Digital Controllers - Anjan Raksit

Digital Controller in a Process Control Loop
AN EL JAN EC R TR AK IC SH JA AL IT D E an KO AV NG d A LK PU INE MI AT R U ER TA A, NI IN VA IN VE G C D R DE HA IA S P T IT A T Y R E TM RJ E N EE T
All the desirable features of a process controller (e.g. anti-integral wind-up, auto/manual modes of operation with bump-less transfer etc.) may be easily incorporated while maintaining the high accuracy and precision of digital systems.
No drift problem is encountered in a digital system.
In a processor based digital controller, rapid switching from one algorithm to another (e.g. a P controller to a PID controller) and automatic tuning of controller parameters are possible.
+ r + Set point _ e Error
ADC
Final mn m Digital DAC Control Controller Element controller Error output Controller sequence sequence output en
Process
ADC equivalent
DAC equivalent
en
mn
Sampler
C (hold capacitor)
Sample & hold
mn
Error
Controller output sequence
0 en
time
0 1 2 3
Error sequence
DAC output m
0 1 2 3
0 2
time
Sampling interval
(ADC conversion time + computation time)
The error computation may be performed digitally if the set-point is available in digital form (say, from digital keyboard) and the measured variable is digitized with the ADC.
Digital System
+ _
rn Set point sequence (digital)
en
Error sequence
Digital Controller
mn
DAC
Final Control Element
Process
Cn ADC output sequence
Proper selection of the sampling interval is necessary for satisfactory operation of the process control loop. A large may lead to unstable operation of the loop (because of the extra lag introduced in the loop), whereas a very small requires a high speed digital hardware (hence high cost) to implement the controller.
The analog controller output is
m = Kp e + b
The controller output at the nth instant is given by
mn = K pen + bn = mn + bn
where, Kp = proportional gain, bn = fixed bias.
Realization of the P controller:
Proportional (P) Controller
bn en
Realization of Digital Controllers (through discrete approximation of analog controllers)
Kp
m'n
mn
Let the integration term be
The integration term at the nth instant is
Proportional-Integral (PI) Controller
1 t m = K p e + o e.dt + b Ti
= m + b, where Ti is the integral time and
I = o e.dt
t
1 t m = K p e + o e.dt Ti
where is the sampling interval
n
I n = o e.dt = o
n
Now,
n o
e.dt = o
( n 1)
e.dt = I n 1
e.dt + n e.dt
for n = 1,2,3,..
Then,
The second term of the above relation represents the area under the curve e for (n 1) t n.
This area may be approximated by the shaded rectangle (called the method of rectangular integration) as
Pl Controller
e
e t =n
e t =( n1)
( n 1)
Thus,
n ( n1)
Using the notation en for
e.dt e t =n
( n 1)
e t = n ,
e t=n
Pl Controller
e t=( n1)
In = In1 + en
Now, output m at the nth instant may be expressed as
Similarly the output at the (n 1)th instant may be expressed as
The difference between these two outputs is
1 mn mn1 = Kp en en1 + ( In In1 ) Ti
Pl Controller
where
In = In1 + en
1 mn mn1 = K p en en1 + ( In In1 ) Ti en mn mn 1 = K p en en 1 + Ti

or,
Substituting the value of (In In-1 ) for rectangular integration, [(In In-1 ) = en ]
mn = en Kp 1+ Kpen1 + mn1 Ti = aoen + a1en1 + mn1, say

ao = K p 1+ Ti
and
where,
Pl Controller
a1 = K p
Controller output at the nth instant: mn = mn + bn
mn= a0en + a1en-1 + mn-1

en-1
Problem: Develop a digital PI Controller using Trapezoidal rule for integration.
Realization of the PI Controller
a1
delay
bn (bias)
en error sequence
ao
m'n
delay
m'n-1
mn Controller output sequence
The controller output is proportional to the derivative of a standard PI controller and it may be expressed as (without bias):
mn = mn mn1 = mn mn1 = ao en + a1en1

Velocity form of controller is useful when the actuator is some kind of adder (integral action), like a stepping motor.
Velocity or incremental form of PI controller
Software Realization of the PI Controller
e1
a1
a0
mp
# include < studio.h> void main (void) { float e = 0, e1, m, mp = 0; float a0, a1, b; float adc (void) ; // digitized error void dac (float m ) ; // analog output a0 = - - - - - - - - ; // Kp [1 + /Ti ] a1 = - - - - -- - - ; // -Kp b = - - - - - - - - - ; // bias
e1
a1
a0
mp
for (;;) // continuous loop { // loop time is the sampling interval e1 = e; e = adc (); mp = mp + a0*e + a1*e1; m = mp + b; // provision for saturation if (m < 0) m = 0; if (m > 100) m = 100; dac (m); } }
e1
a1
a0
mp
float adc (void) // Analog-to-digital conversion { float v; scanf (%f, &v); // for (keyboard) simulation return v; // (to be replaced for actual } // realization) void dac (float m) // Digital-to-analog conversion { printf (%f\n, m); // for (VDU) simulation } // (to be replaced for actual realization)
Let the derivative of error e at the nth instant be
Dn may be approximated using the backward difference algorithm as

e slope = Dn
Dn
Proportional-Derivative (PD) Controller
de m = K p e + Td + b dt
where Td is the derivative time.
de Dn = dt t = n
en en1
en
en-1
approximate slope
(n-1) n time
Thus the controller output at the nth instant is
K p Td Td or m n = K p 1 + e n e n 1 + b n = a o e n + a1 e n 1 + b n
Td ao = K p 1 +
where
PD Controller
e n e n 1 m n = K p e n + Td
+ bn
and
a1 =
K pTd
Problem: Develop a c program for software realization of the PD Controller.
Realization of the PD Controller
en-1
a1
delay
bn (bias)
en (error)
ao
mn Controller output
To avoid derivative action from a sudden change in set-point, the derivative action is generally derived from the measured output. Now, e = r c then,
Thus, the controller output may be expressed as,
dc m = K p e Td +b dt
PD Controller
Provision for anti-derivative kick
de dr dc = , dt dt dt dc = , assuming set-point r is constant. dt
dc m = K p e Td +b dt
using backward difference algorithm, the controller output at the nth instant is
c n c n 1 m n = K p e n Td + bn K p Td K p Td or m n = K p en cn + c n 1 + bn
PD Controller
Provision for anti-derivative kick
ao = K p po = p1 = K pTd
= a o en + p o c n + p1c n 1 + bn
where
= po
K pTd
Realization of PD Controller with anti-derivative kick
Problem: Develop a c program for software realization of the PD Controller with anti-derivative kick
en error
ao
bn (bias)
cn (process output) cn-1
po
mn (Controller output)
delay
p1
1 t de m = K p e + o edt + Td +b dt Ti = m '+ b (say)

where
The controller output (without bias) at the nth instant, using backward difference algorithm, is
Proportional-Integral-Derivative (PID) Controller
1 t de m ' = K p e + o edt + Td dt Ti
In en en 1 mn = K p en + + Td Ti
The controller output (without bias) at the nth instant is
In en en 1 mn = K p en + + Td Ti
The controller output at the (n-1)th instant is
I n1 en1 en2 + Td mn1 = K p en1 + Ti

Subtracting,
I I T mn1 = Kp en en1 + n n1 + d ( en + en2 2en1 ) mn Ti
PID Controller
Using rectangular integration algorithm,
then,
or,
or,
PID Controller
I I T mn1 = Kp en en1 + n n1 + d ( en + en2 2en1 ) mn Ti
I n I n 1 = en
Td mn mn1 = K p en en1 + en + ( en 2en1 + en2 ) Ti
Td K pTd 2Td +1 K p + en2 mn = en 1+ + K p en1 + mn1 Ti

mn = ao en + a1en 1 + a2 en 2 + mn 1
mn = ao en + a1en 1 + a2 en 2 + mn 1
where,
and
PID Controller
Td ao = K p 1 + + Ti 2T a1 = K p d + 1
K pTd
a2 =
and
en
en-2
en-1
a1
a2
mn = mn + bn
ao
m'n
mn = ao en + a1en 1 + a2 en 2 + mn 1
Realization of the PID Controller
m'n-1
+
bn (bias) mn
Problems:
1. Develop a digital PID controller using trapezoidal rule for integration 2. Develop a program in C for software realization of the PID controller 3. Modify the above controller to provide anti-derivative kick feature
PID Controller
Techniques used for anti-integral windup
By saturating or limiting the integral value By resetting the integral value to zero By omitting the integral term By adaptive adjustment of controller parameters
Some anti-integral windup schemes
Saturation 100 r
Stop integration when PI/PID controller internal output (prior to the saturation block) exceeds the saturation limits
+ +
_
Kp
0 100
Process
Multiplier
Kp sTi
hard switching Scheme
1/0
Stop integration if saturated
Scheme for PI Controller
PI control without anti-integral windup
Saturation 100 r +
Performance of anti-integral windup scheme

e n C
Kp
0 100
Process
Multiplier
Kp sTi
hard switching Scheme
1/0
Stop integration if saturated
Scheme for PI Controller
PI control with anti-integral windup
Saturation 100
Reduce integration gradually as PI/PID controller internal output exceeds the saturation limits
+ +
_
Kp
0 100
Process
Kp Ti
1 s
soft switching Scheme
G : a constant
Clegg integrator the integrator is set to zero (reset) when the error crosses zero
reset control reset SW C R error _ +
integrator output
Automatic/Manual modes of Operations
Manual mode means open loop manual control
Manual input or commnd Man e Controller m error Auto Final Control element v
Automatic mode means automatic closed loop operation
Process
Manual input or commnd Man e Controller m error Auto Final Control element
Process
If there is any difference between the controller output and the manual command, a bump occurs in the process output when the switch position is altered. To provide bump-less transfer from auto-to-manual change over, special arrangements may be made for set-point initialization. The manual command is driven to equal the controller output when the loop is in AUTO mode.
Manual input or commnd Man e Controller m error Auto Final Control element
Process
When the loop is in MANual mode, if there is a steady error existing due to any difference between the set-point of the controller and the process output (under manual control), integral term, in case of PI and PID controllers, may wind-up to a large value, and consequently anti-integral wind-up is necessary for such situations. To provide bump-less transfer for all the operating modes, incremental or velocity from of controller is used with an additional integrator.
Scheme for bump-less transfer
Incremental manual command Man Digital incremental controller m'n Auto Integrator m'n
bn (bias) mn
en
The incremental controller output (without bias) at the nth instant may be expressed as
mn = mn mn 1
K pTd Td 2Td = en K p 1 + + en 1 K p + 1 + en 2 Ti = ao en + a1en 1 + a2en 2
for a PID controller
bn (bias) mn
en
The integrator may be represented as
m'n
Integrator
m'n
The integrator output may be represented as
m n = m n m n 1 + m n 1 = m n + m n 1
m'n
m'n
m'n-1
bn (bias) mn
en
The presence of integrator at the output ensures a smooth output variation even when the actual manual command is different from the actual controller output under closed-loop control.
en-2
Realization of the incremental type PID Controller
a2
en-1
a1
Incremental manual Command MAN
bn
en
ao
m'n
AUTO
m'n
mn
Incremental or Velocity Controller
m'n-1
Integrator
Automatic tuning of PID Controllers the Relay autotuner

Suitable for processes with non-zero dead-time.
This is based on a special technique for determining the critical gain Kc and critical time period Tc of the process loop. Kc is the gain margin of the process loop.
Controller parameters Kp, Ti and Td are calculated according to Ziegler Nichols (Z-N) rule for a stable time response. Z-N settings
Controller Kp Ti Td P Controller 0.5 K c PI Controller 0.45 K c 0.6 K c Tc/1.2 Tc/2 PID Controller Tc/8
gain margin: 2

The critical gain Kc and critical time period Tc are determined from an experiment with relay (switching element) feedback.
set point r=0
The relay control provides ON / OFF control of the process. The input r is set to zero.
The output c oscillates around a mean value of zero (limit-cycle oscillations). The process is driven by a square wave of amplitude D.
+ e D
0 -D
Process
Relay
set point r=0
Assuming the process to be a low-pass system, the process output c contains mainly the fundamental component.
Thus the error signal e becomes sinusoidal,
+ e D
0 -D
Process
Relay
e = A sin t

set point r=0 e D v
+
0 -D
Process
e = A sin t
Relay
The relay output v may be found out as follows:

v v D
-D
time
-D
time
-A

v
set point r=0 e D v
+
0 -D
Process
Relay
time
-D
The Fourier series of the relay output (v) may be expressed as:
4D 1 1 v= sin t + sin 3t + sin 5t + ....... 3 5
The process practically attenuates all higher harmonics other than the fundamental. Then the process gain at frequency becomes
output amplitude A A G ( ) = = = 4D 4D input amplitude

set point r=0 e D v
+
0 -D
Process
Process gain at frequency :
G ( ) =
Relay
output amplitude A A = = 4D 4D input amplitude
Now, to maintain steady oscillations at = c, the loop gain is 1 (considering negative feedback). Thus the gain of the relay controller (i.e. the critical gain) at = c is
Kc =
G ( c )
4D A
as K c . G ( c ) = 1
Also G (c) = , as relay phase shift is zero.

set point r=0 e D v
Thus by knowing the relay amplitude D and by measuring the amplitude A of the process output c, critical gain Kc may be determined K c =
+
0 -D
Process
Relay
4D . A
Tc may be estimated by measuring the frequency of the output oscillation Tc =
2 . c
Block diagram of the Relay autotuner
(The Satt Control Autotuner by Satt Control, Sweden)
Manual command M r + + _ e PID Controller A v Process D T Mode switch 0 D Relay
M A T
Manual position Auto position Tune position
At first, the process is brought to equilibrium state ( zero error), by setting a constant control signal in manual mode. The tuning is then activated by pushing the mode switch to tune position.
Modified relay characteristic for a non-zero set-point

v v Controller output for zero error
The relay amplitude D is initially set to 10% of the controller output-range.
This amplitude is adjusted after one and a half period to give oscillation of 2% of the mean output. This ensures minimum disturbance at the process output due to tuning.
This adjustment is done by measuring the change in output during the first one and a half period. The modified relay amplitude is stored for the next tuning operation.
D = 0.1vmax e 0 0 t Manual setting
The system is automatically switched to Auto mode after estimating the critical gain Kc and critical time period Tc during first 5 period of oscillation. The parameters of PID controller (viz. Kp, Ti and Td ) are determined from Kc and Tc according to Z-N rule.
Satt Control Autotuner
Manual command M r + _
PID Controller
Process
T Mode switch
0 D
Relay
M A T
Manual position Auto position Tune position

Digital Controllers - Anjan Raksit

Uploaded by

Document Information

Original Description:

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

Digital Controllers - Anjan Raksit

Uploaded by

Copyright:

Available Formats

Digital Controller in a Process Control Loop

No drift problem is encountered in a digital system.

Sample & hold

Controller output sequence

(ADC conversion time + computation time)

rn Set point sequence (digital)

Final Control Element

Cn ADC output sequence

The analog controller output is

The controller output at the nth instant is given by

Realization of the P controller:

Realization of Digital Controllers (through discrete approximation of analog controllers)

The analog controller output is

Let the integration term be

The integration term at the nth instant is

Realization of Digital Controllers (through discrete approximation of analog controllers)

= m + b, where Ti is the integral time and

Using the notation en for

Now, output m at the nth instant may be expressed as

Similarly the output at the (n 1)th instant may be expressed as

The difference between these two outputs is

1 mn mn1 = Kp en en1 + ( In In1 ) Ti

1 mn mn1 = K p en en1 + ( In In1 ) Ti en mn mn 1 = K p en en 1 + Ti

mn = en Kp 1+ Kpen1 + mn1 Ti = aoen + a1en1 + mn1, say

Controller output at the nth instant: mn = mn + bn

mn= a0en + a1en-1 + mn-1

Problem: Develop a digital PI Controller using Trapezoidal rule for integration.

mn Controller output sequence

mn = mn mn1 = mn mn1 = ao en + a1en1

Velocity or incremental form of PI controller

The analog controller output is

Let the derivative of error e at the nth instant be

Dn may be approximated using the backward difference algorithm as

Realization of Digital Controllers (through discrete approximation of analog controllers)

where Td is the derivative time.

Thus the controller output at the nth instant is

Problem: Develop a c program for software realization of the PD Controller.

Thus, the controller output may be expressed as,

de dr dc = , dt dt dt dc = , assuming set-point r is constant. dt

Realization of PD Controller with anti-derivative kick

cn (process output) cn-1

The analog controller output is

1 t de m = K p e + o edt + Td +b dt Ti = m '+ b (say)

Realization of Digital Controllers (through discrete approximation of analog controllers)

The controller output (without bias) at the nth instant is

I n1 en1 en2 + Td mn1 = K p en1 + Ti

I I T mn1 = Kp en en1 + n n1 + d ( en + en2 2en1 ) mn Ti

Using rectangular integration algorithm,

I I T mn1 = Kp en en1 + n n1 + d ( en + en2 2en1 ) mn Ti

Td mn mn1 = K p en en1 + en + ( en 2en1 + en2 ) Ti

Td K pTd 2Td +1 K p + en2 mn = en 1+ + K p en1 + mn1 Ti

Realization of the PID Controller

hard switching Scheme

Stop integration if saturated

Scheme for PI Controller

PI control without anti-integral windup

Performance of anti-integral windup scheme

hard switching Scheme

Stop integration if saturated

Scheme for PI Controller

PI control with anti-integral windup

soft switching Scheme