(Charles A. Poynton) A Technical Introduction To D PDF

Charles Poynton
Charles Poynton
I am an independent contractor specializing in the physics, mathematics,
and engineering of digital color imaging systems, including digital video,
HDTV, and digital cinema (D-cinema). I do technology forecasting,
systems modelling, algorithm development (including digital filter design),
video signal processing architecture, color characterization and calibration,
and image quality assessment. (More...)
tel: +1 416 413 1377
e-mail: concatenate surname at surname dot com (poynton@poynton.com)
[email protocol]
www.poynton.com
What's new?
Digital Video and

HDTV
Courses, seminars
&c.
Color technology
Video engineering
Digital Signal
Processing
Typography and
design
Hire me!
to 2004-12-09. I'm back from London. I'll present a half-day tutorial HDTV
and Digital Cinema Camera Technology at the HPA Tech Retreat in Palm
Springs, on January 25, 2005.
My book Digital Video and HDTV Algorithms and Interfaces is holding
fairly steady between the 10,000-th and 20,000-th most popular book at
Amazon.com. The Table of Contents is available; the Errata were updated
on 2004-10-19. Today, Amazon indicates that DVAI is #1 (the most
popular book purchase) at Texas Instruments; it's #18 in San Jose, Calif.!
My partner Barbara finds all of this quite frightening.
Upcoming (and past) events.
Includes Frequently Asked Questions(FAQs) about Gamma and Color.
Information concerning technical aspects of video.
Where to find digital filter design packages.
Articles I've written concerning typography, information design, and
presentation in the digital world. Also, archaic information is available
concerning making web pages usable, "This site is best experienced";
archaic information concerning the FrameMaker publication system is also
available.
I'm an independent contractor.
Personal stuff
& biographical data
http://www.poynton.com/ (1 di 2) [15/01/2005 16:22:32]
Charles Poynton
Someone once asked Peggy Lee who she thought was the best jazz singer.
Her answer was, "You mean besides Ella?"
Charles Poynton
Copyright 2005-01-13
http://www.poynton.com/ (2 di 2) [15/01/2005 16:22:32]
Momaku
Momaku
Calligraphy by Judth Dowling of the Japanese kanji for retina: Momaku.
Mo, the upper character, means "net". Maku, the lower character, means
"membrane."
This is adapted from the title page of The Retina: an approachable part of
the brain, by John E. Dowling (Cambridge, Mass.: Belknap Press of
Harvard University Press, 1987). Highly recommended!
Mr. Ogawa of Toshiba pointed out to me, via Greg De Priest, that Judith
Dowling's character is in old-style Kanji. I have added the horizontal bar at
the top of the lower-left character, so as to present the modern character.
Charles Poynton
1998-03-19
http://www.poynton.com/notes/misc/Momaku.html [15/01/2005 16:22:34]
Charles Poynton - Freelance
Charles Poynton Independent contractor
I am an independent contractor - all me a consultant if you must - specializing in the physics,

mathematics, and engineering of digital color imaging systems, including digital video, HDTV, and
digital cinema (D-cinema). I do technology forecasting, systems modelling, algorithm development
(including digital filter design), video signal processing architecture, color characterization and
calibration, and image quality assessment. I live and work in Toronto.
For about a decade, I have been using Mathematica to do analysis, design, modelling, and simulation
of signal, color, and video processing systems. I have recently added MATLAB to my toolkit,
including Simulink, the DSP Blockset, the Image Processing Toolbox, the Optimization Toolbox, and
the Signal Processing Toolbox.
I pay a great deal of attention to communicating my work, both to my clients (by writing technical
reports, proposals, analysis documents, and the like), and to the wider community (by teaching
courses and seminars and by writing books). I execute my own illustrations (using Adobe Illustrator),
and typeset my own work (using Adobe FrameMaker).
I have been working primarily in these areas:
Analyzing color specification, calibration, capture, processing, storage, and display for digital
cinema, computer animation, computer generated imagery (CGI), and broadcast, industrial,
and consumer video.
For two different manufacturers of emerging displays for digital cinema,
I consulted on color science, helped establish image coding standards, and
assisted in the development, testing, and evaluation of signal processing
algorithms.
I have consulted to several film studios on the establishment of image coding
standards, and helped to introduce color characterization, color calibration, and
color management into their production pipelines.
I have an extensive implementation, in Mathematica, of code for color image
encoding, decoding, matrixing, and processing. You can obtain more
information.
Developing, characterizing, modeling, and analyzing algorithms for motion image
http://www.poynton.com/notes/bio/consulting.html (1 di 3) [15/01/2005 16:22:43]
processing at quality levels from digital cinema, through studio broadcast, down to PCs, with
implementation technologies from high-level software, through microcode, to commerciallyavailable VLSI. Generally I perform high-level architectural work, and deliver algorithms as
some combination of textual description, equations, and/or Mathematica code. Sometimes
I provide bit-accurate C-code as well. My clients typically implement these algorithms
themselves in VHDL or Verilog. (I read these languages, but don't [yet] write them.)
For a startup company developing a revolutionary new film scanning
technology, I assisted in color characterization of their highly unusual device,
and in the development of unique image signal processing algorithms.
For a manufacturer of graphics accelerator chips, I designed filtering
algorithms associated with NTSC decoding and encoding, deinterlacing,
reinterlacing, resizing, and frame rate conversion.
For a startup company manufacturing very large scale full-color LED display
systems, I designed the color signal processing architecture including uniformity
correction and color matrixing; great attention was paid to motion artifacts
associated with the interaction between eye tracking and pulse-width modulation
(PWM).
Though this is somewhat old hat, I have an implementation, in Mathematica, of
bit-accurate models of NTSC and PAL encoding and decoding. You can obtain
more information.
Technology assessment and forecasting. I assess the value of intellectual property, to assist
companies to exercise due diligence with respect to licensing or acquisition. I assess how
technological developments in digital video, compression, HDTV, and accurate color
technology are likely to affect companies. Sometimes, this verges into product planning.
For a company that designs and manufactures standard (commercial) integrated
circuits, I consulted on the technological aspects of a business strategy to enter
the domain of consumer electronics.
For a large manufacturer of studio video equipment, I assisted in the evaluation
of the technology of a company that was a candidate for acquisition.
Expert witness in patent litigation. I assess and evaluate patents and patent portfolios, and
occasionally I act as an expert witness in patent litigation.
For a company that designs and manufactures video equipment, I examined the
claims of an inventor that claimed that his patent had been infringed. I gave
a deposition in which I cast doubt on the validity of the claims; this resulted in
an out-of-court settlement.
Writing. I have written several white papers - some attributed, some published anonymously on various topics within my expertise.
For Discreet Logic, a developer of large-scale application software for the
creation of digital media, I wrote a white paper concerning conversion between
R'G'B' (4:4:4) and Y'CBCR (4:2:2) video coding systems. A competitor was
making suspect claims about their own video coding, and implied that Discreet's
R'G'B' system was inferior. Discreet needed an authoritative voice to politely
refute these claims in public.
Teaching courses and seminars. In addition to the many public tutorials, courses, and
seminars that I have presented over the last decade, I have organized and presented many inhouse events for large semiconductor manufacturers, film studios, manufacturers of camera
and display systems, and system and application software companies. Most of my teaching
gigs provide for "open" consulting time for individual contributors and design teams.
Many of the companies that I work for prefer that fact not to be made public. So I cannot tell you
here - or perhaps even at all - who they are.
Sometimes, I charge a per diem rate; sometimes I take fixed-price contracts. Sometimes I perform
contracts without leaving my office in Toronto; sometimes I travel to my clients' facilities. Sometimes
I consult over the telephone, for an hourly rate. If you're interested in having me consult for your
organization, telephone me and we'll discuss your needs, my references, and my rates.
If you're considering hiring me, perhaps you're interested in my medium and long-term goals.
Charles
2005-01-11
Charles Poynton - Courses & seminars (upcoming)
Charles Poynton Courses & seminars (upcoming)
I'll present a half-day tutorial HDTV and Digital Cinema

Camera Technology at the HPA Tech Retreat in Palm Springs.
The seminar takes place in the afternoon of Tuesday,
January 25, 2005; the retreat proper takes place Wednesday
through Friday.
You can review past public courses and seminars that I have
presented, in the 1990s, 2000, 2001, 2002, 2003, 2004, and
2005.
In addition to open, public courses such as the ones listed
above, I frequently develop and teach on-site courses or
seminars for commercial organizations. I have taught courses
for film and animation studios, workstation and PC
manufacturers, companies that design and manufacture
integrated circuits, and companies that make display systems
and subsystems. I have presented courses on digital video,
color science, color management, and other topics. You can
review the syllabus for the 1-day, 2-day, and 3-day versions of
the Digital Video Technology course.
When I teach a course on digital video, I usually arrange for
each participant to be given a copy of my book, and I also
distribute customized course note handouts. If you're
interested in having me teach at your organization, telephone
me, and we'll discuss your needs, my references, and my rates.
I may also be available for consulting (much as I hate the
word).
Charles Poynton
2005-01-11
http://www.poynton.com/notes/events/index.html [15/01/2005 16:22:53]
HDTV and Digital Cinema Camera Technology
Palm Springs, Calif.

Thursday, Jan. 25, 2005, 13h00-17h00
HDTV and Digital Cinema

Camera Technology
Instructor: Charles Poynton
Duration: 1/2 day
In association with 2005 Tech Retreat
Synopsis:
The 3CCD "beamsplitter" camera has ruled video for about a quarter of a century. This is set to
change - single-sensor "mosaic" cameras have already been announced by at least three companies,
and more are surely on the way. In this 1/2-day seminar, Charles Poynton will start by reviewing
optics and lens design for HDTV and digital cinema cameras. He will then outline the task of color
separation with both the tried-and-true prism beamsplitter and emergent color filter array (CFA)
technology that originated in the digital still camera world. He will describe the "demosaicking"
algorithms necessary to reconstruct color in these cameras. He will then discuss the optics, physics,
and electronics of CCD and CMOS image sensors themselves, paying particular attention to the
sources and treatment of noise. He will conclude by discussing the emergent technologies of wide
color gamut and high dynamic range imaging. Much of the material that he will present constitutes
"work in progress" for his next book.
Audience: This seminar is appropriate for technical professionals who are experienced in creating and
manipulating color imagery for SDTV, HDTV, or digital cinema. It is also suitable for programmers
and engineers.
Materials provided: Course handouts will be provided. Portions of the seminar will be based upon
Charles Poynton's book Digital Video and HDTV Algorithms and Interfaces, copies of which will be
available for purchase.
Registration: TBD, Approx. USD 150, including lunch. Register through HPA; contact Hollywood
Post Alliance Executive Director, Eileen Kramer at +1 213 614 0860.
Charles Poynton - Courses & seminars
www.poynton.com/notes/events/
2004-12-09
http://www.poynton.com/notes/events/20050125_PSP_HPA.html [15/01/2005 16:22:55]
Digital Video and HDTV Algorithms and Interfaces
Digital Video and HDTV

Algorithms and Interfaces
by Charles Poynton,
(San Francisco: Morgan Kaufmann Publishers, 2003)
[hardcover, 736 pages, USD 59.94].
available from the publisher, online retailers, and bookstores.
Placing video in the context of computing

Rapidly evolving computer and communications technologies have achieved data transmission rates
and data storage capacities high enough for digital video. But video involves much more than just
pushing bits! Achieving the best possible image quality, accurate color, and smooth motion requires
understanding many aspects of image acquisition, coding, processing, and display that are outside the
usual realm of computer graphics. At the same time, video system designers are facing new demands
to interface with film and computer system that require techniques outside conventional video
engineering.
Charles Poynton's 1996 book A Technical Introduction to Digital Video became an industry favorite
for its succinct, accurate, and accessible treatment of standard definition television (SDTV). In Digital
Video and HDTV, Poynton covers not only SDTV, but also high definition television (HDTV) and
compression systems. With the help of hundreds of high quality technical illustrations, this book
presents the following topics:
Basic concepts of digitization, sampling, quantization, gamma, and filtering

Principles of color science as applied to image capture and display
Scanning and coding of SDTV and HDTV
Video color coding: luma, chroma (4:2:2 component video, 4fSC composite video)
Analog NTSC and PAL
Studio systems and interfaces
Compression technology, including M-JPEG and MPEG-2
Broadcast standards and consumer video equipment
CHARLES POYNTON is an independent contractor specializing in the physics, mathematics, and

engineering of digital color imaging systems, including digital video, HDTV, and digital cinema (Dcinema). He designed and built the digital video equipment used at NASA to convert video from the
Space Shuttle into NTSC, initiated Sun Microsystems' HDTV research project in the early 1990s, and
has taught many popular courses on HDTV and video technologies. A Fellow of the Society of
Motion Picture and Television Engineers (SMPTE), Poynton was awarded the Society's prestigious
David Sarnoff Gold Medal for his work to integrate video technology with computing and
http://www.poynton.com/DVAI/index.html (1 di 2) [15/01/2005 16:23:03]
communications.
The Table of Contents is available. Errata are available.
Sample chapters will soon be available here.
You can order from the publisher, or Amazon, or Barnes & Noble.
2003-09-25b
Charles Poynton
http://www.poynton.com/DVAI/index.html (2 di 2) [15/01/2005 16:23:03]
A Technical Introduction to Digital Video
by Charles Poynton, published by John Wiley & Sons, 1996 (ISBN 0-471-12253-X, hardcover, USD
49.99).
Computers and communication systems have now reached the stage where is it possible to have
photographic-quality color pictures. But smooth motion and accurate color, though easy to achieve in
video equipment, remain beyond the reach of general purpose computers. This book will help
computer system designers, engineers, programmers and technicians to learn the techniques of digital
video, to bring smooth motion and accurate color to computing. If you are a television professional,
this book will help you to understand the technology at the core of digital video.
The book was published in 1996, and reached fifth printing. The superseding edition, Digital Video
and HDTV Algorithms and Interfaces was published in January, 2003, by Morgan Kaufmann.
Because the superseding edition is on the streets, A Technical Introduction to Digital Video will soon
be of print.
If you search on "Poynton," you'll encounter Henry James' book The Spoils of Poynton - that is what
a former boss of mine, Hugh Lawford, used to call my computer programs!
Send e-mail if you have any corrections or suggestions!
Table of Contents
The Table of Contents is available online.
http://www.poynton.com/notes/TIDV/index.html (1 di 2) [15/01/2005 16:23:11]
Sample chapters
Two chapters are available online, in typographic-quality Acrobat PDF format:
Chapter 1, Basic Principles Acrobat PDF format

Chapter 6, Gamma Acrobat PDF format
Errata
Known errors are listed in the Errata. If you discover an error that is not listed, please report it to me
by e-mail.
2003-09-25
Charles Poynton
http://www.poynton.com/notes/TIDV/index.html (2 di 2) [15/01/2005 16:23:11]
A Technical Introduction to Digital Video - Contents
Contents
Acknowledgments
List of figures
List of tables
Preface
vii
xix
xxv
xxvii
Formulas xxviii
Luma vs Luminance xxviii
CCIR vs ITU-R xxviii
Standards xxix
Layout and typography xxix
Further reading xxx
Basic principles
Imaging 1
Digitization 2
Pixel array 3
Spatiotemporal domains 4
Scanning notation 4
Viewing distance and angle 5
Aspect ratio 6
Frame rate, refresh rate 6
Motion portrayal 8
Raster scanning 8
Interlace 11
Scanning standards 12
Sync structure 13
http://www.poynton.com/notes/TIDV/TIDV_toc.html (1 di 8) [15/01/2005 16:23:21]
Data rate 15
Data rate of digital video 15
Linearity 16
Perceptual uniformity 17
Noise, signal, sensitivity 17
Quantization 18
Frequency response, bandwidth 19
Bandwidth and data rate 20
Resolution 21
Resolution in film 21
Resolution in television 22
Resolution in computer graphics 22
Luma 23
The unfortunate term "video luminance" 24
Color difference coding 24
Component digital video, 4:2:2 26
Composite video 27
Composite digital video, 4fSC 28
Analog interface 28
High-definition television, HDTV 29
2 Raster images in computing
33
Introduction 34
Symbolic image description 34
Raster images 35
Dithering 38
Conversion among types 39
Data compression 40
Image compression 40
Lossy compression 40
3 Filtering and sampling
43
Introduction 43
Sampling theorem 44
Sampling at exactly 0.5fS 45
Frequency response 48
Frequency response of a boxcar 49
Frequency response of point sampling 50
Fourier transform pairs 51
Digital filters 53
Impulse response 54
Finite impulse response (FIR) filters 54
Physical realizability of a filter 55
Phase response (group delay) 56

Infinite Impulse Response (IIR) filters 57
Lowpass filter 57
Digital filter design 60
Reconstruction 62
Reconstruction close to 0.5fS 62
(sin x)/x correction 63
Further reading 65
4 Image digitization and reconstruction
67
Comb filtering 68
Frequency spectrum of NTSC 69
Spatial frequency domain 72
Image sampling in computing 76
Image reconstruction 76
Spot size 78
Transition samples 79
Picture center and width 79
5 Luminance and lightness
81
Radiance, intensity 82
Luminance 82
Luminance from red, green, and blue 83
Adaptation 85
Lightness sensitivity 85
Lightness, CIE L* 88
Linear and nonlinear processing 89
6 Gamma
91
Gamma in physics 92
The amazing coincidence! 94
Gamma in film 96
Surround effect 99
Gamma in video 100
Rec. 709 transfer function 102
SMPTE 240M transfer function 103
CRT transfer function details 104
Gamma in computer graphics 107
Gamma in video, computer graphics, SGI, and Macintosh 108
Pseudocolor 110
Halftoning 110
Printing 111
Limitations of 8-bit intensity 113

The future of gamma correction 114
7 Color science for video
115
Fundamentals of vision 116

Color specification 117
Color image coding 118
Definitions 119
Spectral power distribution (SPD) and tristimulus 119
Scanner spectral constraints 120
CIE XYZ tristimulus 122
CIE (x, y) chromaticity 125
Color temperature 127
White 129
Perceptually uniform color spaces 130
Additive mixture, RGB 132
Characterization of RGB primaries 133
NTSC primaries (obsolete) 134
EBU primaries 134
SMPTE RP 145 primaries 136
Rec. 709 primaries 136
CMFs and SPDs 136
Gamut 146
Noise due to matrixing 147
Transformations between RGB and CIE XYZ 147
Transforms among RGB systems 148
Camera white reference 149
Monitor white reference 149
Wide gamut reproduction 150
Scanning colored media 150
"Subtractive" mixture, CMY 151
Unwanted absorptions 153
Further reading 154
8 Luma and color differences
155
Color acuity 155

RGB and Y, B-Y, R-Y color cubes 157
Constant luminance 158
Conventional luma/color difference coding 162
Nonlinear Red, Green, Blue (R'G'B') 164
Rec. 601 luma 165
SMPTE 240M luma 166
Rec. 709 luma 166
Errors due to nonconstant luminance 166
Subsampling 168
Luma/color difference summary 168
9 Component video color coding
171
B'-Y', R'-Y' components 172

PBPR components 173
CBCR components 174
Y'CBCR from computer RGB 177
Kodak PhotoYCC 178
Y'UV, Y'IQ confusion 179
UV components 180
IQ components 182
Color coding standards 183
10 Composite NTSC and PAL
185
Subcarrier regeneration 186

Quadrature modulation 187
Decoder and monitor controls 189
Narrowband I chroma 190
Frequency interleaving 191
PAL encoding 192
PAL-M, PAL-N 194
Incoherent subcarrier 194
Analog videotape recording 195
NTSC-4.43 197
SECAM 197
11 Field, frame, line, and sample rates
199
Field rate 199

Line rate 200
Sound subcarrier 200
Addition of composite color 200
NTSC color subcarrier 201
PAL color subcarrier 202
4fSC sampling 203
Common sampling rate 204
Genlock 205
12 525/59.94 scanning and sync
207
Frame rate 207

Interlace 207
Line sync 210
Field/frame sync 211
Sync distribution 213
Picture center, aspect ratio, and blanking 213
Halfline blanking 214
13 525/59.94 component video
215
RGB primary components 215

Nonlinear transfer function 215
Luma, Y' 216
Component digital 4:2:2 interface 216
Component analog R'G'B' interface 217
Component analog Y'PBPR interface, SMPTE 219
Component analog Y'PBPR interface, industry standard
14 525/59.94 NTSC composite video
221
Subcarrier 221
Two-frame sequence 222
Burst 222
Color differences, U, V 223
Color difference filtering 223
Chroma, C 223
Setup 224
S-video-525, Y' / C 3.58 225
Composite NTSC encoding 225
Composite digital NTSC interface, 4fSC 226
Composite analog NTSC interface 227
15 625/50 scanning and sync
229
Frame rate 229

Interlace 229
Line sync 232
Field/frame sync 232
Sync distribution 233
Aspect ratio 235
16 625/50 component video
237
RGB primary components 237

Nonlinear transfer function 237
Luma, Y' 238
220
Component digital 4:2:2 interface 238

Component analog R'G'B' interface 238
Component analogY'PBPR interface 239
17 625/50 PAL composite video
241
Subcarrier 241
Four-frame sequence 241
Burst 242
Color difference components, U, V 242
Color difference filtering 243
Chroma, C 243
S-video-625, Y'/C 4.43 243
Composite PAL encoding 244
Composite digital PAL interface, 4fSC 244
Composite analog PAL interface 245
18 Electrical and mechanical interfaces
247
Analog electrical interface 247

Analog mechanical interface 247
Parallel digital interface 248
Serial digital interface 249
Fiber optic interfaces 250
19 Broadcast standards
251
ITU-R, former CCIR 252

ITU-R scanning nomenclature 252
M/NTSC (NTSC) 253
M/PAL (PAL-M, PAL-525) 253
N/PAL (PAL-N, PAL-3.58) 253
B,G,H,I/PAL (PAL) 254
D,K/SECAM (SECAM) 254
Summary of parameters 254
20 Test signals
257
Colorbars 257
Frequency response 258
Differential gain, DG 258
Differential phase, DP 260
Pulse signals 260
Modulated 12.5T, 20T pulses 262
21 Timecode
265
Introduction 265
Dropframe timecode 266
Longitudinal timecode, LTC 267
Vertical interval timecode, VITC 268
Editing 269
Flag bits 270
Further reading 270
Appendix
A Glossary of video signal terms
Index
273
301
1997-06-23
Charles Poynton - A Technical Introduction to Digital Video
Basic principles
This is Chapter 1 of the book
A Technical Introduction to
Digital Video, by Charles
Poynton. Copyright 1996
John Wiley & Sons.
This chapter is a summary of the fundamental concepts

of digital video.
If you are unfamiliar with video, this chapter will introduce the major issues, to acquaint you with the
framework and nomenclature that you will need to
address the rest of the book. If you are already knowledgeable about video, this chapter will provide a quick
refresher, and will direct you to specific topics about
which youd like to learn more.
Imaging
The three-dimensional world is imaged by the lens of
the human eye onto the retina, which is populated with
photoreceptor cells that respond to light having wavelengths in the range of about 400 nm to 700 nm. In an
imaging system, we build a camera having a lens and a
photosensitive device, to mimic how the world is
perceived by vision.
Although the shape of the retina is roughly a section of
a sphere, it is topologically two-dimensional. In a
camera, for practical reasons, we employ a flat image
plane, sketched in Figure 1.1 overleaf, instead of a
spherical image surface. Image system theory concerns
analyzing the continuous distribution of power that is
incident on the image plane.
A photographic camera has, in the image plane, film
that is subject to chemical change when irradiated by
Figure 1.1 Scene, lens, image plane.
light. The active ingredient of photographic film is

contained in a thin layer of particles having carefully
controlled size and shape, in a pattern with no coherent
structure. If the particles are sufficiently dense, an
image can be reproduced that has sufficient information for a human observer to get a strong sense of the
original scene. The finer the particles and the more
densely they are arranged in the film medium, the
higher will be the capability of the film to record spatial
detail.
Digitization
Signals captured from the physical world are translated
into digital form by digitization, which involves two
processes. A signal is digitized when it is subjected to
both sampling and quantization, in either order. When
an audio signal is sampled, the single dimension of time
is carved into discrete intervals. When an image is
sampled, two-dimensional space is partitioned into
small, discrete regions. Quantization assigns an integer
to the amplitude of the signal in each interval or region.
1-D sampling
A signal that is a continuous one-dimensional function

of time, such as an audio signal, is sampled through
forming a series of discrete values, each of which represents the signal at an instant of time. Uniform
sampling, where the time intervals are of equal duration, is ubiquitous.
2-D sampling
A continuous two-dimensional function of space is

sampled by assigning, to each element of a sampling
A TECHNICAL INTRODUCTION TO DIGITAL VIDEO
grid, a value that is a function of the distribution of

intensity over a small region of space. In digital video
and in conventional image processing, the samples lie
on a regular, rectangular grid.
Samples need not be digital: A CCD camera is inherently sampled, but it is not inherently quantized.
Analog video is not sampled horizontally, but is
sampled vertically by scanning, and sampled temporally at the frame rate.
Pixel array
A digital image is represented by a matrix of values,
where each value is a function of the information
surrounding the corresponding point in the image.
A single element in an image matrix is a picture
element, or pixel. In a color system, a pixel includes
information for all color components. Several common
formats are sketched in Figure 1.2 below.
In computing it is conventional to use a sampling grid
having equal horizontal and vertical sample pitch
square pixels. The term square refers to the sample
pitch; it should not be taken to imply that image information associated with the pixel is distributed uniformly
throughout a square region. Many video systems use
sampling grids where the horizontal and vertical sample
pitch are not equal.
2
35
64
80
52 280
1
11
20
19
SIF,
82 Kpx
0
24
48
0
60
0
Video, 300 Kpx

PC/Mac, 1 2 Mpx
High-Definition Television (HDTV), 1 Mpx
72
Workstation, 1 Mpx
0
90
High-Definition Television (HDTV), 2 Mpx
Figure 1.2 Pixel array.
CHAPTER 1
8
10
BASIC PRINCIPLES
H
O
(T R
R IZ
A O
N N
SV T
ERAL
SE
)
TEMPORAL
VERTICAL
AL
TI
A
SP
Figure 1.3 Spatiotemporal domains.
Some framebuffers provide a

fourth byte, which may be
unused, or used to convey
overlay or transparency data.
In computing it is usual to represent a grayscale or

pseudocolor pixel as a single 8-bit byte. It is common
to represent a truecolor pixel as three 8-bit red, green,
and blue (RGB ) components totaling three bytes
24 bits per pixel.
Spatiotemporal domains
A digital video image is sampled in the horizontal,
vertical, and temporal axes, as indicated in Figure 1.3
above. One-dimensional sampling theory applies along
each of these axes. At the right is a portion of the twodimensional spatial domain of a single image. Some
spatial processing operations cannot be separated into
horizontal and vertical facets.
Scanning notation
In computing, a display is described by the count of
pixels across the width and height of the image.
Conventional television would be denoted 644 483,
which indicates 483 picture lines. But any display
system involves some scanning overhead, so the total
number of lines in the raster of conventional video is
necessarily greater than 483.
4
Video scanning systems have traditionally been

denoted by their total number of lines including sync
and blanking overhead, the frame rate in hertz, and an
indication of interlace (2:1) or progressive (1:1) scan, to
be introduced on page 11.
525/59.94/2:1 scanning
is used in North America and Japan, with an analog

bandwidth for studio video of about 5.5 MHz.
625/50/2:1 scanning
is used in Europe and Asia, with an analog bandwidth

for studio video of about 6.5 MHz. For both 525/59.94
and 625/50 component digital video according to
ITU-R Rec. BT.601-4 (Rec. 601), the basic sampling
rate is exactly 13.5 MHz. Bandwidth and sampling rate
will be explained in later sections.
1125/60/2:1 scanning
is in use for high-definition television (HDTV), with an

analog bandwidth of about 30 MHz. The basic
sampling rate for 1125/60 is 74.25 MHz. A variant
1125/59.94/2:1 is in use. This scanning system was
originally standardized with a 1920 1035 image
having pixels about 4 percent taller than square.
1920 1080
The square-pixel version of 1125/60 is now commonly

referred to as 1920 1080.
1280 720
A progressive-scan one megapixel image format is

proposed for advanced television in the United States.
Viewing distance and angle

A viewer tends to position himself or herself relative to
a scene so that the smallest detail of interest in the
scene subtends an angle of about one minute of arc
(160), approximately the limit of angular discrimination for normal vision. For the 483 picture lines of
conventional television, the corresponding viewing
distance is about seven times picture height (PH); the
horizontal viewing angle is about 11. For the 1080
picture lines of HDTV, the optimum viewing distance is
3.3 screen heights, and the horizontal viewing angle is
almost tripled to 28. The situation is sketched in
Figure 1.4 overleaf.
CHAPTER 1
BASIC PRINCIPLES
Conventional TV
TV, 640 480
1 480 PH
1 (1 60)
1 PH
7.1 PH
HDTV, 1920 1080

1 1080 PH
1 (1 60)
1 PH
Figure 1.4 Viewing
distance and angle.
distance
3400
PH
lines
3.3 PH
To achieve a viewing situation where a pixel subtends

1 , viewing distance expressed in units of picture
60
height should be about 3400 divided by the number of
picture lines. A computer user tends to position himself
or herself closer than this about 50 to 60 percent of
this distance but at this closer distance individual
pixels are discernible. Consumer projection television is
viewed closer than 7 PH, but at this distance scan lines
become objectionable.
Aspect ratio
Variants of conventional
525/59.94 systems having 16:9
aspect ratio have recently been
standardized, but few are
deployed as I write this.
Aspect ratio is the ratio of image width to height.

Conventional television has an aspect ratio of 4:3.
High-definition television uses a wider ratio of 16:9.
Cinema commonly uses 1.85:1 or 2.35:1. In a system
having square pixels, the number of horizontal samples
per picture width is the number of scanning lines in the
picture height times the aspect ratio of the image.
Frame rate, refresh rate

A succession of flashed still pictures, captured and
displayed at a sufficiently high rate, can create the illusion of motion. The quality of the motion portrayal
depends on many factors.
6
Take care to distinguish

flicker from twitter,
described on page 11.
Most displays for moving images involve a period of

time when the reproduced image is absent from the
display, that is, a fraction of the frame time during
which the display is black. In order to avoid objectionable flicker, it is necessary to flash the image at a rate
higher than the rate necessary to portray motion.
Refresh rate is highly dependent on the ambient illumination in the viewing environment: The brighter the
environment, the higher the flash rate must be in order
to avoid flicker. To some extent the brightness of the
image itself influences the flicker threshold, so the
brighter the image, the higher the refresh rate must be.
Since peripheral vision has higher temporal sensitivity
than central (foveal) vision, the flicker threshold of
vision is also a function of the viewing angle of the
image.
Refresh rate is generally engineered into a system.
Once chosen, it cannot easily be changed. Different
applications have adopted different refresh rates,
depending on the image quality requirements and
viewing conditions of the application.
In the darkness of a cinema, a flash rate of 48 Hz is
adequate. In the early days of motion pictures, a frame
rate of 48 Hz was thought to involve excessive expenditure for film stock, and 24 frames per second were
found to be sufficient to portray motion. So, a conventional film projector flashes each frame twice. Higher
realism can be obtained with specialized cameras and
projectors that operate at higher frame rates, up to 60
frames per second or more.
In a dim viewing environment typical of television
viewing, such as a living room, a flash rate of 60 Hz is
sufficient. Originally, television refresh rates were
chosen to match the local AC power line frequency.
In a bright environment such as an office, a refresh rate
above 70 Hz might be required.
CHAPTER 1
BASIC PRINCIPLES
Motion portrayal
It is conventional in video for each element of an image
sensor device to integrate light from the scene for the
entire frame time. This captures as much of the light
from the scene as possible, in order to maximize sensitivity and/or signal-to-noise ratio. In an interlaced
camera, the exposure time is usually effectively the
duration of the field, not the duration of the frame.
This is necessary in order to achieve good motion
portrayal.
If the image has elements that move an appreciable
distance during the exposure time, then the sampled
image information will exhibit smear. Smear can be
minimized by using an exposure time that is a fraction
of the frame time; however, the method involves
discarding light from the scene and a sensitivity penalty
is incurred.
When the effect of image information incident during a
single frame time persists into succeeding frames, the
sensor exhibits lag. Lag is a practical problem for tubetype cameras, but generally not a problem for CCD
cameras.
Charles Poynton, Motion
portrayal, eye tracking, and
emerging display technology,in Proceedings of the
30th SMPTE Advanced
motion imaging conference,
192202 (White Plains, New
York: SMPTE, 1996).
Flicker is absent in any image display device that

produces steady, unflashing light for the duration of the
frame time. You might think that a nonflashing display
would be more suitable than a device that flashes, and
many contemporary devices do not flash. However, if
the viewers gaze is tracking an element that moves
across the display, a display with an on-time
approaching the frame time will exhibit smearing of
elements that move. This problem becomes more
severe as eye tracking rates increase; for example, with
the wide viewing angle of high-definition television.
Raster scanning
In cameras and displays, some time is required to
advance the scanning operation to retrace from one
line to the next and from one picture to the next. These
intervals are called blanking intervals, because in a
8
525/59.94 SCANNING
Square
Pixel
Component 4:2:2
Rec. 601 (D-1)
525
525
(D-2)
525
780
944
640
768
480
625
576
858
864
714
702
480
Composite 4fsc
625/50 SCANNING
625
576
910
1135 4625
757
922
480
NTSC
625
576
PAL
Figure 1.5 Digital video rasters. The left column shows 525/59.94 scanning, the right column shows
625/50. The top row shows sampling with square pixels. The middle row shows sampling at the
Rec. 601 standard sampling frequency of 13.5 MHz. The bottom row shows sampling at four times
the color subcarrier. Blanking intervals are shown with dark shading.
525/59.94 is colloquially
referred to as NTSC, and
625/50 as PAL, but the terms
NTSC and PAL properly apply
to color encoding standards
and not to scanning standards.
conventional CRT display the electron beam must be

extinguished (blanked) during these time intervals. The
horizontal blanking time lies between scan lines, and
vertical blanking lies between frames (or fields).
Figure 1.5 above shows the raster structure of
525/59.94 and 625/50 digital video systems, including
these blanking intervals. In analog video, sync information is conveyed during the blanking intervals.
The horizontal and vertical blanking intervals required
for a CRT display are quite large fractions of the line
time and frame time: in 525/59.94, 625/50, and
1920 1035 systems, vertical blanking occupies 8
percent of each frame period. Although in principle a
digital video interface could omit the blanking intervals
and use a clock having a lower frequency than the
sampling clock, this would be impractical. Digital video
standards use interface clock frequencies chosen to
CHAPTER 1
BASIC PRINCIPLES
Voltage, mV
700
Code
235
714 S/PW
350
16
300
0H 64
720 S/AL
Sample clocks, at 13.5 MHz
858 S/TL
Figure 1.6 Scan line waveform for 525/59.94 component video, showing luma. The 720 active
samples contain picture information. Horizontal blanking occupies the remaining sample intervals.
match the large blanking intervals of typical display

equipment. Good use is made in digital systems of
what would otherwise be excess data capacity: A digital
video interface may convey audio signals during
blanking; a digital video tape recorder might record
error correction information in these intervals.
In a digital video system it is

standard to convey samples of
the image matrix in the same
order that the image information would be conveyed in an
analog video system: first the
top line (left to right), then the
next lower line, and so on.
In analog video, information in the image plane is

scanned uniformly left to right during a fixed, short
interval of time the active line time and conveyed
as an analog electrical signal. There is a uniform
mapping from horizontal position in the image to time
instant in the electrical signal. Successive lines are
scanned uniformly from the top of the image to the
bottom, so there is also a uniform mapping from
vertical position in the image to time instant in the electrical signal. The fixed pattern of parallel scanning lines
disposed across the image is the raster. The word is
derived from the Greek rake, from the resemblance of a
raster to the pattern left on a newly raked field.
Figure 1.6 above shows the waveform of a single scan
line, showing voltage from 0 V to 700 mV in a component analog system (with sync at --300 mV), and code-
10
word value from code 16 to code 235 in an 8-bit

component digital system.
Interlace
At the outset of television, the requirement to minimize information rate for transmission and later,
recording led to interlaced scanning. Each frame is
scanned in two successive vertical passes, first the odd
field, then the even field, whose scan lines interlace as
illustrated Figure 1.7 below. Total information rate is
reduced because the flicker susceptibility of vision is
due to a wide-area effect. As long as the complete
height of the picture is scanned rapidly enough to overcome wide-area flicker, small-scale picture information
such as that in the alternate lines can be transmitted at a lower rate.
If the information in an image changes vertically at a
scale comparable to the scanning line pitch if a fine
pattern of black-and-white horizontal line pairs is
scanned, for example then interlace can cause the
content of the odd and the even fields to differ markedly. This causes twitter, a small-scale phenomenon
that is perceived as extremely rapid up-and-down
motion. Twitter can be produced not only from degenerate images such as fine horizontal black-and-white
lines, but also from high-amplitude brightness detail in
an ordinary image. In computer generated imagery
(CGI), twitter can be reduced by vertical filtering.
Figure 1.7 exaggerates the slant

of a fraction of a degree that
results when a conventional
CRT either a camera tube or a
display tube is scanned with
analog circuits. The slant is a
real effect in analog cameras
and displays, although it is
disregarded in the design of
equipment.
If image information differs greatly from one field to

the next, then instead of twitter, large-scale flicker will
1
Figure 1.7 Interlaced scanning
forms a complete picture the
2
frame from two fields, each
comprising half the scanning
lines. The second field is
delayed half the frame time
262
from the first.
264
...
...
265
525
263
CHAPTER 1
BASIC PRINCIPLES
11
result. A video camera is designed to avoid introduction of so much vertical detail that flicker could be
produced. In synthetic image generation, vertical detail
may have to be explicitly filtered in order to avoid
flicker.
Scanning standards
Conventional broadcast television scans a picture
whose aspect ratio is 4:3, in left-to-right, top-tobottom order using interlaced scanning.
A scanning system is denoted by its total line count and
its field rate in hertz, separated by a solidus (slash). Two
scanning standards are established for conventional
television: 525/59.94, used primarily in North America
and Japan; and 625/50, used elsewhere. It is obvious
from the scanning nomenclature that the line counts
and frame rates are different. There are other important differences:
System
Picture:Sync ratio
525/59.94 video in Japan
uses 10:4 picture to sync
ratio and zero setup.
Setup, percent
Count of equalization,
broad pulses
Line number 1, and 0V ,
defined at
525/59.94
625/50
10:4
7:3
7.5
First
equalization pulse
First
broad pulse
The two systems have gratuitous differences in other

parameters unrelated to scanning.
Monochrome systems having
405/50/2:1 and 819/50/2:1
scanning were once used in
Britain and France, respectively,
but transmitters for these standards have now been decommissioned.
Systems with 525/59.94 scanning usually employ NTSC

color coding, and systems with 625/50 scanning
usually use PAL, so 525/59.94 and 625/50 systems are
loosely referred to as NTSC and PAL. But NTSC and
PAL properly refer to color encoding. Although
525/59.94/NTSC and 625/50/PAL systems dominate
worldwide broadcasting, other combinations of scanning and color coding are in use in large and important
regions of the world, such as France, Russia, and South
America.
12
The frame rate of 525/59.94 video is exactly

60
1.001 Hz. In 625/50 the frame rate is exactly 50 Hz.
Computer graphics systems have various frame rates
with few standards and poor tolerances.
An 1125/60/2:1 high-definition television production
system has been adopted as SMPTE Standard 240M
and has been proposed to the ITU-R. At the time of
writing, the system is in use for broadcasting in Japan
but no international broadcasting standards have been
agreed upon.
All of these scanning systems are interlaced 2:1, and
interlace is implicit in the scanning nomenclature.
Noninterlaced scanning is common in desktop
computers and universal in computer workstations.
Emerging high-definition television standards have
interlaced and noninterlaced variants.
John Watkinson, The Engineers Guide to Standards
Conversion. Petersfield, Hampshire, England: Snell & Wilcox,
1994.
Standards conversion refers to conversion among scanning standards. Standards conversion, done well, is
difficult and expensive. Standards conversion between
scanning systems having different frame rates, even
done poorly, requires a fieldstore or framestore. The
complexity of standards conversion between
525/59.94 scanning and 625/50 scanning is the reason
that it is difficult for consumers and broadcasters to
convert European material for use in North America or
Japan, or vice versa.
Transcoding refers to changing the color encoding of a
signal, without altering its scanning system.
Sync structure
At a video interface, synchronization (sync) is achieved
by associating, with every scan line, a line sync datum
denoted 0H (pronounced zero-H). In component digital
video, sync is conveyed using digital codes 0 and 255
outside the range of picture information. In analog
video, sync is conveyed by voltage levels blacker than
black. 0H is defined by the 50-percent point of the
leading (falling) edge of sync.
CHAPTER 1
BASIC PRINCIPLES
13
PREEQUALIZATION
BROAD PULSES
POSTEQUALIZATION
0 V,525
Figure 1.8 Vertical sync
waveform of 525/59.94.
These equalization pulses have

no relationship with the process
of equalization that is used to
compensate poor frequency
response of coaxial cable, or
poor frequency or phase
response of a filter.
In digital technology it is more

intuitive to consider the pulses
that are present rather than the
ones that are absent: The term
serration is now unpopular.
In both 525/59.94 and 625/50 video the normal sync

pulse has a duration of 4.7 s. Vertical sync is identified by broad pulses, which are serrated in order for
a receiver to maintain horizontal sync even during the
vertical interval. Narrow equalization pulses, half the
sync pulse duration at twice the line rate, are present
during intervals immediately before and immediately
following the broad pulses.
When analog sync separators comprised just a few
resistors and capacitors, to achieve stable interlacing
required halving the duration of the line syncs and
introducing additional pulses halfway between them.
Originally the equalization pulses were the ones interposed between the line syncs, but the term now refers
to all of the narrow pulses. The absence of sync level
between the end of a broad pulse and the start of the
following sync was called serration. If you think of field
sync as a single pulse asserted for several lines, serration is the negation of this pulse at twice the line rate.
An equalization pulse has half the duration of a normal
sync. The duration of a vertical (broad) pulse is half the
line time, less a full sync width. A 525/59.94 system
has three lines of preequalization pulses, three lines of
vertical sync, and three lines of postequalization pulses.
A 625/50 system has two and one-half lines (five
pulses) of each of preequalization, broad, and postequalization pulses. Figure 1.8 above sketches the
vertical sync component of 525/59.94 analog video.
Monochrome 525-line broadcasting originated with a
line rate of exactly 15.750 kHz. When color was intro-
14
duced to NTSC in 1953, the monochrome horizontal

frequency was multiplied by exactly 10001001 to obtain
the NTSC color line rate of approximately 15.734 kHz.
Details are in Field, frame, line, and sample rates, on
page 199. All 525-line broadcast signals even monochrome signals now employ this rate. The line rate of
625/50 systems has always been exactly 15.625 kHz,
corresponding to a line time of exactly 64 s.
Data rate
b = bit
B = Byte
103
210
k
K
1000
1024
SI, datacom:
M
106 1 000 000
disk:
M
103 210 1 024 000
RAM:
M
220 1 048 576
Data rate of a digital system is measured in bits per

second (b/s) or bytes per second (B/s), where a byte is
eight bits. The formal, international designation of the
metric system is Systme International dUnits, SI. The
SI prefix k denotes 103 (1000); it is often used in data
communications. The K prefix used in computing
denotes 210 (1024). The SI prefix M denotes 106
(1 000 000). Disk storage is generally allocated in units
integrally related to 1024 bytes; the prefix M applied to
disk storage denotes 1 024 000. RAM memory generally has capacity based on powers of two; the prefix M
applied to RAM denotes 220 or 1024 K (1 048 576).
Data rate of digital video

Line rate is an important parameter of a video system:
Line rate is simply the frame rate multiplied by the
number of lines per total frame.
The aggregate data rate is the number of bits per pixel,
times the number of pixels per line, times the number
of lines per frame, times the frame rate.
In both analog and digital video it is necessary to
convey not only the raw image information, but also
information about which time instants (or which
samples) are associated with the start of frame, or the
start of line. This information is conveyed by signal
synchronization or sync elements. In analog video and
composite digital video, sync is combined with video by
being coded at a level blacker than black.
CHAPTER 1
BASIC PRINCIPLES
15
All computer graphics systems and almost all digital

video systems have the same integer number of sample
clock periods in every raster line. In these cases,
sampling frequency is simply the line rate times the
number of samples per total line (S/TL).
In 625/50 PAL there is not an exact integer number of
samples per line: Samples in successive lines are offset
to the left a small fraction, 4625 of the horizontal
sample pitch. The sampling structure is not precisely
orthogonal, although digital acquisition, processing,
and display equipment treat it so.
The data capacity required for the active pixels of a
frame is computed by simply multiplying the number of
bits per pixel by the number of active pixels per line,
then by the number of active lines per frame. To
compute the data rate for the active pixels, simply
multiply by the frame rate.
Standards are not well established in display systems
used in desktop computers, workstations, and industrial equipment. The absence of published data makes it
difficult to determine raster scanning parameters.
Linearity
A video system should ideally satisfy the principle of
superposition; in other words, it should exhibit
linearity. A function f is linear if and only if (iff):
Eq 1.1
f ( a + b) f ( a) + f ( b)
The function f can encompass an entire system:

A system is linear iff the sum of the individual responses
of the system to any two signals is identical to its
response to the sum of the two. Linearity can pertain to
steady-state response, or to the systems temporal
response to a changing signal.
Linearity is a very important property in mathematics,
in signal processing, and in video. But linearity in one
domain cannot be carried across to another domain if
16
a nonlinear function separates the two. An image signal

usually originates in a sensor that has linear response to
physical intensity. And video signals are usually
processed through analog circuits that have linear
response to voltage or digital systems that are linear
with respect to the arithmetic performed on the codewords. But a video camera applies a nonlinear transfer
function gamma correction to the image signal. So
the image signal is in a linear optical domain, and the
video signal is in a linear electrical domain, but the two
domains are not the same.
Sound pressure level, relative
Perceptual uniformity
1
0
0
300
Angle of rotation, degrees
Figure 1.9 Audio taper.
A system is perceptually uniform if a small perturbation to a component value is approximately equally

perceptible across the range of that value. The volume
control on your radio is designed to be perceptually
uniform: Rotating the knob 10 degrees produces
approximately the same perceptual increment in
volume anywhere across the range of the control. If the
control were physically linear, the logarithmic nature of
loudness perception would place all of the perceptual
action of the control at the bottom of its range.
Figure 1.9, in the margin, shows the transfer function
of a potentiometer with standard audio taper.
The CIE L* system, to be described on page 88, assigns
a perceptually uniform scale to lightness. Video signals
are coded according to perceptual principles, as will be
explained in Chapter 6, Gamma, on page 91.
Noise, signal, sensitivity
A distortion product that can

be attributed to a particular
processing step is known as an
artifact, particularly if it has a
distinctive visual effect on the
picture.
CHAPTER 1
Any analog electronic system is inevitably subject to

noise that is unrelated to the signal to be processed by
the system. As signal amplitude decreases, the noise
makes a larger and larger relative contribution. In
analog electronics, noise is inevitably introduced from
thermal sources, and perhaps also from nonthermal
sources of interference.
In addition to random noise, processing of a signal
may introduce distortion that is correlated to the signal
BASIC PRINCIPLES
17
itself. For the purposes of objective measurement of the

performance of a system, distortion is treated as noise.
Depending on its nature, distortion may be more or less
perceptible than random noise.
Signal-to-Noise Ratio (SNR) is the ratio of a specified
signal, often the reference amplitude or largest amplitude signal that can be carried by a system, to the
amplitude of undesired components including noise
and distortion. SNR is expressed in units of decibels
(dB), a logarithmic measure.
Sensitivity refers to the minimum signal power that
achieves acceptable (or specified) SNR performance.
Quantization
To make a 50-foot-long fence

with fence posts every 10 feet
you need six posts, not five!
Take care to distinguish levels
(here, six) from steps (here, five).
Theoretical SNR for

an k-step quantizer:
20 log
10
(k 12 )
A signal whose amplitude takes a range of continuous

values is quantized by assigning to each of a finite set
of intervals of amplitude a discrete, numbered level. In
uniform quantization the steps between levels have
equal amplitude. The degree of visual impairment
caused by noise in a video signal is a function of the
properties of vision. In video, it is ubiquitous to digitize
a signal that is a nonlinear function, usually a 0.45power function, of physical (linear-light) intensity. The
function chosen minimizes the visibility of noise.
The effect of quantizing to a finite number of discrete
amplitude levels is equivalent to adding quantization
noise to the ideal levels of a quantized signal. Quantization has the effect of introducing noise, and thereby
diminishes the SNR of a digital system. Eight-bit quantization has a theoretical SNR limit of about 56 dB (peak
signal to rms noise).
If an input signal has very little noise, then situations
can arise when the quantized value is quite predictable
at some points, but when the signal is near the edge of
a quantizer step, uncertainty in the quantizer is
reflected as noise. This situation can cause the reproduced image to exhibit noise modulation. It is beneficial to introduce roughly a quantizer steps worth of
18
noise (peak to peak) prior to quantization, to avoid this

effect. This introduces a very small amount of noise in
the picture, but guarantees avoidance of patterning
of the quantization.
MIDTREAD
Quantization can be applied to a unipolar signal such as

luma. For a bipolar signal such as a color difference it is
standard to use a mid-tread quantizer, such as the one
sketched in Figure 1.10 in the margin, so that no
systematic error affects the zero value.
Figure 1.10
Mid-tread quantizer.
Frequency response, bandwidth

Figure 1.11 below shows a test signal starting at zero
frequency and sweeping up to some high frequency.
The response of a typical electronic system is shown in
the middle graph; the response diminishes at high
frequency. The envelope of that waveform the
systems frequency response is shown at the bottom.
ID
TH
W
D
1.0
0.1
0
0
CHAPTER 1
N
L
RE IM
SO ITI
LU NG
TI
O
HALF-POWER
(-3 dB)
0.707
BA
N
Frequency response,
relative
Output
Input
Figure 1.11 Frequency response of any electronic or optical system falls as frequency increases.
Bandwidth is measured at the half-power point (-3 dB), where response has fallen to 0.707. Television displays are often specified at limiting resolution, where response has fallen to 0.1.
Frequency, relative
BASIC PRINCIPLES
19
Loosely speaking, bandwidth is the rate at which information in a signal can change from one state to
another. The response of an electronic system deteriorates above a certain information rate. Bandwidth is
specified or measured at the frequency where amplitude has fallen 3 dB from its value at zero frequency
(called DC) that is, to the fraction 0.707 of its value
at DC.
The rate at which an analog video signal can change
from one state to another, say from white to black, is
limited by the bandwidth of the video system. This
places an upper bound on horizontal resolution.
Consumer video generally refers to horizontal resolution, measured as the number of black and white
elements (TV lines) that can be discerned over a horizontal distance equal to the picture height.
Bandwidth and data rate
Data rate does not apply directly to an analog system,
and the term bandwidth does not properly apply to a
digital system. When a digital system conveys a
sampled representation of a continuous signal, as in
digital video or digital audio, the bandwidth represented by the digitized signal is necessarily less than
half typically about 0.45 of the sampling rate.
Figure 1.12
Bandwidth and data rate.
20
When arbitrary digital information is conveyed through

an analog channel, as by a modem, the data rate that
can be achieved depends on bandwidth, noise, and
other properties of the channel. Figure 1.12, in the
margin, shows a simple scheme that transmits two bits
per second per hertz of bandwidth, or 2400 b/s for a
channel having 1200 Hz analog bandwidth. The
bottom sketch shows that if each half-cycle conveys
one of sixteen amplitude levels, providing the channel
has sufficiently low noise, four bits can be coded per
half-cycle. The rate at which the signal in the channel
can change state the symbol rate or baud rate is
constant at 2400 baud, but this modulation method
has a data rate or bit rate of 9600 b/s.
Resolution
As picture detail increases in frequency, the response of
an imaging system will eventually deteriorate. In image
science and in television, resolution refers to the capability of an imaging system to reproduce fine detail in
the picture.
The absolute upper limit to resolution in a digital image
system is the number of pixels over the width and
height of a frame, and is the way the term resolution is
used in computing.
In conventional North American television, 483 scan
lines cover the height of the image. High-definition
television systems use up to 1080 picture lines. The
amount of information that can be captured in a video
signal is bounded by the number of picture lines. But
other factors impose limits more severe than the
number of lines per picture height.
In an interlaced system, vertical resolution must be
reduced substantially from the scan-line limit, in order
to avoid producing a signal that will exhibit objectionable twitter upon display.
Resolution in film
In film, resolution is measured as the finest pattern of
straight, parallel lines that can be reproduced,
expressed in line pairs per millimeter (lp/mm). A line
pair contains a black region and a white region.
Motion picture film is conveyed vertically through the
camera and projector, so the width not the height
of the film is 35 mm. Cinema usually has an aspect
ratio of 1.85:1, so the projected film area is about
21 mm 11 mm, only three-tenths of the 36 mm
24 mm projected area of 35 mm still film.
The limit to the resolution of motion picture film is not
the static response of the film, but judder and weave in
the camera and the projector.
CHAPTER 1
BASIC PRINCIPLES
21
Resolution in television
In video, resolution refers to the number of line pairs
(cycles) resolved on the face of the display screen,
expressed in cycles per picture height (C/PH) or cycles
per picture width (C/PW). A cycle is equivalent to a
line pair of film. In a digital system, it takes at least two
samples pixels, scanning lines, or TV lines to represent a line pair. However, resolution may be substantially less than the number of pixel pairs due to optical,
electro-optical, and electrical filtering effects. Limiting
resolution is defined as the frequency where detail is
recorded with just 10 percent of the systems lowfrequency response.
In consumer television, the number of scanning lines is
fixed by the raster standard, but the electronics of
transmission, recording, and display systems tend to
limit bandwidth and reduce horizontal resolution.
Consequently, in consumer electronics the term resolution generally refers to horizontal resolution. Confusingly, horizontal resolution is expressed in units of lines
per picture height, so once the number of resolvable
lines is measured, it must be corrected for the aspect
ratio of the picture. Resolution in TV lines per picture
height is twice the resolution in cycles per picture
width, divided by the aspect ratio of the picture.
Resolution in computer graphics
In computer graphics, resolution is simply the number
of discrete vertical and horizontal pixels required to
store the digitized image. For example, a 1152900
system has a total of about one million pixels (one
megapixel, or 1 Mpx). Computer graphics is not generally very concerned about whether individual pixels can
be discerned on the face of the display. In most color
computer systems, an image comprising a one-pixel
black-and-white checkerboard actually displays as a
uniform gray, due to poor high-frequency response in
the cable and video amplifiers, and due to rather large
spot size at the CRT.
22
Pixel
72 dpi
0.35 mm
Spot
0.63 mm
Triad
0.31 mm
Figure 1.13
Pixel/spot/triad.
Computer graphics often treats each pixel as representing an idealized rectangular area independent of all
other pixels. This notion discounts the correlation
among pixels that is an inherent and necessary aspect
of image acquisition, processing, compression, display,
and perception. In fact the rather large spot produced
by the electron beam of a CRT and the arrangement of
phosphor triads on the screen, suggested by
Figure 1.13, produces an image of a pixel on the screen
that bears little resemblance to a rectangle. If pixels are
viewed at a sufficient distance, these artifacts are of
little importance. However, imaging systems are forced
by economic pressures to make maximum perceptual
use of the delivered pixels, consequently we tend to
view CRTs at close viewing distances.
Luma
As you will see in Luma and color differences, on
page 155, a video system conveys image data in the
form of a component that represents brightness, and
two other components that represent color. It is important to convey the brightness component in such a way
that noise (or quantization) introduced in transmission,
processing, and storage has a perceptually similar effect
across the entire tone scale from black to white. Ideally,
these goals would be accomplished by forming a true
CIE luminance signal as a weighted sum of linear-light
red, green, and blue; then subjecting that luminance to
a nonlinear transfer function similar to the CIE L* function that will be described on page 88.
There are practical reasons in video to perform these
operations in the opposite order. First a nonlinear
transfer function gamma correction is applied to
each of the linear R, G, and B. Then a weighted sum of
the nonlinear components is computed to form a luma
signal, Y, representative of brightness.
625/50 standards documents
indicate a precorrection of 1 2.8,
approximately 0.36, but this
value is rarely used in practice.
See Gamma on page 91.
In effect, video systems approximate the lightness

response of vision using RGB intensity signals, each
raised to the 0.45 power. This is comparable to the 13
power function defined by L*.
CHAPTER 1
BASIC PRINCIPLES
23
Recommendation ITU-R
BT.601-4, Encoding Parameters
of Digital Television for Studios.
Geneva: ITU, 1990.
Eq 1.2
The coefficients that correspond to the so-called NTSC

red, green, and blue CRT phosphors of 1953 are standardized in Recommendation ITU-R BT. 601-4 of the
ITU Radiocommunication Sector (formerly CCIR). I call
it Rec. 601. To compute nonlinear video luma from
nonlinear red, green, and blue:
601
Y = 0.299 R + 0.587 G + 0.114 B
The prime symbols in this equation, and in those to

follow, denote nonlinear components.
The unfortunate term video luminance
Unfortunately, in video practice, the term luminance
has come to mean the video signal representative of
luminance even though the components of this signal
have been subjected to a nonlinear transfer function. At
the dawn of video, the nonlinear signal was denoted
Y, where the prime symbol indicated the nonlinear
treatment. But over the last 40 years the prime has
been elided and now both the term luminance and the
symbol Y collide with the CIE, making both ambiguous! This has led to great confusion, such as the incorrect statement commonly found in computer graphics
and color textbooks that in the YIQ or YUV color
spaces, the Y component is CIE luminance! I use the
term luminance according to its standardized CIE definition and use the term luma to refer to the video
signal, and I am careful to designate the nonlinear
quantity with a prime symbol. But my convention is not
yet widespread, and in the meantime you must be
careful to determine whether a linear or nonlinear interpretation is being applied to the word and the symbol.
Color difference coding
In component video, the three components necessary
to convey color information are transmitted separately.
The data capacity accorded to the color information in
a video signal can be reduced by taking advantage of
the relatively poor color acuity of vision, providing full
24
RGB 4:4:4
YCB CR 4:4:4
4:2:2
4:1:1
4:2:0 (JPEG/JFIF,
4:2:0
(Rec. 601)
(DVC)
H.261, MPEG-1)
(MPEG-2)
R0 R1
Y0 Y1
Y0 Y1
Y0 Y1 Y2 Y3
Y0 Y1
Y0 Y1
R2 R3
Y2 Y3
Y2 Y3
Y4 Y5 Y6 Y7
Y2 Y3
Y2 Y3
G0 G1
CB CB
CB01
CB03
G2 G3
CB CB
CB23
CB47
B0 B1
CR CR
CR01
CR03
B2 B3
CR CR
CR23
CR47
CB03
CB03
CR03
CR03
Figure 1.14 Chroma subsampling. A 2 2 array of RGB pixels can be transformed to a luma component Y and two color difference components CB and CR ; color detail can then be reduced by subsampling, provided that full luma detail is maintained. The wide aspect of the CB and CR samples indicates
their spatial extent. The horizontal offset of CB and CR is due to cositing. (JPEG, H.261, and MPEG-1
do not use cositing; instead, their CB and CR samples are taken halfway between luma samples.)
luma bandwidth is maintained. It is ubiquitous to base

color difference signals on blue minus luma and red
minus luma (B--Y, R-Y). Luma and (B--Y, R--Y) can
be computed from R, G, and B through a 33 matrix
multiplication. Once luma and color difference or
chroma components have been formed, the chroma
components can be subsampled (filtered).
YCBCR
In component digital video, CB and CR components

scaled from (B--Y, R--Y) are formed.
YPBPR
In component analog video, PB and PR color difference

signals scaled from (B--Y, R--Y) are lowpass filtered to
about half the bandwidth of luma.
4:4:4
In Figure 1.14 above, the left-hand column sketches a

22 array of RGB pixels that, with 8 bits per sample,
would occupy a total of 12 bytes. This is denoted 4:4:4
RGB. YCBCR components can be formed from
RGB, as shown in the second column; without
subsampling, this is denoted 4:4:4 YCBCR.
The use of 4 as the numerical basis for subsampling
notation is a historical reference to a sample rate of
about four times the color subcarrier frequency.
CHAPTER 1
BASIC PRINCIPLES
25
4:2:2
YCBCR digital video according to Rec. 601 uses 4:2:2

sampling: Chroma components are subsampled by a
factor of 2 along the horizontal axis. Chroma samples
are coincident (cosited) with alternate luma samples.
In an 8-bit system using 4:2:2 coding, the 22 array
occupies 8 bytes, and the aggregate data capacity is 16
bits per pixel. For studio digital video, the raw data rate
is 27 MB/s.
4:1:1
A few digital video systems have used 4:1:1 sampling,

where the chroma components are subsampled by a
factor of 4 horizontally.
4:2:0
JPEG, H.261, MPEG-1, and MPEG-2 usually use 4:2:0

sampling. CB and CR are each subsampled by a factor
of 2 both horizontally and vertically; CB and CR are
sited vertically halfway between scan lines. Horizontal
subsampling is inconsistent. In MPEG-2, CB and CR are
cosited horizontally. In JPEG, H.261, and MPEG-1, CB
and CR are not cosited horizontally; instead, they are
sited halfway between alternate luma samples.
H.261, known casually as

p64, denotes a videoconferencing standard
promulgated by the ITU-T.
MAC
A transmission system for analog components Multiplexed Analog Components, or MAC has been
adopted in Europe for direct broadcast from satellite
(DBS). In MAC, the color difference components are
not combined with each other or with luma, but are
time-compressed and transmitted serially. MAC is not
standardized by ITU-R.
Component digital video, 4:2:2
A version of Rec. 601 uses

18 MHz sampling to produce
a picture aspect ratio of 16:9.
The standard interface for 4:2:2 component digital

video is Rec. ITU-R 601-4. It specifies sampling of luma
at 13.5 MHz and sampling of CB and CR color difference components at 6.75 MHz. This interface is
referred to as 4:2:2, since luma is sampled at four times
3.375 MHz, and each of the CB and CR components at
twice 3.375 MHz that is, the color difference signals
are horizontally subsampled by a factor of 2:1 with
respect to luma. Sampling at 13.5 MHz results in an
integer number of samples per total line (S/TL) in both
26
525/59.94 systems (858 S/TL) and 625/50 systems

(864 S/TL). Luma is sampled with 720 active samples
per line in both 525/59.94 and 625/50.
Component digital video tape recorders are widely
available for both 525/59.94 and 625/50 systems, and
have been standardized with the designation D-1. That
designation properly applies to the tape format, not the
signal interface.
Transport, electrical, and
mechanical aspects of 4:2:2
interface are specified in
Rec. 656. See page 248.
Rec. 601 specifies luma coding that places black at

code 16 and white at code 235. Color differences are
coded in offset binary, with zero at code 128, the negative peak at code 16, and the positive peak at
code 240.
Composite video
The terms NTSC and PAL are
often used incorrectly to refer to
scanning standards. Since PAL
encoding is used with both
625/50 scanning (with two
different subcarrier frequencies)
and 525/59.94 scanning (with a
third subcarrier frequency), the
term PAL alone is ambiguous.
The notation CCIR is sometimes
used to refer to 625/50 scanning, but that is confusing
because the former CCIR now
ITU-R standardized all scanning systems, not just 625/50.
In composite NTSC and PAL video, the color difference

signals required to convey color information are
combined by the technique of quadrature modulation
into a chroma signal using a color subcarrier of about
3.58 MHz in conventional NTSC and about 4.43 MHz
in conventional PAL. Luma and chroma are then
summed into a composite signal for processing,
recording, or transmission. Summing combines brightness and color into one signal, at the expense of introducing a certain degree of mutual interference.
The frequency and phase of the subcarrier are chosen
and maintained carefully: The subcarrier frequency is
chosen so that luma and chroma, when they are
summed, are frequency interleaved. Studio signals have
coherent sync and color subcarrier; that is, subcarrier is
phase-locked to a rational fraction of the line rate;
generally this is achieved by dividing both from a single
master clock. In industrial and consumer video, subcarrier usually free-runs with respect to line sync.
SECAM sums luma and chroma

without using frequency interleaving. SECAM has no application in the studio. See page 254.
Transcoding among different color encoding methods

having the same raster standard is accomplished by
luma/chroma separation, color demodulation, and color
remodulation.
CHAPTER 1
BASIC PRINCIPLES
27
Composite digital video, 4fSC

The earliest digital video equipment processed signals in
composite form. Processing of digital composite signals
is simplified if the sampling frequency is an integer
multiple of the color subcarrier frequency. Nowadays, a
multiple of four is used: four-times-subcarrier, or 4fSC .
For NTSC systems it is standard to sample at about
14.3 MHz. For PAL systems the sampling frequency is
about 17.7 MHz.
Composite digital processing was necessary in the early
days of digital video, but most image manipulation
operations cannot be accomplished in the composite
domain. During the 1980s there was widespread
deployment of component digital processing equipment and component videotape recorders (DVTRs),
recording 4:2:2 signals using the D-1 standard.
However, the data rate of a component 4:2:2 signal is
roughly twice that of a composite signal. Four-timessubcarrier composite digital coding was resurrected to
enable a cheap DVTR; this became the D-2 standard.
The D-2 DVTR offers the advantages of digital
recording, but retains the disadvantages of composite
NTSC or PAL: Luma and chroma are subject to crosscontamination, and the pictures cannot be manipulated without decoding and reencoding.
The development and standardization of D-2 recording
led to the standardization of composite 4fSC digital
parallel and serial interfaces, which essentially just code
the raw 8- or 10-bit composite data stream. These
interfaces share the electrical and physical characteristics of the standard 4:2:2 interface, but with about half
the data rate. For 8-bit sampling this leads to a total
data rate of about 14.3 MB/s for 525/59.94 NTSC, and
about 17.7 MB/s for 625/50 PAL.
Analog interface
Video signal amplitude levels in 525/59.94 systems are
expressed in IRE units, named after the Institute of
Radio Engineers in the United States, the predecessor
28
In 525/59.94 with setup,

picture excursion refers to the
range from blanking to white,
even though strictly speaking
the lowest level of the picture
signal is 7.5 IRE, and not 0 IRE.
525/59.94 NTSC video in

Japan employs zero setup.
of the IEEE. Reference blanking level is defined as 0 IRE,

and reference white level is 100 IRE. The range
between these values is the picture excursion.
Composite 525/59.94 systems have a picture-to-sync
ratio of 10:4; consequently, the sync level of a
composite 525/59.94 signal is --40 IRE. In composite
NTSC systems, except in Japan, reference black is setup
the fraction 7.5 percent ( 340 ) of the reference
blanking-to-white excursion: Composite 525/59.94
employs a pedestal of 7.5 IRE. There are exactly
92.5 IRE from black to white: The picture excursion of a
525/59.94 signal is about 661 mV.
Setup has been abolished from component digital video
and from HDTV. Many 525/59.94 component analog
systems have adopted zero setup, and have 700 mV
excursion from black to white, with 300 mV sync. But
many component analog systems use setup, and it is a
nuisance in design and in operation.
625/50 systems have a picture-to-sync ratio of 7:3, and
zero setup. Picture excursion (from black to white) is
exactly 700 mV; sync amplitude is exactly 300 mV.
Because the reference levels are exact in millivolts, the
IRE unit is rarely used, but in 625/50 systems an IRE
unit corresponds to exactly 7 mV.
A video signal with sync is distributed in the studio with
blanking level at zero (0 VDC ) and an amplitude from
synctip to reference white of one volt into an impedance of 75 . A video signal without sync is distributed
with blanking level at zero, and an amplitude from
blanking to reference white of either 700 mV or
714 mV.
High-definition television, HDTV

High-definition television (HDTV) is defined as having
twice the vertical and twice the horizontal resolution of
conventional television, a picture aspect ratio of 16:9, a
frame rate of 24 Hz or higher, and at least two channels of CD-quality sound.
CHAPTER 1
BASIC PRINCIPLES
29
HDTV studio equipment is commercially with

1125/60/2:1 scanning and 19201035 image format,
with about two megapixels per frame six times the
number of pixels of conventional television. The data
rate of studio-quality HDTV is about 120 megabytes
per second. Commercially available HDTV cameras rival
the picture quality of the best motion picture cameras
and films.
NHK Science and Technical
Research Laboratories, High
Definition Television: Hi-Vision
Technology. New York: Van
Nostrand Reinhold, 1993.
SMPTE 274M-1995,
1920 1080 Scanning and
Interface.
Except for their higher sampling rates, studio standards

for HDTV have a close relationship to studio standards
for conventional video, which I will describe in the rest
of the book. For details specific to HDTV, consult the
book from NHK Labs, SMPTE 274M and 296M.
Advanced Television (ATV) refers to transmission
systems designed for the delivery of entertainment to
consumers, at quality levels substantially improved over
conventional television. ATV transmission systems
based on 1125/60/2:1 scanning and MUSE compression have been deployed in Japan. The United States
has adopted standards for ATV based on 19201080
and 1280720 image formats. MPEG-2 compression
can compress this to about 20 megabits per second, a
rate suitable for transmission through a 6 MHz terrestrial VHF/UHF channel.
The compression and digital transmission technology
developed for ATV has been adapted for digital transmission of conventional television; this is known as
standard-definition television (SDTV). MPEG-2
compression and digital transmission allow a broadcaster to place about four digital channels in the bandwidth occupied by a single analog NTSC signal. Digital
television services are already deployed in direct broadcast satellite (DBS) systems and are expected soon in
cable television (CATV).
30
With the advent of HDTV, 16:9 widescreen variants of

conventional 525/59.94 and 625/50 component video
have been proposed and even standardized. In studio
analog systems, widescreen is accomplished by having
the active picture represent 16:9 aspect ratio, but
keeping all of the other parameters of the video standards. Unless bandwidth is increased by the same 4 3
ratio as the increase in aspect ratio, horizontal detail
suffers.
In digital video, there are two approaches to achieving
16:9 aspect ratio. The first approach is comparable to
the analog approach that I mentioned a moment ago:
The sampling rate remains the same as conventional
component digital video, and horizontal resolution is
reduced by a factor of 3 4 . In the second approach, the
sampling rate is increased from 13.5 MHz to 18 MHz.
I consider all of these schemes to adapt conventional
video to widescreen be unfortunate: None of them
offers an increase in resolution sufficient to achieve the
product differentiation that is vital to the success of any
new consumer product.
CHAPTER 1
BASIC PRINCIPLES
31
Gamma
This is Chapter 6 of the book
A Technical Introduction to
Digital Video, by Charles
Poynton, published in 1996
by John Wiley & Sons.
In photography, video, and computer graphics, the

gamma symbol, , represents a numerical parameter
that describes the nonlinearity of intensity reproduction. Gamma is a mysterious and confusing subject,
because it involves concepts from four disciplines:
physics, perception, photography, and video. This
chapter explains how gamma is related to each of these
disciplines. Having a good understanding of the theory
and practice of gamma will enable you to get good
results when you create, process, and display pictures.
This chapter focuses on electronic reproduction of
images, using video and computer graphics techniques
and equipment. I deal mainly with the reproduction of
intensity, or, as a photographer would say, tone scale.
This is one important step to achieving good color
reproduction; more detailed information about color
can be found in Color science for video, on page 115.
A cathode-ray tube (CRT) is inherently nonlinear: The
intensity of light reproduced at the screen of a CRT
monitor is a nonlinear function of its voltage input.
From a strictly physical point of view, gamma correction can be thought of as the process of compensating
for this nonlinearity in order to achieve correct reproduction of intensity.
As explained in Luminance and lightness, on page 81,
the human perceptual response to intensity is distinctly
nonuniform: The lightness sensation of vision is roughly
Copyright 1996 John Wiley & Sons
91
a power function of intensity. This characteristic needs

to be considered if an image is to be coded so as to
minimize the visibility of noise and make effective
perceptual use of a limited number of bits per pixel.
Combining these two concepts one from physics, the
other from perception reveals an amazing coincidence: The nonlinearity of a CRT is remarkably similar
to the inverse of the lightness sensitivity of human
vision. Coding intensity into a gamma-corrected signal
makes maximum perceptual use of the channel. If
gamma correction were not already necessary for physical reasons at the CRT, we would have to invent it for
perceptual reasons.
Photography also involves nonlinear intensity reproduction. Nonlinearity of film is characterized by a
parameter gamma. As you might suspect, electronics
inherited the term from photography! The effect of
gamma in film concerns the appearance of pictures
rather than the accurate reproduction of intensity
values. The appearance aspects of gamma in film also
apply to television and computer displays.
Finally, I will describe how video draws aspects of its
handling of gamma from all of these areas: knowledge
of the CRT from physics, knowledge of the nonuniformity of vision from perception, and knowledge of
viewing conditions from photography. I will also discuss
additional details of the CRT transfer function that you
will need to know if you wish to calibrate a CRT or
determine its nonlinearity.
Gamma in physics
The physics of the electron gun of a CRT imposes a
relationship between voltage input and light output
that a physicist calls a five-halves power law: The
intensity of light produced at the face of the screen is
proportional to the voltage input raised to the power
5 . Intensity is roughly between the square and cube of
2
the voltage. The numerical value of the exponent of
the power function is represented by the Greek letter
92
Light Intensity, cd m-2
120
100
80
60
40
20
0
0
100
200
300
400
500
600
700
Video Signal, mV
Figure 6.1 CRT transfer function involves a nonlinear relationship between video signal and light
intensity, here graphed for an actual CRT at three different settings of the Picture control. Intensity is
approximately proportional to input signal voltage raised to the 2.5 power. The gamma of a display
system or more specifically, a CRT is the numerical value of the exponent of the power function.
(gamma). CRT monitors have voltage inputs that reflect

this power function. In practice, most CRTs have a
numerical value of gamma very close to 2.5.
Figure 6.1 above is a sketch of the power function that
applies to the single electron gun of a grayscale CRT, or
to each of the red, green, and blue electron guns of a
color CRT. The functions associated with the three guns
of a color CRT are very similar to each other, but not
necessarily identical. The function is dictated by the
construction of the electron gun; the CRTs phosphor
has no significant effect.
Gamma correction involves
a power function, which has
the form y = x a (where a is
constant). It is sometimes
incorrectly claimed to be an
exponential function, which
has the form y = a x (where
a is constant).
The process of precompensating for this nonlinearity

by computing a voltage signal from an intensity value
is known as gamma correction. The function required is
approximately a 0.45-power function, whose graph is
similar to that of a square root function. In video,
gamma correction is accomplished by analog circuits at
the camera. In computer graphics, gamma correction is
usually accomplished by incorporating the function into
a framebuffers lookup table.
CHAPTER 6
GAMMA
93
Alan Roberts, Measurement of

display transfer characteristic
(gamma, ), EBU Technical
Review 257 (Autumn 1993),
3240.
The actual value of gamma for a particular CRT may

range from about 2.3 to 2.6. Practitioners of computer
graphics often claim numerical values of gamma quite
different from 2.5. But the largest source of variation in
the nonlinearity of a monitor is caused by careless
setting of the Black Level (or Brightness) control of
your monitor. Make sure that this control is adjusted so
that black elements in the picture are reproduced
correctly before you devote any effort to determining
or setting gamma.
Getting the physics right is an important first step
toward proper treatment of gamma, but it isnt the
whole story, as you will see.
The amazing coincidence!

In Luminance and lightness, on page 81, I described
the nonlinear relationship between luminance and
perceived lightness. The previous section described how
the nonlinear transfer function of a CRT relates a
voltage signal to intensity. Heres the surprising coincidence: The CRT voltage-to-intensity function is very
nearly the inverse of the luminance-to-lightness relationship of vision. Representing lightness information as
a voltage, to be transformed into luminance by a CRTs
power function, is very nearly the optimal coding to
minimize the perceptibility of noise. CRT voltage is
remarkably perceptually uniform.
Suppose you have a luminance value that you wish to
communicate to a distant observer through a channel
having only 8 bits. Consider a linear light representation, where code zero represents black and code 255
represents white. Code value 100 represents a shade of
gray that is approximately at the perceptual threshold:
For codes above 100, the ratio of intensity values
between adjacent codes is less than 1 percent; and for
codes below 100, the ratio of intensity values between
adjacent code values is greater than 1 percent.
For luminance values below 100, as the code value
decreases toward black, the difference of luminance
94
255
201
200
= 0.5%
101
100
=1%
26
25
= 4%
0
Figure 6.2 Fixed-point
linear-light coding.
values between adjacent codes becomes increasingly

visible: At code 25, the ratio between adjacent codes is
4 percent, which is objectionable to most observers.
These errors are especially noticeable in pictures having
large areas of smoothly varying shades, where they are
known as contouring or banding.
Luminance codes above 100 suffer no artifacts due to
visibility of the jumps between codes. However, as the
code value increases toward white, the codes have
decreasing perceptual utility. For example, at code 200
the ratio between adjacent codes is 0.5 percent, well
below the threshold of visibility. Codes 200 and 201 are
visually indistinguishable: Code 201 is perceptually
useless and could be discarded without being noticed.
This example, sketched in Figure 6.2 in the margin,
shows that a linear-luminance representation is a bad
choice for an 8-bit channel.
In an image coding system, it is sufficient, for perceptual purposes, to maintain a ratio of luminance values
between adjacent codes of about a 1 percent. This can
be achieved by coding the signal nonlinearly, as roughly
the logarithm of luminance. To the extent that the log
function is an accurate model of the contrast sensitivity
function, full perceptual use is made of every code.
As mentioned in the previous section, logarithmic
coding rests on the assumption that the threshold function can be extended to large luminance ratios. Experiments have shown that this assumption does not hold
very well, and coding according to a power law is
found to be a better approximation to lightness
response than a logarithmic function.
The lightness sensation can be computed as intensity
raised to a power of approximately the one-third:
Coding a luminance signal to a signal by the use of a
power law with an exponent of between 1 3 and 0.45
has excellent perceptual performance.
CHAPTER 6
GAMMA
95
S. S. Stevens, Psychophysics.
New York: John Wiley &
Sons, 1975.
Incidentally, other senses behave according to power

functions:
Percept
Physical quantity
Power
Loudness
Sound pressure level
0.67
Saltiness
Sodium chloride concentration
1.4
Smell
Concentration of aromatic
molecules
0.6
Heaviness
Mass
1.45
Gamma in film
This section describes gamma in photographic film.
I give some background on the photographic process,
then explain why physically accurate reproduction of
luminance values gives subjectively poor results. Video
systems exploit this gem of wisdom from photography:
Subjectively better images can be obtained if proper
account is taken of viewing conditions.
When film is exposed, light imaged from the scene
onto the film causes a chemical change to the emulsion of the film, and forms a latent image. Subsequent
development causes conversion of the latent image into
small grains of metallic silver. This process intrinsically
creates a negative image: Where light causes silver to
be developed, the developed film absorbs light and
appears dark. Color film comprises three layers of emulsion sensitized to different wavelength bands, roughly
red, green, and blue. The development process
converts silver in these three layers into dyes that act as
colored filters to absorb red, green, and blue light.
Film can be characterized by the transfer function that
relates exposure to the transmittance of the developed
film. When film is exposed in a camera, the exposure
value at any point on the film is proportional to the
luminance of the corresponding point in the scene,
multiplied by the exposure time.
96
Figure 6.3 Tone response

of color reversal film. This
graph is redrawn, with
permission, from Kodak
Publication H-1. It shows
the S-shaped exposure characteristic of typical colorreversal photographic film.
Over the straight-line
portion of the log-log curve,
the density of the developed film is a power function of exposure intensity.
B
3.2
3.0
G
R
2.8
2.6
2.4
2.2
GB
2.0
Density
EASTMAN Professional
Motion Picture Films,
Kodak Publication H-1,
Fourth Edition. Rochester,
NY: Eastman Kodak
Company, 1992. Figure 26.
3.4
1.8
1.6
1.4
1.2
1.0
BG
0.8
0.6
0.4
0.2
D min
0.0
-2.0
-1.0
Exposure, log lux s
P
D = log10 0
PT
D:
P0 :
Density
Incident Power
PT :
Transmitted Power
See Table 7.5, Density

examples, on page 153.
Transmittance is defined as the fraction of light incident on the developed film to light absorbed. Density is
the logarithm of incident power divided by transmitted
power. The characteristic of a film is usually shown by
plotting density as a function of the logarithm of exposure. This D-log E curve was first introduced by Hurter
and Driffield, so it is also called an H&D plot. In terms
of the physical quantities of exposure and transmittance, a D-log E plot is fundamentally in the log-log
domain.
A typical film plotted in this way is shown in the plot in
Figure 6.3 above. The plot shows an S-shaped curve
that compresses blacks, compresses whites, and has a
CHAPTER 6
GAMMA
97
reasonably linear segment in the central portion of the

curve. The ubiquitous use of D-log E curves in film
work and the importance of the linear segment of the
curve in determining correct exposure leads many
people to the incorrect conclusion that film has an
inherently logarithmic luminance response in terms of
physical quantities! But a linear slope on a log-log plot
is characteristic of a power function, not a logarithmic
function: In terms of physical quantities, transmittance
of a typical film is a power function of exposure. The
slope of the linear segment, in the log-log domain, is
the exponent of the power function; in the straight-line
region of the films response curve its numerical value is
known as gamma.
Since development of film forms a negative image, a
second application of the process is necessary to form a
positive image; this usually involves making a positive
print on paper from a negative on film. In the reversal
film used in 35 mm slides, developed silver is removed
by a bleaching process, then the originally unexposed
and undeveloped latent silver remaining in the film is
converted to metallic silver to produce a positive image.
This cascaded process is repeated twice in the
processing of motion picture film. It is important that
the individual power functions at each stage are kept
under tight control, both in the design and the
processing of the film. To a first approximation, the
intent is to obtain roughly unity gamma through the
entire series of cascaded processes. Individual steps may
depart from linearity, as long as approximate linearity is
restored at the end of the chain.
Now, heres a surprise. If a film system is designed and
processed to produce exactly linear reproduction of
intensity, reflection prints look fine. But projected
transparencies slides and movies look flat, apparently lacking in contrast! The reason for this involves
another aspect of human visual perception: the
surround effect.
98
Figure 6.4 Surround effect. The

three gray squares surrounded
by white are identical to the
three gray squares surrounded
by black, but the contrast of the
black-surround series appears
lower than that of the whitesurround series.
5.5
Surround effect
LeRoy E. DeMarsh and Edward

J. Giorgianni, Color Science for
Imaging Systems, in Physics
Today, September 1989, 4452.
As explained in Adaptation, on page 85, human vision

adapts to an extremely wide range of viewing conditions. One of the mechanisms involved in adaptation
increases our sensitivity to small brightness variations
when the area of interest is surrounded by bright
elements. Intuitively, light from a bright surround can
be thought of as spilling or scattering into all areas of
our vision, including the area of interest, reducing its
apparent contrast. Loosely speaking, the vision system
compensates for this effect by stretching its contrast
range to increase the visibility of dark elements in the
presence of a bright surround. Conversely, when the
region of interest is surrounded by relative darkness,
the contrast range of the vision system decreases: Our
ability to discern dark elements in the scene decreases.
The effect is demonstrated in Figure 6.4 above, from
DeMarsh and Giorgianni.
The surround effect has implications for the display of
images in dark areas, such as projection of movies in a
cinema, projection of 35 mm slides, or viewing of television in your living room. If an image is viewed in a
dark or dim surround, and the intensity of the scene is
reproduced with correct physical intensity, the image
will appear lacking in contrast.
CHAPTER 6
GAMMA
99
Film systems are designed to compensate for viewing

surround effects. Transparencies (slide) film is intended
for viewing in a dark surround. Slide film is designed to
have a gamma considerably greater than unity about
1.5 so that the contrast range of the scene is
expanded upon display. Video signals are coded in a
similar manner, taking into account viewing in a dim
surround, as I will describe in a moment.
The important conclusion to take from this section is
that image coding for the reproduction of pictures for
human viewers is not simply concerned with mathematics, physics, chemistry, and electronics. Perceptual
considerations play an essential role in successful image
systems.
Gamma in video
In a video system, gamma correction is applied at the
camera for the dual purposes of coding into perceptually uniform space and precompensating the
nonlinearity of the displays CRT. Figure 6.5 opposite
summarizes the image reproduction situation for video.
Gamma correction is applied at the camera, at the left;
the display, at the right, imposes the inverse power
function.
Coding into a perceptual domain was important in the
early days of television because of the need to minimize the noise introduced by over-the-air transmission.
However, the same considerations of noise visibility
apply to analog videotape recording, and also to the
quantization noise that is introduced at the front end of
a digital system when a signal representing intensity is
quantized to a limited number of bits. Consequently, it
is universal to convey video signals in gamma-corrected
form.
As explained in Gamma in film, on page 96, it is important for perceptual reasons to stretch the contrast
ratio of a reproduced image when viewed in a dim
surround. The dim surround condition is characteristic
100
al
gin
Ori e
n
sce
r/
nne
Sca ra
e
cam
on
issi
nsm
Tra m
te
sys
pla
Dis
ced
du
pro
Re e
n
sce
Figure 6.5 Image reproduction in video.

ver
ser
Intensity from the scene is reproduced at the display,
b
O
with a scale factor to account for overall intensity change.
However, the ability of vision to detect an intensity difference is not
uniform from black to white, but is approximately a constant ratio about
1 percent of the intensity. In video, intensity from the scene is transformed by a
function similar to a square root into a nonlinear, perceptually uniform signal that is transmitted. The camera is designed to mimic the human visual system, in order to see lightness
in the scene the same way that a human observer would; noise introduced by the transmission
system then has minimum perceptual impact. The nonlinear signal is transformed back to linear
intensity at the display, using the 2.5-power function that is intrinsic to the CRT.
of television viewing. In video, the stretching is

accomplished at the camera by slightly undercompensating the actual power function of the CRT to obtain
an end-to-end power function with an exponent of 1.1
or 1.2. This achieves pictures that are more subjectively
pleasing than would be produced by a mathematically
correct linear system.
0 .45 =
1
2.222
= 0 .4545
2.2
0 .45 2.5 1.13
Rec. 709 specifies a power function exponent of 0.45.

The product of the 0.45 exponent at the camera and
the 2.5 exponent at the display produces the desired
end-to-end exponent of about 1.13. An exponent of
0.45 is a good match for both CRTs and for perception. Some video standards have specified an exponent
of 1 2.2.
Emerging display devices such as liquid crystal displays
(LCDs) have nonlinearity different from that of a CRT.
But it remains important to use image coding that is
CHAPTER 6
GAMMA
101
1.0
Power function exponent
0.45
0.8
Video signal
Toe slope
4.5
0.6
0.4
0.2
Figure 6.6 Rec. 709

transfer function.
0.081
0
0.2
0.018
0.4
0.6
0.8
1.0
Light intensity, relative
well matched to perception. Furthermore, image interchange standards using the 0.45 value are very well
established. The economic importance of equipment
that is already built to these standards will deter any
attempt to establish new standards just because they
are better matched to particular devices. We can expect
new display devices to incorporate local correction, to
adapt between their intrinsic transfer functions and the
transfer function that has been standardized for image
interchange.
Rec. 709 transfer function
Figure 6.6 above illustrates the transfer function
defined by the international Rec. 709 standard for
high-definition television (HDTV). It is basically a
power function with an exponent of 0.45. Theoretically a pure power function suffices for gamma correction; however, the slope of a pure power function is
infinite at zero. In a practical system such as a television camera, in order to minimize noise in the dark
regions of the picture it is necessary to limit the slope
(gain) of the function near black. Rec. 709 specifies a
slope of 4.5 below a tristimulus value of +0.018, and
stretches the remainder of the curve to maintain func102
tion and tangent continuity at the breakpoint. In this

equation the red tristimulus (linear light) component is
denoted R, and the resulting gamma-corrected video
signal is denoted with a prime symbol, R709. The
computation is identical for the other two components:
Eq 6.1
R 0.018
4.5 R,
R709 =
0.45
1
.
099
0
.
099
,
0
.018 < R
R
Standards for conventional 525/59.94 video have

historically been very poorly specified. The original
NTSC standard called for precorrection assuming a
display power function of 2.2. Modern 525/59.94 standards have adopted the Rec. 709 function.
Formal standards for 625/50 video call for precorrection for an assumed power function exponent of 2.8 at
the display. This is unrealistically high. In practice the
Rec. 709 transfer function works well.
SMPTE 240M transfer function
SMPTE Standard 240M for 1125/60 HDTV was
adopted several years before international agreement
was achieved on Rec. 709. Virtually all HDTV equipment that has been deployed as I write this uses
SMPTE 240M parameters. The 240M parameters are
slightly different from those of Rec. 709:
Eq 6.2
R 0.0228
4.0 R,
R240 M =
0.45
1
.
1115
0
.
1115
,
0
.0228 < R
R
The difference between the SMPTE 240M and Rec. 709

transfer functions is negligible for real images. It is a
shame that international agreement could not have
been reached on the SMPTE 240M parameters that
were widely implemented at the time the CCIR (now
ITU-R) discussions were taking place.
The Rec. 709 values are closely representative of
current studio practice, and should be used for all but
very unusual conditions.
CHAPTER 6
GAMMA
103
1.0
Intensity, relative
0.8
0.6
0.4
0.2
0
0 mV
54 mV
16
0
Analog video, zero setup

Analog video, 7.5% setup
Rec. 601 digital video code
Typ. computer framebuffer code
700 mV
714 mV
235
255
Figure 6.7 CRT signal levels and intensity. A video signal may be represented as analog voltage, with
zero setup or with 7.5-percent setup. Alternatively, the signal may be represented digitally using
coding from 0 to 255 (for computer graphics), or Rec. 601 coding from 16 to 235 (for studio video).
CRT transfer function details

This section provides technical information concerning
the nonlinearity of a CRT. This section is important if
you wish to determine the transfer function of your
CRT, to calibrate your monitor, or to understand the
electrical voltage interface between a computer framebuffer and a monitor.
Figure 6.7 above illustrates the function that relates
signal input to a CRT monitor to the light intensity
produced at the face of the screen. The graph characterizes a grayscale monitor, or each of the red, green,
and blue components of a color monitor. The x-axis of
the graph shows the input signal level, from reference
black to reference white. The input signal can be
presented as a digital code or an analog voltage
according to one of several standards. The y-axis shows
the resulting intensity.
For analog voltage signals, two standards are in use.
The range 54 mV to 714 mV is used in video systems
104
that have 7.5-percent setup, including composite

525/59.94 systems such as NTSC, and computer video
systems that conform to the levels of the archaic
EIA RS-343-A standard. Computer framebuffer digitalto-analog converters often have 7.5-percent setup;
these almost universally have very loose tolerance of
about 5 percent of full scale on the analog voltage
associated with reference black. This induces black level
errors, which in turn cause serious errors in the intensity reproduced for black. In the absence of a display
calibrator, you must compensate these framebuffer
black-level errors by adjusting the Black Level (or
Brightness) control on your monitor. This act effectively marries the monitor to the framebuffer.
The accuracy of black level reproduction is greatly
improved in newer analog video standards that have
zero setup. The voltage range 0 to 700 mV is used
in zero-setup standards, including 625/50 video in
Europe, and all HDTV standards and proposals.
For the 8-bit digital RGB components that are ubiquitous in computing, reference black corresponds to
digital code 0, and reference white corresponds to
digital code 255. The standard Rec. 601 coding for
studio digital video places black at code 16 and white
at code 235. Either of these digital coding standards
can be used in conjunction with an analog interface
having either 7.5-percent setup or zero setup. Coding
of imagery with an extended color gamut may place
the black and white codes even further inside the
coding range, for reasons having to do with color
reproduction that are outside the scope of this chapter.
The nonlinearity in the voltage-to-intensity function of
a CRT originates with the electrostatic interaction
between the cathode and the grid that controls the
current of the electron beam. Contrary to popular
opinion, the CRT phosphors themselves are quite linear,
at least up to an intensity of about eight-tenths of peak
white at the onset of saturation.
CHAPTER 6
GAMMA
105
Knowing that a CRT is intrinsically nonlinear, and that

its response is based on a power function, many users
have attempted to summarize the nonlinearity of a CRT
display in a single numerical parameter using the relationship:
Eq 6.3
intensity = voltage
This model shows wide variability in the value of

gamma, mainly due to black-level errors that the model
cannot accommodate due to its being pegged at
zero: The model forces zero voltage to map to zero
intensity for any value of gamma. Black-level errors
that displace the transfer function upward can be fit
only by choosing a gamma value that is much smaller
than 2.5. Black-level errors that displace the curve
downward saturating at zero over some portion of
low voltages can get a good fit only by having a
value of gamma that is much larger than 2.5. In effect,
the only way the single gamma parameter can fit a
black-level variation is to alter the curvature of the
function. The apparent wide variability of gamma
under this model has given gamma a bad reputation.
A much better model is obtained by fixing the exponent of the power function at 2.5, and using a single
parameter to accommodate black-level error:
Eq 6.4
intensity = (voltage + )
2.5
This model fits the observed nonlinearity much better

than the variable-gamma model.
William B. Cowan, An Inexpensive Scheme for Calibration of a
Colour Monitor in terms of CIE
Standard Coordinates, in
Computer Graphics, vol. 17,
no. 3 (July 1983), 315321.
If you want to determine the nonlinearity of your

monitor, consult the article by Cowan. In addition to
describing how to measure the nonlinearity, he
describes how to determine other characteristics of
your monitor such as the chromaticity of its white
point and its primaries that are important for accurate color reproduction.
106
Gamma in computer graphics

Computer graphics software systems generally perform
calculations for lighting, shading, depth-cueing, and
antialiasing using intensity values that model the physical mixing of light. Intensity values stored in the framebuffer are gamma-corrected by hardware lookup tables
on the fly on their way to the display. The power function at the CRT acts on the gamma-corrected signal
voltages to reproduce the correct intensity values at the
face of the screen. Software systems usually provide a
default gamma value and some method to change the
default.
The voltage between 0 and 1 required to display a red,
green, or blue intensity between 0 and 1 is this:
Eq 6.5
signal = intensity
In the C language this can be represented as follows:

signal = pow((double)intensity,(double)1.0/gamma);
In the absence of data regarding the actual gamma

value of your monitor, or to encode an image intended
for interchange in gamma-corrected form, the recommended value of gamma is 1 0.45 (or about 2.222).
You can construct a gamma-correction lookup table
suitable for computer graphics applications, like this:
#define SIG_FROM_INTEN(i) \
((int)( 255.0 * pow((double)(i) / 255.0, 0.45)))
int sig_from_inten[256], i;
for (i=0; i<256; i++)
sig_from_inten[i] = SIG_FROM_INTEN(i);
Loading this table into the hardware lookup table at

the output side of a framebuffer will cause RGB intensity values with integer components between 0 and
CHAPTER 6
GAMMA
107
255 to be gamma-corrected by the hardware as if by

the following C code:
red_signal = sig_from_inten[r];
green_signal = sig_from_inten[g];
blue_signal = sig_from_inten[b];
A lookup table at the output of the framebuffer enables

signal representations other than linear-light. If gammacorrected video signals are loaded into the framebuffer,
then a unity ramp is appropriate at the lookup table.
This arrangement will maximize perceptual performance.
The availability of a lookup table at the framebuffer
makes it possible for software to perform tricks, such as
inverting all of the lookup table entries momentarily to
flash the screen without modifying any data in the
framebuffer. Direct access to framebuffer lookup tables
by applications makes it difficult or impossible for
system software to avoid annoyances, such as
colormap flashing, and to provide features such as
accurate color reproduction. To allow the user to make
use of these features, applications should access lookup
tables in the structured ways that are provided by the
graphics system.
Gamma in video, computer graphics, SGI, and Macintosh
Transfer functions in video, computer graphics, Silicon
Graphics, and Macintosh are sketched in Figure 6.8
opposite. Video is shown in the top row. Gamma
correction is applied at the camera, and signals are
maintained in a perceptual domain throughout the
system until conversion back to intensity at the CRT.
What are loosely called JPEG
files use the JPEG File Interchange Format (JFIF). Version
1.02 of that specification states
that linear-light coding (gamma
1.0) is used. That is seldom the
case in practice. Instead, power
laws of 0.45, 1 1.8 0.55, or
1.7
2.5 0.68 are used.
108
Computer graphics systems generally store intensity

values in the framebuffer, and gamma-correct on the
fly through hardware lookup tables on the way to the
display, as illustrated in the second row.
Silicon Graphics computers, by default, use a lookup
table with a 1.7-power function; this is shown in the
third row.
TRANSFER
FUNCTION
FRAMESTORE
(implicit)
Video, PC
INTENSITY
MONITOR
2.5
0.45
RAMP
2.22
Computer
Graphics
(implicit)
FRAMEBUFFER
LOOKUP
TABLE
MONITOR
2.5
INTENSITY
0.45
RAMP
2.22
8-bit Bottleneck
Silicon
Graphics
INTENSITY
LOOKUP
TABLE
FRAMEBUFFER
LOOKUP
TABLE
2.5
1
1.7
1.47
0.68
LOOKUP
TABLE
0.59
FRAMEBUFFER
LOOKUP
TABLE
Macintosh
INTENSITY
MONITOR
MONITOR
2.5
1
1.8
0.56
QuickDraw RGB codes
1.45
0.69
Figure 6.8 Gamma in video, computer graphics, SGI, and Macintosh. In a video system, shown in the
top row, a transfer function in accordance with vision is applied at the camera. The middle row illustrates computer graphics: Calculations are performed in the linear light domain and gamma correction is applied in a lookup table at the output of the framebuffer. Silicon Graphics computers take a
hybrid approach: Part of the correction is accomplished in software, and a 1 1.7 power function is
loaded into the lookup table. The approach used by Macintosh computer sketched in the bottom row.
JFIF files originated on Macintosh ordinarily encode R, G, and

B tristimulus (intensity) values
raised to the 1 1.8 power.
Macintosh computers use the approach shown in the

bottom row. Part of gamma correction is effected by
application software prior to presentation of RGB
values to the QuickDraw graphics subsystem; the
remainder is accomplished in the lookup tables. The
dominance of Macintosh computers in graphic arts and
prepress has made gamma 1.8 a de facto standard.
CHAPTER 6
GAMMA
109
Pseudocolor
In Raster images in computing, on page 1, I described
how pseudocolor systems have lookup tables whose
outputs are directly mapped to voltage at the display. It
is conventional for a pseudocolor application program
to provide, to a graphics system, RGB color values that
are already gamma-corrected for a typical monitor. A
pseudocolor image stored in a file is accompanied by a
colormap whose RGB values incorporate gamma
correction. If these values are loaded into a 24-bit
framebuffer whose lookup table is arranged to gammacorrect intensity values, the pseudocolor values will be
gamma-corrected a second time, resulting in poor
image quality.
If you want to recover intensity from gamma-corrected
RGB values, for example to back-out the gamma
correction that is implicit in the RGB colormap values
associated with an 8-bit colormapped image, construct
an inverse-gamma table. You can employ a lookup
technique as above, building an inverse table
INTEN_FROM_SIG using code similar to the
SIG_FROM_INTEN code on page 107, but with an
exponent of 1 0.45 instead of 0.45. Be aware that the
perceptual uniformity of the gamma-corrected image
will be compromised by mapping into the 8-bit intensity domain: Contouring which I will discuss on page
page 113 will be introduced into the darker shades.
Halftoning
Figure 6.9
110
Continuous-tone (grayscale or color) image data can be

reproduced using a process, such as color photography
or thermal dye transfer printing, where a continuously
variable amount of color material can be deposited at
each point in the image. But some reproduction
processes offset lithographic printing and laserprinters, for example place a fixed density of color at
each point in the reproduced image. Grayscale and
color images are halftoned or screened in order to
be displayed on these devices. Halftoning produces
apparent continuous tone by varying the area of small
dots in a regular array. Viewed at a sufficient distance,
an array of small dots produces the perception of light

gray, and an array of large dots produces dark gray.
Halftone dots are usually placed on a regular grid,
although stochastic screening has recently been introduced, which modulates the spacing of the dots rather
than their size. The screening is less visible because the
pattern is not spatially correlated.
Robert Ulichney, Digital

Halftoning. Boston: MIT
Press, 1988.
Peter Fink, PostScript
Screening: Adobe Accurate
Screens. Mountain View,
CA: Adobe Press, 1992.
In order for halftoning to produce a reasonably good

impression of continuous tone, the individual dots must
not be too evident. If it is known that a reproduction
will be viewed at a certain distance, then the number of
screen lines per inch can be determined. This measurement can be expressed in terms of the angle subtended
by the screen line pitch at the intended viewing
distance: If the screen line pitch subtends any more
than about one minute of arc at the viewers retina,
screen lines are likely to be visible.
The standard reference to halftoning algorithms is
Ulichney, but he does not detail the nonlinearities
found in practical printing systems. For details about
screening for color reproduction, consult Fink.
Printing
An image destined for halftone printing conventionally
specifies each pixel as dot percentage in film. An imagesetters halftoning machinery generates dots whose
areas are proportional to the requested coverage. In
principle, dot percentage in film is inversely proportional to linear-light reflectance.
Figure 6.10 Dot gain

mechanism.
Two phenomena distort the requested dot coverage

values. First, printing involves a mechanical smearing of
the ink that causes dots to enlarge. Second, optical
effects within the bulk of the paper cause more light to
be absorbed than would be expected from the surface
coverage of the dot alone. These phenomena are
collected under the term dot gain, which is the
percentage by which the light absorption of the printed
dots exceeds the requested dot coverage.
CHAPTER 6
GAMMA
111
1.0
Reflectance
0.8
Calc. Reflectance
1.75
0.025+ 100-C
102
0.6
Typ. Reflectance
Dot gain
0.4
Dot Gain
0.2
0.0
100
80
60
40
20
Dot Percentage in Film, C

Figure 6.11 Transfer function in offset printing. Dot gain refers to light absorption in excess of that
predicted by ink coverage, or dot percentage in film, alone. When expressed as intensity instead of
absorption, and as total absorption instead of excess absorption, the standard dot gain characteristic
of offset printing reveals a transfer function roughly similar to that of a CRT.
Standard offset printing produces a dot gain at 50

percent of about 22 percent: When 50 percent absorption is requested, 72 percent absorption is obtained.
The midtones print darker than requested. This results
in a transfer function from code to reflectance that
closely resembles the voltage-to-light curve of a CRT.
Correction of dot gain is conceptually similar to gamma
correction in video: Physical correction of the defect
in the reproduction process is very well matched to the
lightness perception of human vision. Coding an image
in terms of dot percentage in film involves coding into a
roughly perceptually uniform space. The standard dot
gain functions employed in North America and Europe
correspond to intensity being reproduced as a power
function of the digital code, where the numerical value
of the exponent is about 1.75, compared to about 2.2
for video or 3 for CIE L*. The value used in printing is
lower than the optimum for perception, but works well
for the rather low contrast ratio of offset printing.
112
The default power function used in the Macintosh

produces a mapping from code to reflectance that is
similar to that of printing: Raw QuickDraw RGB
codes subtracted from 255 and sent to an imagesetter
produce reflectance that closely matches the intensity
displayed on the Macintosh monitor.
I have described the linearity of conventional offset
printing. Other halftoned devices have different characteristics, and require different corrections.
Limitations of 8-bit intensity
As mentioned in Gamma in computer graphics, on
page 107, computer graphics systems that render
synthetic imagery usually perform computations in the
linear-light (intensity) domain. Graphics accelerators
usually perform Gouraud shading in the intensity
domain, and store 8-bit intensity components in the
framebuffer. Eight-bit intensity representations suffer
contouring artifacts, due to the contrast sensitivity
threshold of human vision discussed in Lightness sensitivity, on page 85. The visibility of contouring is
enhanced by a perceptual effect called Mach bands;
consequently, the artifact is sometimes called banding.
In fixed-point intensity coding where black is code zero,
code 100 is approximately the threshold of visibility at
1 percent contrast sensitivity: Code 100 represents the
darkest gray that can be reproduced without the increments between adjacent codes being perceptible. I call
this value best gray. One of the determinants of the
quality of an image is the ratio of intensities between
brightest white and best gray. In an 8-bit linear-light
system, this ratio is a mere 2.5:1. If an image is
contained within this contrast ratio, then it will not
exhibit banding but the low contrast ratio will cause the
image to appear flat. If an image has a contrast ratio
substantially larger than 2.5:1, then it is liable to show
banding. In 12-bit linear light coding the ratio improves
to 40:1, which is adequate for the office but does not
approach the quality of a photographic reproduction.
CHAPTER 6
GAMMA
113
High-end systems for computer generated imagery

(CGI) usually do not depend on hardware acceleration.
They perform rendering calculations in the intensity
domain, perform gamma correction in software, then
write gamma-corrected values into the framebuffer.
These systems produce rendered imagery without the
quantization artifacts of 8-bit intensity coding.
The future of gamma correction
Work is underway to implement facilities in graphics
systems to allow device-independent specification of
color. Users and applications will be able to specify
colors, based on the CIE standards, without concern for
gamma correction. When this transition is complete, it
will be much easier to obtain color matching across
different graphics libraries and different hardware. In
the meantime, you can take the following steps:
Establish good viewing conditions. If you are using a
CRT display, you will get better image quality if your
overall ambient illumination is reduced.
Ensure that your monitors Black Level (or Brightness)
control is set to correctly reproduce black elements on
the screen.
Use gamma-corrected RGB representations whenever you can. An image coded with gamma correction
has good perceptual uniformity, resulting in an image
with much higher quality than one coded as 8-bit
intensity values.
When you exchange images either in truecolor or
pseudocolor form, code RGB color values using the
Rec. 709 gamma value of 1 0.45.
In the absence of reliable information about your
monitor, display pictures assuming a monitor gamma
value of 2.5. If you view your monitor in a dim
surround, use a lower value of about 2.2.
114
A Technical Introduction to Digital Video - Errata
Errata
This page gives you access to the errata for the book A Technical Introduction to Digital Video, by
Charles Poynton, published by John Wiley & Sons (1996). I provide the Errata in typographic-quality
Acrobat PDF format, including replacement figures, and in HTML format (viewable directly in your
web browser). You can access information about formats.
The book is in its fourth printing. To determine which printing of the book you have, turn to the
copyright page of the front matter, page iv, and examine the bottom line: The rightmost digit of that
line indicates which printing you have. The second, third, fourth, and fifth printings are identical.
Printing
Updated
Acrobat PDF,
Typeset quality
2002-04-24 Acrobat PDF format
2, 3, 4 & 5 2002-04-24 Acrobat PDF format
HTML
Errata for the first printing

Errata for the second and subsequent printings
Charles Poynton - A Technical Introduction to Digital Video

2002-04-24
http://www.poynton.com/notes/TIDV/Errata/index.html [15/01/2005 16:25:23]
Poynton's Document Formats
Charles Poynton Document formats

I make many of my documents available in HTML form. For documents whose content depends on
typographic presentation and/or on illustrations, I distribute Acrobat PDFformat. Details on PDF
format, and a few other formats, is presented here.
PDF - Portable Document Format - .pdf

I'm an Adobe Acrobat fan. Acrobat uses Portable Document Format (.pdf) files.
Freely-distributable Acrobat Readers are available for Macintosh, Windows, MSDOS and UNIX. PDF files can be viewed on-screen and printed to PostScript or
non-PostScript printers without regard for which fonts are installed on your
system. Version 3.0.2 for the Mac is in the file ar302.bin at Adobe's ftp site
(MacBinary format, 5581952 bytes). Version 4.0 for the Mac is in the file
ar40eng.sit.bin at Adobe's ftp site (MacBinary format, 4175232 bytes).
Netscape has integrated PDF viewer technology within version 3 (and higher) of the Netscape
browser, so it is easy to use PDF with the web. Personally, I despise the user interface. Five (!)
window titles - the menu bar, the Netscape window title bar, Netscape's buttons (or icons, or both), the
URL line, then an Acrobat viewer toolbar - all in addition to whatever navigation aids might already
be present in the PDF document! But there's no accounting for style on the web. I disable the Plug-in,
and instead use the Acrobat Reader - actually, Acrobat Exchange, since I am licenced for the
commercial Acrobat version 3 package. Adobe tells you how to configure Netscape to view PDF.
The best way to get Acrobat is to access Adobe's web site. The Adobe server is
very reliable, but there are mirror sites.
If you transfer an Acrobat PDF file to a Windows or MS-DOS machine, transfer in binary mode.
Although a version 1.0 of the PDF format was advertised as being 7-bit ASCII and "platform
independent," in fact a PDF file contains coded byte offsets. If you transfer in ASCII (text) mode you
will alter the lineends from one character (CR or LF) to two (CR/LF) and thereby disturb the offsets.
The Reader will rebuild them, but not before frightening you with a "File damaged" alert. Binary
mode is now formally part of the PDF 1.1 standard (and its successor versions). Binary mode was
formerly deprecated by Adobe, but starting with PDF-1.1 (corresponding to version 2 of the Acrobat
products), Adobe is encouraging the use of binary PDF files. This gives you all the more reason to
transfer in binary mode.
Phil Smith has written an excellent, short note Noddy's Guide to PDF.
http://www.poynton.com/Poynton-formats.html (1 di 2) [15/01/2005 16:25:33]
Poynton's Document Formats
PostScript - .ps
These files can be printed with a PostScript printer and viewed on most UNIX platforms. To learn
more about PostScript, read Allen Braunsdorf's PostScript FAQ. Most of the PostScript files here have
no embedded fonts and require only Times, Helvetica, Palatino and Symbol from the core 35
("LaserWriter Plus") set of fonts that are built-in to almost all PostScript laser printers. The PostScript
files use generic PPDs and are laid out with generous margins for US Letter size paper. I confess I
don't know how well they print to A4.
If you are running Windows, UNIX or VMS, you can view PostScript using GhostScript
or GhostView. You can obtain GhostScript from ftp.cs.wisc.edu.
You can view and print PostScript on a Macintosh using Mac GhostView. Version 5.5 is
available. See my archaic page Viewing PostScript on a Macintosh.
Gnu zip Compressed - .gz

Many of the PostScript files on my ftp server are compressed using gzip ("gnu-zip"), a publicallyavailable compression package. You can obtain various versions of the gzip program, including its
source code, by anonymous ftp from the following locations:
prep.ai.mit.edu:/pub/gnu/
gatekeeper.dec.com:/pub/GNU/
ftp.uu.net:/systems/gnu/
Gnu zip replaces the traditional UNIX compress program. The major advantage of Gzip over
compress (.Z) is that it has no patent infringement strings attached - UNISYS claims a patent on the
LZW compression algorithm upon which compress is based.
If you use a Macintosh, Aladdin's latest StuffIt Expander (freeware) will decode these files. It is
available at all the usual Mac repositories. You can download the freeware StuffIt Expander 5.5 from
Aladdin's ftp site (MacBinary format, 824704 bytes) or the AOL mirror (MacBinary format, 824704
bytes). Aladdin's DropStuff with Expander Expander 5.5 (shareware, $30) adds to StuffIt Expander
additional decoding capabilities such as gnu zip. Obtain it from Aladdin's ftp site (MacBinary format,
2049920 bytes) or the AOL mirror (MacBinary format, 2049920 bytes).
Charles
1999-08-12
http://www.poynton.com/Poynton-formats.html (2 di 2) [15/01/2005 16:25:33]
Viewing PostScript on a Macintosh
This page outlines shareware and freeware methods to view and print PostScript on a Mac. See also
Macintouch's PostScript RIP alternatives.
6.0.1
You can view and print PostScript on a Mac using Thomas Kiffe's MacGhostView, version 2.1. This
port is essentially a stand-alone subset of CMacTEX 3.6, based on GhostScript version 6.0.1; visit
Thomas' MacGhostView page. These files unpack into approximately 8 MB:
macghostview (SEA file, BinHex encoded, approx. 6765 Kbytes)
5.5
Mac GS is port of GhostScript version 5.5 to the Mac, ported by Jeff Schindall in conjunction with
Aladdin Enterprises.
You can consult Aladdin's ghostscript page.
You can consult the manual:
macgsmanual.html (HTML format, 24085 bytes)
Then, you can download the installer:
macgs-550-installer.bin (MacBinary format, 4660864 bytes)
1.0
Mark Lentczner of Glyphic Technology ported GhostScript 3.33 to the Mac; this was released as Mac
GS Viewer 1.0. Glyphic no longer supports this version; it is obsolete.
0.x
http://www.poynton.com/notes/mac/Viewing_postscript.html (1 di 2) [15/01/2005 16:25:37]
Prior to the release of Mac GhostView, I wrote a document that explains how to view and print
PostScript on a Macintosh using Adobe's commercial Acrobat Distiller. The note contains an
introduction to PostScript, information about Printing PostScript on a Mac, a section concerning
Viewing and printing PostScript on a Mac, and an introduction to Encapsulated PostScript (EPS). The
document is now somewhat dated.
PDF format (135,047 bytes)
Charles - Mac
2001-05-09
http://www.poynton.com/notes/mac/Viewing_postscript.html (2 di 2) [15/01/2005 16:25:37]
Charles A. Poynton
56A Lawrence Avenue E
Toronto, ON M4N 1S3
CANADA
tel +1 416 486 3271
fax +1 416 486 3657
poynton@inforamp.net

This note concerns the viewing and printing of PostScript on a Macintosh. The note contains an
Introduction to PostScript, information about Printing PostScript on a Mac,a section concerning
Viewing and printing PostScript on a Mac, and an introduction to Encapsulated PostScript (EPS).
1
Introduction to PostScript
PostScript is a programming language specialized to produce images with type and drawn
elements. It is possible to encode a continuous-tone image into a PostScript file, but TIFF is better
suited to that purpose. A PostScript file is independent of the resolution of the output device: the
interpreter in a particular device produces an image appropriate for that device.
A PostScript file generally contains only characters from the 96-character ASCII set, and accommodates any line-end convention (UNIX, MS-DOS and Mac) so it is relatively easy to move PostScript
files through networks and across e-mail links.
Since PostScript includes a complete programming language, it is generally not possible to convert a
PostScript file into some other format either an object-oriented format like PICT or WMF, or a
bitmapped format like GIF or TIFF without using a full PostScript language interpreter. A few
restricted, specialized dialects of PostScript such as the Adobe Illustrator file format [4] can be
translated without a full interpreter.
PostScript was invented by Adobe Systems, Inc. The specification of the PostScript language [1] and
the PostScript font format [2] have been made public by Adobe, and several PostScript clones are
commercially available. However, the name PostScript is a trademark that belongs to Adobe, and
the copyright to the PostScript Language Reference Manual is the property of Adobe.
A large effort within a community computer programmers has resulted in a free interpreter called
GhostScript. That group takes as their motto, Implementation is the sincerest form of flattery.
2
Printing PostScript on a Mac
Most models of the Apple LaserWriter series or printers have built-in PostScript interpreters. A
Macintosh computer that is attached to a PostScript-equipped LaserWriter, either directly or
through an AppleTalk network, can print PostScript files.
A Macintosh ordinarily generates PostScript on the fly, often from QuickDraw. To print a pre
existing PostScript file you need a program that bypasses the Macs printer driver and downloads
the PostScript file to the printer. You can use Apples LaserWriter Utility, part of Apples system
software. Adobe distributes a comparable program called Downloader. You can also use Rich
Siegels popular and well-respected freeware program DropPS.
The Mac itself has no native capability to interpret a PostScript program and produce an image. If
you have a printer without PostScript capability and but you have access to PostScript files from
the Internet, for example you will have to go to a certain amount of trouble to print the images.
You will also have to go to some trouble to view the images without printing them.
1995/01/13 Charles A. Poynton. All rights reserved.
1 of 3
V IEWIN G PO STSC RIPT O N A M A C INTOSH
Most UNIX workstation vendors have licenced the Display PostScript System (DPS) from Adobe.
Workstations from those vendors can display PostScript directly.
3
Viewing and printing PostScript on a Mac
The GhostScript system and its viewer GhostView work well MS-DOS, Windows, UNIX and OS/2
systems. A version of GhostScript is available for the Macintosh and In theory GhostScript can be
used to view and print PostScript files on a Mac. However, GhostScript for the Macintosh is difficult
to install, difficult to use, and not very robust. It fails on many PostScript files. I cannot recommend
it. A new port is promised soon, so its worth watching GhostScript developments.
I use a highly reliable method that uses commercial Adobe software. My method has two passes.
First, I use the Adobe Acrobat Distiller application to interpret the PostScript code and generate a
Portable Document Format (PDF) file. Then I use Adobes Acrobat Reader to view the resulting
PDF. The method is highly reliable. Both the Distiller and the Reader are commercial-grade software. The Acrobat system has the capability to substitute fonts that were used in the original document but are not available at the viewing Mac. This is accomplished through a limited edition of
SuperATM, which is included in the Reader distribution. Acrobat Reader can also be used to print,
either to a PostScript printer or to a non-PostScript (QuickDraw) printer.
Although the Acrobat Reader is freely available from Adobe, Acrobat Distiller is commercial software. It is available as a self-contained product, and is a component of certain versions of Adobe
Acrobat Pro. The Distiller is bundled with Adobe Illustrator version 5.5; that is how I obtained mine.
For the moment, Adobe is the sole source of PDF technology. Adobe has published the specifications of the PDF file format [3], so we can anticipate PDF becoming available from other vendors.
4
Encapsulated PostScript (EPS)
This section concerns a dialect of PostScript called Encapsulated PostScript (EPS), which is used
extensively in graphic and desktop publishing applications. EPS is standardized by Adobe [5].
A PostScript file may produce no image at all, an image on one page or images on many pages.
Encapsulated PostScript (EPS) is a dialect of PostScript where a single EPS file produces exactly one
image. In addition to being restricted to one image, an EPS file contains information about the
extent of the image on the page the bounding box of the image. This information is included at
the head of the EPS file as PostScript comments that are ignored by a PostScript interpreter but are
recognized by applications and system software. The comments conform to document structuring
conventions (DSC) [6]. Also, the PostScript program in an EPS file has certain restrictions.
Although an EPS file contains all of the PostScript instructions to draw a page, it does not necessarily end with the showpage operator that is required by a printer to print the page. Also, the
image that is drawn may be outside the imaging area of a particular printer. An Apple LaserWriter
cannot print closer than 1 4 inch to the edge of a page, but a very small EPS image might fit entirely
in the non-printable margin. In fact it is common for an EPS image to have its origin at PostScript
coordinates (0, 0), at the extreme corner of the page, which is cropped on a regular LaserWriter. So
an EPS file might print properly when send to a PostScript printer, but it might not. The safest way
to print an EPS file is to import it into an EPS-capable application, then print from that application.
An EPS file may optionally include, within the file, a bitmapped preview image. There are no fewer
than four versions of EPS preview images: PICT, Windows Metafile (WMF), TIFF and ASCII. The
fourth variant is denoted EPSI (EPS interchange). A particular application might recognize one,
two, three or four of these formats. Counting the case of no preview, there are five variants.
V I E W I NG PO S T S CR IP T O N A M A CIN T OSH
Within each variant, the preview may be black and white (one bit) or color (eight-bit). The standard
allows full-color (32-bit) preview images, but applications rarely implement this option.
All EPS-capable applications on a Macintosh recognize the PICT variant, where the preview image is
stored as a PICT resource (ID=256) within the EPS file. Resources are a unique feature of the Mac
file system, and the PICT format is Macintosh-specific. Transferring a Mac EPSF file with a PICT
preview to another platform results in the loss of the preview image. Although the importing application will display the file as a grey rectangle, the EPS will print correctly.
The WMF variant of EPSF is ubiquitous In MS-DOS and Windows systems; the TIFF variant is less
common. In both of these formats the preview image is coded in binary form as part of the file. In
both cases the binary content must be stripped out before transfer to a Macintosh or to a PostScript
interpreter.
In the EPSI format, the preview image is encoded into PostScript comments at the head of the data
portion of the file. This format has a preview, contains no Mac resource and is free of non-PostScript binary data. You would expect the EPSI format to provides a preview image that is usable on
every platform. Regrettably, very few applications and virtually no Mac applications recognize
the EPSI variant. Although Adobe invented the EPSI format, even popular and well-respected
Adobe applications such as Photoshop and Illustrator do not implement EPSI! So in practice EPSI is
nearly useless. The notable exception is FrameMaker, where the EPSI format is implemented on
UNIX, Mac and Windows.
To summarize, printing EPS files takes special measures, and it is very difficult to transport EPS from
one platform to another.
5
References
[1] PostScript language reference manual, Second Edition (The Big Red Book). Adobe Systems
Incorporated, 1990. Addison-Wesley, Reading, MA. ISBN 0-201-18127-4.
[2] Adobe type 1 font format (The Black Book). Adobe Systems Incorporated, 1990. AddisonWesley, Reading, MA. ISBN 0-201-57044-0.
[3] Portable Document Format Reference Manual. Adobe Systems Incorporated, 1993. AddisonWesley, Reading, MA. ISBN 0-201-62628-4.
[4] Adobe Illustrator File Format Specification, Version 3.0 (Draft, 28 October 1992), Adobe
Technical Note LPS5007, available as <ftp://ftp.adobe.com/pub/adobe/DeveloperSupport/
TechNotes/5007AI_Spec_v3.0_Draft.pdf>.
[5] Encapsulated PostScript File Format Specification, Version 3.0 (1 May 1992), Adobe Technical
Note LPS5002, available as <ftp://ftp.adobe.com/pub/adobe/DeveloperSupport/TechNotes/
5002.EPSF_Spec_v3.0.pdf>, also published as Appendix H in PostScript language reference
manual, Second Edition, [1] above.
[6] Document Structuring Conventions Version 3.0, Adobe Technical Note LPS5002, available as <ftp://ftp.adobe.com/pub/adobe/DeveloperSupport/TechNotes/
5001.DSC_Spec_v3.0.ps>, also published as Appendix G in PostScript language reference
manual, Second Edition, [1] above.
Charles Poynton - Macintosh
Charles Poynton Macintosh

On this page you will find links to several archaic documents that I have written concerning the
Macintosh. Several documents are available in Acrobat PDF format.
Before I go any further, if you use Frame (or any high-end DTP product) on a Mac, let me suggest
that you rip out your TrueType fonts !
Macintosh Dialup Internet Access

These notes describe how to configure and use a Macintosh to access the Internet through an Internet
access provider that offers dialup PPP service. A document specific to InfoRamp is available; a
generic version applies to any ISP.
Accessing the IBM Global Network from a

Macintosh
IBM Global Network, a.k.a. IBM Internet Connection (sometimes known in the USA and Canada as
Advantis), is a large and comprehensive supplier of Internet access services. Access to IBM Global
Network is only officially "supported" on Windows, OS/2 and AIX operating systems, but because
the network operates with open, standard protocols, it is possible to gain access using a Macintosh
through a SLIP connection.

You can view and print PostScript on a Mac.
Gamma Correction on the Apple Macintosh

This note details the treatment of gamma correction on Apple Macintosh (tm) computers.
Introduction to RIPping
Raster Image Processing, or RIPping, refers to the conversion of a PostScript file to a high-resolution
bitmap, a necessary process in typographic-quality printing. RIPping is performed by the PostScript
interpreter that resides in a laser printer or filmsetter. You can access several brief documents that
http://www.poynton.com/Poynton-mac.html (1 di 2) [15/01/2005 16:26:02]
Charles Poynton - Macintosh
introduce you to RIPping, in order to get a document from a Macintosh to a commercial printer.
Mac startup keys

Upon powering-up or restarting, various pieces of Mac ROM and System software examine the
keyboard, and take special actions if certain keys are held down. This page sumarizes the actions.
Charles
1998-01-05(a)
http://www.poynton.com/Poynton-mac.html (2 di 2) [15/01/2005 16:26:02]
Rip out your TrueType fonts!
If you use PostScript fonts or ATM,

If you use font-intensive desktop publishing applications on a Mac, and you produce PostScript files,
or you use Type 1 fonts and you print to a non-PostScript printer, you may have experienced spacing
and underlining anomalies. These anomalies often arise when the outline (printer) fonts and bitmap
(screen) fonts are mismatched: You are likely to experience problems if for a particular font name
(such as Palatino), you combine a Type 1 outline (say from Adobe) with a TrueType bitmap (say from
Apple).
I strongly recommend that you avoid this situation by ripping all the TrueType outline fonts out of
your system (except perhaps the ones named for cities). In the following description, I assume that
you are using Mac OS 7.1 or greater.
A suitcase file may contain bitmapped (screen) fonts, and may contain TrueType
outline fonts. Active suitcase files are located in the Fonts subfolder of the System
Folder. If you are running Adobe Type Manager (ATM), Symantec's Suitcase
application, or some other font utility, you might have active fonts in other locations. A
suitcase file has the icon shown at the left.
A bitmapped screen font within a suitcase has this icon. The corresponding name is
listed with a size, in points. Screen fonts are supplied by Apple, Adobe, and other font
vendors.
A TrueType font within a suitcase file has this icon. The three differently-sized letters A
in the icon are suggestive of the capability of the font to be rendered at different sizes.
The corresponding name is listed without a numerical size indication, because the font
can be rendered at any size. If a TrueType font is active, it will mask any like-named
Type 1 font: if you want to use a Type 1 font, then you should remove any like-named
TrueType font.
An Adobe Type 1 font is stored as its own file, with this icon, in the Fonts subfolder.
There are separate files for different styles of the font - Roman, Italic, Bold, Bold Italic,
and so on. It is standard to name these files with the 5-3-3-3 scheme: the first 5
characters are taken from the font name (e.g. Palat for Palatino); groups of three
characters such as Ita or Bol may follow to indicate style variations.
If you have a non-Adobe Type 1 font, it will have a different icon, such as this one.
I recommend that you make inactivate every TrueType fonts that is named after a city, except for
Chicago, Monaco, and Geneva. Inactivate a font by simply moving it out of the Fonts folder.
http://www.poynton.com/notes/mac/Rip_out_TrueType.html (1 di 2) [15/01/2005 16:26:04]
The Chicago, Monaco, and Geneva TrueType fonts are used by system software. I advise you to leave
these installed, in TrueType form. However, I also advise you not to use them in any document! They
are optimized for the screen.
New York is a cheap imitation of Times; Geneva is a cheap imitation of Helvetica. Upon printing to a
PostScript printer, Monaco, Geneva, and New York are normally substituted by Courier, Helvetica,
and Times respectively. If you really want Courier, Helvetica, or Times in a document, make sure you
specify it explicitly: do not rely upon the font substitution feature to to obtain one of these. In fact,
I recommend that you access the PostScript options of Page Setup, and disable Substitute Fonts. This
setting is stored on a per-document basis.
Once you have made any non-city-named TrueType fonts inactive, confirm that for every Type 1
outline font, you have installed the matching screen font, and it is from the same vendor as the Type 1
font. If you Get Info for an Adobe font, the file information dialog includes Adobe's copyright
information.
More detail is available in the document Apple v. Adobe Metrics.
You might also be interested in PostScript vs. TrueType Fonts: Which should you use? by Judy Litt.
There is an archaic option in some older Microsoft products: fractional spacing. Very early Macintosh
applications restricted spacing of characters to the same pitch as pixels on the screen, 1/72 inch. For
some unaccountable reason, Microsoft chose to perpetuate this behaviour into their applications, long
after the LaserWriter was commonplace, and long after the Mac system software allowed essentially
infinite resolution of character positions. To achieve good spacing in these Microsoft products, you
must enable Fractional Spacing (that is, fractions of 1/72 inch) in the Page Setup dialog.
Charles - Mac, Frame, Introduction to RIPping
1998-05-29
http://www.poynton.com/notes/mac/Rip_out_TrueType.html (2 di 2) [15/01/2005 16:26:04]
Apple v. Adobe Metrics
If you use Frame - or any high-end DTP product - on a Mac, let me suggest that you rip out your
TrueType fonts !
Apple's TrueType fonts have metrics - in particular, character widths - that have different values from
the metrics of Adobe fonts of the same name. If you have TrueType fonts installed in your system, as
is the default upon installing system 7, your applications will use the TrueType metrics, and send to
the printer jobs that have character spacing determined by TrueType metrics. If your printer has an
Adobe font whose name matches characters spaced using the TrueType metrics, then the printer will
use the Adobe type 1 font and its metrics, and your spacing will screw up.
The results vary application-by-application, depending on the character quanta (characters, words or
lines) by which the app delivers its PostScript output. The problem can manifest itself as justification
errors, underlines not aligning with their words and text being cropped by its (supposed) bounding
boxes.
The solution is this: for any Adobe font resident in your printer, rip the corresponding TrueType font
out of your system and make sure Adobe's screen (bitmapped) font is installed. Also, make sure you
always un-check Font Substitution, in order to avoid Geneva TrueType being sent to the printer with
its name remapped to Helvetica, invoking the metrics mismatch problem. If you have old, old, old
applications, when printing to a LaserWriter you should always check Fractional Widths, otherwise
your characters will be placed on a rather coarse 1/72-inch grid.
The spacing problem can appear whether or not Adobe Type Manager (ATM) is running. If you are
running ATM then you will want Adobe's outline (printer) fonts as well as the bitmaps. Of course the
presence of a TrueType font in the system file - or in the Fonts folder of System 7.1 and subsequent will render inaccessible the correspondingly-named Adobe type 1 font.
The fundamental causes of this problem is that Apple have not yet learned what "open systems"
means. Apple likes to invent different things, all their own, different from what others before them
have done even if the function is identical. That is why:
Apple uses CR to end a line instead of MS-DOS CR/LF or UNIX LF;

Apple uses a ":" to separate filename components instead of MS-DOS's "\" or UNIX's "/";
Apple invented AFP when NFS would have done just fine;
Apple picked 1904 as the basis of their date system instead of MS-DOS's 1900, to save two or
three clock cycles in the time utilities package;
Apple decided that MPW command names and wildcard characters should be just a little bit
different from their UNIX counterparts.
I like Apple a lot, but it is increasingly difficult to continue being a fan while they keep jerking us
http://www.poynton.com/notes/mac/Rip_out_TT_details.html (1 di 2) [15/01/2005 16:26:06]
around like this.

They know about the TrueType/Adobe spacing problem - they should fix it.
Charles - Mac
1997-07-22
http://www.poynton.com/notes/mac/Rip_out_TT_details.html (2 di 2) [15/01/2005 16:26:06]
bitmap, a necessary process in typographic-quality printing. In high-end printing, RIPping is
performed by the PostScript interpreter that resides in a laser printer or filmsetter.
Here you can access four brief documents that introduce you to RIPping, in order to get a document
from a Macintosh to a commercial printer:
Submitting raw PostScript to a print shop
Preparing to RIP
Preparing to RIP on a DocuTech
Submitting PostScript files to a commercial printer
DocuTechs I've known
You might also want to take a look at Demystifying Service Bureaus, Part I, Part II, Part III, and Part
IV, all by Judy Litt.
I wrote this document in later 1997. Since then, Acrobat PDF has become the format of choice for
submitting jobs to a commercial printer. However, perhaps surprisingly, they're slow to innovate: You
may find that your printer is not confident in handling PDF. If so, hand off PostScript, created as I
described above, or find a printer that is familiar with PDF.
Charles - Mac
2000-03-31
http://www.poynton.com/notes/mac/Intro_to_RIPping.html [15/01/2005 16:26:09]
If you want to have files printed by a commercial printer, it is a good idea to submit raw PostScript
files instead of application (Quark, PageMaker, or FrameMaker) files.
Print shops have a dilemma with accepting raw PostScript: If a PostScript job prints, it is virtually
guaranteed that it will print in exact conformance with the user's expectations when the file was
created. But if the user makes a mistake in the creation of the file that causes a PostScript error upon
interpretation - if the user forgets to include a font, say - it is next to impossible for the print shop to
repair the error in the PostScript file!
The incidence of PostScript errors can be minimized by the print shop if it has access to the customer's
application files. However, submission of application files to the print shop has the problem that a
whole set of files must be submitted: You must submit not only your Quark or PageMaker document,
but also all of its referenced graphics files. Submitting an application file also requires that the print
shop is licensed for, and has installed, that application! Virtually all print shops have Quark and
PageMaker, but they do not have some of the more specialized applications such as FrameMaker.
High-end print shops are generally happy to accept raw PostScript if you are willing to provide it,
because it alleviates them of the need to deal with PostScript errors. But they might be hesitant if they
consider you to be inexperienced with PostScript: They don't want to waste RIP time on failed
PostScript jobs, and they don't want to waste technicians' time explaining things to you. The solution
is to deliver your first PostScript job error-free!
Unfortunately, many imagesetters produce poor error messages, or none at all. Here is an actual error
message that I received upon running a DocuTech job at Kinko's:
http://www.poynton.com/notes/mac/Submitting_raw_ps.html (1 di 2) [15/01/2005 16:26:24]
Try debugging from THAT! Not even a page number is provided! If you take care in preparing your
job, you can prevent nightmares like this.
Adobe is moving rapidly to establish Acrobat PDF as the format of choice for electronic submission
of print jobs. PDF has the advantage of being very robust concerning fonts, and enables use of just
one file for each complete job. The Acrobat Reader is freely and widely available. However, not all
print shops are familiar with Acrobat, and there are one or two potential kinks in the process.
You might be interested in the companion document Preparing to RIP.
Charles - Introduction to RIPping
1998-01-18
http://www.poynton.com/notes/mac/Submitting_raw_ps.html (2 di 2) [15/01/2005 16:26:24]
Preparing to RIP
Preparing to RIP
This note explains how to use a Macintosh to prepare PostScript files for RIPping on a DocuTech, a
filmsetter, or other high-performance PostScript output devices. (You can interpolate from this
document to figure out how to accomplish the same task using a Windows machine, but be aware that
Windows operating system generated notoriously poor PostScript code. Be sure to use Adobe's printer
drivers, not those provided by Microsoft.)
First, obtain the PostScript Printer Description (PPD) file for the intended device. The print shop that
owns the device may be able to provide you with this file. Alternatively, you can access Adobe's
archive of PPD files, or access the web site of the manufacturer of the device. You will have to know
the model number of the device. Place the PPD file into the Printer Descriptions subfolder of
Extensions of your System Folder.
Use the Chooser to select a LaserWriter version 8 driver. I use either Apple's LaserWriter driver,
version 8.5.1, or Adobe's "AdobePS" driver, version 8.5.1. (If you're interested in saving to PDF, or in
virtual printers, consult this note at MacFixIt. If you wish to install both the Adobe and Apple drivers,
install the Apple driver last, or read this note at MacFixIt.) [Apple LaserWriter 8.6 is now available,
as a component of Mac OS 8.5.]
If you are using the desktop printing feature, choose a PostScript printer, access Printing -> Change
Setup ..., and select the appropriate PPD. If you are not using desktop printing, you must have a
PostScript printer online: Open the Chooser, select a PostScript printer, choose Setup ..., then select
the appropriate PPD.
From you application, choose Page Setup and ensure that the printing settings - particularly the page
size settings - are appropriate for your job. If you are using a sophisticated application - Illustrator,
FrameMaker, PageMaker, or Quark XPress, page size is specified in two places: in the Page Setup
dialog, and also in the document itself, through a different dialog that depends upon the particular
application. Double-check this: If you get it wrong, then your job is liable to be printed with the
incorrect cropping or positioning.
I recommend that you disable (deselect) all of the Image & Text PostScript Options - Substitute Fonts,
Smooth Text, Smooth Graphics, Precision Bitmap Alignment, and Unlimited Downloadable Fonts.
When your document is ready, choose Print. Access Print to File, to create a PostScript job file.
(Encapsulated PostScript, EPS, is inappropriate for ripping.) Chose to include All fonts not in PPD.
Many imagesetters, including some early DocuTechs, are PostScript level 1 devices: Choose level 2
or level 3 only if you are quite certain that the device offers the corresponding capability. You can use
either ASCII or binary format; binary leads to more compact files, but can be more difficult to
transport through e-mail. Choose double-sided printing; ensure that you choose the correct binding
edge. Use an extension of ".ps". It will be convenient for your printer if you use a base name that is
http://www.poynton.com/notes/mac/Preparing_to_RIP.html (1 di 2) [15/01/2005 16:26:35]
Preparing to RIP
your last name, or the name of your company. (The filename handout.ps may be unambiguous to you,
but you can imagine that a large printer may receive twenty files daily having this name! On the other
hand, the printer is likely to receive at most one file daily having the name Poynton.ps.)
For any document that you are reproducing in quantity larger than five or ten pieces, you should give
your document a title, a date, and an attribution. (Failure to provide a clear title and author
identification of a document is one of the Ten common mistakes in the typesetting of technical
documents.) If you're going to the trouble to print your document well, then I recommend that you
give your document a title page. If the document comprises more than about 40 pages, arrange with
your printer to bind the document with Wire-O or plastic spiral binding, with your cover page on the
front printed on card stock material, and a blank piece of card stock material as a back cover. Provide
your cover in its own PostScript file, separate from your main document.
In addition to the PostScript job file, I generally provide an Acrobat PDF file to be used in case of
PostScript failures. Not all commercial printing shops are familiar with PDF at the moment, though
Adobe is moving rapidly to establish PDF as the format of choice for electronic submission of print
jobs. You might find that the print shop is happier with Acrobat PDF than with PostScript. Take care,
though, that any critical fonts are embedded in your PDF - otherwise, they will be synthesized at the
printers, and may not appear quite as you want.
Detail specific to the Xerox DocuTech is contained in the companion document, Preparing to RIP on
a DocuTech.
Submit the file according to the companion document, Submitting PostScript files to a commercial
printer.
1998-12-13
http://www.poynton.com/notes/mac/Preparing_to_RIP.html (2 di 2) [15/01/2005 16:26:35]
Ten common mistakes
Ten common mistakes in the

typesetting of technical documents
Charles Poynton
The widespread use of word processors has permitted - even forced - many technical authors to
typeset their own work. But most technical authors have no familiarity with the principles of
typesetting. This note catalogs what I consider to be the ten most common mistakes, in descending
frequency of incidence.
You are reading the HTML version of this document. Typographic-quality versions of this
document are available in Acrobat PDF format (62129 bytes). You can access information
about formats.
Inside these parentheses ( ) should be a single space character. Should you see gobbledigook instead
of a single space, some parts of this document will be garbled. Please contact the developers of your
web browser and explain to them that they should properly implement the non-breaking space code
that is standardized in ISO 8859-1 and HTML-2.0.
1. Thoughtless use of monospace fonts

In the Courier typeface, and other typefaces restricted by the mechanical constraints of typewriters,
the letter i is forced to have the same set width as the letter W. It is quite legitimate to use Courier, or
another monospace font, if you want your document to look like it was produced on a typewriter. But
if you want a document to appear attractive and permanent, choose a proportional font.
2. Lack of consideration for line length and type size
A 66 character line is widely regarded as ideal for readability. With 12-point Courier type - so-called
Pica, with ten characters per inch - 66 characters make a 6.6-inch line. Set on an 8.5-inch page width,
this leaves a reasonable margin of about an inch on each side. But 12-point type, especially Courier, is
too large for all but very unusual cases. With 10-point Palatino, quite suitable for a technical
document, a typical 6.6-inch line has 100 characters, far too many for continuous reading. To use 10point Palatino, a line length of about 4 1/8 inches would be ideal. If you want to use a single column,
consider using 11-point type in a column about 5 inches wide. If your document is almost wholly text,
consider using 9-point type in a layout with two 3-inch columns.
3. Gigantic heads
http://www.poynton.com/notes/typesetting/index.html (1 di 3) [15/01/2005 16:26:40]
Ten common mistakes
It is obviously necessary to distinguish headings (or heads) from body text. Normally, a head is placed
in the margin, or set off by spacing. In these circumstances, it is unnecessary to set the head in a size
larger than the body type, and usually unnecessary to use a different font family. If you are unfamiliar
with typesetting, simply set the head in boldface type of the same size as the body face.
4. Mistaken application of typewriter conventions
Follow the period at the end of a sentence by a single space, not two. Use a single paragraph mark to
terminate a paragraph, not two. To achieve spacing, or to force a paragraph to the top of a page, use
your word processor's spacing and positioning controls instead of blank lines. It is wrong to set two
hyphens in place of a dash: Use an en dash surrounded by normal (word) spaces. Modern word
processors and page layout programs allow you to use typographic characters. Use the straight single
and double quotes ' and " only to denote minutes, seconds, feet, and inches; and then only if there is
insufficient space to spell out the unit. For all other purposes, use typographer's ("curly") quotes.
5. Failure to use italics for emphasis
Underlining is strictly for typewriters. Modern word processors and page layout programs give access
to italic typefaces: Use italics for emphasis. To use boldface would make your page look blotchy. Use
quotation marks only for their intended purpose: quotations. To use quotation marks for emphasis is
disruptive to smooth reading of your text.
6. Failure to control line breaks
Modern word processors and page layout programs allow you to control line breaks and hyphenation.
Use a nonbreaking space to prevent adjacent elements - such as a numerical quantity and its
associated unit - from being separated by a line or page break. Use a nonbreaking hyphen to avoid the
elements of a compound modifier, such as 35-millimeter, from being broken.
7. Failure to clearly identify paragraphs
A surprising number of technical articles fail to make a clear demarcation between paragraphs. This
does a great disservice to the reader. You must choose one of two alternatives to identify paragraphs:
Either indent the first line of every paragraph, leaving all lines on the page with the same spacing, or
use no indentation at all and place a blank line between paragraphs.
8. Careless setting of fractions
Set a fraction using superscript for the numerator and subscript for the denominator. Use the fraction
slash character as a separator - on a Macintosh, access it as Shift-Option-1.
9. Errors in orientation of figures
If your document includes figures oriented differently from the text, it will be necessary for the reader
to rotate your document. You must ensure that he or she has to turn the document only once per page,
Ten common mistakes
and always in the same direction. Arrange the layout so that the reader rotates the body text clockwise
90 degrees to view a figure. In other words, when viewing the text in its natural orientation, place the
top of a rotated figure to the left.
10. Failure to provide clear title and author
A surprising number of technical authors fail to clearly identify their work. If you want your reader to
be certain who wrote the work, to cite it, and to determine whether he or she has the most recent
version, you must take care to provide complete information on the first page.
Charles Poynton
Charles Poynton
tel +1 416 413 1377
fax +1 416 413 1378
poynton @ poynton.com
www.poynton.com
Ten common mistakes in the

typesetting of technical documents
The widespread use of word processors has permitted even forced
many technical authors to typeset their own work. But most technical
authors have no familiarity with the principles of typesetting. This note
catalogs what I consider to be the ten most common mistakes, in
descending frequency of incidence.
1
Thoughtless use of monospace fonts: In the Courier typeface, and

other typefaces designed to accommodate the mechanical constraints
of typewriters, the letter i is forced to have the same set width as the
letter W. It is quite legitimate to use Courier, or another monospace
font, if you want your document to look like it was produced on
a typewriter. But if you want a document to appear attractive and
permanent, choose a font that has proportional spacing.
Gigantic heads: It is obviously necessary to distinguish headings (or

heads) from body text. Normally, a head is placed in the margin, or set
off by spacing. In these circumstances, it is unnecessary to set the head
in a size larger than the body type, and usually unnecessary to use
a different font family. If you are unfamiliar with typesetting, simply set
the head in boldface type of the same size as the body face. If you
must use a larger size, dont use boldface.
Failure clearly identify title and author: A surprising number of technical authors fail to clearly identify their work. If you want your reader
to be certain who wrote the work, to be able to cite it, and to determine whether he or she has the most recent version, you must take
care to provide title, author, and version or date on the first page.
Inappropriate application of typewriter conventions: Follow the period

at the end of a sentence by a single space, not two. Use a single paragraph mark to terminate a paragraph, not two. To achieve spacing, or
to force a paragraph to the top of a page, use your word processors
spacing and positioning controls instead of blank lines. It is wrong to set
two hyphens in place of a dash: Use an en dash surrounded by normal
(word) spaces. Modern word processors and page layout programs
allow you to use typographic characters. Use the straight single and
Copyright 1998-02-08 Charles Poynton
1 of 2
TEN COMMON MISTAKES IN THE TYPESETTING OF TECHNICAL DOCUMENTS
double quotes ' and " only to denote minutes, seconds, feet, and
inches; and then only if there is insufficient space to spell out the unit.
For all other purposes, use typographers (curly) quotes. (Unfortunately, HTML has no typographical quotes.)
5
Lack of consideration for line length and type size: A 66 character line
is widely regarded as ideal for readability. With 12-point Courier type
so-called Pica, with ten characters per inch 66 characters make
a 6.6-inch line. Set on an 8.5-inch page width, this leaves a reasonable
margin of about an inch on each side. But 12-point type, especially
Courier, is too large for all but very unusual cases. With 10-point
Palatino, quite suitable for a technical document, a typical 6.6-inch line
has 100 characters, far too many for continuous reading. To use
10-point Palatino, a line length of about 4 1 8 inches would be ideal. If
you want to use a single column, consider using 11-point type in
a column about 5 inches wide. If your document is almost wholly text,
consider using 9-point type in a layout with two 3-inch columns.
Unsightly indication of emphasis: Underlining is strictly for typewriters.

Modern word processors and page layout programs give access to italic
typefaces: Use italics for emphasis. To use boldface would make your
page look blotchy. Use quotation marks only for their intended
purpose: quotations. To use quotation marks for emphasis is disruptive
to smooth reading of your text.
Lack of control of line breaks: Modern word processors and page

layout programs allow you to control line breaks and hyphenation. Use
a nonbreaking space to prevent adjacent elements such as a numerical quantity and its associated unit from being separated by a line or
page break. Use a nonbreaking hyphen to avoid the elements of a
compound modifier, such as 35-millimeter, from being broken.
Failure to clearly distinguish paragraphs: A surprising number of technical articles fail to make a clear demarcation between paragraphs. This
does a great disservice to the reader. You must choose one of two alternatives to identify paragraphs. If you set all lines on the page tightly
spaced, then indent the first line of every paragraph. Alternatively, place
a blank line between paragraphs and use no indentation.
Careless setting of fractions: Set a fraction using superscript for the

numerator and subscript for the denominator. Use the fraction slash
character as a separator on a Macintosh, access it as Shift-Option-1.
10 Poorly oriented figures: If your document includes figures oriented

differently from the text, it will be necessary for your reader to rotate
the document. You must ensure that he or she has to turn the document only once per page, and always in the same direction. Arrange
the layout so that the reader rotates the body text clockwise 90 to
view a figure. In other words, when viewing the text in its natural
orientation, place the top of a rotated figure to the left.
This note explains how to prepare PostScript files on a Macintosh, for printing on a Xerox DocuTech
135.
"DocuTech" refers to a series of high performance laser printers made by Xerox - whoops, The
Document Company. A DocuTech model 135 with the PostScript option is essentially a high-speed
PostScript laser printer and a high-speed duplicator - or loosely, Xerox machine - melded into a 3,193
lb unit. It prints 135 pages per minute.
If you want to have files printed on a DocuTech, it is a good idea to submit raw PostScript files
instead of application (Quark, PageMaker, or FrameMaker) files. To learn why, see the companion
document Submitting raw PostScript to a print shop.
You'll need the PostScript Printer Description (PPD) file for the DocuTech 6135. Obtain it from
Xerox: xrd61353.ppd (text format, 37045 bytes)
Create a PostScript job file according to the companion document Preparing to RIP.
The DocuTech is a Postcript level 2 printer. Use a file extension of ".ps". It will be convenient for
your priner if you use a base name that is your last name, or the name of your company: I imagine that
a typical DocuTech shop gets three or four files daily that have the same name, handout.ps!
If your document will have a cover, submit that as a separate one-page job. Some shops have a
separate per-file RIPping charge, of perhaps $10 or $15. The extra file will incur an extra RIP charge,
but asking the DocuTech operator to separate the pages into separate jobs is inviting error.
In addition to the PostScript job file, I generally provide an Acrobat PDF file to be used in case of
PostScript failures. Not all DocuTech shops are familiar with PDF at the moment, though Adobe is
moving rapidly to establish PDF as the format of choice for electronic submission of print jobs.
Submit the file according to the companion document, Submitting PostScript files to a commercial
printer.
You will find that the DocuTech images screens considerably lighter than a typical laser printer - in
particular, screens lighter than about 5% will disappear. The nasty solution is to make sure that and
graphic files are adjusted appropriately! A more complicated solution that involves setting PostScript
transfer functions is beyond the scope of this note.
1998-12-13
http://www.poynton.com/notes/mac/DocuTech_prepare.html (1 di 2) [15/01/2005 16:27:12]
http://www.poynton.com/notes/mac/DocuTech_prepare.html (2 di 2) [15/01/2005 16:27:12]
Xerox - whoops, The Document Company!
Xerox - whoops, The Document Company!
As far as I am concerned, Xerox is an eminently suitable name for a company. The name is easily
recognizable, well-respected, and (presumably) free from cross-cultural problems.
However, the corporate strategy experts at Xerox thought that it wasn't good enough, so in about
1994, with the advice of Landor Associates, they changed the name to The Document Company
Xerox. (Or perhaps they just recommended adding a tag line.) In French, the line is Les Gens de
Documents - literally, the gentlemen of documents.
Web pages at the Xerox site explain that a videotape is actually just another form of document. Here's
the quote, from What is the document?
The hard copy manuscript of a Mozart Symphony sitting in a Salzburg Museum is a
document. A CD of the Vienna Philharmonic playing that symphony is an audio
document with the added information of the conductors [sic] interpretation. A video of
the performance is yet another document.
I don't agree. Xerox seems to be arguing for device-independent multimedia, across a pretty wide
range (sheet music, digital audio, video). As far as I'm concerned, there's no such thing as good
multimedia. Any good presentation is necessarily optimized for the presentation medium.
Charles - Preparing to RIP on a DocuTech
1999-01-20
http://www.poynton.com/notes/rants/Xerox_whoops.html [15/01/2005 16:27:13]
Submitting PostScript files

to a commercial printer
To submitting a PostScript file to a commercial printer, first prepare the PostScript job file according
to the companion document, Preparing to RIP. To locate a DocuTech, read about DocuTechs I've
known.
Accompany your file with a "Read me" document that explains to the printer exactly what needs to be
done. In addition to the obvious things such as the number of copies, provide such details as:
the type of paper;

the page format (tall or wide);
whether you have specified single-sided or double-sided printing in the file, and whether you
wish the job printed single-sided or double-sided,
if double-sided, whether you want "head-to-head," suitable for flipping pages right-to-left, or
"head-to-toe" (a.k.a. "tumble"), suitable for flipping pages upwards - see the sketch below;
whether you have included all fonts used by the document (highly recommended), or whether
you expect the print shop to provide some fonts (strongly discouraged);
whether the file contains any intentionally-blank pages, such as a blank verso page ending a
chapter;
any binding or covers that should be supplied or printed; and
how you want the job to be delivered and billed.
If you're going to the trouble to print on DocuTech, instead of printing on a laser printer (and perhaps
photocopying), I recommend that you choose a paper that is more opaque than typical commodity
laserwriter paper. At some shops, ask for "60 pound paper." It doesn't have to be "bond," and in many
cases it won't be anywhere near two and a half times heavier than what they call "24 pound" paper,
but it will prevent your reading through the page to the reverse side.
The preferred physical media for most print shops is a Mac-formatted 100 megabyte Zip cartridge.
Most printing shops also accept 44 MB and 88 MB SyQuest cartridges; some accept Jaz carts.
http://www.poynton.com/notes/mac/DocuTech_submit.html (1 di 2) [15/01/2005 16:27:17]
If you do not have a Zip (or SyQuest or Jaz) drive, you will have to use a floppy disk, and transfer the
file across the net. Alternatively, you could transfer the file to the printing shop by e-mai, or by
modem transfer to a BBS. Call them and ask about their facilities.
Some very small shops prefer files to be supplied on PC-formatted floppy disk; if the DocuTech is not
networked, then this is your only option. In this case, give the file a name in accordance with MSDOS ("8.3") file name conventions. If necessary, use the zip compression scheme to compress the file,
and give the file an extension of ".zip". Zip compression is available through several shareware and
freeware programs on the Mac. You might try zipit, version 1.3.8 (BinHex format, 623284 bytes).
If you are transmitting the job electronically to a shop that uses Macs, you'll probably want to use
StuffIt compression and give the file an extension of ".sit". Transmit the file in binary mode (or, if
available, MacBinary mode).
You might be interested in reviewing some tips from a commercial shop on perparing files:
Submitting Electronic Mechanicals - a wonderful oxymoron!
1998-08-06
http://www.poynton.com/notes/mac/DocuTech_submit.html (2 di 2) [15/01/2005 16:27:17]
This note explains how to find a Xerox DocuTech 135 to print your job. (I assume you're familiar
with my Introduction to RIPping.)
I often use DocuTechs when I'm on the road, away from Toronto. Usually, I use Kinko's. I have used
Kinko's locations in Los Angeles, Las Vegas, Phoenix, Dallas, New York, ..., well, you get the drift.
But I have to say that the service is intermittent, at best. I've frequently found the network down, the
network up but the spooler down, the machine tied up for the next 4 days, etc. Call first, and confirm
that they've got a DocuTech, that it's available, and that everything is working. I know this sounds
pedantic. Trust me. The great thing about DocuTech is that if you prepare the job correctly, and you
jump through all of the hoops to get to the machine, and if it prints at all, then it prints perfectly.
Although Kinko's DocuTech service is generally unreliable but usable, this Palo Alto location is
superb, and the La Jolla (San Diego) and Austin locations are very good:
Kinko's
249 California Ave.
Palo Alto, CA 94306
U.S.A.
tel: +1 650 328 3381
fax: +1 650 328 7518
Kinko's La Jolla (For Canadians: "laah HOY-ah")
8849 Villa La Jolla Drive
La Jolla, CA 94306
U.S.A.
tel: +1 619 457 3775
fax: +1 619 457 0946
Kinko's
9222 Burnett Road, #101
Austin, TX 78758
U.S.A.
tel: +1 512 339 1191
fax: +1 512 339 1625
Aside from these gems, service at Kinko's is spotty. (Who knows, personnel might have changed at
those locations.) In any event Kinko's charges an arm and a leg. Try to find a local, independent shop,
if you can. I think you'll find that service and price will be better by far. Search the Yellow Pages!
Search on the Internet!
http://www.poynton.com/notes/mac/DocuTechs_I_ve_known.html (1 di 2) [15/01/2005 16:27:19]
Here are the guys I use in Toronto:

Print Three (ask for Adrian)
Yonge-Eglinton Centre
2300 Yonge St
Toronto ON M4P 1E4
Toronto, ON M4P 1E4
Canada
tel: +1 416 481 5159
e-mail: print3.ye@sympatico.ca
Here are three particularly great places outside of Toronto:
Techniprint
2545 N. Seventh Street
Phoenix, AZ 85006
U.S.A.
tel: +1 602 257 0686
San Jose Blue
835 West Julian Street
San Jose, CA 94306
U.S.A.
tel: +1 408 295 5770
fax: +1 408 971 3299
e-mail: MacPrinting@SJBlue.com, PCPrinting@SJBlue.com
Call Print (ask for Stuart)
163-167 Great Portland Street
London W1N 5FD
England
tel: +44 171 580 7122
fax: +44 171 580 7083
Try to find a shop is reasonably computer-savvy, so that you are able to submit your file by e-mail.
Beware file size; anything more than about a megabyte could potentially cause trouble. Learn about
MIME attachments.
2001-02-21a
http://www.poynton.com/notes/mac/DocuTechs_I_ve_known.html (2 di 2) [15/01/2005 16:27:19]
Charles Poynton - FrameMaker
Charles Poynton FrameMaker

Here are brief descriptions of several documents that I have written about Adobe's FrameMaker
product.
The documents are available in several formats: PDF, PostScript and in some cases, ASCII text. I
encourage you to use the PDF format.
Before I go any further, if you use Frame (or any high-end DTP product) on a Mac, let me suggest
that you rip out your TrueType fonts !
bitmap, a necessary process in typographic-quality printing. RIPping is performed by the PostScript
interpreter that resides in a laser printer or filmsetter.
Here you can access four brief documents that introduce you to RIPping, in order to get a document
from a Macintosh to a commercial printer. Information about printing from FrameMaker to a Xerox
DocuTech is included.
Making Acrobat Bookmarks using FrameMaker

This note explains how to use FrameMaker 4 on a Mac to make Portable Document Format (PDF)
files suitable for Adobe's Acrobat Reader. The note concentrates on how to produce bookmarks. I
assume that you have access to Adobe's Acrobat Distiller.
FrameMaker 4 does not produce PDF crossreferences on any platform other than Mac, and on a Mac
produces crossreferences only within a single Frame document. These restrictions have been lifted in
Frame 5.
Although Frame 5 can make bookmarks, this capability is limited to bookmarks that are extracted
from paragraph text (say section heads). My technique can be used to make bookmarks from markers.
A forthcoming revision of the note will deal with Frame 5, and will explain how to include - in a
Frame file - information that to cause Acrobat's document information to be set, and information to
instruct Acrobat Reader to open at the size you like and with bookmarks open.
http://www.poynton.com/Poynton-frame.html (1 di 3) [15/01/2005 16:27:21]
FrameMaker Generated Files

Frame has six kinds of generated files. You can generate a file from paragraphs, from markers, or
from references, and you can have the result sorted or not. A sorted file is called an Index (under the
Generated Index category) or an Alphabetical List (under Generated Files category). This table
summarizes the thirteen categories of generated files that are presented in the the Add File to Book
and Generate File dialogs.
Frame EPSF converter

If you are using EPSF files on a Mac, by default they will have PICT previews which cannot be
displayed on UNIX platforms. Adobe has defined the EPSI (interchange) variant of EPS; this variant
has a bitmap preview image coded in ASCII comments at the front of the PostScript and should, in
theory, be useable on any platform -- it is resource-fork-free and (potentially) ASCII-only and lineend safe. Unfortunately, Adobe has not implemented EPSI in its Mac products such as Photoshop or
Illustrator. This causes fuss and bother when you exchange Frame files between Mac and UNIX. To
convert EPSF files to the EPSI interchange format, obtain this program (from Frame's ftp site):
FrameEPSFConverter 3.0 (BinHex/SEA format, 30614 bytes)
Short Subjects
I rummaged around the attic and came up with a handful of short articles that I have written on
diverse subjects. All are in ASCII text format; most are less than 10 KB. Most are copies or direct
adaptations of Usenet news postings.
Advanced paragraph numbers (text format)

Exporting Frame graphics as EPS (text format)
FrameMensa Quiz #1, QUIZ (text format)
FrameMensa Quiz #1, Motivation (text format)
FrameMensa Quiz #1, ANSWER (text format)
Last page number of book (text format)

Not all settings are specd (text format)
No foilio on blank verso (text format)
One or more spaces (text format)
Shrinkwrap leaves a 1 pt gap (text format)
Frame-Acrobat (text format)
FrameMaker last char of para (text format)
Small Caps character format (text format)
publ quality screenshots (text format)
Acrobat_links_FrameMaker_xref(text format)
Other FrameMaker links ...

Charles
1997-07-22
Charles A. Poynton
Toronto, ON M4N 1S3
CANADA
tel +1 416 486 3271
fax +1 416 486 3657
poynton@inforamp.net
Making Acrobat Bookmarks using FrameMaker
This note explains how to use FrameMaker to make Portable Document Format (PDF) files suitable
for Adobes Acrobat Reader. The note concentrates on how to produce bookmarks.This note is
written for Frame 4.0.4 on a Mac, although Windows and UNIX are the same in principle. I assume
that you have access to Adobes Acrobat Distiller.
This document lives at
<ftp://inforamp.net/pub/users/poynton/doc/Frame/Frame_Acro_Bookmarks.pdf>
1
Acrobat Overview
Portable Document Format (PDF) is a file format invented by Adobe for the electronic distribution
of documents. A PDF document is not PostScript, but sort of its second cousin. You can think of it
as embodying the imaging operations of PostScript without the programming language. A document in PDF form has the same potential print quality as a PostScript file, but it can be viewed or
printed without involving the full complexity of a PostScript interpreter. A PDF file contains only the
7-bit ASCII printable character set, and accommodates any line-end convention, so PDF is freely
transportable across MS-DOS, Windows, Macintosh and UNIX platforms.
Font handling in the Acrobat system is quite sophisticated. If a PDF file references fonts that are not
present during reading, Acrobat Reader uses ATM and MultiMaster technology to synthesize, onthe fly, a font that has the same metrics and roughly the same look.
The PDF file format is documented in Portable Document Format Reference Manual, Adobe
Systems Incorporated (Addison-Wesley, 1993, ISBN 0-201-62628-4).
Adobe distributes free Acrobat Readers in MS-DOS, Windows, Mac and Sun UNIX versions. Find
these at <ftp://ftp.adobe.com/pub/adobe/Applications/Acrobat/>.
2
Generating PDF
To generate a PDF file, you use either Adobe PDFWriter or Adobe Distiller.
PDFWriter is a component of a commercial software package from Adobe called Acrobat Exchange.
PDFWriter takes the place of a print driver, and writes a PDF file to disk. PDFWriter offers no PostScript capability to the application. In the Mac environment, it operates as a QuickDraw printer.
Although this is quite satisfactory for many uses, EPS files that are incorporated into a document
will have their preview bitmaps written to the PDF file.
Acrobat Distiller, a commercial software package available from Adobe, incorporates a PostScript
Level 2 interpreter. It can take any PostScript file, including embedded EPS drawings or pictures,
and generate PDF. Adobe Illustrator version 5.5 for the Mac comes with a copy of Distiller.
1 of 4
MA K I N G A C R OB A T B OOK MA R K S U S I N G F R A ME M AKER
Bookmarks
Bookmarks are a feature of PDF files that make it easy for a reader to navigate through a document. You can think of a bookmark as being an executable element of a table of contents.
Unless special measures are taken, a PDF file produced by either PDFWriter or Distiller has no bookmarks. Bookmarks can be added to a PDF file using Adobes Acrobat Exchange product, but it is a
big nuisance to generate bookmarks manually. The remainder of this note explains how bookmarks
can be generated more or less automatically by Frame.
4
Distiller operators
Adobes Distiller implements a PostScript operator pdfmark which inserts a bookmark into a generated PDF file. This operator is documented in Adobes Tech Note LPS 0172, pdfmark Reference
Manual. You will use Frames Generate/Update facility to extract bookmarks information from your
document, and generate a file of lines with the pdfmark operator.
You will make a small batch file that issues the Distillers RunFile operator to pull your bookmark
file, then your documents PostScript file, from disk and present them to the interpreter in turn.
RunFile is documented in Adobes Tech Note 4402, Using the RunFile Procedure to Combine PostScript Files.
5
Inserting the bookmarks
You could generate bookmarks from paragraphs having a certain tag, for example Heading1 for
first-level headings. However the reader of your document may have a restricted amount of screen
area that she wants to devote to bookmarks when she is using the Acrobat Reader, and your document may have longish section headings. Also, Acrobats bookmarks allow no special character
formatting, but your document may have headings that include character formatting. I recommend
that you generate the bookmarks from markers having short versions of the text of the section
headings, instead of the from the section headings themselves. I use Marker Type 11.
Go through your document adding markers of type 11, with the text that you wish to be collected
into bookmarks. If you add a section later, remember to add a marker. If you want to include a
parenthesis or a backslash in bookmark text, you must escape it with backslash: to get (A), use
$A$. Use only characters in the ASCII set, because bookmark characters in the upper 128 are
apparently not preserved across platforms.
Make sure that your System Variables and Cross-References are up-to-date. I find it frustrating that
Frame does not invite me to update these before I Print or Generate, but its better than Quark or
PageMaker which have no cross-reference capability at all!
Print your document to a PostScript file.
Frame inserts into its PostScript code an advertisement that appears only when the PostScript is
processed by the Distiller. Use a text editor to search the PostScript for a line containing the word
created, like the line below. Then delete the line.
187 776 M (This document was created with FrameMaker 4.0.4) FmPT
Generating the Bookmarks
Compiling the bookmarks involves generating a file, containing PostScript code, from the markers
that you have inserted in your document. While your document is open, you will choose Generate
from Frames File menu. Frame will find a file with the name of the document and the suffix (extension) that you specify, and will insert or replace the body text of that document by generated text.
4
MAKING ACRO BAT BO O KM ARKS US I NG F R A ME MA K E R
Frame obtains the format of lines in a generated file from a flow on a reference page in that file.
The name (tag) of the flow is the same as the suffix of the generated file. The special flow is
normally on a reference page of the same name. That flow has a paragraph whose tag is named
according to the tag or marker type and the filename suffix.
For bookmarks, you will generate a List Of Markers (LOM) of type 11. Create a new empty document with a single text column on its body page. Add a reference page named LOM, with a flow
named LOM, with a single line having the tag Type11LOM. The line contains this text:
[/Page <$pagenum> /View [/XYZ null null null] /Title (<$paranum> <$markertext>) /OUT pdfmark
This line contains the arguments and the operator that the Distiller needs to create a bookmark. The
items in angle-brackets will be substituted by Frame during the course of file generation. Save this
as a Frame document with the name of the main document and the suffix LOM.
Now open your document file. Choose Generate/Update, specify List of Markers, Type 11, and
indicate an suffix LOM. Frame will discover the file that you just saved, so it will use the formats in
that file instead of generating a new file. The generated text will look something like this:
[/Page 3 /View [/XYZ null null null] /Title (4 Gamma) /OUT pdfmark
[/Page 4 /View [/XYZ null null null] /Title (5 Gamma correction) /OUT pdfmark
[/Page 5 /View [/XYZ null null null] /Title (6 NTSC gamma 2.2) /OUT pdfmark
Save this document as text, with the suffix .txt. Giving it the suffix txt emphasizes that it contains
text, but more importantly it avoids overwriting the LOM file whose format information will be
required the next time you generate.
When Frame generates a document, if it cant find an appropriate reference page with an appropriate flow and an appropriate paragraph tag, it creates whatever is missing. So if you have trouble
generating, see if Frame has created a new reference page or a new paragraph. Then you can diagnose what caused Frame to generate it.
7
Batch file
You will present to the Distiller a single small batch file that specifies your bookmark text file and
the PostScript file from your document. The file looks like this:
%!
% collect named .ps files into a single Distiller run. See Adobe Tech Note 4402.
%
/prun {/mysave save def RunFile clear cleardictstack mysave restore} def
(Macintosh HD:Frame Work:Manifesto.LOM.txt) prun
(Macintosh HD:Frame Work:Manifesto.doc.ps) prun
%EOF
Create this file in Frame and Save As ... Text, or use your favourite text editor. Replace Macintosh
HD and Frame Work with the full path to your documents. I use the suffix LOM.txt for the bookmark file and doc for the document file, but this is just my convention. Save the batch file with the
name you wish to give the completed document, and include the suffix .PS.
8
Distill
Finally, drop the batch file onto the Distiller. If anything is wrong, youll hear about it. But if all is
well, the Distiller will produce a single file named for the batch file and with the suffix .PDF.
Test the result by viewing it with the Acrobat Reader. The Reader normally starts up with bookmarks out of sight, so grab the window pane bar at the lower left and drag it to make the bookmarks visible, or choose Bookmarks from the View menu.
4
MA K I N G A C R OB A T B OOK MA R K S U S I N G F R A ME M AKER
You can now distribute your bookmarked PDF file.

When you update your main document, to make a new PDF file take these steps:
Update system variables and cross-references,
Print your document to a PostScript file,
Generate from your document file,
Save the generated file As ... Text, and
Drop the batch file onto the Distiller.
Send me e-mail if you like the results! Better still, send me your PDF file!
9
Bookmarks for a Book
The technique of generating bookmarks for a single Frame document can be applied to a Frame
book. If you are familiar with Frame books you should also be familiar with Frame Tables of
Content (TOCs). If you are not interested in books or TOCs, skip this section!
You will start with a document in one or more chapters of a Frame book file, including a TOC.
Add to your book a generated file for bookmarks, giving an suffix of LOM and one or more
markers to be collected. Be sure to put a dot in front of the suffix.
You will use a batch file to present all of your PostScript files to the Distiller in one session. This is
necessary due to a series of three circumstances. First, Frame puts any generated file on its own
page or pages. Second, when Frame prints a book to a PostScript file, it makes each chapter into a
self-contained PostScript file. Third, the Distiller starts page numbering at 1 for each PostScript file it
processes. This combination of circumstances means that you must present a single file to the
Distiller, to preserve the page numbers of your bookmarks.
You want your bookmarks to be included in the PDF file, but not the PostScript code that generates
them! The batch file bypasses the bookmark chapter, but the first actual chapter in your book
must reset page numbering to 1. Acrobat accommodates only page numbering starting at 1 and
incrementing throughout the PDF file, so you must not restart page numbering anywhere else.
Since the bookmark file has the suffix txt, you can print the entire book at once into separate .ps
files, even the bookmark document, without disturbing the bookmark PostScript file.
Frame does not as of version 4.0.4 provide direct access to the total page count of a book. If
you wish to reference the total page count in the front matter, you must work around this by using
a cross-reference to the variable <$lastpagenum> on a paragraph of its own in the last chapter of
the book. This messy business is documented in the Frame manual (page count, in book file).
10 Bookmarks from paragraphs
Your bookmarks could be taken from paragraph tags instead of from markers. For example you
could collect paragraphs having tags Heading1 for both the Table of Contents and the bookmark
file. But for the reasons that I mentioned earlier in Inserting the bookmarks, I recommend that you
generate the bookmarks from markers.
If you choose to collect paragraph tags instead, rename the reference page, rename its flow, and
change the paragraph tag of your formatting string. Also, change <$markertext> to <$paratext> in
the format string.
4
Charles A. Poynton
Toronto, ON M4N 1S3
CANADA
tel +1 416 486 3271
fax +1 416 486 3657
Poynton@inforamp.net
Frame Generated Files

This table summarizes the characteristics of generated lists and indexes in FrameMaker.
Default
Suffix
Inrernal
Code
Name
Source
Default
Type
Ordering
Generated List
TOC
TOC
Table of Contents
Para
Collate
LOF
LOF
List of Figures
Para
Collate
LOT
LOT
List of Tables
Para
Collate
LOP
LOP
List of Paragraphs
Para
Collate
LOM
LOM
List of Markers
Markers
Collate
AML
AML
Alphabetical Marker List
Markers
Sort
APL
APL
Alphabetical Paragraph List Para
Sort
REF
LR
List of Referencesd
References
Collate
Generated Index
IX
IDX
Standard Index
Marker
Index
Sort
AIX
IOA
Index of Authors
Marker
Author
Sort
SIX
IOS
Subject Index
Marker
Subject
Sort
IOM
IOM
Index of Markers
Marker
Sort
IREF
IR
Index of References
References
Sort
Paragraphs can be chosen from the paragraph tags in the document.

References can be chosen from these:
Marker types can be chosen from these:
Condition Tags
External Cross-Refs
Fonts
Imported Graphics
Publishers and Subscribers
Unresolved Cross-Refs
Header/Footer $1
Header/Footer $2
Index
Comment
Subject
Author
Glossary
Equation
Hypertext
Cross-Ref
Conditional Text
Type 11 through Type 25
1 of 1
http://www.poynton.com/notes/short_subjects/FrameMaker/Advanced_paragraph_numbers
Subject: Re: "Advanced" paragraph numbering?

Date: Sun, 13 Aug 1995 11:47:48 +0000
In article <40fj91$itc_001@news.unit.no>, ohansen@elkraft.unit.no
(Oddbjorn Hansen) asks:
>
>
>
>
I'm currently writing my Ph.D. thesis using Frame 3.1.

Is it possible to put a reference to the paragraph number of the
main chapter into the paragraph numbering of figures, tables and
equations?
Yes, this is possible - even easy - but you must meet a few
conditions. The first condition is that your chapter title, figures,
tables and equations must be in the same Flow. In practice this means that
ALL of the content of your chapter must be in the same flow. That turns
out to be easy if you are willing to use anchored frames to contain your
figures. If you want to use UNanchored frames - frames that are glued to
the page - then what you ask will be a nuisance. In Frame 3.1 or Frame 4
you are forced to use unanchored frames if you have figures that span
columns, say in a two-column layout. I assume your thesis has a one-column
layout. In Frame 5, this restriction is removed, since in Frame 5 an
anchored frame can span columns.
> F:Fig. <n+>.<n=0>.
> T:Table <n+>.<n=0>.
> E:(<n+>.<n=0>)
The second condition is that your chapter title, figures, tables and
equations must be numbered with the same series label. You are forcing
different series labels for these elements. This is sort of OK for a
single-chapter document, but is not necessary, and precludes putting
elements of numbering sequences together as you wish to do.
Instead, take a different approach. Place all of your numbering elements
into a single "autonumber vector" (my term), comprising chapter, section,
figure, table and equation elements. I recommend that you use no series
label at all.
> Now I initiate the numbering by putting 3 paragraphs at the
> beginning of each chapter (using white fonts):
Forget the white stuff. Make a Chapter autonumber like this:
<n+>< =0>< =0>< =0>< =0>
Make figure, table and equation autonumbers like this:
Figure <n>.< ><n+>< >< >
Table <n>.< >< ><n+>< >
Eqn <n>.< >< >< ><n+>
It is my experience that whitewash is OK for a really quick and dirty
workaround, but eventually produces big trouble in a document that has to
be maintained. Most tempations to use white can be overcome by reading the
manual (or consulting the Framers) to find a better way to do things, even
if it's poorly documented. I confess that there are a few situations where
it is necessary, but thankfully yours is not one of them.
C.
Charles Poynton
poynton@poynton.com [Mac Eudora/MIME/BinHex]
<http://www.inforamp.net/~poynton/>
http://www.poynton.com/notes/short_subjects/FrameMaker/Advanced_paragraph_numbers [15/01/2005 16:27:34]
http://www.poynton.com/notes/short_subjects/FrameMaker/exporting_Frame_graphics_as_EPS
In article <DE8KH5.1x3 @news.cern.ch>, jferrerp@dsy-srv3.cern.ch wrote:

> Anyone knows what's the best way to export just a graphic made in
> FrameMaker with the FrameMaker tools to then inport it in Word for Windows
I don't know if this is the "best" way, but here's the only realible way
that I know. [I assume you're using a PostScript printer -- if not, just
take a screen shot and paste it in.]
In Frame, place your graphic on its own page with no headers or footers.
Then Print the single page containing the graphic to an Encapsulated
PostScript (EPS) file. In Word, Insert File or Picture, and choose your
EPS file. If you saved the EPS with a preview, you'll see the preview on
the screen, otherwise you'll see a grey box. But in any case the EPS will
print.
"Saving as EPS" is platform dependent, and I can't speak for Windows. On
my Mac, Frame insists on writing an EPS file that comprises the entire
page, not just the portion that is used. To save a snippet of a Frame page
in EPS, I must both (1) use a text editor to edit the bounding box saved
in the EPS file, and (2) save with no preview, because the preview would
cover the whole page, and would be rendered incorrect after I tighten the
bounding box.
To make an accurate bounding box, place all of your graphic elements
within a frame having a border of None, then access Object Properties of
that frame to determine its position and the dimensions (in points). Take
the height of your page and subtract from it the Offset from Top of the
frame, then subtract the Height of the frame. Add the Width of the frame
to its Offset from Left to get the offset of its right edge. These steps
are necessary to convert from Frame's coordinate system (position from
upper left, and size) to that of PostScript (positiion from bottom left,
and position of opposite corner).
Now use a text editor to open the EPS file and mess with the bounding box.
Find the %%BoundingBox: line early in the file. The four numbers that
follow are the distances of the left, bottom, right and top, in points,
from the PostScript origin at the bottom left corner of the page. The
bounding box saved from Frame on my Mac is 0 0 612 792 for a USletter full
page. Change these to the offset of the lower left corner of your frame
(x, y) and the offset of your upper right corner (x, y). [In the unlikely
event that you find a %%HighResBoundingBox, consult the Adobe Red Book
and/or the PostScript FAQ: I don't want to complicate things further
here.]
Save your edited EPS file, then import it into its destination.
If you want to avoid the BoundingBox editing, or to keep a preview, you
could try an alternate approach of using Frame to scale your graphic up to
the full page size, then importing that as EPS with a preview. You can
then scale it back down in the importing application. This is a nuisance
because you have to force the aspect ratio to fit the selected page size,
then undo that step on import. It's also a nuisance because when first
imported Word will want to position the graphic within Word's page
margins, and it will spill over the sides of the page. And the preview,
stored at 72 dpi, will be big and slow to render, even when scaled back
down.
Because of the difficulty of saving EPS from Frame, and because of the
lack of Frame graphics import and export tools, I use Frame's drawing
tools only for the simplest of things like rulings. For serious graphics
work I use Adobe Illustrator, and I Import its EPS by reference into
Frame. AI's bounding box is computed and stored correctly, so none of the
messing around is necessary. I vote for Frame to cease investing in
FrameMath and drawing tools -- tasks for which excellent tools are easily
available -- and to concentrate on documents, books and typography. But I
http://www.poynton.com/notes/short_subjects/FrameMaker/exporting_Frame_graphics_as_EPS (1 di 2) [15/01/2005 16:27:36]
http://www.poynton.com/notes/short_subjects/FrameMaker/exporting_Frame_graphics_as_EPS
digress.
C.
http://www.poynton.com/notes/short_subjects/FrameMaker/exporting_Frame_graphics_as_EPS (2 di 2) [15/01/2005 16:27:36]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CQUIZ
FrameMensa Quiz #1 -- Paragraph Spacing

This sounds easy:
Work out how to format a document that comprises two kinds of
paragraphs, so that ordinary body text and headings align with grid
lines spaced vertically, four to the inch.
Each body line is to take a quarter of an inch of vertical space.
Consecutive paragraphs are to be separated by a single blank line.
Each heading line is to take two lines' worth of space, with the heading
text in the middle of that space -- that is, the headings are displaced
half a line from the rest of the text, spaced with a quarter line above
and a quarter line below.
It's not a trick question -- at least, I don't think it is.
Hint:
If you put a 0.125 inch Space Before a normal paragraph, and 0.125 inch
Space After, Frame overlaps these spaces and the paragraphs are
separated by 0.125 inches, not 0.250.
If you put 0.250 inches either Before or After a normal paragraph then
you will get the desired space between normal paragraphs, but an
intervening Heading paragraph then has to contend with a minimum of
0.250 inches of space above or below. An additional complexity is that
Frame disregards Space Below a paragraph that falls at the top of a
column. The formatting has to continue to function in this case.
Use of baseline synchronization is prohibited. But you can use Frame
Above and Frame Below if you have to.
C.
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CQUIZ [15/01/2005 16:27:41]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CMotivation
Path: inforamp.net!ts4-05.inforamp.net!user
From: poynton@ poynton.com (Charles Poynton)
Newsgroups: comp.text.frame
Subject: Re: FrameMensa Quiz #1 -- Paragraph Spacing
Date: Mon, 21 Aug 1995 14:33:35 +0000
Lines: 82
Message-ID: <Poynton-2108951433350001@ts4-05.inforamp.net>
References: <Poynton-1908951756560001@ts4-13.inforamp.net>
NNTP-Posting-Host: ts4-05.inforamp.net
In article <Poynton-1908951756560001@ts4-13.inforamp.net>, I presented a quiz.
A correspondent writes,
> This Quiz. What is the practical use of such paragraph tags?
The quiz is not arbitrary. To explain why I want such a format, I have to
start a long way back.
A book is easiest to read when the paper is very white and the ink is very
black. Technically, in addition to fairly "bright" paper with a
reflectance of 0.75 or more, we also want a high "contrast ratio" between
white and black, that is, a high ratio between reflectance of the paper
(say 0.75) and the reflectance of the ink (say 0.02).
But paper is somewhat transparent. The paper used for a book is thinner
than the paper used in a laserprinter, so even more transparent. Open your
photocopier's cover and take a copy of the empty glass -- if you like,
take a copy of the ceiling of your photocopy room, out of focus. This will
produce a piece of paper entirely covered with black toner. Place this
sheet immediately behind the page of the book that you are reading. The
page will become quite dark.
The ink on the reverse of the page you are reading has the same effect -it absorbs light and darkens the face of the page.
You may have once made a photocopy from a newspaper, to find that in the
copy the area between the lines was darkened somewhat by the content on
the reverse of the page. You can alleviate this effect when making a
photocopy by backing such a page with the black sheet that I explained how
to make a moment ago. This is not a feasible solution for reading, but the
photocopier machinery is insensitive to the absolute reflectance of white,
and responds only to the contrast ratio.
Take a book of the highest quality, a book on sculpture or painting say.
(Do not try this with a "computer" book). Hold an opened, single page up
to the light. You will see that the lines of type facing you are set
exactly on top of the lines of type facing the other way -- the lines are
on a fixed grid. Even if a section heading or some other element
intervenes, in a really well composed book, the grid is maintained
throughout the height of every page. This arrangement makes sure that the
dark type on the reverse side of a page has the minimum impact on the
contrast ratio.
Another reason for the grid is more esoteric. If you read the works of the
classic typographers, you find that they all agree that a regular vertical
division of space lends a pleasing rhythm to a piece. Here's Bringhurst:
Time is divisible into any number of increments. So is space. But for
working purposes, time in music is divided into a few proportional
intervals: halves, quarters, eighths, sixteenths and so on. And time in
music is measured. Add a quarter note to a bar whose time is already
accounted for and, somewhere nearby, the equivalent of the quarter note
must come out. Phrasing and rhythm can move in and out of phase -- as
they do in the singing of Billie Holiday and the trumpet solos of Miles
Davis -- but the force of blues phrasing and syncopation vanishes if
the beat is lost.
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CMotivation (1 di 2) [15/01/2005 16:27:43]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CMotivation
Space in typography is like time in music. ... Vertical space is

metered ... You must choose not only the basic measure -- the depth of
the column or page -- but also a basic rhythmical unit. This unit is the
leading, which is the distance from one baseline to the next.
This is from Robert Bringhurst, The Elements of Typograhic Style, Hartley &
Marks, Vancouver BC, 1992.
My "quiz" might seem frivolous at first glance. But I am quite serious: I
want to use FrameMaker to achieve excellent typography. I have found a
very clumsy way to do what I want, but I hope that a Framer has an easier
way.
By the way I received a private reply that gives an amazing account of detail
concerning exactly where FrameMaker sets characters, determined by a feat
of reverse engineering. I do not have permission to post this, however.
C.
Charles Poynton
<mailto:poynton@poynton.com> [Mac Eudora/MIME/BinHex]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CMotivation (2 di 2) [15/01/2005 16:27:43]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CANSWER
FrameMensa Quiz #1 -- ANSWER (spoiler)

A few days ago I proposed this quiz:
Work out how to format a document that comprises two kinds of paragraphs,
so that ordinary body text and headings align with grid lines spaced
vertically, on 18-point spacing.
Each body line is to take 18 points of vertical space.
Consecutive paragraphs are to be separated by a single blank line.
Each heading line is to take two lines' worth of space, with the heading
text in the middle of that space -- that is, the headings are displaced
half a line from the rest of the text, spaced with a quarter line above
and a quarter line below.
Hint:
If you put a 18 point Space Before a normal paragraph, and 18 points of
Space After, Frame overlaps these spaces.
If you put 18 points of Space Before or After a normal paragraph the you
will get the desired space between normal paragraphs, but an intervening
Heading paragraph then has to contend with a minimum of 18 points of
space above or below. An additional complexity is that Frame disregards
Space Below a paragraph that falls at the top of a column. The formatting
has to continue to function in this case.
One solution is to use a column arrangement -- on a master page, say -that has thirty vertically-stacked text frames each 18 points high, each
designed to accommodate a single line. Assign to normal paragraphs a
leading of 18 points; this will allow only one line per text frame. In the
Advanced properties of the heading format, place a Frame Before that refers
to an empty frame 9 points in height. This will displace the headings down
the required amount. You will have to include after every heading paragraph
an empty paragraph, to occupy the text frame immediately below in order to
achieve the required spacing below a heading. But this solution is clumsy
and inefficient.
A better solution uses a character attribute that has almost no
documentation and almost no user interface: the badly-named "vertical
kerning", more properly called "baseline shift".
Select a character and, on a Mac, hold Option and hit Down Arrow in the
inverted-T cursor arrow cluster. The character will be assigned a 1 point
baseline offset, downwards. If you hit Shift-Option-Down Arrow, the
character will be offset 6 points. If you use the Up Arrow key instead,
then you will obtain an upward shift. Remove the baseline offset by
hitting Option-Keypad 5.
The quiz is solved by:
- defining a normal paragraph to have 18 points space before, zero space
after, and 18 point leading,
- defining a heading paragraph to have no space before or after, 18 point
leading, and "keep with next", and
- giving heading paragraphs a downward baseline offset of 9 points.
Baseline offsets cannot be entered into the character or paragraph
catalogs using Frame's user interface, and cannot be copied from one
paragraph or character to another or from one entry in a catalog to
another. However, the attribute is saved in MIF as the FDY attribute,
which specifies downward shift as a percentage of font size. You can cheat
a little and edit the MIF file: apply a downward offset to a character and
save the document as MIF. Move the line containing the FDY attribute from
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CANSWER (1 di 2) [15/01/2005 16:27:45]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CANSWER
its location as an override to a particular character into the default

font section of the heading's paragraph format. When you import the
modified heading format, paragraphs using that default character format
will shift downward the required amount. This makes the solution is quite
automatic.
C.
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMensa_Quiz_1%2CANSWER (2 di 2) [15/01/2005 16:27:45]
http://www.poynton.com/notes/short_subjects/FrameMaker/Last_page_number_of_book
> Does anyone know of a way to [refer to the number of pages in a book,
> like page 37 of 224], apart from a cross-reference to something in the
> last page.
The consensus of the Framers is that there is no automatic way, but the
cross-reference doesn't have to target the last page. Insert a
paragraph containing the variable <$lastpagenum> anywhere in the last
book component. Some people put a paragraph on the first page of the
last chapter and paint it white so as noit to appear in the finished
product, but a better approach is to put the paragraph on a master
page. In any book component, obtain the page number of the last page in
the book through a cross-reference to the text of that paragraph.
Many requests have been made for Frame to introduce a variable for this
purpose. Frame's reluctance to implement this feature may stem from a
potential user-interface trap if chapters in the book restart page
numbering or use prefixes in page numbers. If a book has pages i
through xvi of front matter and body pages 1 through 48, it seems to
me that <$booklastpagenum> should return 48 and not 64. But if the last
page of the book is G-19, what should <$booklastpagenum> return? I
propose that the variable simply return the actual page number of the
last page of the book, G-19, not the count of pages in the book. My
proposed variable name makes this interpretation clear. It would be
guaranteed valid if no pages have been inserted or deleted since
Generate/Update. Some people might find <$bookpagecount> useful.
C.
http://www.poynton.com/notes/short_subjects/FrameMaker/Last_page_number_of_book [15/01/2005 16:27:47]
http://www.poynton.com/notes/short_subjects/FrameMaker/not_all_settings_are_specd
From: poynton@poynton.com (Charles Poynton)
Subject: Re: Renaming P-graph Styles
Date: Wed, 23 Aug 1995 14:13:57 +0000
Lines: 49
References: <41dnat$7f4@internal-dns.peerlogic.com> <ceej-2208952123080001@n2-1414.dynamac.genmagic.com>
In article <41dnat$7f4@internal-dns.peerlogic.com>, Kevin Walsh
<kwalsh@peerlogic.com> wrote:
> When I try to rename a paragraph style, I get the message:
>
>
Not all setting are specified.
>
Fill in the settings, and try again.
>
> All settings looked fine to me. What's happening?
In article <ceej-2208952123080001@n2-14-14.dynamac.genmagic.com>,
ceej@genmagic.com (C J Silverio) replied:
>
>
>
>
>
>
>
I have encountered this problem often while renaming

paragraph styles. I
was unable to determine WHY Frame 5 was inflicting this
particular horror on me. I couldn't figure out which setting
I had left unspecified, or why it complained THIS time and
not the time BEFORE. I eventually decided it was just another
dreary bug.
It's not a bug, but it is dreary.

A Character Format can be cataloged with one or more attributes set "As
Is". When a character format with As Is settings is applied, the "As Is"
attributes remain unchanged.
Paragraph formats do not work the same way. A paragraph format must have
all settings specified in order to be entered into the catalog -character formats accumulate, but paragraph formats stand alone. An
attempt to define a paragraph format with some setting specified As Is
will be rejected by Frame, accompanied by this unhelpful alert.
If you make a selection that includes different paragraph formats, then
examine the paragraph designer, the settings that differ will appear "As
Is". In this state, you cannot catalog the format.
To avoid this situation, define a format from the paragraph designer after
selecting only ONE paragraph -- that way, there can be no conflicting
settings
C.
Charles Poynton
<http://www.inforamp.net/~poynton>
<mailto:poynton@poynton.com> [Mac Eudora, MIME, BinHqx]
tel:
416 486 3271
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/not_all_settings_are_specd [15/01/2005 16:27:49]
http://www.poynton.com/notes/short_subjects/FrameMaker/no_foilio_on_blank_verso
Subject: Re: Q: *Empty* Pages Between Chapters
Date: Sun, 03 Sep 1995 03:11:45 +0000
Lines: 66
References: <41pmll$k9o@hera.easynet.de> <susan-2808950939580001@semiramis.gordian.com>
<41sth7$5ou@bug.rahul.net> <42550l$g5e@jabba.ess.harris.com>
Dear Frame,
Request for "enhancement":
Repair Generate/Update so that pages are neither added nor deleted from
book components in order to satisfy odd/even page conditions between
components. Instead, simply maintain page numbers appropriately. When
Printing a whole book, insert completely empty pages as necessary.
This is important because standsard practice in book publishing is "no
folio on blank verso" pages -- Frame Technology itself adheres to this
convention in its own manual Using FrameMaker, and few of your users would
suspect that you have to intervene manually at every edit cycle to get
this behavior from your product.
C.
Dear Framers,
In article <42550l$g5e@jabba.ess.harris.com>, Cheryl Woodside <caw>
[sic]wrote:
> ... when you generate the book, and it reads that a chapter must start on
> a right page ("Read from File"), it adds the extra page to the previous
> chapter and changes the chapter's setting to "Even Pages."
I've confirmed this to be Frame's behavior, and I don't like it.
As far as I'm concerned, Frame has no business messing with the settings
of Chapter 5 just because Chapter 6 starts on a right page. There's no
logical reason for it. Chapter 5's Delete Empty Pages and Chapter 6's
Start on Right Page are in no way inconsistent.-- Frame just needs to keep
track of the page numbers, and needs to make sure that appropriate empty
pages are imaged (a PostScript showpage and nothing else) when a book file
is printed.
But Frame changes Chapter 5's settings upon Generate/Update, depending on
Chapter 6's settings. Yecch.
If the implementation of Frame requires that a page be added to Chapter 5
in order to properly accommodate the intervening empty page, OK, but that
page should have a master of None and it should evaporate of its own
accord if Chapter 5 grows.
If I WANT headers and footers on blank versos this is easy -- I can set
each chapter to "Make Page Count Even". That will cause a Left master to
be used without any intervention on my part, and "the right thing" happens
on insertion and deletion of pages.
> [Frame's] own Users Manual has blank last pages when page count is odd.
This is the convention in book publishing: "No folio on a blank verso."
I'm about to deliver a 350 page book to Wiley, and I'm going to be annoyed
to make a pass through to manually set all those blank versos to None to
conform to Wiley's house style. Then I'll be annoyed again when I re-edit,
because those damn None pages stick around unless explicitly deleted. I
confess I had assumed that Frame would work correctly, and my suspicion
http://www.poynton.com/notes/short_subjects/FrameMaker/no_foilio_on_blank_verso (1 di 2) [15/01/2005 16:27:51]
http://www.poynton.com/notes/short_subjects/FrameMaker/no_foilio_on_blank_verso
was confirmed -- in error, it seems -- by a glance at the Frame manual.

C.
Charles Poynton
tel:
416 486 3271
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/no_foilio_on_blank_verso (2 di 2) [15/01/2005 16:27:51]
http://www.poynton.com/notes/short_subjects/FrameMaker/one_or_more_spaces
Doris Chen writes to the Framers mailing list,

<<
Your comments only touched upon another series of typographic problems which
revolves around the handling of the space character in Frame. There are six
situations in which the space character should never be allowed:
At the start of a paragraph
At the end of a paragraph
Before a tab stop
After a tab stop
Before a hard return
After a hard return
All of these items stand in the way of precision type setting. If you think
the anomylies that you discussed concerning the edge of a text column are bad
you should see what these situations do to the type setting of tables.
If any of the preceeding situations occur in a table it is impossible to get
columns of data to line up even when you are using fixed width fonts. As a
minimum the smart spaces option should prevent you from ever creating any of
the preceeding situations. Further the extra spaces optionn of the spell
checker should report and correct any of these situations which occur.
Another related problem is allowing typographic font changes (font, size, bold,
etc.) for blocks of text which contain only spaces. As a document ages, a
problem occurs as a result of subsequent editting.
Say you had some text that was originally bold.
You later deleted or made unbold the text, but did not include all the
spaces in the font change. (Most common occurrance is leading or trailing
blanks).
You now have a bold blank between two words in normal font. The
typesetting algorythms in Frame treat the bold blank as having a differrent
width (as they should). The results can be really strange and difficult for a
user to detect on the screen.
The solution is for Frame at least at open and save time to change spaces which
do not have the typographic attributes of their surrounding text set to match
the surrounding text. This should not be done to much more often then open or
save since it is a very compute intensive task.
The problem is so severe we had to write an API program to go through documents
and clean out these problems before final printing.
------------------------------------------------------------------------------For more information contact:
Doris Chen
Softline International, Inc.
voice (510) 849-9817
fax
(510) 849-0156
email doris@softline.com
------------------------------------------------------------------------------>>
(A copy of this message has also been posted to the following newsgroups:
comp.text.frame)
In the Framers note forwarded by me
<Poynton-1308952222020001@ts6-15.inforamp.net>, Doris Chen wrote,
> There are six situations in which the space character should never be allowed:
[numbers mine]
> 1 At the start of a paragraph
> 2 At the end of a paragraph
> 3 Before a tab stop
http://www.poynton.com/notes/short_subjects/FrameMaker/one_or_more_spaces (1 di 3) [15/01/2005 16:27:54]
> 4
> 5
> 6
>
> All
After a tab stop

Before a hard return
After a hard return
of these items stand in the way of precision type setting.
I agree that Frame suffers from all of these problems.

Since version 1.05 on the Mac, eight years ago (!), Microsoft Word has not
suffered from problems 2 or 5. These spaces are allowed in a document, but
they properly make no contribution to linebreaks or wordspacing of the
line.
You could argue that cases 2 and 5 should not be permitted, but here's an
argument in favor of Word's treatment: it makes sense to treat the space
following a word as an entity to be edited along with the word. In the
English language a word is frequently followed by a space! Consider moving
a word or two from one place to another in a document, surely a common
occurrence. I double-click to select a word (perhaps I then drag to select
a few more), then cut, then place the insertion point somewhere else, then
paste. I probably want the space character to come along with the word.
This is precisely what happends in Word. And it extends beautifully into
drag'n'drop editing: double-click (maybe drag), let go (see the
selection), mouse-down on it, then drag'n'drop it. The spaces work out as
a function of the design of the interaction.
In Frame, there's no drag'n'drop. But in the select-cut-click-paste
scenario, with smart spaces off, two spaces are left at the source and
none are provided at the destination. With so-called "smart spaces", the
two spaces at the source close up into one, but no space is provided at
the destination: the inserted word butts up against the word at the point
of insertion.
What I've described is what I consider to be a class-I human-interface bug
in the "smart spaces" logic. In Frame 3.1 this all worked fine, as in
Word, but it was broken upon the release of 4.0 and remains broken in 5.0,
despite my continued pleas to tech support. I wrote and submitted a
one-pager on this more than a year ago, and followed up with two or three
phone calls. But no one that I spoke with at Frame considered this to be
misbehavior on the part of Frame, despite the behavior being demonstrably
different from 3.1 to 4.0.
This relates to spaces before paragraph marks and spaces before hard
returns: if sentences are to be treated as entities during editing, they
too should be permitted to be followed by spaces, without interfering with
layout of lines or table cells. I ought to be enabled to drag across a
sentence (complete with period and following space), cut, click and paste
-- and produce an intact paragraph, without having to manually trim the
spaces. In Word, I can easily do what I want. In Frame, if a space
immediately precedes a paragraph mark, the line layout (or table cell
layout) algorithm breaks.
Word follows a typewriter model, so in cases 1, 2, 4 and 6 it makes sense
for Word to act on the space. Otherwise the Word users would get too
confused. I concur with your view that in FrameMaker these should be
suppressed. I would argue to leave one of them in the document, but have
it occupy zero-width, like a marker does today: simply have the layout
machinery ignore them, as it should ignore the spaces in cases 2 and 5
above.
> Another related problem is allowing typographic font changes (font,
size, bold,
> etc.) for blocks of text which contain only spaces.
I too have been burned by this one.
C.
Charles Poynton
tel:
416 486 3271
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/Shrinkwrap_leaves_a_1_pt_gap
Subject: Re: Shrink-wrap leaves a 1 pt border
Date: Sun, 13 Aug 1995 13:04:04 +0000
Lines: 114
Distribution: world
References: <40e6t0$9pp@yuggoth.ucsb.edu>
In article <40e6t0$9pp@yuggoth.ucsb.edu>, steve@tweedledee.ucsb.edu (Steve
Trainoff) wrote:
> when I import my graphics [in 3.0.1] I use the "shrink-wrap" command to
> shrink the enclosing frame to the size of my graphic. Unfortunately FM
> leaves a 1 point border around the graphic.
This behavior is evident in Frame 4 and Frame 5 as well. In this situation
you have two frames -- the frame with the equation or the graphic, and its
surrounding anchored frame, which FrameMaker has made two points wider and
two points higher. In your post you use the term "border" loosely -- I'll
call it the "slop zone". Strictly speaking, Border refers to what is
painted by the Pen around a frame (unless the Pen is set to None). Border
doesn't affect positioning or size.
I assume that Frame introduced the slop zone to avoid the confusion that
would result among inexperienced users if the two frames were put right on
top of each other -- the newbie wouldn't know how to select one and not
the other, he might not even realize that two frames were involved. I'm
sure that the Frame designer or programmer that decided to do this meant
well, but the two point discrepancy is a big nuisance if you are trying to
achieve precision, simplicity and good typography. A similar technique is
used by Frame when it imports a graphic onto the page or into a text
column: the graphic appears at the center of a frame that's not one but
TWELVE points wider and higher than the graphic.
You can defeat the slop zone by manually setting Offset from Top and
Offset from Left of the interior (equation or imported graphic) frame to
zero, and then setting the Height and Width of the exterior (surrounding)
frame to be exactly the same as the Height and Width of the equation or
imported graphic. But what a pain. On my Mac, I already have my finger on
the Option key to invoke Shrink-Wrap (Option-Keypad *), otherwise I would
suggest to Frame that Option-Shrink-Wrap should produce an exact fit. But
some mechanism should be available to produce a fit that isn't sloppy. How
about a preference setting for "Slop zone" that would default to 1 point
for the newbies, 6 points for imported graphics, but that could be set to
zero for precision work?
Related to the issue of surrounding frames, FrameMaker enforces a minimum
0.015 point (1/4800 inch) width on any border. This setting is
inconsequential if Pen is set to to None, but sometimes I wish I could
just set it to zero.
Here's another problem. High-quality typefaces are designed so that the
tops of the characters all appear, optically, to have the same elevation.
The top of a character like "T" is located along the top boundary of the
em-square. But the top of the curve in a capital "O" (or S or Q or C) must
extend beyond the em-square in order to achieve an optical match. If the
font designer makes the top of the O precisely match the top of the T,
when they're set together it looks like the O came from a smaller font.
In FrameMaker, when you put a text frame directly on the page, the tops of
these characters are rendered correctly, slightly outside the text frame.
But if the text is set within in an anchored or unanchored frame, with an
Offset from Top of zero, the excursions outside the em-square are cropped
http://www.poynton.com/notes/short_subjects/FrameMaker/Shrinkwrap_leaves_a_1_pt_gap (1 di 2) [15/01/2005 16:27:57]
http://www.poynton.com/notes/short_subjects/FrameMaker/Shrinkwrap_leaves_a_1_pt_gap
by Frame -- when printed, your characters will be missing those

components. I find it a highly visible artifact, even in ten-point type on
a 300 dpi laserprinter, because the chopping is quite abrupt. To prevent
this behaviour, you must sacrifice alignment and offset the frame downward
slightly, say a point or so (for typical body type size). This means that
you can't achieve good typography by simply dragging a figure caption to
the top of its anchored frame and letting it snap to the top boundary -if you do that, your characters will be clipped.
There is an analogous problem in the horizontal dimension. It is common
for a font to have characters whose shape extends horizontally outside
what a typographer calles the "body" of the character. In Times-Italic for
example, the "fi" and "fl" ligatures, and the beta characters
have "negative side-bearings": portions of these characters extend to the
left of the body. When placed in a text frame directly on the page, the
edges of these characters appear cropped on-screen, but print properly.
However, when enclosed in a FrameMaker anchored frame with an Offset from
Left of zero, these characters are cropped on display and upon printing.
In Times-Italic, the beta character is the worst case: to avoid Frame's
poor behavior you must manually introduce an Offset from Left of about 1/5
of the point size.
Now let me tie these problems together: the 1-point slop zone upon
shrink-wrapping masks the cropping of excursions that would occur if the
element were the same size as its surround. A cynic might say that the
1-point zone was introduced to avoid cropping, treating the symptom rather
than the disease.
I have collected a list of several word-spacing problems
one: even when word spacing is supposed to be absolutely
100%, optimum 100%, maximum 100%), when the last line of
full, FrameMaker shrinks the word spacing in a line if a
immediately precedes the paragraph mark.
in Frame. Here's
fixed (minimum
a paragraph is
space character
None of these issues materially affects the use of Frame to produce

technical manuals, but they all deter high quality work. I'm a fan of
Frame Technology, but Frame's lack of typographic awareness explains why
serious work is done with PageMaker and Quark.
Recommendations to Frame:
1. Provide an easily-accessible Shrink-Wrap with exact fit.
2. Remove the border restrictions -- allow me to set a border of zero.
3. Repair the mistreatment of excursions of a character's shape outside
its "body" -- when you set a character, set the whole character, not just
most of it.
4. Forgive my being blunt, Frame: learn about typography.
C.
Charles Poynton
tel:
416 413 1377
http://www.poynton.com/notes/short_subjects/FrameMaker/Shrinkwrap_leaves_a_1_pt_gap (2 di 2) [15/01/2005 16:27:57]
http://www.poynton.com/notes/short_subjects/FrameMaker/Frame_Acrobat
Path: inforamp.net!woody07.inforamp.net!user
Subject: Re: Frame-Acrobat
Date: Thu, 02 Feb 1995 22:40:53 -0500
Organization: Poynton Vector
Lines: 134
Message-ID: <poynton-0202952240530001@woody07.inforamp.net>
References: <Larry.Dybala.4.0014E919@psl-online.com>
NNTP-Posting-Host: coach13.inforamp.net
Steve Weyer writes,
> Charles checks this group, but I'll send him email ...
You guys are talking about me while my back's turned! First, to answer Lisa,
> Can anyone tell me how to get a copy of Charles Poynton's tech note?
> Or any other such tech notes?
A page of descriptions and links is at
<http://www.inforamp.net/~poynton/Poynton-frame.html>
If you don't have a web browser, get one! Seriously, the readers of this group
are by definition specialists in technical communication and the distribution of
information. The web is vital in that respect. I have written a document about
using a Mac to access the net using a dial-up PPP service provider. You can find
its description and location in the README at the top level of my ftp
directory. It's also on the web linked from my home page (in the signature
below), but if you're reading it from the web, you don't need it unless you
want to learn about dialup PPP, or you have a flaky MacTCP/MacPPP connection
and want to fix it.
Now, down to the business of these Acrobat links. As Steve correctly
points out, my procedure merely adds PostScript code to the PostScript
that you pull from Frame. The procedure just adds bookmarks. It will
work in any scenario below. The previous version of the note omitted
the to mention a few limitations of current Frame implementations with
respect to PostScript generation and pdfmark inclusion. I will
summarize these limitations. None of these are directly implicated in
the bookmark procedure, although there's an indirect connection.
First of all, Frame 4.0.2 or later on Mac produces in its PostScript "pdfmark"
operators corresponding to hypertext crossreferences. When distilled, those
operators generate hypertext links in the resulting PDF file. It's very nice!
BUT -- Frame on a Mac cannot produce a single PostScript file from a book. It
produces a SEPARATE PostScript file from ECAH Frame file in a book (call them
"chapters"). Each of these PostScript files appears to start with page one, and
all of the internal crossreference links (call them intrachapter links) are
referenced to page one. When you concatenate for distilling in a single job, the
pages will all image correctly, but the hypertext links internal to each chapter
will all collide at the front of the book. "Known bug." Furthermore, Frame's
Postscript includes code for intrachapter links, but has no PostScript code
(pdfmark operators) for crossreferences outside the chapter (interchapter
links).
Now, Frame on UNIX and Windows produces a single PostScript file for a book.
BUT -- Frame on UNIX and Windows versions don't produce any pdfmark
operators. The current versions offer no possibility of Acrobat links.
"Known problem." This is contrary to Steve's note (which is otherwise a
very good summary and has many good tips) -- I too shared this
misconception until Rick Oliver corrected it.
SO you can have (i) a one-chapter book with fully functional links [Mac],
http://www.poynton.com/notes/short_subjects/FrameMaker/Frame_Acrobat (1 di 2) [15/01/2005 16:28:00]
http://www.poynton.com/notes/short_subjects/FrameMaker/Frame_Acrobat
or (ii) a multi-chapter book with faulty intrachapter links and no

interchapter links [Mac], or (iii) a single- or multi-chapter book with
no links at all [UNIX, Windows].
You can take individual Mac files and Insert them one-by-one using Acrobat
Exchange to make a single composite PDF file. This will maintain the
intrachapter links, but due to being performed behind the back of the
bookmark procedure, will defeat the bookmarks.
I have explored writing a program to rejeuvenate the page numbers in Mac
Frame's individual PostScript files, just to get the
multi-chapter-book-with-intrachapter-links case working. Much to my
surprise, I discovered that the page numbers in the ultimate book are
not found in machine-readable form anywhere in the generated
PostScript, except for the place where they are presented for imaging
onto the page. This makes it pretty difficult to parse them out in an
automated way, because the page numbers can be anywhere on the page, or
not there, or in roman numerals, etc. Even if this correction could be
made to each chapter's internal crossreference links, the it would be a
case of diminishing returns because Frame has not included in its
PostScript code the interchapter crossreferences.
In my Color FAQ, I hand-edited the PostScript to fix the crossreference
links. No fun. And mine is in effect only a two-chapter book (TOC and
main body). You could also manually fix them up in Exchange, but that's
no fun either (I did that in the previous version of the Color FAQ). My
document is MUCH shorter and has many fewer crossreferences than the
documents that most of you are working on.
So that's where we stand. For now, if you need Acrobat links, stick to
single-chapter books printed to PostScript on a Mac.
Larry Dybala's idea of putting the TOC at the end of the PDF is a very good
trick -- diabolical, may I say? And Steve notes that if a generated index
is pasted at the end of a document, its links can be made to work.
Finally, if you're still with me, a word about the framers mailing list
reflector. I enjoy reading the comp.text.frame newsgroup once every two
or three days, except when I'm really busy when it's once a week. I
guess its content is partially-overlapping with the framers mailing
list. I receive those e-mails, but not at my main e-mail address. I
couldn't stand it -- getting interrupted twelve or fifteen times daily,
having non-essential mail dropping into my in basket. I prefer to read
news in a more controlled way, like reading the newspaper instead of
reading the letters left by my postman. So I am seriously contemplating
dropping my Framers subscription. But if I do so I will miss lots of
nice juicy little morsels and the odd free lunch. If anyone out there
has news-to-mail or even better mail-to-news gateway capability, or
even knows a guy who said he has a friend who heard of someone that
could, please drop me a line and give some thought to how we could
coerce that person to start up a gateway service.
That way, we can all stick together.
Thanks for listening,
C.
-Charles Poynton
tel:
416 486 3271
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/Frame_Acrobat (2 di 2) [15/01/2005 16:28:00]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMaker_last_char_of_para
From: poynton@inforamp.net (Charles A. Poynton)
Subject: Re: How to prevent changing the standard font of a paragraph?
Date: Sun, 04 Dec 1994 14:09:38 -0500
Organization: InfoRamp inc., Toronto, Ontario (416) 363-9100
Lines: 85
Distribution: world
References: <3bk0di$24k@sunserver.lrz-muenchen.de>
NNTP-Posting-Host: woody03.inforamp.net
In article <3bk0di$24k@sunserver.lrz-muenchen.de>,
sta@mmk.e-technik.tu-muenchen.de wrote:
> [FM4/Solaris] If I change the font format ... while writing AT THE END
of a paragraph,
> and I change it back to the standard format ... then the standard font
> of the WHOLE paragraph will change to the previous selected format.
This is apparently a bug. On 4.0.4p1 on my Mac Quadra, changing the size
or weight or angle of just the last character of a paragraph -- the
character immediately preceding the paragraph mark -- would cause the
entire paragraph to change, just as Holger observed. It was clearly a bug
and not a feature because Undo would only undo the change to the intended
character and not to the rest of the paragraph. Same deal for the last
character of a table entry, the character before the end-of-flow mark.
Changing the character format of just that one character would make the
whole entry change.
But it's flaky -- today, I can't recreate it.
How about this one -- a display bug where any character with a decender
gets displayed propped up on an x-height or so worth of pedestal. Printing
was fine. The bug was zoom-factor dependent -- changing zoom factor
sometimes made it go away. But usually only for a while - it would come
back. Quitting Frame and restarting would make it go away for a while,
anything from twenty minutes to six hours of heavy use. But eventually it
would come back. This on a system with all the TrueType fonts ripped out
-- which is the recommended situation -- running SuperATM 3.6. I could
reproduce it quite reliably with all extensions disabled except ATM.
The problem disappeared when I upgraded to System 7.5 and ATM 3.8.1LE.
Anyone else seen it?
If you have read this far, you must be concerned about bugs and
idiosynchracies. Does it bug anyone else how whitespace at the end of a
line affects wordspace and letterspace calculations? In PageMaker, even in
Microsoft Word, a space character at the end of a line makes no
contribution whatsoever to justification or wordspacing -- it even floats
off past the end of the column if necessary. This is as it should be.
Frame has no typographic sense here -- every space character at the end of
a line contributes to squeezing wordspace or even letterspace to its left.
In a document with Smart Spaces, the extent of the damage is limited an en
or so, but if you run with smart spaces off, amazing things happen. As you
add spaces immediately left of the line end (paragraph mark, end of flow,
or line break), the rest of the line starts squeezing into less and less
space. The line-breaking machinery of Frame refuses to allow the line
formatter to make a new break solely because of space characters, so it
squeezes out to infinity if it has to, well past violating the word space
parameters that you have set in the Advanced properties of the Paragraph
Designer.
It's even funnier in a Centered paragraph. As you add space at the end of
a centered line, with default wordspace parameters (min 90, opt 100, max
110), first the text in the line gets leterspaced a little wider, then
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMaker_last_char_of_para (1 di 2) [15/01/2005 16:28:07]
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMaker_last_char_of_para
stabilizes, then starts squishing.

The design principle here should be, the reader does not know or care how
many spaces were typed at the end of a line, so he should not see their
effects either. I would argue that they should be swallowed at the start
of a line as well, but I admit that there is a certain marketing necessity
to enable Frame to be used by people who grew up with typewriters.
PageMaker and Word concede this point also.
There is a very practical, down-to-earth aspect of this problem. I put a
space character at the end of every sentence. This allows me to cut and
paste sentences, even from the middle of a paragraph to an end or from an
end to a middle. But Frame loses its typographic sense unless I manually
post-process to rip out the spaces from paragraph ends. I also habitually
put a space at the end of a word, even a word in a table entry. This
allows me to cut and paste words without having to manually repair spacing
afterwards. But if leave a space character as the last character of a
table entry, the other characters in the entry are unjustifiably squeezed.
C.
-Charles Poynton
http://www.poynton.com/notes/short_subjects/FrameMaker/FrameMaker_last_char_of_para (2 di 2) [15/01/2005 16:28:07]
http://www.poynton.com/notes/short_subjects/FrameMaker/Small_Caps_character_format
Subject: Re: Small Caps character formatting
Date: Thu, 02 Feb 1995 19:23:16 -0500
Lines: 46
References: <3gpmb1$cog@vixen.cso.uiuc.edu>
NNTP-Posting-Host: woody07.inforamp.net
In article <3gpmb1$cog@vixen.cso.uiuc.edu>, George Thiruvathukal
<gt2727@coewl.cen.uiuc.edu> wrote:
>
>
>
>
Does anyone know how to format text as small caps in FrameMaker 4?

If I am not using the correct Frame lingo, I'm using small caps to
describe text in which the lower case letters look like capitals,
except that they are slightly smaller than true capital letters.
You accurately describe word processor and even page layout lingo for
small caps but they generally look terrible, because they are produced in
exactly the way you describe: the smaller caps are just shrunk big caps.
The problem is that the stem width -- line weight, if you like -- changes
in proportion to the shrinkage. At 72 dpi you can't tell the difference
but it doesn't cut it as typography, not even on a 300 dpi LaserWriter.
You can do this in Frame, if you like, by clicking the Small Caps
attribute in the Character Designer, or the comparable attribute in the
Default Font panel of the Paragraph Designer.
If you want typographic-quality small caps, then you have to buy an
"expert" font that includes, in addition to the uppercase alphabet, a
small-caps alphabet whose stems have the right weight. These "real" small
caps can really make a polished document. You get them simply by choosing
the font -- the big ones are where you expect on the keyboard, and the
little ones are .. well, they're where you expect, too. .
Expert fonts usually also include old-style figures (OSF), in other words
lower-case figures, with descenders. In non-technical matter -- and even
technical matter set classically -- these can look very sharp.
This all paraphrased from Robert Bringhurst, "The Elements of Typographic
Style".
Want more info? Ask in comp.fonts.
C.
Charles Poynton
tel:
416 486 3271
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/Small_Caps_character_format [15/01/2005 16:28:11]
http://www.poynton.com/notes/short_subjects/FrameMaker/publ_quality_screenshots
>From poynton@poynton.com Tue Jan 31 23:54 EST 1995

Date: Tue, 31 Jan 1995 23:55:55 -0500
To: MIKE.NELSON%MSFC26PO@x400gw.msfc.nasa.gov (Michael A. Nelson),
framers@uunet.uu.net
Subject: Re: HELP! Need good screen capture for users guides!
In article <MIKE.NELSON-3101951115020001@128.158.44.231>,
MIKE.NELSON%MSFC26PO@x400gw.msfc.nasa.gov (Michael A. Nelson) wrote:
>
>
>
>
We need something that will translate these screen

captures into a higher resolution so that the quality will
be maintained when these images are printed out on a 300
dpi or greater printer.
This will be an interesting project. Here are a few thoughts.

There's a screen capture utility on the Mac that captures the PICT drawing
commands into a PICT file instead of capturing the bits on the screen.
That should improve certain kinds of screen captures. In particular, you
get a PICT with real characters and fonts, instead of coarse bitmaps. In
other words the bits and lines look like bits and lines, but the
characters are smooth. I guess this is not much help if your app is
running on UNIX but I thought I'd mention it.
Bitmapped screen captures often capture dithering patterns used on screen.
You must not print these - they'll alias into Moire checkerboards in your
book. You mention "commercially-produced guides" - there's examples of
screenshot checkerboard Moire in IBM's OS/2 WARP manual, and when Hayden
Books produced The Tao of AppleScript, 2nd edition they JPEG-compressed
their screenshots for gawd's sake! There's fringes and ringing around the
edges. Looks like bad video. So it seems to me that you and I are on the
leading edge as much as they are. Even if you avoid systemic dithering,
some ancient software still lingering in contemporary operating systems
has dither built-in. For example, on a Mac the elevator bars are dithered
no matter what screen resolution you are using. Use Photoshop or something
similar to touch up the screen captures to fill dithered areas with
uniform grey levels. This is the approach thatt I took in a document that
I wrote that has a lot of screen shots - check out the Mac-Internet
docs linked from the Mac section of my home page. A similar problem is
software that writes closely-spaced lines to the screen, as in the
racing stripes of an active window's titlebar on the Mac. These are OK
at 72 dpi, and reproduce without arifact at 144, 216 and 288 dpi. But
reproduction at other dpi settings - 300, say! - produces uneven line
spacing. You have to choose scale factors carefully, and in awareness
of the resolution capability of the output device.
If you avoid dithering in the screen shot - perhaps by capturing the app
running in 24 bit colour - then you could use Adobe Streamline to trace
the bitmap. Then you could scale to any size, any resolution. You would
have to coerce Streamline to turn sharp corners at every pixel. I have no
experience with this but I am certain that you could force it to do this,
even if you had to pass it a bitmap image resixed to sixteen times. If you
resize a bitmapped screenshot in Photoshop, choose Nearest Neighbor
Interpolation so as to replicate pixels, in one of the rare instances
where this is the right thing to do.
Please report what you find back to the rest of us!
C.
Charles Poynton
tel:
416 486 3271
http://www.poynton.com/notes/short_subjects/FrameMaker/publ_quality_screenshots (1 di 2) [15/01/2005 16:28:14]
http://www.poynton.com/notes/short_subjects/FrameMaker/publ_quality_screenshots
fax:
416 486 3657
http://www.poynton.com/notes/short_subjects/FrameMaker/publ_quality_screenshots (2 di 2) [15/01/2005 16:28:14]
http://www.poynton.com/notes/short_subjects/FrameMaker/Acrobat_links_FrameMaker_xref
(A copy of this message has also been posted to the following newsgroups:
comp.text.frame, comp.text.pdf)
An anonymous correspondent unfamiliar with FrameMaker asks about making
Acrobat links. I explain the basics, then segu into some Frame esoterica.
>
>
>
>
Do you know of any automated process to introduce hypertext links into

a FrameMaker 5 document? For instance, a script would look at a
document and whenever it saw a phrase like "see section 3.4", a
hypertext link would be created. This link would point to section 3.4.
I see that Microsoft Word has traumatized you, locked you in the 80's.
In Frame, you can arrange for this all to happen more or less
automatically. Don't type the literal text "see section 3.4". Instead,
when you want to insert a crossreference, choose Cross-Reference from the
Special menu (or hit Command-K for "Kross-reference").
A dialog box invites you to choose a Source Document -- select an open
document, perhaps the document you're in, perhaps a different document.
Define a Format -- in this case, use
see section <$paranum>
Choose a Source Type -- select the paragraph tag that you use for section
headings. Finally, pick a Reference Source from a scrolling list that
displays all the numbers and titles of your section "paragraphs".
Click OK. Your desired "see section 3.4" now appears at the insertion
point. But it hasn't been inserted as editable text, it's an atomic
crossreference ("marker"). Should you insert or delete a few sections
ahead of 3.4, just Update Cross-References and the reference will be
renumbered.
These atomic crossreferences will Distill automatically into Acrobat PDF
links. No muss, no fuss, no bother.[1]
I personally prefer to use a crossreference that displays the section
title (italicized) and its page number, instead of the section number. The
section title gives the reader a sense of the content of the referenced
section without forcing him to go there. The page number allows the reader
of a paper copy to immediately turn to the indicated page rather than
having to leaf through the document scanning through the sections. I omit
the word "see" from my format definitions, to allow me to use the same
format for references that differ grammatically. Instead of "See section
3.4", I may choose in a particular passage to write "Section 3.4 gives
details of this procedure." The grammatical variety will be appreciated by
your reader.
Once you start exploiting crossreferences, you will be tempted to invent
many specialized formats, one for Sections, one for Chapters, one for
Figures and so on. But I define just a few formats that are useful for
almost all types of reference. I make sure to use style names (tags) that
are sensible words like Section, Chapter, Figure, Table and so on. I
defined this crossreference format, which I call "on_page":
<Italic><$paratext><Default P Font> on page\ <$pagenum>
As you may guess, the codes <Italic> and <Default P Font> turn italics on
and off. The code <$paratext> inserts the text contained in the referenced
item, in this case, the title of section 3.4. When you redefine a format,
just Update Cross-References -- in this example, a reference that
previously appeared "see section 3.4" will take the form "see Delivering
Babies on page 121" (where the two words Delivering Babies are
italicized). I place a nonbreaking space between the word page and the
page number; this is entered on a Mac as the two character sequence
backslash-space.
http://www.poynton.com/notes/short_subjects/FrameMaker/Acrobat_links_FrameMaker_xref (1 di 2) [15/01/2005 16:28:23]
http://www.poynton.com/notes/short_subjects/FrameMaker/Acrobat_links_FrameMaker_xref
I have four additional formats, which I named "above", "below", "opposite"

and "overleaf". They are all nearly identical; here's "above":
<$paratag>\ <$paranum> above
When I need to refer to a Figure or Table or Equation that lies above,
below, opposite or overleaf from the reference, I simply choose the
appropriate format. The number is picked up from the (autonumbered) item;
the number is updated automatically should I insert, delete or rearrange
figures, tables or equations. In case the referenced item lies far away, I
have a fifth format "distant" defined as
<$paratag>\ <$paranum> on page\ <$pagenum>
This expands to something like "Equation 4-13 on page 65".
Ideally I would like one of these five types -- above, below, opposite,
overleaf or distant -- to be chosen by Frame based on the relative
positions of the crossreference and the referenced item upon layout. But
for now I am happy to manually change the format type depending on how
things fall on the page. You can easily search for crossreferences to
validate that appropriate pointers have been used throughout a document,
after editing.[2] One minor stylistic problem with my approach is that
Section, Chapter, Figure, Table and so on always appear with an initial
capital.[3]
To see how all this works in a real document[4], check the PDF version of
my Colour FAQ, linked through my home page on the web.
C.
[1] If you want the crossreferences to show up the Acrobat Reader coloured
green, just include <Green> in the format string, and define the Green
character format accordingly. But I confess there is then some fuss if you
want Green to change to Black when you print to paper.
[2] Of course, you can eliminate the need for this checking by using rigid
reference formats that are the same no matter where the referenced item is
located, but I find that my scheme leads to a more fluid, polished
document.
[3] 1995/07/13. Dear Frame Technology; Re: Request for Enhancement. Please
implement a building-block <$Paratag> that behaves identically to
<$paratag>, but if the first character of the tag is a lowercase letter,
force it to uppercase. Sincerely, /s/ Charles A. Poynton [Frame 4.0.4p1 on
Mac Quadra 900 running System 7.5.1]
[4] Incidentally, I never use footnotes in a real document. Hate them.
Despise them.
----Charles Poynton
http://www.poynton.com/notes/short_subjects/FrameMaker/Acrobat_links_FrameMaker_xref (2 di 2) [15/01/2005 16:28:23]
Charles Poynton - FrameMaker links
Charles Poynton FrameMaker links

This page lists resources available on the Internet concerning the FrameMaker page layout and
publication system.
Poynton's Documents about FrameMaker
See also Poynton's Typography and Design, links to FrameMaker Programs.
Adobe Technology maintains a web site for Frame, and ftp sites for FrameMaker,
FrameMaker+SGML, FrameReader, and FrameViewer.
External pages:
"framers" and other online forums

FrameMaker FAQ at FrameUsers.com
RPI Frame help (for beginners only!)
FrameMaker filters at CERN
FrameMaker links at Yahoo, HTML info,
MML Reference by SoftLine
Creating Drop Caps in FrameMaker 5
Commercial pages:
Frame Technology
Frank Stearns Associates
Charles - links
2000-01-13
http://www.poynton.com/notes/links/framemaker-links.html [15/01/2005 16:28:27]
Charles Poynton - Typography and Design
Charles Poynton Typography and Design

Standardization, instead of individualization.
Cheap books, instead of private-press editions.
Active literature, instead of passive leather bindings.
- Jan Tschichold, 1930
(tacked up near the site formerly known as Todd Fahrner's comfy chair).
I have written a few documents about typography, and about issues of information design and
presentation in the digital world.
Ten Common Mistakes in the Typesetting of Technical Documents
Writing SI Units and Symbols
Technical, production nitty-gritty: Introduction to RIPping
Making web pages usable
My thinking about typography and graphic design has been greatly influenced by these three books.
I consider them to be mandatory reading for any aspiring typographer or illustrator:
Robert Bringhurst, The Elements of Typographic Style. Second edition (Vancouver: Hartley &
Marks, 1996).
Jan Tschichold, The Form of the Book (Vancouver: Hartley & Marks, 1991). Originally
published in German in 1975.
Edward R. Tufte, Envisioning Information (Cheshire, CT: Graphic Press, 1990).
For the first two,
Hartley & Marks Publishers Inc.
3661 West Broadway,
Vancouver, BC V6R 2B8
Canada
+1 604 739 1771
There's apparently a branch office, probably for taking orders from U.S. addreses:
http://www.poynton.com/Poynton-typo-design.html (1 di 2) [15/01/2005 16:28:29]
Charles Poynton - Typography and Design
Box 147
Point Roberts, WA 98281
U.S.A.
I also recommend this classic:
Jan Tschichold, The New Typography (Berkeley, Calif.: University of California Press, 1995).
The book is now in paperback, available at Amazon.
I am engaged in a campaign to modernize the archaic habit, recommended in the Chicago Manual of
Style and many other places, of typesetting a dash as an em dash with no spaces. Here's a lovely quote
from Bringhurst:
"The em dash is the nineteenth-century standard, still prescribed in many editorial style
books, but the em dash is too long for use with the best text faces. Like the oversized
space between sentences, it belongs to the padded and corseted aesthetic of Victorian
typography."
Bringhurst suggests - and I concur - that an appositional phrase should be set off by spaced en-dashes.
Other Typography links ...
Charles
1999-08-03
http://www.poynton.com/Poynton-typo-design.html (2 di 2) [15/01/2005 16:28:29]
Writing SI units and symbols

Charles Poynton
This note explains how to write quantities and units of the Systme international d'units (SI),
colloquially known as the metric system. I catalog the power-of-ten prefixes, and I list some important
units.
Write a numeric value with units in either the journalistic style, using prefix and unit names (four
kilohertz); or the scientific style, using prefix and unit symbols (4 kHz). Don't mix these styles: Do not
mix a prefix name with a unit symbol (WRONG: kiloHz), or a prefix symbol with a unit name
(WRONG: kHertz). Avoid "abbreviations" for units (WRONG: sec., amp); use the unit names or
symbols instead.
If you are writing for an international audience, express values in the metric (SI) system used by the
majority of the world's population. If appropriate, follow an SI value with the equivalent Imperial value
in parentheses. Express the Imperial value with an accuracy comparable to the original: write 5 m
(16 feet), not 5 m (16.4042 feet). Spell out inch, foot, pound and so on: Do not abbreviate to in, ft, and lb
unless space is an overriding concern. Do not use " and ' symbols for inch and foot: These symbols are
easily lost in reproduction, and they are unfamiliar to a large fraction of the world's population.
You are reading the HTML version of this document. The HTML has has crude representations of
some of the characters in the typographic original; it also suffers from lack of line-break control.
A typographic-quality version of this document is available in Acrobat PDF format, in US letter size
optimized for printing (PDF format, 81289 bytes).
Journalistic Style
In free text, use journalistic style for units and measurements: Spell out numbers one through ten in
words; express numbers larger than that in numerals. Follow a number by a space, then the prefix name
and unit name spelled out entirely in lower case and without spaces: four megahertz, 2.2 microfarads,
3.5 megahertz, 75 ohms.
Use hundred, thousand, million, and so on, only for pure numbers. For a number with a unit, spell out
the SI prefix: four kilowatts (not four thousand watts). Avoid using words for extreme quantities larger
than a million, because billion, trillion, and so on, have different numerical values in different countries.
If you absolutely must use words, follow the example of the BBC World Service: say thousand million
or million million.
http://www.poynton.com/notes/units/index.html (1 di 6) [15/01/2005 16:28:54]
Use a hyphen between a numeral and its unit only when necessary to form a compound modifier, and
only with a unit name, not a unit symbol: 3.5-inch diskette, 35-millimeter film. To avoid the confusion
of two hyphens when a negative number is involved, as in -12-volt power, use a space instead of a
second hyphen.
In many countries a comma indicates the decimal: in these countries the notation 10,000 indicates
precisely ten, not ten thousand! Some of your readers will find it ambiguous if you use a comma as a
separator between three-digit groups. In a numeric value having four or more consecutive digits, use a
space to separate groups of three digits, both left and right of the decimal point.
Scientific Style
In a table, an illustration or a technical text, use the scientific style for measurements and units. Write
the number in figures, followed by a nonbreaking space. Then write the prefix symbol and the unit
symbol with appropriate capitalization and no spaces: 4 MHz,
,
. Separate the last digit
from the unit with a nonbreaking space; this will prevent clumsy line breaks.
SI prefix symbols are capitalized for multipliers
and smaller.
and larger, and lower case for multipliers
A unit symbol is written in lower case, except that its initial letter is capitalized if the unit is named after
a person. These are symbols, not abbreviations or contractions: Do not use periods or other punctuation.
To avoid confusion with math symbols ("variables"), do not italicize unit symbols.
Use appropriate capitalization. The symbol k for kilo - a multiplier of 1000 - combines with hertz as
kHz; the symbol for decibel is written dB. A popular computer in 1987 had a nameplate stating its
memory capacity as 1 mb. In fact it had a megabyte of memory, properly written as 1 MB, not a
millibit!
When you write a negative sign, use a nonbreaking hyphen instead of a regular hyphen. This prevents
the sign from being left stranded at the end of the line: 400 V power results from using a standard hyphen,
-400 V power results from a nonbreaking hyphen. The former is, at the very least, confusing to your
reader. At its worst, it could compromise personal safety.
Dates
Different countries have different conventions for writing dates. A reader in the U.S.A. takes 08/04/50
to be August 4th, but a U.K. reader takes it to be the 8th of April. In the next century, will 01/02/03 be
the first, second or third day of the month? Avoid ambiguity. Write dates in the ISO/IEC 8824 form:
1996-06-07.
Unit Combinations
Use a raised dot between units combined by multiplication, to avoid ambiguity. N m for newton meter
avoids potential confusion with nanometer, nm.
per
Use the per notation for everyday units formed by division, such as miles per hour, mph; revolutions per
minute, rpm; and dots per inch, dpi.
slash
In a scientific or engineering unit formed by division, set off a single-element denominator with a slash:
write m/s for meters per second. I write b/s for bits per second, although some people use bps.
exponents
For a compound unit having a complex denominator, use exponent notation: write
per second squared (NOT m/s/s).
for meters
ohm
Use ohm when the
symbol is unavailable (as in ASCII character code).
degrees
The temperature unit kelvin, K, properly has no degree sign. The non-SI symbols for Celsius ( C) and
Fahrenheit ( F) have degree signs in order to avoid ambiguity with coulomb C and farad F. The term
centigrade is obsolete; the proper term is Celsius.
Computing
b, B
Use little b for bit, big B for Byte. Spell these out where necessary to avoid ambiguity.
k
Little k - pronounced KEY-loh or kill-oh, spelled-out kilo - is the standard SI prefix for
not often used in computing.
(1000). It is
K
(1024) common in computing. Do not write or pronounce big K as
Use big K for the multiplier
kilo; to do so invites confusion with little k, 1000. Simply write it as upper-case K and pronounce it kay.
baud
The term baud does not apply to data rate, but to symbol rate. When you see the unit baud used in
computing, the unit b/s (bit per second) is nearly always meant.
mega, giga
When applied to a base unit other than bit, byte or pixel, M (mega) and G (giga) refer to the SI power-often multipliers
and
. Standard data communication rates are based on powers of ten and use
the SI multipliers, not power-of-two multipliers: 1.544 Mb/s denotes 1 544 000 bits per second; 19 200
bits per second is properly written 19.2 kb/s (not 19.2 Kb/s).
disk storage
When applied to bytes of disk storage capacity:
M (mega) denotes
G (giga) denotes
(1000 K); and

(1 000 000 K).
bits, bytes or pixels

When applied to raw bits, bytes or pixels:
M (mega) denotes
G (giga) denotes
(1024 K); and

.
In computing, M (mega) and G (giga) are ambiguous. M could denote 1 000 000, 1 024 000, or
1 048 576. G could denote 1 000 000 000, 1 024 000 000, 1 048 576 000, or 1 073 741 824. The value
of the giga prefix in computing varies more than 7 percent depending on its context. If an exact value is
important, write out the whole number!
SI Prefix Names, Symbols and

Multipliers
This table contains a complete list of SI prefix multiplier names, symbols, and power-of-ten values,
standardized by the Bureau International des Poids et Measures (BIPM, www.bipm.fr). The symbol
alone, and the term micron, have been abolished: Use
for micrometer. Use lower-case u for
if the micro symbol is unavailable.
Basic SI Unit Names and Symbols

This table includes some important SI units and their derivations, and the names of a few individuals
whose names have been given to units. The seven base SI units are m, kg, s, A, K, mol, and cd; the other
units are derived.
Further information
Information is available at BIPM, http://www.bipm.fr/enus/3_SI/.
Information is available at NIST, http://physics.nist.gov/cuu/Units/. See Guide for the Use of the
International System of Units (SI) [NIST Special Publication 811] (Acrobat PDF format, 400 KB),
Typefaces for symbols in scientific manuscripts (Acrobat PDF format, 62 KB), and SI Unit rules and
style conventions - Check List for Reviewing Manuscripts.
Charles Poynton

Charles Poynton
www.inforamp.net/~poynton
This note explains how to write quantities and units in the Systme
international dunits (SI), loosely called the metric system. I catalog
the power-of-ten prefixes, and I list some important units.
Journalistic style
page 1
Scientific style
page 2
Dates
page 3
Unit combinations
page 3
Computing units
page 3
SI prefix names, symbols and multipliers
page 5
Basic SI unit names and symbols
page 6
Write a numeric value with units in either the journalistic style, using
prefix and unit names (four kilohertz); or the scientific style, using prefix
and unit symbols (4 kHz). Dont mix these styles: Do not mix a prefix
name with a unit symbol (WRONG: kiloHz), or a prefix symbol with
a unit name (WRONG: kHertz). Avoid abbreviations for units
(WRONG: sec., amp); use the unit names or symbols instead.
An inch is exactly 25.4 mm,

by definition.
If you are writing for an international audience, express values in the

metric (SI) system used by the majority of the worlds population. If
appropriate, follow an SI value with the equivalent Imperial value in
parentheses. Express the Imperial value with an accuracy comparable to
the original: write 5 m (16 feet), not 5 m (16.4042 feet). Spell out inch,
foot, pound and so on: Do not abbreviate to in, ft, and lb unless space
is an overriding concern. Do not use and symbols for inch and foot;
these symbols are unfamiliar to a large fraction of the worlds population, and they are easily lost in reproduction.
Journalistic style
In free text, use journalistic style for units and measurements: Spell out
numbers one through ten in words; express numbers larger than that in
numerals. Follow a number by a space, then the prefix name and unit
name spelled out entirely in lower case and without spaces:
four megahertz, 2.2 microfarads, 3.5 megahertz, 75 ohms. (The C in
Celsius is capitalized.)
1 of 6
WRITING SI UNITS AND SYMBOLS
Use hundred, thousand, million, and so on, only for pure numbers. For
a number with a unit, spell out the SI prefix: four kilowatts (not four
thousand watts). Avoid using words for extreme quantities larger than
a million, because billion, trillion, and so on, have different numerical
values in different countries. If you absolutely must use words, avoid
ambiguity by following the example of the BBC World Service: Say
thousand million or million million.
Use a hyphen between a numeral and its unit only when necessary to
form a compound modifier, and only with a unit name, not a unit
symbol: 3.5-inch diskette, 35-millimeter film (WRONG: 35-mm film).
To avoid the confusion of two hyphens when a negative number is
involved, as in -12-volt power, use a space instead of a second hyphen.
In many countries a comma indicates the decimal: In these countries
the notation 10,000 indicates precisely ten, not ten thousand! Some of
your readers will find it ambiguous if you use a comma as a separator
between three-digit groups. In a numeric value having four or more
consecutive digits, use a space to separate groups of three digits, both
left and right of the decimal point.
Scientific style
In a table, an illustration or a technical text, use the scientific style for
measurements and units. Write the number in figures, followed by
a nonbreaking space. Then write the prefix symbol and the unit symbol
with appropriate capitalization and no spaces: 4 MHz, 286 mV, 2.2
F, 75 . Using a nonbreaking space prevents clumsy line breaks, such
as the break between 2.2 and F.
SI prefix symbols are capitalized for multipliers 106 and larger, and
lower case for multipliers 103 and smaller.
A unit symbol is written in lower case, except that its initial letter is
capitalized if the unit is named after a person. These are symbols, not
abbreviations or contractions: Do not use periods or other punctuation.
To avoid confusion with math symbols (variables), do not italicize
unit symbols.
Use appropriate capitalization. The symbol k for kilo a multiplier of
1000 combines with hertz as kHz; the symbol for decibel is
written dB. A popular computer in 1987 had a nameplate stating its
memory capacity as 1 mb. In fact it had a megabyte of memory, properly written as 1 MB, not a millibit!
When you write a negative sign, use a nonbreaking hyphen instead of
a regular hyphen. This prevents the sign from being left stranded: 400 V power results from using a standard hyphen, -400 V power
results from a nonbreaking hyphen. The former, at the very least, is
confusing to your reader; at its worst, it could compromise personal
safety.
Dates
Different countries have different conventions for writing dates.
A reader in the U.S.A. takes 08/04/50 to be August 4th, but a U.K.
reader takes it to be the 8th of April. In the next century, will 01/02/03
be the first, second or third day of the month? Avoid ambiguity. Write
dates in the ISO/IEC 8824 form: 1999-06-30.
Unit combinations
Use a raised dot between units combined by multiplication, to avoid ambiguity. N . m for newton . meter
avoids potential confusion with nanometer, nm.
per Use the per notation for everyday units formed by division, such as miles per hour, mph; revolutions per
minute, rpm; and dots per inch, dpi.
slash In a scientific or engineering unit formed by division,
set off a single-element denominator with a slash:
write m/s for meters per second. I write b/s for bits per
second, although some people use bps.
exponents For a compound unit having a complex denominator,
use exponent notation: write m.s -2 for meters per
second squared (NOT m/s/s).
ohm Use ohm when the symbol is unavailable (as in ASCII
character code).
degrees The temperature unit kelvin, K, properly has no degree
sign. The symbols for the non-SI units celsius (C) and
fahrenheit (F) have degree signs in order to avoid
ambiguity with SI units coulomb C and farad F. The
term centigrade is obsolete; the proper term is celsius.
Computing units
b, B Use little b for bit, big B for Byte. Spell these out where
necessary to avoid ambiguity.
k Little k pronounced KEY-loh or kill-oh, spelled-out
kilo is the standard SI prefix for 103 (1000). It is not
often used in computing.
K Use big K for the multiplier 210 (1024) common in

computing. Do not write or pronounce big K as kilo; to
do so invites confusion with little k, 1000. Simply write
it as upper-case K and pronounce it kay. (This usage
conflicts with K for kelvin, the unit of absolute temperature.)
baud The term baud does not apply to data rate, but to
symbol rate. When you see the unit baud used in
computing, the unit b/s (bit per second) is nearly
always meant.
mega, giga When applied to a base unit other than bit, byte or
pixel, M (mega) and G (giga) refer to the SI power-often multipliers 10 6 and 10 9. Standard data communication rates are based on powers of ten and use the SI
multipliers, not power-of-two multipliers: 1.544 Mb/s
denotes 1 544 000 bits per second; 19 200 bits per
second is properly written 19.2 kb/s (not 19.2 Kb/s).
disk storage When applied to bytes of disk storage capacity:
M (mega) denotes 103 . 210 (1000 K); and
G (giga) denotes 106 . 210 (1 000 000 K).
bits, bytes or pixels When applied to raw bits, bytes or pixels:
M (mega) denotes 220 (1024 K); and
G (giga) denotes 230.
In computing, M (mega) and G (giga) are ambiguous.
M could denote 1 000 000, 1 024 000, or 1 048 576.
G could denote 1 000 000 000, 1 024 000 000, or
1 073 741 824. The value of the giga prefix in
computing varies more than 7 percent depending on
its context. If an exact value is important, write out the
whole number!
SI prefix names, symbols and multipliers

This table contains a complete list of SI prefix multiplier names,
symbols, and power-of-ten values, standardized by the Bureau International des Poids et Measures (BIPM, www.bipm.fr). The symbol
alone, and the term micron, have been abolished: Use m for
micrometer. Use lower-case u for 106 if the micro symbol is
unavailable.
lower case prefix

symbols
upper case prefix

symbols
prefix name
prefix
symbol
power-of-ten
yocto
zepto
atto
femto
pico
nano
micro
milli
y
z
a
f
p
n
1024
1021
1018
1015
1012
109
106
103
centi
deci
[unity]
deka
hecto
c
d
[none]
da
h
102
101
10 0
10+1
10+2
kilo
10+3
mega
giga
tera
peta
exa
zetta
yotta
M
G
T
P
E
Z
Y
10+6
10+9
10+12
10+15
10+18
10+21
10+24
The prefixes centi (0.01), deci (0.1),

deka (10) and hecto (100) are commonly
applied to everyday units such as liter
and meter, but are generally
inappropriate for engineering use, with
the exception of decibel, dB.
Basic SI unit names and symbols

This table includes some important SI units and their derivations, and
the names of a few individuals whose names have been given to units.
The seven base SI units have blank in the derived from column; other
units are derived as indicated. A more complete list is found in the SI
brochure of the BIPM.
all lower case

unit symbols
leading capital letter

in unit symbol
unit
name
unit
symbol
derived
from
meter
length
kilogram
kg
mass
second
time
candela
cd
luminous intensity
mole
mol
quantity
named after
amount of substance
-3 .
liter
l, L
10
ohm
W . A -2
ampere
volume
resistance
Georg Simon Ohm
electric current
Henri Ampre
kelvin
thermodynamic
temperature
William Thomson
(Lord Kelvin)
hertz
Hz
frequency
Heinrich Hertz
force
Sir Isaac Newton
-1
-2
newton
joule
kg . m . s
N.m
energy
James Joule
watt
J . s -1
power
James Watt
volt
W . A -1
voltage
Alessandro Volta
Further information
Information is available at BIPM,

<http://www.bipm.fr/enus/3_SI/>.
Information is available at NIST,
<http://physics.nist.gov/cuu/Units/>.

Charles Poynton
This note describes issues of information design that you should understand if you wish to make web
pages that are a pleasure for your visitors to use.
To put pages up on the World Wide Web (www), you need to have access to a web server (perhaps
run by an Internet service provider). You need to use conversion or authoring tools to create the
HTML files of the web. Finally, you need to access the machinery to transfer the files and make them
accessible through your web server. Those tasks are complicated, but they are all more or less
mechanical. They are described well elsewhere, and I assume that you are familiar with them.
In order for the information that you provide to be useful and pleasurable for your web visitors, you
must write well, and if you use images, you must present them well. These are aesthetic issues, and I
cannot tell you too much about them here.
There is an intermediate area between the mechanical and the aesthetic that you must also consider in
making web pages: the area of information design. This domain is fairly well understood for the
production of books, posters, maps and so on, but is rather undeveloped in the electronic domain.
Information design for the web involves technical elements put into service to convey information.
The mechanics will get easier as HTML tools become available, but respecting a carefully-chosen set
of content conventions will remain important, and that work will never be done by machines.
Typographic-quality versions of this document are available in Acrobat PDF format. I provide
a version optimized for onscreen display (PDF format, 142646 bytes), and a version optimized
for printing on US-letter size paper (PDF format, 153498 bytes).
See also, This site is best experienced ...
See also, Ten Common Mistakes in the typesetting of technical documents.
Don't say, "Click here!"

The style guides deprecate this phrase, but it is so rampant on the web that it deserves special
mention: It is a mistake to make a link that is labelled click here.
http://www.poynton.com/Making_web_pages_usable.html (1 di 9) [15/01/2005 16:29:23]
There are several reasons to avoid this phrase. First, some web browsers do not have mice: they
activate links through means other than clicking. Second, web pages are often printed, or saved textonly! The notion of clicking on a phrase in its printed form is quite absurd; the absurdity reflects on
the author. Third, it is good information design and human interface design to put in a link an
indication of what is under the link. A sodapop machine doesn't have a button labelled Press here, it
has a button labelled 7-up!
Say, You can access a list of articles, or A list of articles is available. These phrases make perfect
sense, even when printed or displayed in text-only form without hypertext links.
Don't say, "Under Construction."

The web is so naturally dynamic that it seems to me redundant at best - and cute at worst - to draw
attention to this. So please, abolish those silly Under Construction icons. If it weren't under
construction, it wouldn't be the web.
Don't say, "Coming soon."

Your visitor will be very frustrated to access a link that says, Application Notes are available, only to
be presented with a content-free page that says Coming soon. Put the Coming soon or the not yet
available one level back, on the page with the link, so as to avoid wasting your visitor's time accessing
a useless page.
Strive for visual consistency

Give your pages a consistent design, so your visitor maintains a sense of continuity while at your site.
Apply typographic wisdom: if you center some elements on a page, center all the elements.
Don't put blue text on a black

background
Remember the ransom-note days of desktop publishing, when people thought that because they had
access to forty fonts, they had to use them all in a single document? Thankfully those days have
passed in DTP, but Netscape has implemented codes that allow an HTML document to specify the
colors it wants to be displayed in. Don't succumb to ransom-note color choice. If you are not
competent to choose colors, don't choose: let them default.
If you choose your own colors, choose a light background color and dark text: the larger the
differential between these two, the better the contrast ratio, and the more legible your page. I consider
it a mistake to use a black background, because color CRT displays have much higher contrast ratio
for black text on white than for white text on black. You must choose link colors carefully: many
people who set Netscape to use black background and white text neglect to change the link colors:
Their links display dark blue, which renders them virtually invisible against the black.
You should be aware that web technology cannot yet guarantee accurate color reproduction across
different platforms, so you have no guarantee of consistent color.
If you decide to use a background image, you should do so only with a good understanding of graphic
issues. A poorly-chosen background image can destroy the readability of your pages.
Restrain yourself from using <BLINK> just because Netscape implemented it. None of us will benefit
if we turn the web into a poor imitation of Las Vegas, and your visitor is unlikely to be impressed by a
page that is reminiscent of those TAKE A COUPON! blinking lights in sleazy supermarkets.
Compose to HTML, not to a particular

browser
Netscape browsers dominate the web at the moment, and it is tempting to compose HTML that looks
good when viewed with Netscape. You may find that Netscape's Heading 1 lines, in <H1> style, are
displayed too large. You may choose Heading 3 <H3> instead. I have fallen victim to this temptation,
but it's a bad idea.
HTML is designed to encapsulate the structure of a document, leaving the presentation to the browser.
If you tune a document to a particular browser, your page is almost certain to appear a mess to a
different browser. Even if your visitor is using the same browser that you use, if he has customized the
fonts and sizes in his browser, your document is likely to be poorly presented.
We can expect browser capability to improve, but it is unlikely that you will be inspired to go back
and retune your pages. If you stick to the standard HTML structure, your pages will look no worse
today than anyone else's, and they will look better and better as browsers improve. If you tune your
pages, today they will look better some of the time and worse some of the time, and they will age very
poorly as browsers improve.
If you have a document that begs to be presented typographically, consider distributing it in Acrobat
PDF format instead of - or in addition to - HTML. Acrobat Reader is freely available for the major
platforms: Windows, MS-DOS, Mac, and UNIX. Acrobat Exchange integrates well with the web.
Use 7-bit character set and "escaped

entities"
The assignment of glyphs - or shapes - to character codes between 0 and 127 is established by the ISO
646 standard, which is essentially the international version of ANSI X3.64 (ASCII). This standard
guarantees that 7-bit codes produce the same glyphs on different platforms.
The ISO 8859-1 Latin-1 standard conforms to ISO 646 for codes 0 to 127, but assigns additional
glyphs - mainly accented characters - to codes in the range 128-255. The Macintosh and Windows
operating systems do not respect the ISO 8859 standard, so codes in the range 128 to 255 produce
different glyphs when transported between these platforms. Most applications pay no special attention
to character sets, and inherit the character set native to the underlying operating system.
Web technology allows transport of 8-bit characters coded according to the ISO 8859-1. Web
browsers implement platform-dependent translation so that 8-bit characters received in a web page are
displayed correctly. Some browsers have an option setting to enable the translation; Netscape 1.1 for
the Mac comes with a setting that is not ISO Latin-1. Set your browser to conform to the standards of
the web: Set its character set to ISO Latin-1.
Few text editors implement the ISO 8859-1 character set directly, so creation of web pages using
characters in the range 128-255 is difficult. If you create a web page using a text editor that allows
insertion of codes in the range 128-255, you have two options: You must either take care to avoid or
remove characters in that range, or you must arrange to have those characters translated.
If you remove characters in the range 128-255 by stripping the eighth bit, the result is guaranteed to
comprise just 7-bit ASCII characters. But in stripping the eighth bit, you may inadvertently turn
characters into ASCII codes that you don't intend. On a Macintosh, if your document uses a bullet
character (Option-8), it will turn into a percent sign. It is a better idea to translate, and many utilities
are available to translate from a platform's native character code to ISO 8859-1.
Although eight-bit characters are handled well by the web browsers, transport of 8-bit characters by
other means - e-mail, ftp and physical media - remains problematic. In HTML there is provision to
convey accented characters and other characters of ISO 8859-1 using escaped entities that comprise
an ampersand, a short sequence of 7-bit ASCII letters, and a terminating semicolon. I recommend that
instead of translating to 8-bit ISO 8859-1 you translate to 7-bit ASCII with the escaped entities. This
will assure that your pages are transported easily and displayed correctly on any conformant browser.
A few important characters are not accommodated by ISO 8859-1. The most glaring omission is
typographic (curly) quotes. Your translator will turn these into straight quotes. The trademark sign
((TM)) is absent from ISO 8859-1. Provision has been made in HTML 3.0 for an escaped entity
™ but most browsers in use today do not conform to HTML 3.0 and would display ™
instead of the symbol that you want. Write that one out, (tm).
A handful of escaped entities are not handled properly by Macintosh browsers: avoid the superior
figures, fractions, y-acute, thorn, eth, and the so-called times symbol. If you don't know what these
are, you're probably not using them!
Establish a context for the reader (and

for the robots)
Make sure that the first several lines of text on your page describe the content of that page.
You can include in a web page a link to any other page on the web; part of the power of the web lies
in jumping from site to site. But the flip side of this situation is that your page may be accessed from
places different from what you anticipate. By providing a short outline of the content of your page,
you establish the context for a visitor who has come to your page from somewhere else.
There is another reason for the description to be short, and to be located at the top of the page: Many
automatic programs - the crawlers, wanderers, robots, harvesters and spiders - traverse the web,
extracting and indexing pages. Many of these programs index all of the words in a page, but save only
the first several lines for display in a search result. In order for the user of a search service to
recognize your page as useful when it is returned as a search result, you need a useful description in
the first few lines.
You will find many web pages that have adopted cutesy elements like spaces between the letters of
the page title. People do this in an attempt to create a distinctive look, and sometimes it succeeds in
attracting the viewer's attention. On the other hand, it defeat the robots' attempts to index the page. If
potential readers never access the page, what good is a distinctive style?
Provide a page title

Include a title - the <TITLE> element - on every page. Limit your title to about 40 characters, to avoid
overflowing the your visitor's screen width. Help your visitor to navigate by making the structure of
your titles consistent among your pages. The search engines usually display the page title along with a
search result. If your page has no title it is displayed alongside a message like No Title Provided,
which makes you look unprofessional.
Plan for an international audience

The first w in WWW is for world. Expect the audience for your web pages to be international.
If you write a date in the form 08/04/50, will your visitor think it April or August? In the next century,
will 01/02/03 be the first, second or third day of the month? Banish this confusion for once and for all
by writing dates in the ISO/IEC 8824 form, 1995-10-12.
To respect my international colleagues, in front of any telephone number I place a plus sign and the
country code: +81 for Japan, +44 for UK, +1 for Canada, +1 for the U.S.A. I delimit the area code (or
in other parts of the world, city code) using spaces instead of parentheses: parentheses are not
particularly computer-friendly, and many people handle telephone numbers using computers. In
Europe, do not indicate 0 in front of a city code: people who need it know to dial it, but if a person
unfamiliar with the convention dials the zero, his call will fail.
Keep web pages small

Many people use the web through dialup modem connections capable of transfer rates of only 1000 or
2000 characters per second. If you link to an exceptionally large page, larger than 50 KB or so, you
should provide at the point of a link an indication of the size of the referenced object. Your home
page, including its images, should be no larger than this.
If a link accesses an ftp file, then provide at the point of the link an indication of the format of the file
and the size of the file (no matter how small). This indicates to the reader that accessing the link will
transfer the file. Avoid notations like download here and download now, for the same reasons that you
avoid click here.
If you link to an ftp directory, as opposed to a file, include a trailing slash at the end of the URL. This
indicates to your visitor (and to his web browser or ftp client) that the item is a directory.
Code images correctly

Include WIDTH and HEIGHT information in image (IMG) links. This allows a browser to complete
page layout before accessing the image, and avoids flashing due to re-layout. Choose the WIDTH and
HEIGHT of the actual image file; do not arbitrarily choose them expecting the browser to scale the
image, because not all browsers have that capability, and in any case a scaled bitmap reproduces
poorly.
If your image forms part of a link, include an ALT tag describing the image in words. You will be
thanked by visitors without image display capability, and by visitors who have disabled image display
(perhaps for reasons of speed).
If you have an inline image, make it small (10 KB or less), and save it in GIF format (until PNG
format is widespread). If you want to provide for your visitor an image larger than that, make a small
GIF version of it - a proxy - and place the proxy on your page. Make the proxy a link to the large
image. If the large image is full color or continuous-tone, save it in JPEG/JFIF format.
You can process a GIF bitmapped image so as to make some of its pixels transparent. The opaque
pixels will then be displayed against the background color that was chosen by a preference set in your
visitor's browser. If your visitor has a modern browser and you have specified the appropriate codes in
your HTML, it will display against a background that you have chosen. If you choose to specify
transparency, be aware that the less-sophisticated browsers will display your image entirely opaque.
Choose a background color appropriate for those browsers, say [192, 192, 192] for a light gray. If you
use a custom background color or image, be aware that it will be ignored by less-sophisticated
browsers.
Sign your work

It is frustrating to find a page on the web whose authorship is unknown, especially when there are no
other links on the page to establish where it lives or what it relates to. Sign your pages.
If a user comes to a page from a foreign link, give him the opportunity to explore your home page or
the rest of your pages: make your signature a link, direct or indirect, to your home page.
At the bottom of every page, my signature is a link up within my tree of pages. For a page other than
the index.html file in a directory, I place a signature that names the directory and a link to index.html
in that directory. At the bottom of each index.html file I refer to the title of the next level up, and place
a link to ../index.html. This enables my visitor to ascend the whole tree back to my home.
At the bottom of my home page, my signature is a MAILTO link. If my visitor hasn't discovered the
information he wants in his traversal of my pages, this invites him to send e-mail to me.
I include at the bottom of every page the date that I last modified the page.
Conform to server conventions

Learn from your Internet service provider how to make your files accessible to his web server.
If you use a UNIX server, include the lowercase L at the end of the .html extension when you transfer,
even if your local filenames are limited by MS-DOS or Windows. Use UNIX (LF) line ends in text
files (including HTML) stored at a UNIX server.
Make sure every directory has a file index.html. If you do not do this, then a visitor who manually
enters the path to a directory will be presented with a list of all of the files in that directory, perhaps
including some files that you do not want to advertise.
Your HTML pages include whatever file names and paths you need for your links. The robots and
wanderers will harvest filenames from your HTML code, and add these referenced files to their
indices. If you want a file to be indexed, you should include its name in another file that is indexed
already: The robots will eventually find your new page!
You can place in your web directory a file whose name is not referenced in any of your pages. The
robots will not discover this file. But if a visitor guesses a name, index.bak or index.old for example,
there is no method to prevent the visitor from retrieving that file. The only way to be absolutely
certain that a visitor will not have access to a file is to remove that file from your web directory.
Maintain the hierarchy

My home page is accessible at the URL <http://www.poynton.com/>. If I wish to direct someone to a
page other than my home page, say by e-mail, I specify the full URL of that page: my page of
Macintosh information is located at <http://www.poynton.com/Poynton-mac.html>. However, within
my home page, I use a relative pathname such as Poynton-mac.html. Using a relative pathname makes
it easier for you to maintain pages and links, and makes it easier for your visitor to make local copies
of your pages while maintaining the function of the links.
If you have created a hierarchy of pages, the easiest way to manually create a new page is to copy,
then edit, a page at the same level of the hierarchy.
Choose filenames that are mnemonic. When a visitor decides to save one of your pages, the name you
choose will be presented as his default name.
Once you've chosen the name of a file (or page or directory), stick to it. Other sites may have made
links to your page (or directory). If you change a name, you will break those links.
Test locally
Your pages will be no pleasure for your visitor if they do not work as you yourself intend. Make sure
that your pages work for you before you subject someone else to them!
Test your pages locally, use the Open File capability of your favourite browser. Use two or three
different browsers, to see how they present things differently. Test your pages in black-and-white, to
preview how they will appear to a user who has only black-and-white display capability.
Validate
When you have finished making a page, that you run it through an HTML validation service to ensure
that it conforms to the technical requirements of HTML. If you do not do this, you cannot be sure that
it will work reliably on other browsers and other platforms than yours.
If you have manually created your HTML, you can fix it by hand. If you have used automated
conversion tools, you may have little scope to repair failures in validation. In this case, take the
validation report to the provider of your conversion tools.
Copyright 1997-09-01 (c)

Charles Poynton

1996-06-17
Charles A. Poynton
tel +1 416 486 3271
fax +1 416 486 3657
www.poynton.com/~poynton
This note describes issues of information design that

you should understand if you wish to make web
pages that are a pleasure for your visitors to use.
To put pages up on the World Wide Web (www),
you need to have access to a web server (perhaps
run by an Internet service provider). You need to use
conversion or authoring tools to create the HTML
files of the web. Finally, you need to access the
machinery to transfer the files and make them accessible through your web server. Those tasks are
complicated, but they are all more or less mechanical. They are described well elsewhere, and
I assume that you are familiar with them.
In order for the information that you provide to be
useful and pleasurable for your web visitors, you
must write well, and if you use images, you must
present them well. These are aesthetic issues, and
I cannot tell you too much about them here.
There is an intermediate area between the mechanical and the aesthetic that you must also consider in
making web pages: the area of information design.
This domain is fairly well understood for the production of books, posters, maps and so on, but is rather
undeveloped in the electronic domain. Information
1 of 14
design for the web involves technical elements put

into service to convey information. The mechanics
will get easier as HTML tools become available, but
respecting a carefully-chosen set of content conventions will remain important, and that work will never
be done by machines.
Dont say, Click here!
The style guides deprecate this phrase, but it is so
rampant on the web that it deserves special
mention: It is a mistake to make a link that is
labelled click here.
There are several reasons to avoid this phrase. First,
some web browsers do not have mice: they activate
links through means other than clicking. Second,
web pages are often printed, or saved text-only!
The notion of clicking on a phrase in its printed form
is quite absurd; the absurdity reflects on the author.
Third, it is good information design and human interface design to put in a link an indication of what is
under the link. A sodapop machine doesnt have a
button labelled Press here, it has a button labelled
7-up!
Say, You can access a list of articles, or A list of articles is available. These phrases make perfect sense,
even when printed or displayed in text-only form
without hypertext links.
Dont say, Under Construction.
The web is so naturally dynamic that it seems to me
redundant at best, and cute at worst, to draw attention to this. So please, abolish those silly Under
Construction icons. If it werent under construction,
it wouldnt be the web.
MAKING WEB PAGES USABLE
Dont say, Coming soon.

Your visitor will be very frustrated to access a link
that says, Application Notes are available, only to
be presented with a content-free page that says
Coming soon. Put the Coming soon or the not yet
available one level back, on the page with the link,
so as to avoid wasting your visitors time accessing a
useless page.
Give your pages a consistent design, so your visitor
maintains a sense of continuity while at your site.
Apply typographic wisdom: if you center some
elements on a page, center all the elements.
Dont put blue text on a black background
Remember the ransom-note days of desktop
publishing, when people thought that because they
had access to forty fonts, they had to use all of
them in a single document? Thankfully those days
have passed in DTP, but Netscape has implemented
codes that allow an HTML document to specify the
colors it wants to be displayed in. Dont succumb to
ransom-note color choice. If you are not competent
to choose colors, dont choose: let them default.
If you choose your own colors, choose a light background color and dark text: the larger the differential between these two, the better the contrast ratio,
and the more legible your page. I consider it a
mistake to use a black background, because color
CRT displays have much higher contrast ratio for
black text on white than for white text on black.
You must choose link colors carefully: many people
who set Netscape to use black background and
white text neglect to change the link colors. Their
links display dark blue, which renders them virtually
invisible against the black.
You should be aware that web technology cannot

yet guarantee accurate color reproduction across
different platforms, so you have no guarantee of
consistent color.
If you decide to use a background image, you
should do so only with a good understanding of
graphic issues. A poorly-chosen background image
can destroy the readability of your pages.
Restrain yourself from using <BLINK>. None of us
will benefit if we turn the web into a poor imitation
of Las Vegas, and your visitor is unlikely to be
impressed by a page that is reminiscent of those
TAKE A COUPON! blinking lights in sleazy supermarkets.
Compose to HTML, not to a particular browser
Netscape browsers dominate the web at the
moment, and it is tempting to compose HTML that
looks good when viewed with Netscape. You may
find that Netscapes Heading 1 lines, in <H1> style,
are displayed too large. You may choose Heading 3
<H3> instead. I have fallen victim to this temptation, but its a bad idea.
HTML is designed to encapsulate the structure of a
document, leaving the presentation to the browser.
If you tune a document to a particular browser, your
page is almost certain to appear a mess to a
different browser. Even if your visitor is using the
same browser that you use, if he has customized the
fonts and sizes in his browser, your document is
likely to be poorly presented.
We can expect browser capability to improve, but it
is unlikely that you will be inspired to go back and
retune your pages. If you stick to the standard
4
HTML structure, your pages will look no worse

today than anyone elses, and they will look better
and better as browsers improve. If you tune your
pages, today they will look better some of the time
and worse some of the time, and they will age very
poorly as browsers improve.
If you have a document that begs to be presented
typographically, consider distributing it in Acrobat
PDF format instead of or in addition to HTML.
Acrobat Reader is freely available for the major platforms: Windows, MS-DOS, Mac, and UNIX.
Acrobat Exchange integrates well with the web.
Use 7-bit character set and escaped entities
The assignment of glyphs or shapes to character
codes between 0 and 127 is established by the
ISO 646 standard, which is essentially the international version of ANSI X3.64 (ASCII ). This standard
guarantees that 7-bit codes produce the same
glyphs on different platforms.
The ISO 8859-1 Latin-1 standard conforms to
ISO 646 for codes 0 to 127, but assigns additional
glyphs mainly accented characters to codes in
the range 128-255. The Macintosh and Windows
operating systems do not respect the ISO 8859 standard, so codes in the range 128 to 255 produce
different glyphs when transported between these
platforms. Most applications pay no special attention to character sets, and inherit the character set
native to the underlying operating system.
Web technology allows transport of 8-bit characters
coded according to the ISO 8859-1. Web browsers
implement platform-dependent translation so that
8-bit characters received in a web page are
displayed correctly. Some browsers have an option
setting to enable the translation; Netscape 1.1 for

the Mac comes with a setting that is not ISO
Latin-1. Set your browser to conform to the standards of the web: Set its character set to ISO Latin-1.
Few text editors implement the ISO 8859-1 character set directly, so creation of web pages using
characters in the range 128-255 is difficult. If you
make a web page using a text editor that allows
insertion of codes in the range 128-255, you have
two options: You must either take care to avoid or
remove characters in that range, or you must
arrange to have those characters translated.
If you remove characters in the range 128-255 by
stripping the eighth bit, the result is guaranteed to
comprise just 7-bit ASCII characters. But in stripping
the eighth bit, you may inadvertently turn characters into ASCII codes that you dont intend. On a
Macintosh, if your document uses a bullet character
(, Option-8), it will turn into a percent sign. It is a
better idea to translate, and many utilities are available to translate from a platforms native character
code to ISO 8859-1.
Although 8-bit characters are handled well by the
web browsers, transport of 8-bit characters by other
means e-mail, ftp and physical media remains
problematic. In HTML there is provision to convey
accented characters and other characters of
ISO 8859-1 using escaped entities that comprise an
ampersand, a short sequence of 7-bit ASCII letters,
and a terminating semicolon. I recommend that
instead of translating to 8-bit ISO 8859-1 you translate to 7-bit ASCII with the escaped entities. This
will assure that your pages are transported easily
and displayed correctly on any conformant browser.
A few important characters are not accommodated

by ISO 8859-1. The most glaring omission is typographic (curly ) quotes. Your translator will turn
these into straight quotes. The trademark sign () is
absent from ISO 8859-1. Provision has been made
in HTML 3.0 for an escaped entity ™ but most
browsers in use today do not conform to HTML 3.0
and would display ™ instead of the symbol
that you want. Write that one out, (tm).
A handful of escaped entities are not handled properly by Macintosh browsers: avoid the superior
figures, fractions, y-acute, thorn, eth, and the socalled times symbol. If you dont know what these
are, youre probably not using them!
Establish a context for the reader (and for the robots)
Make sure that the first several lines of text on your
page describe the content of that page.
You can include in a web page a link to any other
page on the web; part of the power of the web lies
in jumping from site to site. But the flip side of this
situation is that your page may be accessed from
places different from what you anticipate. By
providing a short outline of the content of your
page, you establish the context for a visitor who has
come to your page from somewhere else.
There is another reason for the description to be
short, and to be located at the top of the page:
Many automatic programs the crawlers,
wanderers, robots, harvesters and spiders traverse
the web, extracting and indexing pages. Many of
these programs index all of the words in a page, but
save only the first several lines for display in a search
result. In order for the user of a search service to
recognize your page as useful when it is returned as
a search result, you need a useful description in the

first few lines.
You will find many web pages that have adopted
cutesy elements like spaces between the letters of
the page title. People do this in an attempt to create
a distinctive look, and sometimes it succeeds in
attracting the viewers attention. On the other hand,
it defeat the robots attempts to index the page. If
potential readers never access the page, what good
is a distinctive style?
Include a title the <TITLE> element on every
page. Limit your title to about 40 characters, to
avoid overflowing the your visitors screen width.
Help your visitor to navigate by making the structure of your titles consistent among your pages. The
search engines usually display the page title along
with a search result. If your page has no title, it will
be displayed alongside a message like No Title
Provided, which makes you look unprofessional.
The first w in WWW is for world. Expect the audience for your web pages to be international.
If you write a date in the form 08/04/50, will your
visitor think it April or August? In the next century,
will 01/02/03 be the first, second or third day of the
month? Banish this confusion for once and for all by
writing dates in the ISO/IEC 8824 form, 1996-06-17.
To respect my international colleagues, in front of
any telephone number I place a plus sign and the
country code, for example, +81 for Japan or +44 for
the UK. The country code +1 indicates the United
States, Canada, Mexico, and parts of the Caribbean.
8
Delimit the area code or, in other parts of the

world, the city code using spaces instead of parentheses. Parentheses are not particularly computerfriendly, and many people handle telephone
numbers using computers.
Write out only digits that are properly part of the
telephone number: Omit local access codes. In
particular, do not indicate the access code 0 in front
of a city code used in Europe: people who need it
know to dial it, but if a person unfamiliar with the
convention dials the zero, his call will fail. In some
countries, such as Russia, the digit zero can be a
legitimate part of a city code, and must be dialed.
Many people use the web through dialup modem
connections capable of transfer rates of only 1000
or 2000 characters per second. If you link to an
exceptionally large page, larger than 50 KB or so,
you should provide at the point of a link an indication of the size of the referenced object. Your home
page, including its images, should be no larger than
this.
If a link accesses an ftp file, then provide at the
point of the link an indication of the format of the
file and the size of the file (no matter how small).
This indicates to the reader that accessing the link
will transfer the file. Avoid notations like download
here and download now, for the same reasons that
you avoid click here.
If you link to an ftp directory, as opposed to a file,
include a trailing slash at the end of the URL. This
indicates to your visitor and to his web browser or
ftp client that the item is a directory.

Include WIDTH and HEIGHT information in image
(IMG) links. This allows a browser to complete page
layout before accessing the image, and avoids
flashing due to re-layout. Choose the WIDTH and
HEIGHT of the actual image file; do not arbitrarily
choose them expecting the browser to scale the
image, because not all browsers have that capability,
and in any case a scaled bitmap reproduces poorly.
If your image forms part of a link, include an ALT
tag describing the image in words. You will be
thanked by visitors without image display capability,
and by visitors who have disabled image display
If you have an inline image, make it small (10 KB or
less), and save it in GIF format (until PNG format is
widespread). If you want to provide for your visitor
an image larger than that, make a small GIF version
of it a proxy and place the proxy on your page.
Make the proxy a link to the large image. If the
large image is full color or continuous-tone, save it
in JPEG/JFIF format.
You can process a GIF bitmapped image so as to
make some of its pixels transparent. The opaque
pixels will then be displayed against the background
color that was chosen by a preference set in your
visitors browser. If your visitor has a modern
browser and you have specified the appropriate
codes in your HTML, it will display against a background that you have chosen. If you choose to
specify transparency, be aware that the less-sophisticated browsers will display your image entirely
opaque. Choose a background color appropriate for
those browsers, say [192, 192, 192] for a light gray.
If you use a custom background color or image, be
10
aware that it will be ignored by less-sophisticated

browsers.
Sign your work
It is frustrating to find a page on the web whose
authorship is unknown, especially when there are no
other links on the page to establish where it lives or
what it relates to. Sign your pages.
If a user comes to a page from a foreign link, give
him the opportunity to explore your home page or
the rest of your pages: make your signature a link,
direct or indirect, to your home page.
At the bottom of every page, my signature is a link
up within my tree of pages. For a page other than
the index.html file in a directory, I place a signature
that names the directory and a link to index.html in
that directory. At the bottom of each index.html file
I refer to the title of the next level up, and place a
link to ../index.html. This enables my visitor to
ascend the whole tree back to my home.
At the bottom of my home page, my signature is a
MAILTO link. If my visitor hasnt discovered the
information he wants in his traversal of my pages,
this invites him to send e-mail to me.
I include at the bottom of every page the date that
I last modified the page.
Learn from your Internet service provider how to
make your files accessible to his web server.
If you use a UNIX server, include the lowercase L at
the end of the . html extension when you transfer,
even if your local filenames are limited by MS-DOS
11
or Windows. Use UNIX (LF) line ends in text files

(including HTML) stored at a UNIX server.
Make sure every directory has a file index.html. If
you do not do this, then a visitor who manually
enters the path to a directory will be presented with
a list of all of the files in that directory, perhaps
Your HTML pages include whatever file names and
paths you need for your links. The robots and
wanderers will harvest filenames from your HTML
code, and add these referenced files to their indices.
If you want a file to be indexed, you should include
its name in another file that is indexed already: The
robots will eventually find your new page!
You can place in your web directory a file whose
name is not referenced in any of your pages. The
robots will not discover this file. But if a visitor
guesses a name, index.bak or index.old for
example, there is no method to prevent the visitor
from retrieving that file. The only way to be absolutely certain that a visitor will not have access to a
file is to remove that file from your web directory.
My home page is accessible at the URL
<http://www.inforamp.net/~poynton/>. If I wish to
direct someone to a page other than my home
page, say by e-mail, I specify the full URL of that
page: my page of Macintosh information is located
at <http://www.inforamp.net/~poynton/Poyntonmac.html>. However, within my home page, I use a
relative pathname such as Poynton-mac.html. Using
a relative pathname makes it easier for you to maintain pages and links, and makes it easier for your
12
visitor to make local copies of your pages while

maintaining the function of the links.
If you have created a hierarchy of pages, the easiest
way to manually create a new page is to copy, then
edit, a page at the same level of the hierarchy.
Choose filenames that are mnemonic. When a
visitor decides to save one of your pages, the name
you choose will be presented as his default name.
Once youve chosen the name of a file (or page or
directory), stick to it. Other sites may have made
links to your page (or directory). If you change a
name, you will break those links.
Test locally
Your pages will be no pleasure for your visitor if they
do not work as you yourself intend. Make sure that
your pages work for you before you subject
someone else to them!
Test your pages locally, use the Open File capability
of your favourite browser. Use two or three different
browsers, to see how they present things differently.
Test your pages in black-and-white, to preview how
they will appear to a user who has only black-andwhite display capability.
Validate
When you have finished making a page, that you
run it through an HTML validation service to ensure
that it conforms to the technical requirements of
HTML. If you do not do this, you cannot be sure
that it will work reliably on other browsers and other
platforms than yours.
13
If you have manually created your HTML, you can

fix it by hand. If you have used automated conversion tools, you may have little scope to repair failures in validation. In this case, take the validation
report to the provider of your conversion tools.
Provide alternate access paths
The online services such as AOL and CIS are slow to
adapt to new technology, and millions of their
subscribers have no access to the web. If you want
to make information available to online service
subscribers you must use file transfer protocol (ftp).
Modern web browsers accommodate ftp access, so
your files will be available not only to those
subscribers, and to ftp users on the Internet, but
also to web users on the Internet. You will need
your Internet access providers help to set up an ftp
directory at his site.
My technique is to store the bulk of my information
in my ftp directory, and to use the web as an attractive, functional and interactive interface to that information. All of the common web browsers include
ftp capability as a subset, so your web visitor will see
no break in continuity but your information is stored
only once, and CIS and AOL visitors have full
access. My web pages simply contain links that use
ftp protocol to access directories and files. You can
view the source (HTML) of my pages to see how
these links work.
When I create new information to publish, I store
the files themselves in my ftp directories. Then
I create a web page with the descriptions and links.
I open this page in my web browser and Save As
ASCII text-only format. I use that text file as the
README file for the ftp directory.
14

1996-06-17
Charles A. Poynton
tel +1 416 486 3271
fax +1 416 486 3657
www.poynton.com/~poynton
This note describes issues of information design that you should

understand if you wish to make web pages that are a pleasure for
your visitors to use.
To put pages up on the World Wide Web (www), you need to have
access to a web server (perhaps run by an Internet service
provider). You need to use conversion or authoring tools to create
the HTML files of the web. Finally, you need to access the
machinery to transfer the files and make them accessible through
your web server. Those tasks are complicated, but they are all more
or less mechanical. They are described well elsewhere, and I assume
that you are familiar with them.
In order for the information that you provide to be useful and pleasurable for your web visitors, you must write well, and if you use
images, you must present them well. These are aesthetic issues, and
I cannot tell you too much about them here.
There is an intermediate area between the mechanical and the
aesthetic that you must also consider in making web pages: the
area of information design. This domain is fairly well understood for
the production of books, posters, maps and so on, but is rather
undeveloped in the electronic domain. Information design for the
web involves technical elements put into service to convey information. The mechanics will get easier as HTML tools become available,
but respecting a carefully-chosen set of content conventions will
remain important, and that work will never be done by machines.
Dont say, Click here!

The style guides deprecate this phrase, but it is so rampant on the
web that it deserves special mention: It is a mistake to make a link
that is labelled click here.
1 of 10
There are several reasons to avoid this phrase. First, some web
browsers do not have mice: they activate links through means other
than clicking. Second, web pages are often printed, or saved textonly! The notion of clicking on a phrase in its printed form is quite
absurd; the absurdity reflects on the author. Third, it is good information design and human interface design to put in a link an indication of what is under the link. A sodapop machine doesnt have a
button labelled Press here, it has a button labelled 7-up!
Say, You can access a list of articles, or A list of articles is available.
These phrases make perfect sense, even when printed or displayed
in text-only form without hypertext links.
Dont say, Under Construction.
The web is so naturally dynamic that it seems to me redundant at
best, and cute at worst, to draw attention to this. So please, abolish
those silly Under Construction icons. If it werent under construction, it wouldnt be the web.
Dont say, Coming soon.
Your visitor will be very frustrated to access a link that says, Application Notes are available, only to be presented with a content-free
page that says Coming soon. Put the Coming soon or the not yet
available one level back, on the page with the link, so as to avoid
wasting your visitors time accessing a useless page.
Give your pages a consistent design, so your visitor maintains a
sense of continuity while at your site. Apply typographic wisdom: if
you center some elements on a page, center all the elements.
Dont put blue text on a black background
Remember the ransom-note days of desktop publishing, when
people thought that because they had access to forty fonts, they
had to use all of them in a single document? Thankfully those days
have passed in DTP, but Netscape has implemented codes that
allow an HTML document to specify the colors it wants to be
displayed in. Dont succumb to ransom-note color choice. If you are
not competent to choose colors, dont choose: let them default.
If you choose your own colors, choose a light background color and
dark text: the larger the differential between these two, the better
the contrast ratio, and the more legible your page. I consider it a
mistake to use a black background, because color CRT displays
have much higher contrast ratio for black text on white than for
white text on black. You must choose link colors carefully: many
people who set Netscape to use black background and white text
neglect to change the link colors. Their links display dark blue,
which renders them virtually invisible against the black.
You should be aware that web technology cannot yet guarantee
accurate color reproduction across different platforms, so you have
no guarantee of consistent color.
If you decide to use a background image, you should do so only
with a good understanding of graphic issues. A poorly-chosen background image can destroy the readability of your pages.
Restrain yourself from using <BLINK>. None of us will benefit if we
turn the web into a poor imitation of Las Vegas, and your visitor is
unlikely to be impressed by a page that is reminiscent of those
TAKE A COUPON! blinking lights in sleazy supermarkets.
Compose to HTML, not to a particular browser
Netscape browsers dominate the web at the moment, and it is
tempting to compose HTML that looks good when viewed with
Netscape. You may find that Netscapes Heading 1 lines, in <H1>
style, are displayed too large. You may choose Heading 3 <H3>
instead. I have fallen victim to this temptation, but its a bad idea.
HTML is designed to encapsulate the structure of a document,
leaving the presentation to the browser. If you tune a document to
a particular browser, your page is almost certain to appear a mess
to a different browser. Even if your visitor is using the same browser
that you use, if he has customized the fonts and sizes in his
browser, your document is likely to be poorly presented.
We can expect browser capability to improve, but it is unlikely that
you will be inspired to go back and retune your pages. If you stick
to the standard HTML structure, your pages will look no worse
today than anyone elses, and they will look better and better as
browsers improve. If you tune your pages, today they will look
better some of the time and worse some of the time, and they will
age very poorly as browsers improve.
If you have a document that begs to be presented typographically,
consider distributing it in Acrobat PDF format instead of or in
addition to HTML. Acrobat Reader is freely available for the major
platforms: Windows, MS-DOS, Mac, and UNIX. Acrobat Exchange
integrates well with the web.
Use 7-bit character set and escaped entities

The assignment of glyphs or shapes to character codes between
0 and 127 is established by the ISO 646 standard, which is essentially the international version of ANSI X3.64 (ASCII ). This standard
guarantees that 7-bit codes produce the same glyphs on different
platforms.
The ISO 8859-1 Latin-1 standard conforms to ISO 646 for codes 0
to 127, but assigns additional glyphs mainly accented characters
to codes in the range 128-255. The Macintosh and Windows operating systems do not respect the ISO 8859 standard, so codes in
the range 128 to 255 produce different glyphs when transported
between these platforms. Most applications pay no special attention to character sets, and inherit the character set native to the
underlying operating system.
Web technology allows transport of 8-bit characters coded
according to the ISO 8859-1. Web browsers implement platformdependent translation so that 8-bit characters received in a web
page are displayed correctly. Some browsers have an option setting
to enable the translation; Netscape 1.1 for the Mac comes with a
setting that is not ISO Latin-1. Set your browser to conform to the
standards of the web: Set its character set to ISO Latin-1.
Few text editors implement the ISO 8859-1 character set directly,
so creation of web pages using characters in the range 128-255 is
difficult. If you make a web page using a text editor that allows
insertion of codes in the range 128-255, you have two options: You
must either take care to avoid or remove characters in that range,
or you must arrange to have those characters translated.
If you remove characters in the range 128-255 by stripping the
eighth bit, the result is guaranteed to comprise just 7-bit ASCII characters. But in stripping the eighth bit, you may inadvertently turn
characters into ASCII codes that you dont intend. On a Macintosh,
if your document uses a bullet character (, Option-8), it will turn
into a percent sign. It is a better idea to translate, and many utilities
are available to translate from a platforms native character code to
ISO 8859-1.
Although 8-bit characters are handled well by the web browsers,
transport of 8-bit characters by other means e-mail, ftp and physical media remains problematic. In HTML there is provision to
convey accented characters and other characters of ISO 8859-1
using escaped entities that comprise an ampersand, a short
sequence of 7-bit ASCII letters, and a terminating semicolon.
I recommend that instead of translating to 8-bit ISO 8859-1 you

translate to 7-bit ASCII with the escaped entities. This will assure
that your pages are transported easily and displayed correctly on
any conformant browser.
A few important characters are not accommodated by ISO 8859-1.
The most glaring omission is typographic (curly ) quotes. Your translator will turn these into straight quotes. The trademark sign () is
absent from ISO 8859-1. Provision has been made in HTML 3.0 for
an escaped entity ™ but most browsers in use today do not
conform to HTML 3.0 and would display ™ instead of the
symbol that you want. Write that one out, (tm).
A handful of escaped entities are not handled properly by Macintosh browsers: avoid the superior figures, fractions, y-acute, thorn,
eth, and the so-called times symbol. If you dont know what these
are, youre probably not using them!
Establish a context for the reader (and for the robots)
Make sure that the first several lines of text on your page describe
the content of that page.
You can include in a web page a link to any other page on the web;
part of the power of the web lies in jumping from site to site. But
the flip side of this situation is that your page may be accessed from
places different from what you anticipate. By providing a short
outline of the content of your page, you establish the context for a
visitor who has come to your page from somewhere else.
There is another reason for the description to be short, and to be
located at the top of the page: Many automatic programs the
crawlers, wanderers, robots, harvesters and spiders traverse the
web, extracting and indexing pages. Many of these programs index
all of the words in a page, but save only the first several lines for
display in a search result. In order for the user of a search service to
recognize your page as useful when it is returned as a search result,
you need a useful description in the first few lines.
You will find many web pages that have adopted cutesy elements
like spaces between the letters of the page title. People do this in
an attempt to create a distinctive look, and sometimes it succeeds
in attracting the viewers attention. On the other hand, it defeat the
robots attempts to index the page. If potential readers never access
the page, what good is a distinctive style?

Include a title the <TITLE> element on every page. Limit your
title to about 40 characters, to avoid overflowing the your visitors
screen width. Help your visitor to navigate by making the structure
of your titles consistent among your pages. The search engines
usually display the page title along with a search result. If your page
has no title, it will be displayed alongside a message like No Title
Provided, which makes you look unprofessional.
The first w in WWW is for world. Expect the audience for your web
pages to be international.
If you write a date in the form 08/04/50, will your visitor think it
April or August? In the next century, will 01/02/03 be the first,
second or third day of the month? Banish this confusion for once
and for all by writing dates in the ISO/IEC 8824 form, 1996-06-17.
To respect my international colleagues, in front of any telephone
number I place a plus sign and the country code, for example, +81
for Japan or +44 for the UK. The country code +1 indicates the
United States, Canada, Mexico, and parts of the Caribbean.
Delimit the area code or, in other parts of the world, the city
code using spaces instead of parentheses. Parentheses are not
particularly computer-friendly, and many people handle telephone
numbers using computers.
Write out only digits that are properly part of the telephone
number: Omit local access codes. In particular, do not indicate the
access code 0 in front of a city code used in Europe: people who
need it know to dial it, but if a person unfamiliar with the convention dials the zero, his call will fail. In some countries, such as
Russia, the digit zero can be a legitimate part of a city code, and
must be dialed.
Many people use the web through dialup modem connections
capable of transfer rates of only 1000 or 2000 characters per
second. If you link to an exceptionally large page, larger than 50 KB
or so, you should provide at the point of a link an indication of the
size of the referenced object. Your home page, including its images,
should be no larger than this.
If a link accesses an ftp file, then provide at the point of the link an
indication of the format of the file and the size of the file (no
matter how small). This indicates to the reader that accessing the
link will transfer the file. Avoid notations like download here and
download now, for the same reasons that you avoid click here.
If you link to an ftp directory, as opposed to a file, include a trailing
slash at the end of the URL. This indicates to your visitor and to
his web browser or ftp client that the item is a directory.
Include WIDTH and HEIGHT information in image (IMG) links. This
allows a browser to complete page layout before accessing the
image, and avoids flashing due to re-layout. Choose the WIDTH
and HEIGHT of the actual image file; do not arbitrarily choose them
expecting the browser to scale the image, because not all browsers
have that capability, and in any case a scaled bitmap reproduces
poorly.
If your image forms part of a link, include an ALT tag describing the
image in words. You will be thanked by visitors without image
display capability, and by visitors who have disabled image display
If you have an inline image, make it small (10 KB or less), and save
it in GIF format (until PNG format is widespread). If you want to
provide for your visitor an image larger than that, make a small GIF
version of it a proxy and place the proxy on your page. Make
the proxy a link to the large image. If the large image is full color or
continuous-tone, save it in JPEG/JFIF format.
You can process a GIF bitmapped image so as to make some of its
pixels transparent. The opaque pixels will then be displayed against
the background color that was chosen by a preference set in your
visitors browser. If your visitor has a modern browser and you have
specified the appropriate codes in your HTML, it will display against
a background that you have chosen. If you choose to specify transparency, be aware that the less-sophisticated browsers will display
your image entirely opaque. Choose a background color appropriate for those browsers, say [192, 192, 192] for a light gray. If
you use a custom background color or image, be aware that it will
be ignored by less-sophisticated browsers.
Sign your work
It is frustrating to find a page on the web whose authorship is
unknown, especially when there are no other links on the page to
establish where it lives or what it relates to. Sign your pages.
If a user comes to a page from a foreign link, give him the opportunity to explore your home page or the rest of your pages: make
your signature a link, direct or indirect, to your home page.
At the bottom of every page, my signature is a link up within my
tree of pages. For a page other than the index.html file in a directory, I place a signature that names the directory and a link to
index.html in that directory. At the bottom of each index.html file
I refer to the title of the next level up, and place a link to ../
index.html. This enables my visitor to ascend the whole tree back
to my home.
At the bottom of my home page, my signature is a MAILTO link. If
my visitor hasnt discovered the information he wants in his
traversal of my pages, this invites him to send e-mail to me.
I include at the bottom of every page the date that I last modified
the page.
Learn from your Internet service provider how to make your files
accessible to his web server.
If you use a UNIX server, include the lowercase L at the end of the
. html extension when you transfer, even if your local filenames are
limited by MS-DOS or Windows. Use UNIX (LF) line ends in text
files (including HTML) stored at a UNIX server.
Make sure every directory has a file index.html. If you do not do
this, then a visitor who manually enters the path to a directory will
be presented with a list of all of the files in that directory, perhaps
Your HTML pages include whatever file names and paths you need
for your links. The robots and wanderers will harvest filenames
from your HTML code, and add these referenced files to their
indices. If you want a file to be indexed, you should include its
name in another file that is indexed already: The robots will eventually find your new page!
You can place in your web directory a file whose name is not referenced in any of your pages. The robots will not discover this file.
But if a visitor guesses a name, index.bak or index.old for example,
there is no method to prevent the visitor from retrieving that file.
The only way to be absolutely certain that a visitor will not have
access to a file is to remove that file from your web directory.

My home page is accessible at the URL <http://www.inforamp.net/
~poynton/>. If I wish to direct someone to a page other than my
home page, say by e-mail, I specify the full URL of that page: my
page of Macintosh information is located at <http://
www.inforamp.net/~poynton/Poynton-mac.html>. However,
within my home page, I use a relative pathname such as Poyntonmac.html. Using a relative pathname makes it easier for you to
maintain pages and links, and makes it easier for your visitor to
make local copies of your pages while maintaining the function of
the links.
If you have created a hierarchy of pages, the easiest way to manually create a new page is to copy, then edit, a page at the same
level of the hierarchy.
Choose filenames that are mnemonic. When a visitor decides to
save one of your pages, the name you choose will be presented as
his default name.
Once youve chosen the name of a file (or page or directory), stick
to it. Other sites may have made links to your page (or directory). If
you change a name, you will break those links.
Test locally
Your pages will be no pleasure for your visitor if they do not work
as you yourself intend. Make sure that your pages work for you
before you subject someone else to them!
Test your pages locally, use the Open File capability of your favourite browser. Use two or three different browsers, to see how they
present things differently. Test your pages in black-and-white, to
preview how they will appear to a user who has only black-andwhite display capability.
Validate
When you have finished making a page, that you run it through an
HTML validation service to ensure that it conforms to the technical
requirements of HTML. If you do not do this, you cannot be sure
that it will work reliably on other browsers and other platforms than
yours.
If you have manually created your HTML, you can fix it by hand. If
you have used automated conversion tools, you may have little
scope to repair failures in validation. In this case, take the validation
report to the provider of your conversion tools.
10
Provide alternate access paths

The online services such as AOL and CIS are slow to adapt to new
technology, and millions of their subscribers have no access to the
web. If you want to make information available to online service
subscribers you must use file transfer protocol (ftp). Modern web
browsers accommodate ftp access, so your files will be available not
only to those subscribers, and to ftp users on the Internet, but also
to web users on the Internet. You will need your Internet access
providers help to set up an ftp directory at his site.
My technique is to store the bulk of my information in my ftp directory, and to use the web as an attractive, functional and interactive
interface to that information. All of the common web browsers
include ftp capability as a subset, so your web visitor will see no
break in continuity but your information is stored only once, and
CIS and AOL visitors have full access. My web pages simply contain
links that use ftp protocol to access directories and files. You can
view the source (HTML) of my pages to see how these links work.
When I create new information to publish, I store the files themselves in my ftp directories. Then I create a web page with the
descriptions and links. I open this page in my web browser and
Save As ASCII text-only format. I use that text file as the README
file for the ftp directory.
This site is best experienced ...
... using your choice of web browser, and your choice of settings.
The pages are not optimized for any particular browser. Instead, to the extent that is reasonably easy
to achieve with commercially available tools, the pages are compliant with HTML standards, to assure
best possible performance across a wide range of browsers.
You do not have to download any "preview," alpha, beta, or newly-revised browser software, or plugins, before viewing this site.
The site uses inline graphic elements judiciously, so as to make loading of the pages as speedy as
possible.
Colors and backgrounds at this site have been chosen for high contrast. The scheme is this: black text,
white backgrounds. That leads to easy readability.
There are no empty, distracting animations.
There are no <BLINK> codes.
I do not require that you "resize your browser window" - you do not even have to know what that
means!
No gratuitous audio will download and play when you access any page at this site.
No page at this site will "refresh itself" without your asking it to.
Your interaction with this site will comprise retrieval of the pages - it's as simple as that. There is no
saved state. You will not be forced to register, provide passwords, eat cookies, edit your MIME types,
or say OK to arcane alerts about site security.
In order to avoid confusing your navigation, this site uses no frames.
There are no gratuitous JavaScripts that scroll jerky text across your status bar and mess up your
ability to see the destination URLs of links.
There are no <FONT FACE> codes based on assumptions about what fonts I think might be installed
on your system; I do not assume that you are using a particular operating system. In addition, there are
no <FONT SIZE> codes that make assumptions about what text sizes you like to see. I do not assume
that I know what text size is best for your display and for your vision; instead, I assume that you know
http://www.poynton.com/notes/rants/Netscape_rant.html (1 di 2) [15/01/2005 16:30:37]
how to set your browser for your own choice.

These pages are designed to inform, not to impress.
I hope that you won't tell your friends that this site is "kewl," and I hope that that you won't call me
"dood."
See also, Making web pages usable.

See also, Ten common mistakes in the typesetting of technical documents.
Charles
Modified 1998-03-19
http://www.poynton.com/notes/rants/Netscape_rant.html (2 di 2) [15/01/2005 16:30:37]
Charles Poynton - Articles
Charles Poynton Articles

This page has abstracts of technical articles and papers that I have written concerning topics in digital
video, high definition television, and color reproduction. You can access the full original articles.
The abstracts are on the web (in HTML); the papers themselves are available in Adobe Acrobat PDF
format (served up through http protocol).
YUV and luminance considered harmful:

A plea for precise terminology in video
The notation YUV, and the term luminance, are widespread in digital video. However, digital video
almost never uses Y'UV color difference components, and never directly represents the luminance of
color science. The common terms are almost always wrong. This note explains why. I urge video
engineers and computer graphics specialists to use the correct terms, almost always Y'CBCR and luma
Acrobat PDF format, 54864 bytes
Merging Computing with Studio Video: Converting Between R'G'B'

and 4:2:2
In this "white paper" that I wrote for Discreet Logic, I explain the R'G'B' and Y'CBCR 4:2:2
representations, and explain the technical aspects of conversion between the two. I conclude by
suggesting steps that can be taken during production and post-production to avoid difficulty with the
conversion.
Luminance, luma, and the migration to DTV

On February 6, I presented a paper, Luminance, luma, and the migration to DTV, at the 32nd SMPTE
Advanced Motion Imaging Conference in Toronto. For now, just the abstract is available. The
technical note Errors due to nonconstant luminance contains information on this topic. I am in the
proces of preparing a published version of the paper. Some information concerning the Principle of
Constant Luminance is available in this IS&T paper ...
The rehabilitation of gamma

I presented this paper at the SPIE/IS&T Conference in San Jose, Calif., Jan. 26 - 30, 1998. The paper
is published in the Proceedings of that conference, B. E. Rogowitz and T. N. Pappas (eds.),
http://www.poynton.com/Poynton-articles.html (1 di 3) [15/01/2005 16:30:41]
Proceedings of SPIE 3299, 232-249 (Bellingham, Wash.: SPIE, 1998).
Motion portrayal, eye tracking, and emerging display technology

I presented this paper at the SMPTE conference in Seattle, February 1-3, 1996; it is published in The
convergence continues ... Computer technology and television: Proceedings of the 30th SMPTE
Advanced motion imaging conference, 192-202 (White Plains, New York: SMPTE, 1996).
Outlook for home use video terminals

International Broadcasting Symposium '95, Broadcasting in the Multimedia Age, Tokyo, Japan, Nov.
1995.
A Guided Tour of Color Space

This article describes the theory of color reproduction in video, and some of the engineering
compromises necessary to make practical cameras and practical coding systems. I presented this paper
at the SMPTE Advanced Television and Electronic Imaging Conference, San Francisco, Feb. 1995.
This is an edited version of the paper published in the proceedings of that conference, New
Foundations for Video Technology (pages 167-180).
Wide Gamut Device-Independent Colour Image Interchange

Proceedings of International Broadcasting Convention, 1994 (Amsterdam, 16-20 September 1994),
IEE Conference Publication No. 397, pages 218-222.
"Gamma" and its Disguises - The Nonlinear Mappings of Intensity

in Perception, CRTs, Film and Video
Published in SMPTE Journal, Vol. 102, No. 12 (December 1993), 1099-1108. This article has
effectively been superseded by Chapter 6, Gamma, of my book A Technical Introduction to Digital
Video. That chapter is available online, in Acrobat (PDF) format.
High Definition Television and Desktop Computing

International Technical Workshop on Multimedia Technologies in the HDTV Age, IEEE CES Tokyo
Chapter, July 20, 1993.
Color Management Technology for Workstations

Sun Expo '92, Manchester, U.K., Sepember 10, 1992.
RISC/UNIX Workstations in Desktop Color Prepress

Youngblood/IBEC's DeskTop PrePress Today - Colour Conference 92, Toronto, March 26, 1992.
A Tutorial on Magic Numbers for High Definition Electronic

Production
132nd SMPTE Technical Conference, New York, October 13-17, 1990.
Charles
1999-06-19
YUV and luminance considered harmful

Abstract
The notation YUV, and the term luminance, are widespread in digital video. However, digital video
almost never uses Y'UV color difference components, and never directly represents the luminance of
color science. The common terms are almost always wrong. This note explains why. I urge video
engineers and computer graphics specialists to use the correct terms, almost always Y'CBCR and luma
Complete paper
Acrobat PDF format, 73838 bytes
Charles - Articles
1999-06-19
http://www.poynton.com/papers/YUV_and_luminance_harmful.html [15/01/2005 16:30:45]
Charles Poynton
tel +1 416 413 1377
fax +1 416 413 1378
www.inforamp.net/~poynton

The notation YUV, and the term luminance, are widespread in digital
video. In truth, digital video almost never uses YUV color difference
components, and never directly represents the luminance of color
science. The common terms are almost always wrong. This note
explains why. I urge video engineers and computer graphics specialists
to use the correct terms, almost always YCBCR and luma.
Cement vs. concrete
Ill demonstrate by analogy why it is important to use correct terms.
Next time youre waiting in line for a bus, ask the person next to you in
line what building material is used to construct a sidewalk. Chances are
that person will answer, cement.
The correct answer is concrete. Cement is calcined lime and clay, in the
form of a fine, gray powder. Cement is one ingredient of concrete; the
other ingredients are sand, gravel, and water.
In an everyday situation, you need not be precise about which of these
terms are used: If you refer to a bridge as being constructed of
cement, people will know what you mean. Lay people are not confused by the term cement. Interestingly, experts are not confused
either. If a bridge superintendent yells out to his foreman, Get me 500
pounds of cement ! the foreman understands immediately from context whether the superintendent actually wants concrete. However, if
you place an order with a building material supplier for 500 pounds
of cement, you will certainly not receive 500 pounds of concrete! Lay
people have no trouble with the loose nomenclature, and the experts
have little trouble. It is the people in the middle who are liable to
become confused by loose nomenclature. Worse still, they are liable to
use a term without realizing that it is ambiguous or wrong!
True CIE luminance
The principles of color science dictate that true CIE luminance
denoted Y is formed as a weighted sum of linear (tristimulus) RGB
2001-03-08 Charles Poynton
1 of 4
YUV AND LUMINANCE CONSIDERED HARMFUL
601
Y = 0.299 R
+ 0.587 G
+ 0.114 B
Poynton, Charles, A Technical

Introduction to Digital Video
(New York: Wiley, 1996).
components. If CIE luminance were transmitted in a video system, the

system would conform to the Principle of Constant Luminance. But in
video we implement an engineering approximation that departs from
this principle. It was standardized for NTSC in 1953, and remains standard for all contemporary video systems, to form luma, denoted Y, as
a weighted sum of nonlinear (gamma-corrected) RGB components.
The nonlinear transfer function is roughly comparable to a square root.
To form luma, we use the theoretical coefficients of color science, but
we use them in a block diagram different from that prescribed by color
science: As detailed in my book, gamma correction is applied before
forming the weighted sum, not after. The order of operations is
reversed from what you might expect from color science.
The misinterpretation of luminance

Video engineers in the 1950s recognized that the video quantity Y was
very different from CIE luminance, and that it needed to be distinguished from luminance. They described it by the phrase the quantity
representative of luminance. They used the symbol Y, but augmented
it with a prime to denote the nonlinearity: Y. Obviously the qualifier
quantity representative of was cumbersome, and over the decades, it
was elided. And over time, the prime symbol was elided as well. Unfortunately, no new word was invented to supplement luminance, to reinforce the distinction between the color science quantity and the video
quantity. Most video engineers nowadays are unfamiliar with color science, and most do not understand the distinction. Engineers today
often carelessly use the word luminance, and the symbol Y, to refer to
the weighted sum of nonlinear (gamma-corrected) RGB components.
Pritchard, D.H., U.S. Color
Television Fundamentals
A Review, in SMPTE Journal,
v. 86 (Nov. 1977), 819828.
Smith, A.R., Color Gamut
Transform Pairs, in Computer
Graphics, v. 12, n. 2
(Aug. 1978, Proc.
SIGGRAPH 78), 1219.
Foley, James D., and Andries
van Dam, Fundamentals of
Interactive Computer Graphics
(Reading, Mass.: AddisonWesley, 1984).
Foley, James D., Andries van
Dam, Steven Feiner, and John
Hughes, Computer Graphics:
Principles and Practice, Second
Edition (New York: AddisonWesley, 1990).
The sloppy nomenclature made its way into ostensibly authoritative

video references, such as Pritchards SMPTE paper published in 1977.
The computer graphics pioneer Alvy Ray Smith encountered the word
luminance in his quest to adapt video principles to computer graphics.
Smith apparently correlated the use of the term luminance with his
knowledge of color science, and understandably though mistakenly
concluded that video luminance and color science luminance were
identical. Consequently, video YIQ was introduced to computer graphics, having its Y component alleged to be identical to CIE luminance.
That incorrect interpretation propagated into authoritative computer
graphics textbooks. Computer Graphics: Principles and Practice, on
page 589, Section 13.3.3, The YIQ Color Model, states:
The Y component of YIQ is not yellow but luminance,
and is defined to be the same as the CIE Y primary.
(The emphasis is in the original. Yellow refers to CMY, which was
mentioned in the immediately preceding section. CIE Y primary
would be more accurately denoted CIE Y component.)
As you have seen, the so-called Y component of video more properly

designated with a prime symbol, Y is not the same as CIE luminance.
Video Y cannot even be computed from CIE Y, unless two other color
components are also available. The quoted passage is quite wrong.
About 300,000 copies of various editions and adaptations of CG:PP
have been printed. Confusion is rampant.
Pratt, William K., Digital Image
Processing, Second Edition
(New York: Wiley, 1991). p. 64.
The error also propagated into the digital image processing community.
A widely used book in that field states:
N.T.S.C. formulated a color coordinate system for transmission
composed of three tristimulus values YIQ. The Y tristimulus value
is the luminance of a color.
The video quantities are certainly not tristimulus values, which are, by
CIEs definition, proportional to intensity.
Loose nomenclature on the part of video engineers has misled
a generation of digital image processing, computer software, and
computer hardware engineers.
The enshrining of luma

I campaigned for adoption of the term luma to designate the nonlinear
video quantity. The term had no pre-existing meaning, and by virtue of
its being different from luminance, it invites readers from other
domains to investigate fully before drawing conclusions about its relationship with luminance.
With the help of Fred Kolb, my campaign succeeded: In 1993, SMPTE
adopted Engineering Guideline EG 28, Annotated Glossary of Essential
Terms for Electronic Production. EG 28 defines the term luma, and clarifies the two conflicting interpretations of the term luminance. While
a SMPTE EG is not quite a SMPTE Standard, at long last the term has
received official recognition. Theres no longer any excuse for sloppy
use of the term luminance by the authors of video engineering papers.
It is a shame that todays SMPTE and ITU-R standards for digital video
persist in using the incorrect word luminance, without ever mentioning
the ambiguity even conflict with the CIE standards of color science.
Color difference scale factors
When I say NTSC and PAL,

I refer to color encoding, not
scanning: I do not mean
525/59.94 and 625/50.
To represent color, luma is accompanied by two color difference or

chroma components, universally based on blue minus luma and red
minus luma, where blue, red, and luma have all been subject to gamma
correction: B--Y and R--Y. Different scale factors are applied to the
basic B--Y and R--Y components for different applications. YPBPR
scale factors are optimized for component analog video. YCBCR scale
factors are optimized for component digital video such as 4:2:2 studio
video, JPEG, and MPEG. Kodaks PhotoYCC (YC1C2) uses scale factors
optimized to record the gamut of film colors. YUV and YIQ use scale
factors optimized to form composite NTSC and PAL video.
ITU-R Rec. BT.601, Studio

encoding parameters of digital
television for standard 4:3 and
wide-screen 16:9 aspect ratios
(Geneva: ITU).
YCBCR scaling as defined by Rec. 601 is appropriate for component

digital video. YCBCR chroma is almost always subsampled using one of
three schemes: 4:2:2, or 4:2:0, or 4:1:1.
YUV scaling is properly used only as an intermediate step in the formation of composite NTSC or PAL video signals. YUV scaling is not
appropriate when the components are kept separate. However, the
YUV nomenclature is now used rather loosely, and sometimes particularly in computing it denotes any scaling of B--Y and R--Y.
Digital disk recorders (DDRs) are generally able to transfer files across
Ethernet. Abekas introduced the convention of using an extension
.yuv for these files. But the scale factors in Abekas equipment,
at least actually correspond to YCBCR . Use of the .yuv extension
reinforces the misleading YUV nomenclature.
Chroma components are properly ordered B--Y then R--Y, or

CB then CR . Blue is associated
with U, and red with V. U and
V are in alphabetic order.
Hamilton, Eric, JPEG File Interchange Format, Version 1.02

(Milpitas, Calif.: C-Cube Microsystems, 1992).
Subsampling is properly performed only on component digital video,

that is, on YCBCR . Subsampling is inappropriate for YUV. If you see
a system described as YUV 4:2:2, you have a dilemma. Perhaps the
person who wrote the description is unfamiliar with the principles of
component video, and the scale factors actually implemented in the
equipment (or the software) are correct. But you must allow for the
possibility that the engineers who designed or implemented the system
used the wrong scale factors! If the wrong equations were used, then
color accuracy will suffer; however, this can be difficult to diagnose.
Proper YCBCR scaling is usual in Motion-JPEG, and in MPEG. However,
the YCBCR scaling used in stillframe JPEG/JFIF in computer applications usually uses full-range luma and chroma excursions, without
any headroom or footroom. The chroma excursion is 254255 of the
luma excursion. The scaling is almost exactly that of YPBPR , but is
unfortunately described as YCBCR : Now even YCBCR is ambiguous!
I am hopeful that proper YCBCR scaling will be incorporated into the
next revision of JFIF, so that compressed stillframe and motion imagery
in computing can be combined without suffering a conversion process.
Except for very limited use in the encoding and decoding of composite
4fSC (or loosely, D-2) studio video, Y IQ coding is obsolete.
Conclusion: A plea
Using the term luminance for video Y is tantamount to using the word
cement instead of concrete to describe the main construction material
of a bridge. Lay people dont care, and experts can live with it, but people in the middle in this case, the programmers and engineers who
are reimplementing video technology in the computer domain are liable to draw the wrong conclusions from careless use of terms. Users
suffer from this, because the exchange of images is compromised.
I urge video engineers and computer graphics specialists to avoid YUV
and luminance, and to use the correct terms, YCBCR and luma.
Merging Computing with Studio Video
Merging Computing with Studio

Video:
Converting Between R'G'B' and 4:2:2
Abstract
In this paper, I explain the R'G'B' and Y'CBCR 4:2:2 representations, and explain the technical aspects
of conversion between the two. I conclude by suggesting steps that can be taken during production
and post-production to avoid difficulty with the conversion.
Film, video, and computer-generated imagery (CGI) all start with red, green, and blue (RGB) intensity
components. In video and computer graphics, a nonlinear transfer function is applied to RGB
intensities to give gamma corrected R'G'B'. This is the native color representation of video cameras,
computer monitors, video monitors, and television.
The human visual system has poor color acuity. If R'G'B' is transformed into luma and chroma, then
color detail can be discarded without the viewer noticing. This enables a substantial saving in data
capacity - in "bandwidth", or in storage space. Because studio video equipment has historically
operated near the limit of realtime recording, processing, and transmission capabilities, the
subsampled Y'CBCR 4:2:2 format has been the workhorse of studio video for more than a decade.
The disadvantage of 4:2:2 is it's lossy compression. Upon conversion from 8-bit R'G'B' to 8-bit
Y'CBCR, three-quarters of the available colors are lost. Upon 4:2:2 subsampling, half the color detail is
discarded. But production staff are facing increasing demands for quality, and increasing demands to
integrate video production with film and CGI. The lossy compression of 4:2:2 is becoming a major
disadvantage.
Owing to the enormous computing and storage capacity of general-purpose workstations, it is now
practical to do production directly in R'G'B' (or as it's known in studio video terminology, 4:4:4). To
integrate traditional studio video equipment into the new digital studio, conversion between R'G'B'
and 4:2:2 is necessary.
This paper depends heavily upon graphics, so it is not provided in HTML/GIF format. A
typeset version of the paper is available in Acrobat PDF format (1662989 bytes), at two
alternate sites:
Discreet Logic (sorry about the outrageous URL), or
http://www.poynton.com/papers/Discreet_Logic/index.html (1 di 2) [15/01/2005 16:31:41]
Merging Computing with Studio Video
Poynton's site.
You can access information about document formats.
Charles - Articles
1998-07-28a
http://www.poynton.com/papers/Discreet_Logic/index.html (2 di 2) [15/01/2005 16:31:41]
Charles Poynton
Charles Poynton
Momaku
I am an independent contractor specializing in the physics, mathematics,

and engineering of digital color imaging systems, including digital video,
HDTV, and digital cinema (D-cinema). I do technology forecasting,
systems modelling, algorithm development (including digital filter design),
video signal processing architecture, color characterization and calibration,
and image quality assessment. (More...)
tel: +1 416 413 1377
poynton@poynton.com [Mac Eudora/MIME/BinHex/uu] [email protocol]
www.poynton.com
Digital Video and

HDTV
My book Digital Video and HDTV Algorithms and Interfaces is holding

fairly steady between the 10,000-th and 20,000-th most popular book at
Amazon.com. The Table of Contents and the newly-updated Errata have
been updated at 2004-03-04.
Courses, seminars
&c.
Upcoming (and past) events. I'm presenting several events this summer,
including events at Cine Gear Expo and Panavision in Los Angeles, at
JTS2004 in Toronto, and SIGGRAPH in LA.
Color technology
Video engineering
Digital Signal
Processing
Typography and
design
Hire me!
Includes Frequently Asked Questions(FAQs) about Gamma and Color.

Information concerning technical aspects of video.
Where to find digital filter design packages.
Articles I've written concerning typography, information design, and
presentation in the digital world. Also, a note about making web pages
usable, "This site is best experienced", and information concerning the
FrameMaker publication system.
I'm an independent contractor.
Personal stuff
& biographical data
"... experience proves that anyone who has studied geometry is infinitely
quicker to grasp difficult subjects than one who has not."
Plato - The Republic, Book 7, 375 B.C.
PZ internet
Charles Poynton
http://www.poynton.com/papers/index.html [15/01/2005 16:31:51]
Discreet Logic
10, Rue Duke
Montral, Qubec
Canada H3C 2L7
Tel: (514) 393-1616
Fax: (514) 393-0110
www.discreet.com
Merging computing with studio video:

Converting between RGB and 4:2:2
Charles Poynton
www . poynton . com
Abstract
In this paper, I explain the RGB and YCBCR 4:2:2 representations,
and explain the technical aspects of conversion between the two.
I conclude by suggesting steps that can be taken during production
and post-production to avoid difficulty with the conversion.
Film, video, and computer-generated imagery (CGI) all start with red,
green, and blue (RGB ) tristimulus components proportional to
intensity linear light. A nonlinear transfer function is applied to RGB
to give gamma corrected RGB. This is the native color representation
of video cameras, computer monitors, video monitors, and television.
The human visual system has poor color acuity. If RGB is transformed
into luma and chroma, then color detail can be discarded without the
viewer noticing. This enables a substantial saving in data capacity in
bandwidth, or in storage space. Because studio video equipment has
historically operated near the limit of realtime capture, recording,
processing, and transmission capabilities, the subsampled YCBCR 4:2:2
format has been the workhorse of studio video for more than a decade.
The disadvantage of 4:2:2 is its lossy compression. Upon matrixing
from 8-bit RGB to 8-bit YCBCR , three-quarters of the available colors
are lost. Upon 4:2:2 subsampling, half the color detail is discarded.
However, production staff are facing increasing demands for quality,
and increasing demands to integrate video production with film and
CGI. The lossy compression of 4:2:2 is becoming a major disadvantage.
Owing to the enormous computing and storage capacity of generalpurpose workstations, it is now practical to do production directly in
RGB (or as it is known in studio video terminology, 4:4:4). To integrate traditional studio video equipment into the new digital studio,
conversion between RGB and 4:2:2 is necessary.
1 of 8
CONVERTING BETWEEN RGB AND 4:2:2
Introduction
Linear light RGB is the native color coding of CGI. In computing, the
gamut of colors comprises the volume bounded by the unit RGB cube:
See Figure 1 opposite. In video and computer graphics, a nonlinear
transfer function is applied to RGB tristimulus signals to give gamma
corrected RGB, often in 8 bits each. See Figure 2, on page 4.
ITU-R Rec. BT.601, Studio

encoding parameters of digital
television for standard 4:3 and
wide-screen 16:9 aspect ratios
(Geneva: ITU).
If RGB is transformed into luma and color difference components,

YCBCR , then color detail can be subsampled (lowpass filtered) without
the viewer noticing. This leads to a substantial saving in data capacity
in bandwidth, or in storage space. Subsampling in YCBCR involves
a visually lossless lossy compression system. The 4:2:2 scheme has
a compression ratio of 1.5:1, and the 4:2:0 and 4:1:1 schemes have
compression ratios of 2:1. The subsampled YCBCR 4:2:2 representation of Rec. 601 is standard in studio digital video. However, YCBCR
has several problems in the digital studio:
Codeword utilization in YCBCR is very poor. RGB coding with 8 bits
per component allows every one of the 224 combinations, or 16 million
codewords, to represent a color. Theoretically, 3 4 or more of the
legal YCBCR code combinations do not represent colors! In 8-bit
Rec. 601 standard YCBCR , only 17% of the codewords represent
colors. YCBCR has fewer colors or equivalently, more quantization
noise, or poorer signal-to-noise ratio (SNR) than RGB.
The designation D-1 is sometimes loosely applied to 4:2:2.

However, D-1 properly refers
to a particular DVTR format,
not to an interface standard.
Filtering and subsampling operations that form the 4:2:2 signal remove
chroma detail. If subsampling is accomplished by simply dropping or
averaging alternate CB and CR samples, then filtering artifacts (such as
aliasing) will be introduced. Artifacts can accumulate if filtering is
repeated many times. Subsampling using a sophisticated filter gives
much better results than simply dropping or averaging samples.
However, even sophisticated filters can exhibit fringing on certain color
edges, if conversion between RGB and 4:2:2 is repeated many times.
Loss of color detail makes it more difficult to pull bluescreen or greenscreen mattes from 4:2:2 than from RGB.
Test signals characterize the electrical performance of a video system.
Standard video test signals include elements that are synthesized electronically as sine waves, and injected onto the signal. Many of these
elements have no legitimate RGB representation. Since these signals
can be conveyed through YCBCR without incident, some people claim
YCBCR to have an advantage. However, in my opinion, it is more
important to allocate bits to picture information than to signals that
cannot possibly represent picture information.
In general, YCBCR is optimized for realtime video, at the expense of
more difficult interface with film, CGI, and general-purpose computer
tools. RGB does not exploit chroma subsampling, so it has somewhat
higher data capacity requirements than YCBCR .
G AXIS
+1
Yl
Wt
Gray axis
(R = G = B)
Cy
18% Gray
Mg
+1
IS
AX
B
0
Bk
B AXIS
+1
Figure 1 RGB unit cube encompasses linearly-coded RGB tristimulus

values, each proportional to intensity. This scheme is poorly matched to
the lightness sensitivity of vision.
Computing gamut
STEP (riser)
Linear light coding is used in CGI, where physical light is simulated.

However, linear light coding performs poorly for images to be viewed.
The best perceptual use is made of the available bits by using nonlinear
coding that mimics the nonlinear lightness response of human vision.
In the storing and processing of images, linear light coding is rarely
used. In the display of images, linear light coding is never used. In
video, computing, and many other domains, a nonlinear transfer function is applied to RGB tristimulus signals to give nonlinearly-coded (or
gamma corrected ) components, denoted with prime symbols: RGB.
LEVEL (tread)
0
In an 8-bit system with nonlinear coding, each of R, G, and B ranges

from 0 through 255, inclusive. Each component has 255 steps (risers)
and 256 levels: A total of 224 colors that is, 16777216 colors are
representable. Not all of them can be distinguished visually; not all are
perceptually useful; but they are all colors. See Figure 2 overleaf.
RGB in video
Studio video RGB standards provide footroom below the black code,
and headroom above the white code. The primary purpose of footroom and headroom is to accommodate the transients that result from
filtering in either the analog or digital domains. Their secondary
purpose is to provide some margin to handle level variations in signals
originated in the analog domain. (Additionally, the headroom provides
a marginal improvement in highlight handling and exposure latitude.)
Wt
Gray axis
(R = G = B)
Cy
G COMPONENT
255
Yl
18% Gray
Mg
R
55
T2
N
NE
MP
CO
R
Bk
B COMPONENT
B
255
Figure 2 RGB cube represents nonlinear (gamma corrected) RGB typical of

computer graphics. Though superficially similar to the RGB cube of Figure 1, it is
dramatically different in practice owing to its perceptual coding.
254 +238
HEADROOM
INTERFACE
PROCESSING
235 +219
16
-15
FOOTROOM
Charles Poynton, Concerning

legal and valid video
signals, www.poynton.com
Eight-bit Rec. 601 coding has an excursion of 219 codes from black to
white. For no good technical reason, footroom and headroom are
assigned asymmetrically: Footroom has 15 levels, but headroom has 19.
An offset of +16 is added at an 8-bit interface. (Hardware engineers
say that black is at code 16, and white is at code 235.) The sketch in
the margin shows abstract levels in bold, and hardware levels in italics.
Interface codes 0 and 255 are reserved for synchronization purposes,
and are prohibited from appearing in video or ancillary data.
The so-called valid colors encompass the volume that is spanned when
each RGB component ranges from reference black to reference white.
In Rec. 601, each component has 219 steps (risers) that is, 220 levels.
That gives 220 220 220, or 10648000 colors: About 64% of the total
volume of codewords is valid.
Linear light RGB is the basis for color representation in film and CGI,
but linear light coding is a poor match to human perception. Greatly
improved results are obtained by using nonlinear RGB coding that
mimics the lightness sensitivity of vision. We can use another more
subtle application of the properties of vision to code video signals:
Vision has poor acuity to color detail, compared to its acuity for lightness. Providing that lightness detail is maintained, color detail can be
discarded. Owing to the nature of the visual system, if subsampling is
done correctly, it will not be noticed. Subsampling has two steps: First,
a lightness component and two color components are formed. Then,
detail is discarded from the two color components.
219
REFERENCE WHITE
Yl
Cy
G
Y AXIS
Mg
IS
AX
R
CR
2
11
-112
0
12
-1
B
112
CB AXIS
REFERENCE BLACK
Figure 3 YCBCR cube is formed when gamma-corrected RGB are transformed

to luma and chroma signals, which are then then scaled. Only about 14 of the
available YCBCR volume represents colors; the rest is wasted. This transform is
performed before 4:2:2, 4:2:0, or 4:1:1 chroma subsampling.
YCBCR video
To exploit the poor color acuity of vision, luma is formed as a properlyweighted sum of nonlinear R, G, and B. It is standard to use the coefficients of Rec. 601. Two color difference or chroma components
are then formed as blue minus luma and red minus luma, where blue,
red, and luma incorporate gamma correction. (Luma, B--Y, and R--Y
can be formed simultaneously from R, G, and B through a 33 matrix
multiplication.)
601
Y = 0.299 R
+ 0.587 G
+ 0.114 B
Charles Poynton, YUV and

luminance considered
harmful: A plea for precise
terminology in video,
www.poynton.com
Various scale factors, and various notations, are applied to the basic
B--Y and R-Y color differences. The correct scaling and nomenclature for component digital systems is YCBCR (not YUV). The correct
term for the lightness component is luma (not luminance).
If each of the Y, CB, and CR components has 8 bits of precision, then
obviously the entire YCBCR cube has the same number of codewords
as 8-bit RGB. However, it is immediately obvious from the appearance of the transformed RGB unit cube in Figure 3 above that only
a small fraction of the total volume of the YCBCR coordinate space is
occupied by colors! The number of colors accommodated is computed
as the determinant of the transform matrix. In Rec. 601 YCBCR , only
about 1 4 of the Rec. 601 studio video RGB codes are used.
1
220 2252
4
3
220
2784375
10648000
= 0.261
Of the 16.7 million colors available in studio RGB, only about

2.75 million are available in YCBCR . If RGB is transcoded to YCBCR ,
RGB 4:4:4
YCBCR 4:4:4
4:2:2
4:1:1
4:2:0 (JPEG/JFIF,
4:2:0
(Rec. 601)
(480i DV25; D-7)
H.261, MPEG-1)
(MPEG-2 fr)
R0 R1
Y0 Y1
Y0 Y1
Y0 Y1 Y2 Y3
Y0 Y1
Y0 Y1
R2 R3
Y2 Y3
Y2 Y3
Y4 Y5 Y6 Y7
Y2 Y3
Y2 Y3
G0 G1
CB0 CB1
CB01
CB03
G2 G3
CB2 CB3
CB23
CB47
B0 B1
CR0 CR1
CR01
CR03
B2 B3
CR2 CR3
CR23
CR47
CB03
CB03
CR03
CR03
Figure 4 Chroma subsampling. Providing full luma detail is maintained, visions poor color acuity enables color detail to
be reduced by subsampling. A 2 2 array of RGB pixels is matrixed to a luma component Y and color difference
(chroma) components CB and CR . CB and CR are then filtered (averaged). Here, CB and CR samples are drawn wider or
taller than the luma samples to indicate their spatial extent. The horizontal offset of CB and CR is due to cositing. (In
4:2:0 in JPEG/JFIF, MPEG-1, and H.261, chroma samples are sited interstitially, not cosited.)
then transcoded back to RGB, the resulting RGB cannot have any
more than 2.75 million colors!
Izraelevitz, David, and Joshua L.
Koslov, Code Utilization for
Component-coded Digital
Video, in Tomorrows Television, Proceedings of 16th
Annual SMPTE Television
Conference (White Plains, New
York: SMPTE, 1982), 2230.
The color difference components are bipolar. Unscaled, they range from
roughly --1 to +1. For analog engineers, the doubled excursion represents a 6 dB SNR penalty for the chroma components. Digital engineers should consider the sign to consume an extra bit in each of CB
and CR . This codeword utilization issue represents a serious limitation of
8-bit YCBCR performance. It necessitates techniques such as Quantels
patented dynamic rounding.
In addition to this obvious problem of codeword utilization, transforms
between YCBCR and RGB must have carefully-chosen matrix coefficients. If the product of the encoding matrix and the decoding matrix is
not very nearly an identity matrix, then roundoff errors will accumulate
every time an image is transcoded. High-end manufacturers take great
care in choosing these matrix coefficients; however, the entire problem
is circumvented by operating in RGB.
Chroma subsampling
Once color difference components have been formed, they can be
subsampled (filtered). The data compression that results from subsampling is the justification for using YCBCR in the first place! To
subsample by simply dropping samples leads to aliasing, and consequent poor image quality. It is necessary to perform some sort of averaging operation. The various subsampling schemes in use are sketched
in Figure 4 above.
Some systems implement 4:2:0 subsampling with minimum computation by simply averaging CB over a 22 block, and averaging CR over
the same 22 block. Simple averaging causes subsampled chroma to
take an effective position centered among a 22 block of luma
4
4
4
4
Figure 6 Interstitial 4:2:0 filter

for subsampling may be implemented using simple averaging. The rectangular outline
indicates the subsampled
YCBCR block; the black dot
suggests the effective siting of
the computed chroma sample.
4
4
samples, what I call interstitial siting. Low-end decoders simply replicate the subsampled 4:2:0 CB and CR to obtain the missing chroma
samples, prior to conversion back to RGB. This technique, sketched in
Figure 6 in the margin, is used in JPEG/JFIF stillframes in computing,
MPEG-1, and ITU-R Rec. H.261 videoconferencing.
Simple averaging causes subsampled chroma to take an effective position halfway between two luma samples, what I call interstitial siting.
This approach is inconsistent with standards for studio video and
MPEG-2, where CB and CR are cosited horizontally.
Weights of [ 1 4 , 1 2 , 1 4 ] can be used to achieve horizontal cositing as
required by Rec. 601, while still using simple computation, as sketched
at the top of Figure 7 in the margin. A [ 1 4 , 1 2 , 1 4 ] filter can be
combined with [ 1 2 , 1 2 ] vertical averaging, so as to be extended to
4:2:0 used in MPEG-2, as sketched at the bottom of Figure 7.
8
1
8
Figure 7 Cosited filters for

subsampling use weights that
cause each computed chroma
sample to be horizontally
aligned with a luma sample.
Simple averaging filters exhibit poor image quality. Providing the

weights are carefully chosen, a filter combining a large number of
samples that is, a filter with a larger number of taps will always
perform better than a filter with a smaller number of taps. (This fact is
not intuitive, because high frequency information is only apparent
across a small scale.) High-end digital video and film equipment uses
sophisticated subsampling filters, where the subsampled CB and CR of
a 21 pair in 4:2:2, or 22 quad of 4:2:0, take contributions from
many surrounding samples.
Sample aspect ratio, square sampling

In computing, it is a de facto standard to have samples equally-spaced
horizontally and vertically (square sampling). In conventional video,
various sample aspect ratios are in use: Sample aspect ratios differ
between 525/59.94 and 625/50, and neither has equally-spaced
samples. In high-definition television (HDTV), thankfully, square
sampling has been adopted.
In certain adaptations of YCBCR for film, the nonsquare sample aspect
ratio of conventional 625/50 video has been maintained. This forces
a resampling operation when that imagery is imported into the CGI
environment, and another resampling operation when it is exported.
If resampling is done well, it is intrinsically expensive. If resampling is
done poorly, or done often (in tandem), it introduces artifacts.
RGB and YCBCR characterization
Charles Poynton, The rehabilitation of gamma, in Human
Vision and Electronic
Imaging III, Proc. SPIE/IS&T
Conf. 3299, ed. B.E. Rogowitz
and T.N. Pappas (Bellingham,
Wash.: SPIE, 1998).
RGB is completely characterized by four technical parameters: white

point, primary chromaticities, transfer function, and coding range.
(A fifth rendering intent parameter is implicit; see my SPIE/IS&T paper.)
White point, primary chromaticities, and transfer function are all standardized by Rec. 709. The parameters of Rec. 709 closely represent
current practice in video and in computing. We have, in effect, reached
worldwide consensus on RGB coding. This is highly significant.
Coding range in computing has a de facto standard excursion,

0 to 255. Studio video accommodates footroom and headroom; its
range is standardized from 16 to 235. (In ITU-R Rec. BT.1361, the
coding range of Rec. 709 is extended to achieve a wider gamut.)
709
Y = 0.2126 R
+ 0.7122G
+ 0.0722 B
0
64 720
601YC C
B R
48
19
SDTV
709YC C
B R
10
80
HDTV
20
YCBCR is characterized by all of the parameters of RGB, plus a set of

luma coefficients. The coefficients of Rec. 601 are ubiquitous in
conventional 525/59.94 video, 625/50 video, and computing. But
according to recently-adopted SMPTE and Advanced Television Systems
Committee (ATSC) standards, HDTV will use a new, different set: the
luma coefficients of Rec. 709. This introduces a huge problem: There
will be one flavor of YCBCR for small, standard-definition television
(SDTV) pictures, and another for big (HDTV) pictures. YCBCR data
cannot be accurately exchanged between these flavors of coding
without undergoing a mathematical transform of comparable
complexity and comparable susceptibility to artifacts as resampling
for the correction of pixel aspect ratio. (If the mathematical transform is
not performed, then dramatic color errors result.)
Practical suggestions
To maximize performance at the interface of computing and video,
I recommend that you take these steps:
Acquire RGB 4:4:4 images wherever possible, instead of acquiring
images already subjected to the YCBCR transform and 4:2:2 subsampling. For realtime transfer, use the dual SDI link.
Stay in RGB if your production situation permits. The first conversion
to YCBCR will cause an unrecoverable loss of 75% of the available
RGB codewords, and the first subsampling to 4:2:2 will cause an
unrecoverable loss of half the color detail.
Avoid repeated conversions back and forth between RGB and 4:2:2.
Conversions after the first are liable to accumulate rounding errors, and
are liable to accumulate filtering artifacts such as aliasing.
Retain intermediates in RGB 4:4:4 format where possible. Use DLT or
Exabyte computer media, instead of videotape. Where intermediate or
archival work must be recorded on video equipment, use 10-bit D-5
recording, instead of 8-bit D-1.
Minimize resampling. To the extent possible, avoid changing from one
sample structure to another for example, from square sampling to
nonsquare, or from nonsquare to square.
Establish and maintain accurate black levels. Establish the correct black
level for a scene or an element upon entry to the digital domain. When
possible, perform this adjustment using video playback equipment.
(Establishing and maintaining white level is not quite so important.)
Presented at the 32nd SMPTE

Advanced Motion Imaging Conference,
Toronto, Feb. 6, 1998
Abstract
Since 1953, we have been using the wrong block diagram for color video! The principles of color
science dictate that we mix linear RGB to make true luminance, denoted Y. This is known as the
Principle of Constant Luminance. But in video we depart from that principle, and implement an
engineering approximation: We mix nonlinear ("gamma corrected") R'G'B' to make what I call luma,
denoted Y'. (Many video engineers carelessly call this luminance.) To form luma, we use the
theoretical coefficients of color science, but we use them in the wrong block diagram: We apply
gamma correction before the mixing, instead of after. This alteration in the block diagram is more or
less inconsequential in practice, though the departure from theory is apparent in the dark band seen
between the green and magenta color bars of the standard video test pattern.
The Rec. 709 HDTV standard has, wrongly in my opinion, adopted a new set of luma coefficients.
There is an incremental benefit in using the "right" coefficients, even in the wrong block diagram. The
change would be sensible if HDTV was a closed system. But changing the coefficients introduces a
second flavor of Y'CBCR - effectively, there is now one flavor of Y'CBCR for small pictures (SDTV),
and a different flavor of Y'CBCR for big pictures (HDTV).
HDTV originated in the studio will be downconverted to SDTV for simulcast. Consumer ATV
receivers will display conventional video sources through upconversion. Even studios are likely to use
upconversion, to allow the conventional studio equipment as a cheap way to originate "HDTV" in the
early phases of deployment. No studio upconverters or downconverters today do the necessary
processing to accommodate the different flavors of luma coefficients - instead, production personnel
are faced with having to perform manual colour correction. In consumer equipment, it seems highly
unlikely that the correction will ever be properly implemented in hardware or software: It is
impractical to perform a 3 x 3 matrix multiplication for every pixel, especially when the DCT of the
decompression requires only 3 multiplies per pixel! Instead, the consumer will be faced with poor
colour reproduction.
The computer and communications industries find it hard to understand why color coding parameters
should change as a function of image size. The different flavor means that it is impossible to cut and
paste Y'CBCR data from a big image to a little one.
The MPEG-2 standard provides, in its data stream, an indication of the luma coefficient set (flavor)
used in encoding. I advocate adoption of a recommended practice for ATV that calls for the use of
http://www.poynton.com/papers/SMPTE_98_YYZ_Luma/index.html (1 di 2) [15/01/2005 16:33:27]
Rec. 601 luma coefficients - in other words, that HDTV, ATV, and DTV use the same flavor of luma
coefficients that have been used since 1953. This will aid the transition to high definition, ATV, and
DTV, by delivering the high colour quality that studio engineers, program producers, and consumers
expect.
No printed version of this paper is available at the moment, though I plan to write it up eventually and
submit it for publication. An audiotape recording of the conference presentation is available from
SMPTE.
See also, Constant Luminance.
Charles Poynton - Video engineering
1998-03-26
http://www.poynton.com/papers/SMPTE_98_YYZ_Luma/index.html (2 di 2) [15/01/2005 16:33:27]
Constant Luminance
Constant Luminance
Since 1953, we have been using a block diagram for color video that is different from the one that a
color scientist would prefer to use. The principles of color science dictate that we mix linear RGB
(tristimulus signals) to make true luminance, denoted Y. If a video system were to operate in this way,
it would adhere to the Principle of Constant Luminance. But in video we depart from that principle,
and implement an engineering approximation: We mix nonlinear ("gamma corrected") RGB to make
what I call luma, denoted Y'. (Many video engineers carelessly call this luminance.) To form luma, we
use the coefficients that a color scientist would use to form luminance, but we use them in a different
block diagram than the color scientist expects: We apply gamma correction before the mixing, instead
of after. This alteration in the block diagram introduces a few image artifacts that are usually fairly
minor. The departure from the theoretically correct order of operations is apparent in the dark band
seen between the green and magenta color bars of the standard video test pattern.
Details are available in Chapter 8 of Digital Video and HDTV Algorithms and Interfaces.
The issue of constant luminance (or lack of it) is intimately intertwined with gamma correction.
Gamma has unjustifiably acquired a bad reputation. I presented a paper on the topic, The
rehabilitation of gamma, at a SPIE/IS&T conference in 1998.That paper outlines the Principle of
Constant Luminance. As you can deduce from its title, that paper concentrates on the reproduction of
lightness (which is related to luminance, which is related to luma). It merely outlines the color issues.
I presented the related issue of choosing luma coefficients for conventional video, DTV, ATV, and
HDTV, in a SMPTE paper in 1998: Luminance, luma, and the migration to DTV. The so-called paper
is virtual at this moment, having not been actually finished in that medium! However, the abstract of
the presentation is available:
For the truly courageous, an audiotape of the session is available through SMPTE. The opening
paragraph of this note is the first paragraph of that paper's abstract.
Some fragments of the paper-in-progress are available. Start with the brief technical note Errors due to
nonconstant luminance. If you STILL want to keep going, access the links at the bottom of that page.
All of this will be tied together within a month or two, and then (eventually) released as the written
version of the SMPTE paper.
Related documents, typeset, available in Acrobat PDF format:
The rehabilitation of gamma (Acrobat PDF format, 1223508 bytes)
http://www.poynton.com/notes/video/Constant_luminance.html (1 di 2) [15/01/2005 16:33:29]
Constant Luminance
The magnitude of nonconstant luminance errors (Acrobat PDF format, 89246 bytes)
Transforms between luma coefficient sets (Acrobat PDF format, 87907 bytes)
Charles Poynton - Video engineering

2004-02-25
http://www.poynton.com/notes/video/Constant_luminance.html (2 di 2) [15/01/2005 16:33:29]
No browsing!
Sorry, no browsing!
http://www.poynton.com/notes/index.html [15/01/2005 16:33:31]

by Charles Poynton,
available from the publisher, online retailers, and bookstores.
Placing video in the context of computing

Rapidly evolving computer and communications technologies have achieved data transmission rates
and data storage capacities high enough for digital video. But video involves much more than just
pushing bits! Achieving the best possible image quality, accurate color, and smooth motion requires
understanding many aspects of image acquisition, coding, processing, and display that are outside the
usual realm of computer graphics. At the same time, video system designers are facing new demands
to interface with film and computer system that require techniques outside conventional video
engineering.
Charles Poynton's 1996 book A Technical Introduction to Digital Video became an industry favorite
for its succinct, accurate, and accessible treatment of standard definition television (SDTV). In Digital
Video and HDTV, Poynton covers not only SDTV, but also high definition television (HDTV) and
compression systems. With the help of hundreds of high quality technical illustrations, this book
presents the following topics:
Basic concepts of digitization, sampling, quantization, gamma, and filtering

Principles of color science as applied to image capture and display
Scanning and coding of SDTV and HDTV
Video color coding: luma, chroma (4:2:2 component video, 4fSC composite video)
Analog NTSC and PAL
Studio systems and interfaces
Compression technology, including M-JPEG and MPEG-2
Broadcast standards and consumer video equipment
CHARLES POYNTON is an independent contractor specializing in the physics, mathematics, and

engineering of digital color imaging systems, including digital video, HDTV, and digital cinema (Dcinema). He designed and built the digital video equipment used at NASA to convert video from the
Space Shuttle into NTSC, initiated Sun Microsystems' HDTV research project in the early 1990s, and
has taught many popular courses on HDTV and video technologies. A Fellow of the Society of
Motion Picture and Television Engineers (SMPTE), Poynton was awarded the Society's prestigious
David Sarnoff Gold Medal for his work to integrate video technology with computing and
http://www.poynton.com/notes/DVAI/index.html (1 di 2) [15/01/2005 16:33:37]
communications.
The Table of Contents is available. Errata are available.
Sample chapters will soon be available here.
You can order from the publisher, or Amazon, or Barnes & Noble.
This book was to been published by Wiley as Digital Video and HDTV: Pixels, Pictures, and
Perception, but the title and publisher have changed. (Why?)
2003-02-14
Charles Poynton
http://www.poynton.com/notes/DVAI/index.html (2 di 2) [15/01/2005 16:33:37]

by Charles Poynton,
to be available from the publisher on or about December 2, 2002
Contents
Foreword by Jim Blinn
Foreword by Mark Schubin
List of figures
List of tables
Preface
Acknowledgements
Part 1 - Introduction
1 - Raster images
2 - Quantization
3 - Brightness and contrast controls
4 - Raster images in computing
5 - Image structure
6 - Raster scanning
http://www.poynton.com/notes/DVAI/DVAI_TOC.html (1 di 5) [15/01/2005 16:33:39]
7 - Resolution
8 - Constant luminance
9 - Rendering intent
10 - Introduction to luma and chroma
11 - Introduction to component SDTV
12 - Introduction to composite NTSC and PAL
13 - Introduction to HDTV
14 - Introduction to video compression
15 - Digital video interfaces
Part 2 - Principles
16 - Filtering and sampling
17 - Resampling, interpolation, and decimation
18 - Image digitization and reconstruction
19 - Perception and visual acuity
20 - Luminance and lightness
21 - The CIE system of colorimetry
22 - Color science for video
23 - Gamma
24 - Luma and color differences

25 - Component video color coding for SDTV
26 - Component video color coding for HDTV
27 - Video signal processing
28 - NTSC and PAL chroma modulation
29 - NTSC and PAL frequency interleaving
30 - NTSC Y'IQ system
31 - Frame, field, line, and sample rates
32 - Timecode
33 - Digital sync, TRS, ancillary data, and interface
34 - Analog SDTV sync, genlock, and interface
35 - Videotape recording
36 - 2-3 pulldown
37 - Deinterlacing
Part 3 - Video compression

38 - JPEG and motion-JPEG (M-JPEG) compression
39 - DV compression
40 - MPEG-2 video compression
Part 4 - Studio standards

40 - 480i component video
40 - 480i NTSC composite video
43 - 576i PAL composite video
45 - SDTV test signals
46 - 1280 x 720 HDTV
47 - 1920 x 1080 HDTV
Part 5 - Broadcast and consumer standards

48 - Analog NTSC and PAL broadcast standards
49 - Consumer analog NTSC and PAL
50 - Digital television broadcast standards
Appendices
A - YUV and luminance considered harmful
B - Introduction to radiometry & photometry
Glossary of video signal terms
Index
2002-11-06
DVAI

by Charles Poynton,
to be available from the publisher on or about December 2, 2002
Contents
List of figures
List of tables
Preface
Acknowledgements
1 - Raster images
Imaging
Aspect ratio
Digitization
Pixel array
Visual acuity
Lightness terminology
Nonlinear image coding
Linear and nonlinear
Luma and color difference components
http://www.poynton.com/notes/DVAI/DVAI_TOC_full.html (1 di 16) [15/01/2005 16:33:45]
SDTV/HDTV
2 - Quantization
Decibels
Quantization error
Linearity
Headroom and footroom
Brightness and contrast controls in desktop graphics
Symbolic image description

Raster images
Conversion among types
Display modes
Image files
"Resolution" in computer graphics
5 - Image structure
Image reconstruction
Sampling aperture
Spot profile
Box distribution
Gaussian distribution
6 - Raster scanning
Flicker, refresh rate, and frame rate

Introduction to scanning
Scanning parameters
Interlaced scanning
Twitter
Interlace in analog systems
Interlace and progressive
Scanning notation
Interlace artifacts
Motion portrayal
Segmented frame (24PsF)
Video system taxonomy
Conversion among systems
7 - Resolution
Magnitude frequency response and bandwidth

Kell effect
Resolution
Resolution in video
Viewing distance
Interlace revisited
The principle of constant luminance

Compensating the CRT
Departure from constant luminance
"Leakage" of luminance into chroma
Surround effect
Tone scale alteration
Incorporation of rendering intent
Rendering intent in desktop computing
Luma
Sloppy use of the term luminance
Color difference coding (chroma)
Chroma subsampling
Chroma subsampling notation
Chroma subsampling filters
Chroma in composite NTSC and PAL
Scanning standards
Widescreen (16:9) SDTV
Progressive SDTV (480p/483p)
Square and nonsquare sampling

Resampling
NTSC and PAL encoding

NTSC and PAL decoding
S-video interface
Frequency interleaving
Composite digital SDTV (4fSC)
Composite analog SDTV
Comparison of aspect ratios

HDTV scanning
The 1035i (1125/60) system
Color coding for Rec. 709 HDTV
Data compression
Image compression
Lossy compression
JPEG
Motion-JPEG
MPEG
Picture coding types (I, P, B)
Reordering
MPEG-1
MPEG-2
Component digital SDTV interface (Rec. 601, "4:2:2")

Composite digital SDTV (4fSC) interface
Serial digital interface (SDI)
Component digital HDTV HD-SDI
Interfaces for compressed video
SDTI
DVB ASI and SMPTE SSI
IEEE 1394 (FireWire, i.LINK)
Switching and mixing
Timing in analog facilities
Timing in composite analog NTSC and PAL

Timing in digital facilities
Part 2 - Principles
Sampling theorem
Sampling at exactly 0.5fS
Magnitude frequency response
Magnitude frequency response of a boxcar
The sinc weighting function
Frequency response of point sampling
Fourier transform pairs
Analog filters
Digital filters
Impulse response
Finite impulse response (FIR) filters
Physical realizability of a filter
Phase response (group delay)
Infinite impulse response (IIR) filters
Lowpass filter
Digital filter design
Reconstruction
Reconstruction close to 0.5fS
(sin x)/x correction
Further reading
2:1 downsampling
Oversampling
Interpolation
Lagrange interpolation
Lagrange interpolation as filtering
Polyphase interpolators
Polyphase taps and phases
Implementing polyphase interpolators
Decimation
Lowpass filtering in decimation
Spatial frequency domain

Comb filtering
Spatial filtering
Image presampling filters
Image reconstruction filters
Spatial (2-D) oversampling
Retina
Adaptation
Contrast ratio
Contrast sensitivity
Contrast sensitivity function (CSF)
Radiance, intensity
Luminance
Relative luminance
Luminance from red, green, and blue
Lightness (CIE L*)
Fundamentals of vision
Definitions
Spectral power distribution (SPD) and tristimulus
Scanner spectral constraints
CIE XYZ tristimulus
CIE [x, y] chromaticity
Blackbody radiation
Color temperature
White
Perceptually uniform color spaces
CIE L*u*v*
CIE L*a*b*
CIE L*u*v* and CIE L*a*b* summary
Color specification
Color image coding
Further reading
Additive reproduction (RGB)

Characterization of RGB primaries
CIE RGB primaries
NTSC primaries (obsolete)
EBU Tech. 3213 primaries
SMPTE RP 145 primaries
Rec. 709/sRGB primaries
CMFs and SPDs
Luminance coefficients
Transformations between RGB and CIE XYZ
Noise due to matrixing
Transforms among RGB systems
Camera white reference
Monitor white reference
Gamut
Wide-gamut reproduction
Further reading
23 - Gamma
Gamma in CRT physics

The amazing coincidence !
Gamma in video
Optoelectronic transfer functions (OETFs)
sRGB transfer function
Transfer functions in SDTV
Bit depth requirements
Gamma in emerging display devices
Gamma in video, CGI, SGI, and Macintosh
Gamma in pseudocolor
Limitations of 8-bit linear coding
Linear and nonlinear coding in CGI
Color acuity
RGB and R'G'B' color cubes
Conventional luma/color difference coding
Luminance and luma notation
Nonlinear red, green, blue (R'G'B')
Rec. 601 luma

Rec. 709 luma
SMPTE 240M-1988 luma
Chroma subsampling, revisited
Luma/color difference summary
SDTV and HDTV luma chaos
Luma/color difference component sets
B'-Y', R'-Y' components for SDTV

PBPR components for SDTV
CBCR components for SDTV
Y'CBCR from studio RGB
Y'CBCR from computer RGB
"Full-range" Y'CBCR
Y'UV, Y'IQ confusion
B'-Y', R'-Y' components for Rec. 709 HDTV

PBPR components for Rec. 709 HDTV
components for Rec. 709 HDTV
CBCR components for Rec. 1361 HDTV
Conversions between HDTV and SDTV
Color coding standards
Transition samples
Edge treatment
Picture lines
Choice of SAL and SPW parameters
Video levels
Setup (pedestal)
Rec. 601 to computing
Enhancement
Median filtering
Coring
Chroma transition improvement (CTI)
Scan-velocity modulation (SVM)

Mixing and keying
UV components
NTSC chroma modulation
NTSC chroma demodulation
PAL chroma modulation
Subcarrier regeneration
S-video interface
Decoder controls
Notch filtering
Frequency interleaving in NTSC
Cross-luma and cross-color
Frequency interleaving in PAL
Spatial frequency spectra of NTSC
Spatial frequency spectra of PAL
One-dimensional frequency spectrum of NTSC
One-dimensional frequency spectrum of PAL
Narrowband Q
IQ components
Y'IQ encoding
Field rate
Line rate
Sound subcarrier
Addition of composite color
NTSC color subcarrier
576i PAL color subcarrier
4fSC sampling
Common sampling rate
Numerology of HDTV scanning
Audio rates
32 - Timecode
Introduction
Dropframe timecode
Editing
Linear timecode (LTC)
Vertical interval timecode (VITC)
Timecode structure
Further reading
TRS in 4:2:2 component SDTV

TRS in HD-SDI
TRS-ID in 4fSC composite video
Digital to analog timing relationships
Ancillary data
SDI coding
HD-SDI coding
Summary
Analog sync
Odd/even, first/second, top/bottom
Sync distribution
Genlock
Analog horizontal blanking interval
Sync separation
Component analog levels
Composite analog levels
Analog electrical interface
Analog mechanical interface
S-video electrical and mechanical interface
Playback in shuttle
Recording
Playback
Editing
Digital VTRs
Timebase error
Channel coding
Analog VTR signal processing

Analog videotape formats
Digital VTR signal processing
Digital videotape formats
DV family
DV recording
Studio adaptation of DV technology
HDTV videotape formats
Consumer bitstream recording - DV ATV, DV DVB
Further reading
36 - 2-3 pulldown
Conversion of film to different frame rates

Native 24 Hz coding
Conversion to other rates
37 - Deinterlacing
Spatial domain
Vertical-temporal domain
Motion adaptivity

JPEG blocks and MCUs

JPEG block diagram
Level-shifting
Discrete cosine transform (DCT)
JPEG encoding example
JPEG decoding
Compression ratio control
JPEG/JFIF
Motion-JPEG (M-JPEG)
Further reading
39 - DV compression
DV chroma subsampling
DV frame/field modes
Picture-in-shuttle in DV
DV overflow scheme
DV quantization
Consumer DV variants - SD, LP, SDL, HD
Professional DV variants
DV digital interface (DIF)
Sony Digital Betacam compression
Sony Betacam SX compression
D-5 HD compression
D-11 (HDCAM) compression
MPEG-2 profiles and levels

Picture structure
Frame rate and 2-3 pulldown in MPEG
Luma and chroma sampling structures
Macroblocks
Picture coding types - I, P, B
Prediction
Motion vectors (MVs)
Coding of a block
Frame and field DCT types
Zigzag and VLE
Refresh
Motion estimation
Rate control and buffer management
Bitstream syntax
Transport
Further reading

Frame rate
Interlace
Line sync
Field/frame sync
RGB primary components
Nonlinear transfer function
Luma (Y')
Picture center, aspect ratio, and blanking
Halfline blanking
Component digital 4:2:2 interface
Component analog R'G'B' interface

Component analog Y'PBPR interface, EBU N10
Subcarrier
NTSC two-frame sequence
NTSC burst
Color differences (U, V)
Color difference filtering
Chroma (C)
Setup
S-video-525 (Y'/C3.58)
Composite NTSC encoding
Composite digital NTSC interface (4fSC)
Composite analog NTSC interface
43 - 576 i component video
Frame rate
Interlace
Line sync
Analog field/frame sync
Luma (Y')
Component analog Y'PBPR interface
Subcarrier
PAL four-frame sequence
PAL burst
Color difference components (U, V)
Chroma (C)
S-video-625 (Y'/C4.43)
Composite PAL encoding
Composite digital PAL interface (4fSC)
Composite analog PAL interface
Colorbars
Colorbar notation
Frequency response
Differential gain (DG)
Differential phase (DP)
Pulse signals
Modulated 12.5T, 20T pulses
46 - 1280 x 720 HDTV
Scanning
Analog sync
Luma (Y')
Pre- and postfiltering characteristics
47 - 1920 x 1080 HDTV
Scanning
Analog sync
Relationship to SMPTE 240M (1035i) scanning
Luma (Y' )

ITU-R (former CCIR)
ITU-R scanning nomenclature

M/NTSC (NTSC)
Audio in NTSC
B,G,H,I/PAL (PAL)
Audio in PAL
PAL-M, PAL-N
SECAM
Multiplexed analog components (MAC)
Summary of parameters
Multistandard consumer equipment

Degenerate analog NTSC and PAL
Coherent subcarrier
Incoherent subcarrier
Nonstandard scanning
SCART interface
Heterodyne (color-under) recording
VHS trick mode playback
Timebase correction (TBC)
Japan
United States
ATSC modulation
Europe
Further reading
Appendices
Cement vs. concrete

True CIE luminance
Conclusion: A plea
B - Introduction to radiometry and photometry
Radiometry
Photometry
Image science
Units
Further reading

Index
2002-11-06
DVAI
Moved ... redirecting ...
http://www.poynton.com/notes/DVAI/errata/index.html [15/01/2005 16:34:01]

by Charles Poynton,
Contents
List of figures
List of tables
Preface
Acknowledgements
1 - Raster images
2 - Quantization
5 - Image structure
6 - Raster scanning
http://www.poynton.com/DVAI/DVAI_TOC.html (1 di 5) [15/01/2005 16:34:06]
7 - Resolution
Part 2 - Principles
23 - Gamma

32 - Timecode
36 - 2-3 pulldown
37 - Deinterlacing

39 - DV compression

46 - 1280 x 720 HDTV
47 - 1920 x 1080 HDTV

Appendices
B - Introduction to radiometry & photometry
Index
2003-09-25
DVAI

by Charles Poynton,
Contents
List of figures
List of tables
Preface
Acknowledgements
1 - Raster images
Imaging
Aspect ratio
Digitization
Pixel array
Visual acuity
Lightness terminology
Nonlinear image coding
Linear and nonlinear
Luma and color difference components
SDTV/HDTV
http://www.poynton.com/DVAI/DVAI_TOC_full.html (1 di 16) [15/01/2005 16:34:15]
2 - Quantization
Decibels
Quantization error
Linearity
Headroom and footroom
Brightness and contrast controls in desktop graphics
Symbolic image description

Raster images
Conversion among types
Display modes
Image files
"Resolution" in computer graphics
5 - Image structure
Image reconstruction
Sampling aperture
Spot profile
Box distribution
Gaussian distribution
6 - Raster scanning
Flicker, refresh rate, and frame rate

Introduction to scanning
Scanning parameters
Interlaced scanning
Twitter
Interlace in analog systems
Interlace and progressive
Scanning notation
Interlace artifacts
Motion portrayal
Segmented frame (24PsF)
Video system taxonomy

Conversion among systems
7 - Resolution
Magnitude frequency response and bandwidth

Kell effect
Resolution
Resolution in video
Viewing distance
Interlace revisited
The principle of constant luminance

Compensating the CRT
Departure from constant luminance
"Leakage" of luminance into chroma
Surround effect
Tone scale alteration
Incorporation of rendering intent
Rendering intent in desktop computing
Luma
Sloppy use of the term luminance
Color difference coding (chroma)
Chroma subsampling
Chroma subsampling notation
Chroma subsampling filters
Chroma in composite NTSC and PAL
Scanning standards
Widescreen (16:9) SDTV
Progressive SDTV (480p/483p)
Square and nonsquare sampling
Resampling
NTSC and PAL encoding

NTSC and PAL decoding
S-video interface
Frequency interleaving
Composite digital SDTV (4fSC)
Composite analog SDTV
Comparison of aspect ratios

HDTV scanning
The 1035i (1125/60) system
Color coding for Rec. 709 HDTV
Data compression
Image compression
Lossy compression
JPEG
Motion-JPEG
MPEG
Picture coding types (I, P, B)
Reordering
MPEG-1
MPEG-2
Component digital SDTV interface (Rec. 601, "4:2:2")

Composite digital SDTV (4fSC) interface
Serial digital interface (SDI)
Component digital HDTV HD-SDI
Interfaces for compressed video
SDTI
DVB ASI and SMPTE SSI
IEEE 1394 (FireWire, i.LINK)
Switching and mixing
Timing in analog facilities
Timing in composite analog NTSC and PAL
Timing in digital facilities
Part 2 - Principles
Sampling theorem
Sampling at exactly 0.5fS
Magnitude frequency response
Magnitude frequency response of a boxcar
The sinc weighting function
Frequency response of point sampling
Fourier transform pairs
Analog filters
Digital filters
Impulse response
Finite impulse response (FIR) filters
Physical realizability of a filter
Phase response (group delay)
Infinite impulse response (IIR) filters
Lowpass filter
Digital filter design
Reconstruction
Reconstruction close to 0.5fS
(sin x)/x correction
Further reading
2:1 downsampling
Oversampling
Interpolation
Lagrange interpolation
Lagrange interpolation as filtering
Polyphase interpolators
Polyphase taps and phases
Implementing polyphase interpolators
Decimation
Lowpass filtering in decimation
Spatial frequency domain

Comb filtering
Spatial filtering
Image presampling filters

Image reconstruction filters
Spatial (2-D) oversampling
Retina
Adaptation
Contrast ratio
Contrast sensitivity
Contrast sensitivity function (CSF)
Radiance, intensity
Luminance
Relative luminance
Luminance from red, green, and blue
Lightness (CIE L*)
Fundamentals of vision
Definitions
Spectral power distribution (SPD) and tristimulus
Scanner spectral constraints
CIE XYZ tristimulus
CIE [x, y] chromaticity
Blackbody radiation
Color temperature
White
Perceptually uniform color spaces
CIE L*u*v*
CIE L*a*b*
CIE L*u*v* and CIE L*a*b* summary
Color specification
Color image coding
Further reading
Additive reproduction (RGB)

Characterization of RGB primaries
CIE RGB primaries
NTSC primaries (obsolete)

EBU Tech. 3213 primaries
SMPTE RP 145 primaries
Rec. 709/sRGB primaries
CMFs and SPDs
Luminance coefficients
Transformations between RGB and CIE XYZ
Noise due to matrixing
Transforms among RGB systems
Camera white reference
Monitor white reference
Gamut
Wide-gamut reproduction
Further reading
23 - Gamma
Gamma in CRT physics

The amazing coincidence !
Gamma in video
Optoelectronic transfer functions (OETFs)
sRGB transfer function
Transfer functions in SDTV
Bit depth requirements
Gamma in emerging display devices
Gamma in video, CGI, SGI, and Macintosh
Gamma in pseudocolor
Limitations of 8-bit linear coding
Linear and nonlinear coding in CGI
Color acuity
RGB and R'G'B' color cubes
Conventional luma/color difference coding
Luminance and luma notation
Nonlinear red, green, blue (R'G'B')
Rec. 601 luma
Rec. 709 luma
Chroma subsampling, revisited

Luma/color difference summary
SDTV and HDTV luma chaos
Luma/color difference component sets
B'-Y', R'-Y' components for SDTV

PBPR components for SDTV
CBCR components for SDTV
"Full-range" Y'CBCR
Y'UV, Y'IQ confusion
B'-Y', R'-Y' components for Rec. 709 HDTV

PBPR components for Rec. 709 HDTV
components for Rec. 709 HDTV
CBCR components for Rec. 1361 HDTV
Conversions between HDTV and SDTV
Color coding standards
Transition samples
Edge treatment
Picture lines
Choice of SAL and SPW parameters
Video levels
Setup (pedestal)
Rec. 601 to computing
Enhancement
Median filtering
Coring
Chroma transition improvement (CTI)
Scan-velocity modulation (SVM)
Mixing and keying
UV components
NTSC chroma modulation
NTSC chroma demodulation
PAL chroma modulation
Subcarrier regeneration
S-video interface
Decoder controls
Notch filtering
Frequency interleaving in NTSC
Cross-luma and cross-color
Frequency interleaving in PAL
Spatial frequency spectra of NTSC
Spatial frequency spectra of PAL
One-dimensional frequency spectrum of NTSC
One-dimensional frequency spectrum of PAL
Narrowband Q
IQ components
Y'IQ encoding
Field rate
Line rate
Sound subcarrier
Addition of composite color
NTSC color subcarrier
576i PAL color subcarrier
4fSC sampling
Common sampling rate
Numerology of HDTV scanning
Audio rates
32 - Timecode
Introduction
Dropframe timecode
Editing
Linear timecode (LTC)
Vertical interval timecode (VITC)
Timecode structure
Further reading
TRS in 4:2:2 component SDTV

TRS in HD-SDI
TRS-ID in 4fSC composite video
Digital to analog timing relationships
Ancillary data
SDI coding
HD-SDI coding
Summary
Analog sync
Odd/even, first/second, top/bottom
Sync distribution
Genlock
Analog horizontal blanking interval
Sync separation
Component analog levels
Composite analog levels
Analog electrical interface
Analog mechanical interface
S-video electrical and mechanical interface
Playback in shuttle
Recording
Playback
Editing
Digital VTRs
Timebase error
Channel coding
Analog VTR signal processing
Analog videotape formats
Digital VTR signal processing
Digital videotape formats

DV family
DV recording
Studio adaptation of DV technology
HDTV videotape formats
Consumer bitstream recording - DV ATV, DV DVB
Further reading
36 - 2-3 pulldown
Conversion of film to different frame rates

Native 24 Hz coding
Conversion to other rates
37 - Deinterlacing
Spatial domain
Vertical-temporal domain
Motion adaptivity

JPEG blocks and MCUs

JPEG block diagram
Level-shifting
Discrete cosine transform (DCT)
JPEG encoding example
JPEG decoding
Compression ratio control
JPEG/JFIF
Motion-JPEG (M-JPEG)
Further reading
39 - DV compression
DV chroma subsampling
DV frame/field modes
Picture-in-shuttle in DV
DV overflow scheme
DV quantization
Consumer DV variants - SD, LP, SDL, HD
Professional DV variants
DV digital interface (DIF)
Sony Digital Betacam compression
Sony Betacam SX compression
D-5 HD compression
D-11 (HDCAM) compression
MPEG-2 profiles and levels

Picture structure
Frame rate and 2-3 pulldown in MPEG
Luma and chroma sampling structures
Macroblocks
Picture coding types - I, P, B
Prediction
Motion vectors (MVs)
Coding of a block
Frame and field DCT types
Zigzag and VLE
Refresh
Motion estimation
Rate control and buffer management
Bitstream syntax
Transport
Further reading

Frame rate
Interlace
Line sync
Field/frame sync
Luma (Y')
Halfline blanking
Component analog Y'PBPR interface, EBU N10
Subcarrier
NTSC two-frame sequence
NTSC burst
Color differences (U, V)
Chroma (C)
Setup
S-video-525 (Y'/C3.58)
Composite NTSC encoding
Composite digital NTSC interface (4fSC)
Composite analog NTSC interface
43 - 576 i component video
Frame rate
Interlace
Line sync
Analog field/frame sync
Luma (Y')
Subcarrier
PAL four-frame sequence
PAL burst
Color difference components (U, V)
Chroma (C)
S-video-625 (Y'/C4.43)
Composite PAL encoding
Composite digital PAL interface (4fSC)
Composite analog PAL interface

Colorbars
Colorbar notation
Frequency response
Differential gain (DG)
Differential phase (DP)
Pulse signals
Modulated 12.5T, 20T pulses
46 - 1280 x 720 HDTV
Scanning
Analog sync
Luma (Y')
47 - 1920 x 1080 HDTV
Scanning
Analog sync
Relationship to SMPTE 240M (1035i) scanning
Luma (Y' )

ITU-R (former CCIR)

ITU-R scanning nomenclature
M/NTSC (NTSC)
Audio in NTSC
B,G,H,I/PAL (PAL)
Audio in PAL
PAL-M, PAL-N
SECAM
Multiplexed analog components (MAC)
Summary of parameters
Multistandard consumer equipment

Degenerate analog NTSC and PAL
Coherent subcarrier
Incoherent subcarrier
Nonstandard scanning
SCART interface
Heterodyne (color-under) recording
VHS trick mode playback
Timebase correction (TBC)
Japan
United States
ATSC modulation
Europe
Further reading
Appendices
Cement vs. concrete

True CIE luminance
Conclusion: A plea
B - Introduction to radiometry and photometry
Radiometry
Photometry
Image science
Units
Further reading

Index
2003-09-25
DVAI

by Charles Poynton,
Errata
This page gives you access to the errata for the book, in typographic-quality Acrobat PDF format,
including replacement figures.
The book is currently in its third printing. To determine which printing of the book you have, turn to
the copyright page of the front matter, page iv, and examine the line commencing 2007 2006 ...: The
rightmost digit of that line indicates which printing you have.
Printing
Updated
Acrobat PDF, Typeset quality
2004-10-19
Errata for the first printing
2, 3
2004-10-19
Errata for the second and third printings
The fourth printing hasn't been printed yet! If you see one, or if you suspect an error in the book or in
any Errata document, please inform me by email e-mail.
Charles Poynton - Digital Video and HDTV Algorithms and Interfaces
2005-01-11
http://www.poynton.com/DVAI/errata/index.html [15/01/2005 16:34:18]
Charles Poynton
www.poynton.com
Errata to the first printing,

Digital Video and HDTV Algorithms
and Interfaces
This note contains errata to the first printing of the book Digital
Video and HDTV Algorithms and Interfaces, by Charles Poynton (San
Francisco: Morgan Kaufmann, 2003). I encourage you to make these
corrections in your copy of the book.
This note contains errata for the first printing. To determine which
printing of the book you have, turn to the copyright page of the front
matter (page iv) and examine the line starting 2007 2006 near the
bottom of the page. The rightmost digit of that line indicates which
printing you have. If you have the second printing or a subsequent
printing, I suggest that you obtain the Errata for that printing.
Although all corrections known at 2004-10-19 are reported here, if
you use the appropriate Errata you will avoid having to review errors
reported here that are correctly printed in your copy.
www.poynton.com/DVAI/errata
I revise this note as I discover errors, and I tag each entry with the
date it was posted. Prior to your making these corrections, I suggest
that you check to see if a more recent Errata document has been
posted. When you make the corrections herein, I suggest that you
annotate, on the copyright page of your book, the Errata revision date
that you find at the bottom of this page. Then when you check future
revisions of this Errata document, you can easily identify any additional corrections that need to be made.
In the entries below, I acknowledge individuals who have reported
errors. Entries without attribution are mine, except that numerous
corrections dated 2002-09-22 are thanks to Eric Garci, and several
corrections dated 2002-09-23 are thanks to Alain Fontaine.
2003-02-14: Page xlii of the front matter. Halfway into the second
paragraph, correct Nelson to Nelsen. Sorry, Don!
2003-09-22: Page 7, Figure 1.7. Change PC/Mac VGA to
PC/Mac SVGA.
2004-10-19 Charles Poynton
1 of 10
ERRATA TO DIGITAL VIDEO AND HDTV ALGORITHMS AND INTERFACES
2003-09-23: Page 14. In the first two lines of the caption to

Figure 1.18, exchange middle and bottom.
2003-10-15: Page 24. Among the y-axis legends of Figure 2.5, replace
235 by 240. Also, in the fifth line of the text, correct the spelling of
representation. Thanks to Kunio Kawaguchi.
2003-09-22: Page 59, Table 6.2. Under QXGA, replace 1365 by 1536.
2003-09-22: Page 74. In the paragraph next to the marginal note, in
the fourth line replace columns by rows, and in the fifth line replace
rows by columns.
2003-09-22: Page 84. In Figure 9.3, change E = 0.4 to E = 0.5.
2003-10-15: Page 90. In Figure 10.1, the sketch for 4:1:1 is in error.
A replacement figure, with an augmented caption, is provided on
page 7 of this document. On page 91, in the third sentence of the first
paragraph, change two to three and insert the two words vertical and
in front of horizontal. In the blank line between the first and second
paragraphs, insert this sentence:
In 4:2:0 DV, CB and CR alternate line by line.
Overleaf, on page 92, the lower right corner of Figure 10.2 shows
incorrect chroma subsampling for 4:2:0 DV and 4:1:1 DV. A replacement figure is provided on page 8 of this document. Thanks to Guy
Bonneau, Don Craig, Patrick Law, and Adam Wilt for helping to accurately document this subsampling mess.
2003-09-22: Page 113. In the last line, change (1080 i24, 1080 i30) to
(1080 p24, 1080 p30).
2003-09-24: Page 134. In the second point, change CAT 5 coaxial to
CAT 5e or CAT 6 unshielded twisted pair (UTP). Thanks to Steve
Lampen and Alain Fontaine.
2004-10-15: Page 137. In the last line of the top paragraph, change
measureed from with to measured from. Thanks to Xingbo Wang.
2003-09-23: Page 152. In the penultimate line, and the last line,
delete the minus signs in front of each of the two occurrences of 12 .
2003-09-24: Page 172. In the penultimate line of the second paragraph, the mathematicians name is Lagrange. Thanks to Alex Ball.
2003-02-14: Page 196, Figure 19.1. This figure is mistakenly a duplicate of Figure 19.2. The correct figure is provided below, on page 8 of
this document.
2003-09-23: Page 201. In the penultimate line, change 90 Td to 9 Td.
2003-02-14: Page 202. In the marginal note, correct Robsons first

initial from V to J (for John). You may also wish to correct the index,
on page 684. Thanks to Tom Robson, Johns son.
2003-02-14: Page 205, Figure 20.1. The scotopic curve should be
labelled V(): insert the prime symbol.
_
2004-08-22: Page 205. Historically, y() denoted the luminous efficiency function. In the book, I used the notation Y(), which was
under consideration by the CIE at the time the book was being
written. Subsequent
to publication of the book, the CIE decided to
_
retain the y() notation. In the bottom paragraph of page 205, in the
bottom marginal note on that page, in Figure
20.1, and in the second
_
paragraph of page 206, change Y() to y().
2004-05-07: Page 207. In the line immediately above Equation 20.1,

change STDV to SDTV. Thanks to Xingbo Wang.
2004-08-22: Page 208. In the forthcoming revision of Publ. 15.2 to
15.3, the CIE has changed the coefficients in the equation for L* to
achieve exact C0 and C1 continuity at the breakpoint between the
linear and power-function segments of the function. Insert this
marginal note to the left of the paragraph preceding Equation 20.2:
The fraction (24116)3 is approximately 0.008856; the fraction
(11612)3 is approximately 903.3.
The approximate values were
used in CIE Publ. 15.2 (1986).
2004-08-22: Page 208. In Equation 20.2, change 0.008856 to

(24116)3, and change 903.3 to (11612)3. The new equation is below:
3
116 Y ;
12 Yn
L* =
1
Y 3
116 16;
Yn
Y 24
Yn 116
Eq 20.2
24
Y
116 < Y
2004-08-22: Page 209. In the second line of the first full paragraph of
the page, change 0.008856 or less to (24116)3 or less, that is, less
than about 0.008856.
_
_
_
2004-08-22: Page 216. Historically, the notation x(), y(), and z()
was used for color matching functions. In the book, I used the notation X(), Y(), and Z() that was under consideration by the CIE at
the time
_ the_ book was
_ being written. Ultimately the CIE decided to
retain x(), y(), and z(). In the bottom paragraph of page 216, in the
middle marginal note on that page, in Figure 21.4, in several places on
page 217, and in the caption paragraph
to Figure
21.5 on page 218,
_
_
_
change X(), Y(), and Z() to x(), y(), and z() respectively.
2003-10-12: Page 221. In fourth line of the fourth paragraph, replace

Plankian with Planckian. (The scientists name is spelled Max Planck.)
2004-08-22: Page 225. In Equation 21.3, change 0.008856 to
(24116)3, and change 903.3 to (11612)3; the new equation is below.
(See the notes above for page 208.)
3
116 Y ;
12 Yn
L* =
1
Y 3
116 16;
Yn
Y 24
Yn 116
Eq 21.3
24
Y
116 < Y
2004-08-22: Page 228. In the text immediately below Equation 21.12,

change 0.008856 to (24116)3. To the left of that paragraph, insert this
marginal note:
The fraction (24116)3 is approximately 0.008856; the fraction
841
108 is approximately 7.787.
The approximate values were
used in CIE Publ. 15.2 (1986).
In Equation 21.13, change 7.787 to 841108. The new equation is here:

Eq 21.13
841
16
t+
108
116
2004-10-19: Page 231. In the the first paragraph under the heading
Further reading, and in the first marginal note, change Styles to Stiles.
Thanks to Xingbo Wang.
2003-09-23: Page 250. In second line of the paragraph under
Equation 22.6, replace rows with columns.
2003-02-14: Page 262. In the marginal note at the top of the page,
replace the first D by E .
2003-12-16: Page 266, In Equation 23.7, append the digit 5 to the
end of the range of applicability of the first line of the equation: The
range should read -0.25 L < -0.0045.
2003-09-24: Page 291. In the second line of Equation 24.5, replace
two instances of R with G; in the third line, replace two instances of R
with B. Thanks to Lindsay Steele.
2004-07-03: Page 310. Replace Equation 25.12. Thanks to Andrew

Murray and Masaki Kato. Beware that a previous correction to this
matrix gave incorrect values:
Eq 25.12
601Y '
76.245 149.685
29.07 255 R'
255

1
CB =
43.366 85.136 128.502 255 G'
256
128.502 107.604 20.898

CR
255 B'
2003-09-24: Page 319. Replace Equation 26.9:

219R
Eq 26.9
219G
219B
256
1
256
256
256
0
46.885
464.430
394.150
117.165
0
709
219Y
CB
CR
16
128
128
Replace Equation 26.10:
Eq 26.10
709 Y 16
15.874 255 R
46.742 157.243
219
1
=
128
C
+
25
.
765
86
.
674
112
.439 255 G
B
256
C 128
112.439 102.129 10.310 255 B
In the line immediately below Equation 26.10, replace 601 by 709.

Thanks to James Tyson and Mike Meyers.
2003-09-24: Page 320. Equation 26.12 is in error; replace it with this:
601Y '
219
CB =
CR
0
0
0.099312
0.191700 709
219Y '

0.989854 0.110653 CB
0.072453
0.983398 CR
Eq 26.12
Equation 26.13 is in error; replace it as follows:

709Y '
219
CB =
CR
0
0
601
0.115550 0.207938 219
Y'

1.018640
0.114618 CB
0.075049
1.025327 CR
Eq 26.13
Thanks to Victor Duvanenko for discovering both of these errors.

2003-02-14: Page 363, paragraph 2, line 3. Delete the italicized a.
2003-02-14: Page 379, paragraph 2, line 4. Change 16 seconds to
33.367 seconds. Thanks to Eric Garci.
2004-03-04: Page 383. The paragraph adjacent to Figure 32.2 should
reference that figure, not Figure 31.2. Thanks to Don Orofino.
2003-09-22: Page 432. In the second line of the last paragraph,
change video frame rate to video field rate.
2003-09-22: Page 437. In the bottom paragraph, delete the word in at
the start of the fourth line.
2003-09-24: Page 440. The first word of the caption to Figure 37.12
should read Intrafield. If youre a stickler for detail, change the List of
Figures (page xxxii) accordingly. Thanks to David Salotti.
2003-09-24: Page 442. Figure 37.15 mistakenly has three stages
instead of four; a replacement figure is provided on page 9 of this
document. In the third line of the top paragraph of the page, replace
three by four in two places. Thanks to Mike O'Connell and Billy Biggs.
2003-09-23: Page 448. In the caption to Figure 38.1, replace a luma
block with four luma blocks.
2003-02-14: Page 448, third line from bottom. Insert of betwen array
and the spatial.
2003-10-15: Page 462. Replace the middle paragraph and the associated marginal note with this:
SMPTE 314M defines DV25 and
DV50 for studio use. The Blue Book,
and IEC standards, use the word
decimated instead of discarded.
IEC 61834-1, cited in the margin of
page 422, prescribes the subsampling schemes for consumer DV.
SMPTE 314M declares that in subsampling 4:2:2 to

4:1:1, every other pixel is discarded. Obviously, high
image quality requires that proper filtering be
performed before discarding samples. In DV, CB and CR
samples coincide with luma both horizontally and vertically. However, in the 4:2:0 scheme used in 576i
consumer equipment, CR samples are not sited at the
same locations as CB samples. Instead, CB and CR
samples are sited in line-alternate vertical positions
throughout each field: Each CB sample is centered two
image rows below an associated CR sample.
2003-09-22: Page 506. In second marginal equation, change the first
and third minus signs to plus:
41.259
63.555 858 732 + 2

+
2
13.5
2003-09-22: Page 535. In the third line of the second paragraph,

replace 100% by 75%.
2004-10-15: Page 559. In Table 47.2, change the penultimate table
heading from Contents, left half to Contents, progressive line or interlace left half. Change the rightmost heading from Contents, right half
to Contents, interlace right half. In the table entry at the lower right
corner, change tri/none to none. Thanks to Andrew Steer.
2004-07-06: Page 561. The first two lines of the fourth paragraph are
printed correctly. However, in a previous Errata document, I wrongly
suggested changing five to ten in the first line and one to two in the
second. The proper correction is to Figure 47.2, as noted in the entry
below. (I thought I was wrong once, but I was mistaken!)
2004-07-06: Page 564, Figure 47.2. In the progressive system, change

successive pairs of broad pulses to a single pulse. A replacement figure
is provided on page 10 of this document. Thanks to Eric Garci, Jason
Griffin, and Pierre Berthet.
2004-07-03: Page 573. In the last line of the first paragraph under the
heading Audio in NTSC, change 25 to 75. Thanks to William Hooper.
2003-09-22: Page 582. In the penultimate line, change 262p60.05 to
262/60.05/1:1, and in the last line, change 312p50.08 to
312/50.08/1:1.
2003-09-22: Page 589. In the top marginal note, change 704 to 720.
Adjacent to the second paragraph, add this marginal note:
As an alternative to downsampling, analog scanning can cover 34
of the height of the 4:3 screen to
yield a picture aspect ratio of 16:9.
2004-05-07: Page 601. In the first line of the caption of Table B.1,
change In radiometry to In photometry. Thanks to Xingbo Wang.
Replacement figures
4:4:4
RGB
R0 R1
4:4:4
YCBCR
Y0 Y1
4:2:2
4:1:1
4:2:0 JPEG/JFIF,
4:2:0
4:2:0
Rec. 601
480i DV25; D-7
H.261, MPEG-1
MPEG-2 fr
576i cons. DV
Y0 Y1
Y0 Y1 Y2 Y3
Y0 Y1
Y0 Y1
Y0 Y1
R2 R3
Y2 Y3
Y2 Y3
Y4 Y5 Y6 Y7
Y2 Y3
Y2 Y3
Y2 Y3
G0 G1
CB0 CB1
CB01
CB03
G2 G3
CB2 CB3
CB23
CB47
B0 B1
CR0 CR1
CR01
CR03
B2 B3
CR2 CR3
CR23
CR47
CB03
CB03
CR03
CR03
CR
CB
Figure 10.1 Chroma subsampling. A 22 array of RGB pixels is matrixed into a luma component Y and two color difference components CB and CR . Color detail is reduced by subsampling CB
and CR ; providing full luma detail is maintained, no degradation is perceptible. In this sketch,
samples are shaded to indicate their spatial position and extent. In 4:2:2, in 4:1:1, and in 4:2:0
used in MPEG-2, CB and CR are cosited (positioned horizontally coincident with a luma sample). In
4:2:0 used in JPEG/JFIF, H.261, and MPEG-1, CB and CR are sited interstitially (midway between
luma samples). In the 4:2:0 variant used in consumer 576i DV, CB and CR are vertically sited in
line-alternate fashion in each field (starting with a CR sample sited over the top left luma sample.)
CYR Y CYR Y CYR Y CYR Y

4:2:0 DV
interlaced
CYB Y CYB Y CYB Y CYB Y

CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
Figure 10.2 (lower
right-hand portion)
CY Y Y Y CY Y Y Y
CY Y Y Y CY Y Y Y
100
10
100 m
10 m
1m
Cone cells (3 types)

Photopic vision
1k
Rod cells (1 type)

Scotopic vision
TWILIGHT SUNLIGHT
STARLIGHT
10 k
MOONLIGHT
Absolute scene
luminance, cd m-2
100
Figure 19.1 Luminance

range of vision
4:1:1 DV
interlaced
IN
Figure 37.15 [1, 4, 6, 4, 1] transverse filter implementation
OUT
560
(bottom)
Figure 47.2 1080i and 1080p vertical blanking interval

561
1123
(bottom)
1121
562
1124
22H
20H
5H
INTERLACED SYSTEM, FIRST FIELD/SEGMENT
0V
563
564
565
566
23H
201 2H
6H
5H
567
568
569
8 ...
8 ...
570 ... 582
INTERLACED SYSTEM, SECOND FIELD/SEGMENT
1125
1122 ...1125
45H
41H
5H
PROGRESSIVE SYSTEM, FRAME
0V
20
41
583
21
584
top
image
row
42
top
image
row
560
561
1124
562
1125
1122 ...1125
bottom
image
row
1121
585 ... 1123
22 ...
43 ...
bottom
image
row
10

(Charles A. Poynton) A Technical Introduction To D PDF

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

(Charles A. Poynton) A Technical Introduction To D PDF

Uploaded by

Copyright:

Available Formats

Charles Poynton

Digital Video and

http://www.poynton.com/ (1 di 2) [15/01/2005 16:22:32]

http://www.poynton.com/ (2 di 2) [15/01/2005 16:22:32]

http://www.poynton.com/notes/misc/Momaku.html [15/01/2005 16:22:34]

Charles Poynton - Freelance

Charles Poynton Independent contractor

I am an independent contractor - all me a consultant if you must - specializing in the physics,

Developing, characterizing, modeling, and analyzing algorithms for motion image

http://www.poynton.com/notes/bio/consulting.html (1 di 3) [15/01/2005 16:22:43]

Charles Poynton - Freelance

http://www.poynton.com/notes/bio/consulting.html (2 di 3) [15/01/2005 16:22:43]

Charles Poynton - Freelance

http://www.poynton.com/notes/bio/consulting.html (3 di 3) [15/01/2005 16:22:43]

Charles Poynton - Courses & seminars (upcoming)

Charles Poynton Courses & seminars (upcoming)

I'll present a half-day tutorial HDTV and Digital Cinema

http://www.poynton.com/notes/events/index.html [15/01/2005 16:22:53]

HDTV and Digital Cinema Camera Technology

Palm Springs, Calif.

HDTV and Digital Cinema

http://www.poynton.com/notes/events/20050125_PSP_HPA.html [15/01/2005 16:22:55]

Digital Video and HDTV Algorithms and Interfaces

Digital Video and HDTV

Placing video in the context of computing

Basic concepts of digitization, sampling, quantization, gamma, and filtering

CHARLES POYNTON is an independent contractor specializing in the physics, mathematics, and

Digital Video and HDTV Algorithms and Interfaces

http://www.poynton.com/DVAI/index.html (2 di 2) [15/01/2005 16:23:03]

A Technical Introduction to Digital Video

A Technical Introduction to Digital Video

http://www.poynton.com/notes/TIDV/index.html (1 di 2) [15/01/2005 16:23:11]

A Technical Introduction to Digital Video

Chapter 1, Basic Principles Acrobat PDF format

http://www.poynton.com/notes/TIDV/index.html (2 di 2) [15/01/2005 16:23:11]

A Technical Introduction to Digital Video - Contents

A Technical Introduction to Digital Video

A Technical Introduction to Digital Video - Contents

A Technical Introduction to Digital Video - Contents

Phase response (group delay) 56

A Technical Introduction to Digital Video - Contents

Limitations of 8-bit intensity 113

Fundamentals of vision 116

Color acuity 155

A Technical Introduction to Digital Video - Contents

B'-Y', R'-Y' components 172

Subcarrier regeneration 186

Field rate 199

Frame rate 207

A Technical Introduction to Digital Video - Contents

RGB primary components 215

Frame rate 229

RGB primary components 237

A Technical Introduction to Digital Video - Contents

Component digital 4:2:2 interface 238

Analog electrical interface 247

ITU-R, former CCIR 252

http://www.poynton.com/notes/TIDV/TIDV_toc.html (7 di 8) [15/01/2005 16:23:21]

A Technical Introduction to Digital Video - Contents

http://www.poynton.com/notes/TIDV/TIDV_toc.html (8 di 8) [15/01/2005 16:23:21]

This chapter is a summary of the fundamental concepts

Figure 1.1 Scene, lens, image plane.

light. The active ingredient of photographic film is