digital control 1
TRANSCRIPT
Digital Control SystemsCCS-532
Lecture - 10
Ref: Chapter 13: Nise, N. S. Control System Engineering
Chapter 13: Dorf, R. C. & Bishop, R. H. , Modern Control Systems
Chapter 2, 11: Benjamin C. Kuo. , Automatic Control Systems
Dr Pavan ChakrabortyIIIT-Allahabad
Indian Institute of Information Technology - Allahabad
Digital Control Systems
Digital/Discrete Control
• More useful for computer systems• Time is discrete
– denoted k instead of t
• Main tool is z-transform
– f(k) ® F(z) , where z is complex– Analogous to Laplace transform for s-domain
• Root-locus analysis has similar flavour– Insights are slightly different
0
)()()]([k
kzkfzFkfZ
Figure 13.1Conversion of antenna azimuth position control system from:a. analog control tob. digital control
Indian Institute of Information Technology - Allahabad
1. Reduced Cost,
2. Flexibility in response to design changes
3. Noise immunity
Figure 13.2a. Placement of the digital computer within the loop;b. Detailed block diagram showing placement of A/D and D/A converters
Indian Institute of Information Technology - Allahabad
Figure 13.3 Digital-to-analog conver ter (DAC)
Indian Institute of Information Technology - Allahabad
ADC – 2 step process and is not instantaneous.1. Converted to a sampled signal.2. Converted to a sequence of binary numbers
DAC & ADC
c. conversion ofsamples to digitalnumbers
Figure 13.4Steps in ADC
a. analog signal;
Sampling Rate must be at least twice the bandwidth of the signal, or else there will be distortion.
This minimum sampling frequency is called the “Nyquist”sampling rate.
Indian Institute of Information Technology - Allahabad
b. analog signal after sample-and-hold;
Indian Institute of Information Technology - Allahabad
Figure 13.5 Two views of uniform-rate sampling:
a. Switch opening and closing;
k
WT TkTtukTtutftstftfW
)()()(*
b. Product of time waveform and sampling waveform
Indian Institute of Information Technology - Allahabad
k
WT TkTtukTtutftstftfW
)()()(*
interval. sampling during constant )(tfTTW
])()([)(*
k
WT TkTtukTtutftstftfW
Since the Eq. (above) is a product of 2 time fn., taking the LAPLACE TRANSFORM in order to find a TRANSFER FUNCTION is not simple.
Simplification )()( kTftf
Small TW
kTs
k
sT
k
sTkTskTs
T es
ekTf
s
e
s
ekTfsF
ww
W
1)(*
Indian Institute of Information Technology - Allahabad
kTs
k
sT
k
sTkTskTs
T es
ekTf
s
e
s
ekTfsF
ww
W
1)(*
kTs
k
ww
T es
sTsT
kTfsFW
......!2
11
)(
2
*
expantion, series with Replacing sTwe
Small TW
kTsw
k
kTs
k
wT eTkTfe
s
sTkTfsF
W
)(*
Inverse Laplace
Indian Institute of Information Technology - Allahabad
Fig13.6 b. Product of
time waveform and sampling waveform
Figure 13.6Model of sampling with a uniform rectangular pulse train
Indian Institute of Information Technology - Allahabad
Figure 13.7 Ideal sampling and the zero-order hold
Indian Institute of Information Technology - Allahabad
Basic Concepts• Consider a sequence of values: {xk : k = 0,1,2,... }• These may be samples of a function x(t), sampled at
instants t = kT; thus xk = x(kT).
• The Z transform is simply a polynomial in z having the xk as coefficients:
0
)(k
kzkfkfZzF
Indian Institute of Information Technology - Allahabad
Z Transform
Uses of Z-transform in this course
• Analysis of Discrete & Sampled-data Systems
• Transfer Function & Block Diagram Representations
• Exploit the z-plane– Dynamic Analysis– Root Locus Design
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Fundamental Functions
• Define the impulse function: {k} = {1, 0, 0, 0,....}
1)( kZz
• Define the unit step function: {uk} = {1, 1, 1, 1,....}
11
11
0
z
z
zzuZzU
k
kk
(Convergent for |z| < 1)Indian Institute of Information Technology - Allahabad
Z-transform of
Consider: constant = and , akau kk ,2,1,0
az
z
azazaz
azzauZk
k
k
kkk
for 11
1 11
0
1
0
Recall:Geometric SeriesRecall:Geometric Series
11
1
0
xx
xk
k for
ka
Indian Institute of Information Technology - Allahabad
Z-transform of
Consider: constant = and , akkau kk ,2,1,0
221
1
121
11
1
1
11
1}{
}{}{}{
az
az
az
az
zaza
azda
daaZ
da
da
ada
daZkaaZkaZuZ
k
kkkk
MATLAB: See ztrans and iztrans commands.
We will use these two functions as generating functions.
MATLAB: See ztrans and iztrans commands.
We will use these two functions as generating functions.
kk kau
Indian Institute of Information Technology - Allahabad
Generating common functions using
• a=1 yields the unit step function
• a=exp(+/-bT) yields the exponential function
ka
1z
z
bTez
z
Indian Institute of Information Technology - Allahabad
Generating the Z-transform of common functions
• yields sinusoids
)}sin({)}cos({
cos2
sincos22
22
*
*
*
*
kTrjZkTrZ
rzTrz
TjrTrzz
rrerez
rezz
rez
rez
rez
zTjTj
Tj
Tj
Tj
Tj
Note: To ensure unambiguous poles on z-plane
wT<pi or
2pi/T=ws > 2w
the “sampling theorem”
Note: To ensure unambiguous poles on z-plane
wT<pi or
2pi/T=ws > 2w
the “sampling theorem”
Tjera
Indian Institute of Information Technology - Allahabad
Generating common functions
using• a=1 yields the unit ramp function
•
21
z
TzkTZkTZ
kk kau
See (Appendix G of Dorf and Bishop) for the Properties of z-Transforms See (Appendix G of Dorf and Bishop) for the Properties of z-Transforms
2}{bT
bTbkT
ez
TzekTeZ
bTea
Indian Institute of Information Technology - Allahabad
Z-Transform: Forward Shift or1-stage advance
0
001
10
)1(1
0
1)1(1
011
)(
}{
zuzzU
zuzuzuz
zuzzuz
zzuzuuZ
m
mm
m
mm
k
kk
k
kk
k
kkk
Indian Institute of Information Technology - Allahabad
Z-Transform: Forward Shift or2-stage advance
02
12
10
112
)(
)(
}{}{
uzzuzUz
zuzuzzUz
zuuzZuZ kk
In general:In general:
1
0
)(}{n
m
mm
nnnk zuzzUzuZ
Indian Institute of Information Technology - Allahabad
Delay/Shift Property• Let y(t) = x(t-T) (delayed by T and truncated at t = T)
yk = y(kT) = x(kT-T) = x((k-1)T) = xk-1 ; y0 = 0
1
11
)(k
kk
k
kkk zxzyyZzY
• Let j = k-1 ; k = j + 1
)()( 1
0
1
0
1 zXzzxzzxzYj
jj
j
jj
• The values in the sequence, the coefficients of the polynomial, slide one position to the right, shifting in a zero.
Indian Institute of Information Technology - AllahabadIndian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Some Important Teorems of the z-Transform
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
No PolesNo Poles
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Table 13.2 z-transform theorems
Indian Institute of Information Technology - Allahabad
The Laplace Connection
• Consider the Laplace Transforms of x(t) and y(t): sXeTtxLtyLsY Ts
• Equate the transform domain delay operators:Tsez 1 Tsez
• Examine s-plane to z-plane mapping . . .
Indian Institute of Information Technology - Allahabad
zeTs
Indian Institute of Information Technology - Allahabad
S-Plane to Z-Plane Mapping
Anything in the Alias/Overlay region in the S-Plane will be overlaid on the Z-
Plane along with the contents of the strip between s=± j/T. In order to avoid aliasing, there must be nothing in this region, i.e. there must be no signals present with radian frequencies higher than f /T, or cyclic frequencies higher than f = 1/2T. Stated another way, the sampling frequency must be at least twice the highest frequency present (Nyquist rate).
Figure 13.13Mapping regions of the s-plane onto the z-plane
Indian Institute of Information Technology - Allahabad
Mapping Poles and ZerosA point in the Z-plane r e jwill map to a point in the S-plane according to:
T
rs
lnRe
Ts
Im
Conjugate roots will generate a real valued polynomial in s of the form:
22 2 nnss
2ln
1
r
Tn
T
r
n ln
Indian Institute of Information Technology - Allahabad
Example 1: Running Average Algorithm
Transfer Function
4
23321
4
1
4
1
z
zzzzX
zzzzXzY
4
23
4
1
z
zzzz
X
Y
4321
kkkkk
xxxxy
Note: Each [Z-1] block can be thought of as a memory cell, storing the previously applied value.
(Non-Recursive)
Z Transform
Indian Institute of Information Technology - Allahabad
Block Diagram
Example 2: Trapezoidal Integrator
211
Txxyy kkkk
2
11 TzXzzXzYzzY
21
1
21
11
1 T
z
zzX
T
z
zzXzY
(Recursive)
Z Transform
1
1
2 z
zT
zX
zY
Indian Institute of Information Technology - Allahabad
Block Diagram Transfer Function
Ex. 2 (cont) Block Diagram Manipulation
Intuitive Structure
Equivalent Structure
Explicit representation of xk-1 and yk-1 has been lost, but memory element usage has been reduced from two to one.
Indian Institute of Information Technology - Allahabad
Ex. 2 (cont) More Block Diagram Manipulation
1
1
2 z
zT
zX
zY
Indian Institute of Information Technology - Allahabad
Note that the final form is equivalent to a rectangular integrator with an additive forward path. In a PI compensator, this path can be absorbed by the proportional term, so there is no advantage to be gained by implementing a trapezoidal integrator.
Table 13.1 Partial table of z- and s-transforms
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Inverse z-Transform
Y(z) is y(kT); not y(t)
1. Partial-fraction expansion.
2. Power-series / Synthetic Division method.
3. The inverse formula
Common approaches for taking the inverse z-transform
Partial Fraction Expansion: decompose U(z) into a linear combination of easily inverted z-transforms.
• Power-series method / Synthetic Division: divide polynomials to obtain a power series in z-1.
MATLAB’s “residue” command computes the partial fraction expansion.
MATLAB’s “residue” command computes the partial fraction expansion.
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
1.Partial-fraction expansion.
If Y(z) has at least one zero at z=0, the paritial-fraction expansion of Y(z)/z should be performed.
Partial fraction expansion method
.1
)(
)()(
1
rr
r
m
jj
r
rj
i i
i
mz
a(z)r
a(z)i
Czz
zB
zz
zA
za
zbzU
r
ty multiplici has root
of roots repeated all includes
of roots distinct all includes
where
The coefficients (residues) {A}, {B} and C can be calculated by hand for low-order systems or by expanding U(z)/z in a partial fraction expansion using “residue” and multiplying by z in high-order systems.
The coefficients (residues) {A}, {B} and C can be calculated by hand for low-order systems or by expanding U(z)/z in a partial fraction expansion using “residue” and multiplying by z in high-order systems.
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Power-series method / Synthetic division method
Represent U(z) as a power series in z-1:
Since,
22
110)( zuzuuzU
otherwise
for
0
1][1 mk
mkzZ m
]2[]1[][ 210 kukukuuk
N.B.: Synthetic division method is useful for a limited number of values in a sequence or as a quick check on other methods
N.B.: Synthetic division method is useful for a limited number of values in a sequence or as a quick check on other methods
Indian Institute of Information Technology - Allahabad
2. Power-series method.
2-171
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
3. The inverse formula
Indian Institute of Information Technology - Allahabad
Ap
plicati
on
of
the z
-Tra
nsfo
rm t
o t
he S
olu
tion
of
Lin
ear
Diff
ere
nce
Eq
uati
on
Indian Institute of Information Technology - Allahabad
The Transfer FunctionDefinition: The transfer function of a system is the ratio of the transform of the output of the system to to that of its input with zero initial conditions.
nnn
nn
nn
nn
nn
nn
azaz
bzbzb
zaza
zbzbb
zE
zU
zEzbzEzbzEb
zUzazUzazU
11
210
11
110
110
11
1)(
)(
)()()(
)()()(
Using z-1 as the delay operator,
nknkknknkk ebebebuauau 11011
Indian Institute of Information Technology - Allahabad
Relation between transfer function and pulse response
Consider H(z)=U(z)/E(z) with E(z) =1that is e[k]=1 for k=0 and zero elsewhere.Then, U(z) = H(z)
The z-transform of the unit-pulse response of a discrete system is its transfer function.
Indian Institute of Information Technology - Allahabad
General input-output relation for linear, stationary discrete systems
U(z) = H(z) E(z)
H(z) is a rational function (ratio of two polynomials) of
the complex variable z.
For H(z) = a(z)/b(z):• Values of z for which a(z) = 0 are called zeros of H(z).
• Values of z for which b(z) = 0 are called poles of H(z).
• If z0 is a pole of H(z) and (z-z0)p.H(z) has neither a
pole or zero at z0, H(z) is said to have a pole of order p
at z0.
Indian Institute of Information Technology - Allahabad
Figure 13.8Sampled-datasystems:a. Continuous;b. Sampled input;c. sampled inputand output
Indian Institute of Information Technology - Allahabad
Pulse Transfer Function
nTtnTrtrn
0
* nTtgnTrtcn
0
k
k
zkTczC
0
Indian Institute of Information Technology - Allahabad
nTtnTrtrn
0
*
nTtgnTrtcn
0
k
k
zkTczC
0
nTkTgnTrkTcn
0
k
k n
zTnkgnTrzC
0 0
)(
)()(00
0 0
zRzGznTrzmTg
zmTgnTrzC
n
mm
m
nm
nm n
Letting m=k-n we find
Figure 13.9 Sampled-data systems and their z-transforms
Indian Institute of Information Technology - Allahabad
Block Diagram Reduction
Figure 13.10Steps in blockdiagram reductionof a sampled-data system
Indian Institute of Information Technology - Allahabad
Figure 13.11 Digital system for Skill-Assessment Exercise 13.4
Indian Institute of Information Technology - Allahabad
Problem: Find T(z)=C(z)/R(z) Digital system for Skill-Assessment Exercise 13.4
Indian Institute of Information Technology - Allahabad
Block diagram descriptions of discrete systemsTransfer function representations of discrete systems permit the use of block diagrams to describe discrete systems in a manner analogous to continuous system representation.
• Add transfer functions in parallel.
• Multiply transfer functions in series (cascade).
• A simple feedback loop reduces to the forward path transfer function divided by one minus the open-loop transfer function.
• Block diagram manipulation and Mason’s rule apply without change.
Indian Institute of Information Technology - Allahabad
The discrete convolution sumConsider a linear, stationary discrete system.
• If its response to a unit pulse is h[k], then its response to a pulse of amplitude e0 is e0*h[k] since the system is linear.
• Since the system is stationary a delay of the input will cause an equal delay in the response.
• The effect of a sequence of pulses is the sum of their individual effect. For an infinite sequence:
k. sample at output the on i sample at
pulse input an of effect the is where ik
iikik
h
heu
N.B. k >i for a
causal system
N.B. k >i for a causal system
Indian Institute of Information Technology - Allahabad
State space descriptions of discrete systems
Any linear, stationary discrete system can be represented in the following format:
dimension. eappropriat of matrices are J and H,,
and vector, state the is x
vectors output and input the are and where
yu
kuJkxHky
kukxkx
][][][
][][]1[
N.B. The matrices A,B,C, and D are often used to describe the state space representation of a system.
N.B. The matrices A,B,C, and D are often used to describe the state space representation of a system.
Indian Institute of Information Technology - Allahabad
Example of a state space representation
e[k][k] x] [kx
[k] x] [kx
][k xe[k] [k] x] [kx
e[k][k]]-x[k x )-(z
(z)/E(z) X
12
21
211
1121
1
1
12
21
1 or
Consider G(z)=U(z)/E(z)=K(z+1)/(z2-1)Let:
Then,
Indian Institute of Information Technology - Allahabad
[k]xK[k]x K
[k]xK][kxKu[k]
)K(z(z)U(z)/X
**
**
21
11
1
1
1
or
Example of a state space representation (cont.)
][0][][
][1
0][
01
10]1[
2
1
2
1
2
1
kekx
xKKku
kekx
xk
x
x
Indian Institute of Information Technology - Allahabad
Figure 13.12Computer-controlled torches cut thicksheets of metal used in construction
© BlairSteitz/Photo Researchers.
Indian Institute of Information Technology - Allahabad
Correspondence with Continuous Signals
Consider the discrete signal to be generated by sampling a continuous signal, We can then exploit our knowledge of the s-plane
features by transferring them to the z-plane.
Indian Institute of Information Technology - Allahabad
Mapping from the s-plane to the z-plane
Consider an arbitrary point on the s-plane;
js Which maps onto the z-plane as:
TjTTj eeez
We will consider several lines on the s-plane and their
map onto the z-plane. Recall ωT < prevents ambiguity, I.e., permits a one-to-one mapping between the two planes.
Indian Institute of Information Technology - Allahabad
Figure 13.13Mapping regions of the s-plane onto the z-plane
Indian Institute of Information Technology - Allahabad
S-plane real axis s z e T
• For Re{s}>0 : Re{z}>1
• For Re{s}<0 : 0<Re{z}<1
• For s=0: z=1
Real axis of s-plane maps to positive-real axis of z-planeReal axis of s-plane maps to positive-real axis of z-plane
Indian Institute of Information Technology - Allahabad
S-plane imaginary axis &arbitrary point in primary strip
• For s=j ω : z=exp{j ω T}A line of unit length making an angle of wT radians withthe real axis.
• For s = j ω s/2=jπ/T : z =e xp{jπ} = -1
• For s=a+j ω : z = exp{aT+j ω T} = exp{aT}*exp{j ω T}
A line of length, exp{aT}, making an angle of ω T radianswith the real axis.
Note that - π < ω T<π for there to be a one-to-onemap between the s-plane and the z-plane, I.e., -π/T< ω <π/Tor - ω s/2< ω < ω s/2. (Sampling Theorem) The region on the s-plane for ω s/2< ω < ω s/2 is called the primary strip.
Note that - π < ω T<π for there to be a one-to-onemap between the s-plane and the z-plane, I.e., -π/T< ω <π/Tor - ω s/2< ω < ω s/2. (Sampling Theorem) The region on the s-plane for ω s/2< ω < ω s/2 is called the primary strip.
Indian Institute of Information Technology - Allahabad
S-plane line of constant damping
s j
z e e e
n n
sT T j Tn n
1 2
1 2
11
2
2
1 2
n n
j
TT
z e e
Consider:
Let:
Then,Logarithmic spirals on the z-plane for constant damping ratio.
Logarithmic spirals on the z-plane for constant damping ratio.
Indian Institute of Information Technology - Allahabad
S-plane line of constant natural frequency
s j
z e e e
n n
sT T j Tn n
1 2
1 2
z e e for
T
n nT j T
n
1 2
0 1 and
Again consider:
Contours of constant damping ratio and natural frequency on the z-plane are drawn by the MATLAB command ‘zgrid’.
Contours of constant damping ratio and natural frequency on the z-plane are drawn by the MATLAB command ‘zgrid’.
Indian Institute of Information Technology - Allahabad
-20 -18 -16 -14 -12 -10 -8 -6 -4 -2 0-20
-15
-10
-5
0
5
10
15
20S-PLANE USING "SGRID" - lines of constant damping ratio & natural frequency
LINES OF CONSTANT DAMPING RATIO
0.1
0.9
LINES OF CONSTANTNATURAL FREQUENCY
10
5
Indian Institute of Information Technology - Allahabad
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1Z-PLANE USING "ZGRID" - lines of constant damping ratio & natural frequency
LINES OFCONSTANTNATURAL FREQUENCY
LINES OF CONSTANTDAMPING RATIO
PI/10T 9PI/10T
0.1
0.3
Upper plane is divided into 10 segments so wn*T=pi/10 or
wn=pi/10T per segment.
Upper plane is divided into 10 segments so wn*T=pi/10 or
wn=pi/10T per segment.
Indian Institute of Information Technology - Allahabad
Figure 13.14Finding stability ofa missile controlsystem:a. missile;b. conceptual blockdiagram;c. block diagram;d. block diagramwith equivalent singlesampler
Indian Institute of Information Technology - Allahabad
Figure 13.15Digital system for
Example 13.7
Indian Institute of Information Technology - Allahabad
Internal & external stability
• Internal stability is concerned with the responses of all the internal (state) variables of a system.
• External stability is concerned with the response of the output variables of a system such as described by the transfer function or impulse response model.
• They differ in that some of the internal modes of the system may not be connected to the input and output of a given system.
Indian Institute of Information Technology - Allahabad
Bounded-Input Bounded-Output (BIBO) stability
For external stability a common definition of an appropriate response is that for every bounded input the output should also be bounded.
A necessary and sufficient condition for BIBO stability is
iikh
Note: A rational transfer function can be expanded in a partial fraction expansion so its pulse response will be the sum of its terms. Thus, if all poles are inside the unit circle, the system is stable. If at least one pole is on or outside the unit circle the system is not BIBO stable.
Note: A rational transfer function can be expanded in a partial fraction expansion so its pulse response will be the sum of its terms. Thus, if all poles are inside the unit circle, the system is stable. If at least one pole is on or outside the unit circle the system is not BIBO stable.
Indian Institute of Information Technology - Allahabad
Proof:
.
.
i-k
k
i
h if bounded is output the Thus,
u Then,
all for e Let
ikikiiki hMhehe
iM1) Sufficiency
2) Necessity
BIBO not is system the true is condition the unless Thus,
u
:is at output The input. thisApply
h for
h for e :input bounded the Consider
0
i-
i-i
ii
i i
i
iii
i
i
hh
hhe
k
h
h
2
0
00
0
Indian Institute of Information Technology - Allahabad
MATLAB commands defining discrete systems as objects
• tfSYS = tf(NUM,DEN,TS) creates a discrete-time transfer function with
sample time TS (set TS=-1 if the sample time is undetermined).
• zpk SYS = zpk(Z,P,K,Ts) creates a discrete-time ZPK model with sample
time Ts (set Ts=-1 if the sample time is undetermined).
• ss SYS = ss(A,B,C,D,Ts) creates a discrete-time SS model with sample
time Ts (set Ts=-1 if the sample time is undetermined).
Indian Institute of Information Technology - Allahabad
Table 13.3Routh table for Example 13.8
Indian Institute of Information Technology - Allahabad
Figure 13.16Digital system forSkill-Assessment
Exercise 13.5
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Figure 13.18Constant damping ratio, normalizedsettling time, and normalized peak time plots on the z-plane
Indian Institute of Information Technology - Allahabad
Figure 13.19The s-plane
sketch of constant
percentovershoot line
Indian Institute of Information Technology - Allahabad
Figure 13.20Generic digital
feedback controlsystem
Indian Institute of Information Technology - Allahabad
Figure 13.21Digital feedback
control forExample 13.10
Indian Institute of Information Technology - Allahabad
Root Locus analysis of Discrete Systems
• Stability boundary: |z|=1 (Unit circle)
• Settling time = distance from Origin
• Speed = location relative to Im axis– Right half = slower– Left half = faster
Effect of discrete poles
|z|=1
Longer settling time
Re(s)
Im(s)
Unstable
Stable
Higher-frequencyresponse
Tsez :Intuition
Figure 13.22Root locus
for the systemof Figure
13.21
Indian Institute of Information Technology - Allahabad
Figure 13.23Root locus for the system of Figure
13.21 with constant
0.7 damping ratiocurve
Indian Institute of Information Technology - Allahabad
Figure 13.24Sampled step
response of thesystem of
Figure 13.21 withK = 0.0627
Indian Institute of Information Technology - Allahabad
Figure 13.25a. Digital controlsystem showing
the digital computerperforming
compensation;b. continuous
systemused for design;c. transformed
digitalsystem
Indian Institute of Information Technology - Allahabad
Figure 13.26Closed-loop
responsefor the
compensatedsystem of Example
13.12 showing effect
of three differentsampling
frequencies
Indian Institute of Information Technology - Allahabad
Figure 13.27Block diagram
showing computeremulation of a digital
compensator
Indian Institute of Information Technology - Allahabad
Figure 13.28 Flowchart for a second-orderdigital compensator
Indian Institute of Information Technology - Allahabad
Figure 13.29 Flowchart to implement
Indian Institute of Information Technology - Allahabad
Figure 13.30 Antenna control system:a. analog implementation;b. digital implementation
Indian Institute of Information Technology - Allahabad
Figure 13.31Analog antenna azimuth position
control system converted to a digital system
Indian Institute of Information Technology - Allahabad
Figure 13.32Root locussuperimposed
over constant damping ratio curve
Indian Institute of Information Technology - Allahabad
Figure 13.33Sampled step
response of theantenna azimuthposition control
system
Indian Institute of Information Technology - Allahabad
Figure 13.34Simplified block
diagram of antennaazimuth control
system
Indian Institute of Information Technology - Allahabad
Figure 13.35Closed-loop digitalstep response forantenna controlsystem with a leadcompensator.
Indian Institute of Information Technology - Allahabad
Figure 13.36Flowchart for digital lead compensator
Indian Institute of Information Technology - Allahabad
Figure P13-1 (p. 839)
Figure P13.2
Figure P13-3 (p. 840)
Figure P13-4 (p. 840)
Indian Institute of Information Technology - Allahabad
Figure P13.5
Indian Institute of Information Technology - Allahabad
Figure P13.6
Indian Institute of Information Technology - Allahabad
Figure P13.7
Indian Institute of Information Technology - Allahabad
Figure P13.8
Indian Institute of Information Technology - Allahabad
Figure P13.9Simplified blockdiagram for robot
swing motion
Indian Institute of Information Technology - Allahabad
Figure P13.10Simplified block
diagram of a floppydisk drive
Indian Institute of Information Technology - Allahabad
z-Transforms of Common Functions
Name f(t) F(z)
Impulse
Step
Ramp
Exponential
Sine
1
1z
z
2)1( z
z
aez
z
1)(Cos2
Sin2 zaz
az
1)( tf
ttf )(
atetf )(
)sin()( ttf
F(s)
1
s
1
2
1
s
as 1
22
1
s
00
01)(
t
ttf
System ID for Admission Control
M(z)
G(z) N(z) S(z)
Controller Notes Server
Sensor
R(z)+
E(z)U(z)
-
Q(z)
ARMA Models
Control Law
Transfer Functions
zz
zK
cz
dzd
az
zbzGzSzN i 1
1)()()(
1
10
1
0
Open-Loop:
zz
zKzG
cz
dzdzS
az
zbzN
i 1
1)(
)(
)(
1
10
1
0
)()1()(
)1()()1()(
)()1()(
101
01
teKtutu
tqdtqdtmctm
tubtqatq
i
Root Locus Analysis of Admission Control
Predictions:•Ki small => No controller-induced oscillations•Ki large => Some oscillations•Ki v. large => unstable system (d=2)•Usable range of Ki for d=2 is small
Experimental Results
Control(MaxUsers)
Response(queue length)
Good
Slow
Bad
Useless
Advanced Control Topics
• Robust Control– Can the system tolerate noise?
• Adaptive Control– Controller changes over time (adapts)
• MIMO Control– Multiple inputs and/or outputs
• Stochastic Control– Controller minimizes variance
• Optimal Control– Controller minimizes a cost function of error and control energy
• Nonlinear systems– Neuro-fuzzy control– Challenging to derive analytic results
Issues for Computer Science
• Most systems are non-linear– But linear approximations may do
• eg, fluid approximations
• First-principles modeling is difficult– Use empirical techniques
• Control objectives are different– Optimization rather than regulation
• Multiple Controls– State-space techniques– Advanced non-linear techniques (eg, NNs)
Selected Bibliography
• Control Theory Basics– G. Franklin, J. Powell and A. Emami-Naeini. “Feedback Control of Dynamic Systems, 3rd ed”.
Addison-Wesley, 1994.– K. Ogata. “Modern Control Engineering, 3rd ed”. Prentice-Hall, 1997.– K. Ogata. “Discrete-Time Control Systems, 2nd ed”. Prentice-Hall, 1995.
• Applications in Computer Science– C. Hollot et al. “Control-Theoretic Analysis of RED”. IEEE Infocom 2001 (to appear).– C. Lu, et al. “A Feedback Control Approach for Guaranteeing Relative Delays in Web Servers”.
IEEE Real-Time Technology and Applications Symposium, June 2001.– S. Parekh et al. “Using Control Theory to Achieve Service-level Objectives in Performance
Management”. Int’l Symposium on Integrated Network Management, May 2001– Y. Lu et al. “Differentiated Caching Services: A Control-Theoretic Approach”. Int’l Conf on
Distributed Computing Systems, Apr 2001– S. Mascolo. “Classical Control Theory for Congestion Avoidance in High-speed Internet”.
Proc. 38th Conference on Decision & Control, Dec 1999– S. Keshav. “A Control-Theoretic Approach to Flow Control”. Proc. ACM SIGCOMM, Sep
1991– D. Chiu and R. Jain. “Analysis of the Increase and Decrease Algorithms for Congestion
Avoidance in Computer Networks”. Computer Networks and ISDN Systems, 17(1), Jun 1989
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad
Effect of additional zero or pole on discrete step response
• The major effect of an additional zero to the step response of a discrete system is to substantially change the percent overshoot. See FP&W Figures 4.29 through 4.31.• The major effect of an additional pole is to increase the rise time. See Figure 4.32.• These effects are conveniently demonstrated using MATLAB’s ltiview GUI. Consider the system:theta=18*pi/180; zeta=0.5; %Figure 4.29 r=exp(-zeta*theta/sqrt(1-zeta^2))sys=tf(1,[1 -2*r*cos(theta) r^2],1);sys=sys/dcgain(sys),ltiview
Indian Institute of Information Technology - Allahabad
Discrete frequency response•Let: G(z)=Y(z)/U(z) with U(z) a sinusoid.
•Then, U(z)=z/(z-exp{jwT)) and Y(z)=G(z)z/(z-exp(jwT))=N(z)z/D(z)(z-exp(jwT)) Y(z) = G(exp(jwT))*z/(z-exp(jwT)+Transient response
•For a stable system, the transient response will die out withincreasing time and the steady-state output will be:
y_steady-state(kT) = G(z=exp(jwT))*exp(jkwT)= |G(exp(jwT))|*exp(jwT+arg(G(exp(jwT)))
• G(exp(jwT)) with -pi<wT<=pi is the discrete system’s frequency response. G(exp(jwT)) for 0=<wT=<pi is computed or measured and G(exp(-jwT)) =conj(G(exp(jwT))).
Indian Institute of Information Technology - Allahabad
Indian Institute of Information Technology - Allahabad