free

Periodic "Mach Wave" Radiations from a Small

Ambient and Induced Pressure Fluctuations in

Density Fluctuations and Radiated Noise for a

Acoustic Far Field of a Point Source in Parallel

Supersonic Jet Noise Suppression with

Minimization of Jet and Core Noise of a Turbojet

Jet and Suppressor Correlation Measurements 399

Cross-Correlation of Noise Produced inside a Hot

The Effects of Forward Speed on a Number of

Turbojet Exhaust Silencers 439

Panel Discussion: Jet Noise 458

Chapter II Combustions and Core Engine Noise 465

Convergence of Theory and Experiment in Direct

Further Experimental Results on the Structure

Combustion Intensity and Distribution Relation