You are on page 1of 51

KOM 3781

Discrete-Time Control Systems

Veysel Gazi
Lecture 5: System Time-
Response Characteristics
Outline
• System Time Response
• System Characteristic Equation
• Mapping the s-Plane into the z-Plane
• Closed-loop z-plane poles and the system
transient response
• Steady-State Accuracy
• Simulation of continuous-time and discrete-
time systems
Example
Consider the system:
R(s) + X(s) X*(s) 1 − e −Ts X (s) 4 Y(s)
T=0.1 s s+2

Let’s calculate its step response.


The open loop Transfer Function G(z) is calculated as
G( z)
Y ( z) = R( z ) with:
1 + G( z)
1 − e −Ts 4  z − 1  4 
G( z) =   =  s ( s + 2) 
 s s + 2  z  
z − 1  2(1 − e −2T ) z  0.3625
=  −2T 
= T = 0.1sec
z  ( z − 1)( z − e )  z − 0.8187
Example
Then, the closed loop Transfer Function and the system
response are calculated as: R(z) + X(z) Y(z)
G( z)
 

Y ( z) G( z) 0.3625  1  0.3625
= =  =
R( z ) 1 + G ( z ) z − 0.8187  1 + 0.3625  z − 0.4562
 z − 0.8187 
0.3625 0.3625 z 0.667 z 0.667 z
Y ( z) = R( z ) = = −
z − 0.4562 ( z − 0.4562)( z − 1) ( z − 1) ( z − 0.4562)
and : y (kT ) = −1
Y ( z ) = y(kT ) = 0.667 1 − (0.4562)k 
The steady-state value of the response can be found from
the final value theorem:
lim y ( n) = lim ( z − 1)Y ( z )
n → z →1

Or, by inspection of the output equation: [y(kT)]ss = 0.667


(i.e. (0.4562 ) k → 0 as k →  )
Example
R(s) + X(s) X*(s) 1 − e −Ts X (s) 4 Y(s)
T=0.1 s s+2

>> T = 0.1; Gs = tf([4],[1 2]); Gz = c2d(Gs,T);


Tz = feedback(Gz,1); Rz = tf([1 0],[1 -1],T);
Yz = zpk(Rz*Tz) R(z) + X(z) G ( z ) Y(z)
Yz =
0.3625 z −
----------------
(z-1) (z-0.4562)

>> % Partial fraction expansion of C(z)/z


[r,p,k]=residue([0.3625],[1 -1.4562 0.4562])

r =
0.6666
-0.6666

p =
1.0000
0.4562

k =
[]
Example
SettlingMax (after risen) = Steady State
102% Steady State
Peak = Steady State
98% Steady State 90%
>> T = 0.1;
Gs = tf([4],[1 2]); SettlingMin (after risen)
Gz = c2d(Gs,T);
Tz = feedback(Gz,1);
step(Tz) RiseTime
stepinfo(Tz)

RiseTime: 0.2000
SettlingTime: 0.5000
SettlingMin: 0.6034
SettlingMax: 0.6667
Overshoot: 0
Undershoot: 0 10%
Peak: 0.6667
PeakTime: 4.6000

SettlingTime
Example
Consider the continuous time system:
R(s) + X(s) 4 Y(s)
s+2
Calculate its transfer function: −
4
G p (s)
= s+2 =
Y ( s) 4
=
R( s) 1 + G p ( s) 1+
4 s+6
s+2
Calculate its step response:
4  4 1 0.667 0.667
Yc ( s ) = R( s) =   = −
s+6  s+6 s s s+6
yc (t ) = 0.667(1 − e −6t )  [ yc (t )]ss = 0.667
or (Final Value Th.) lim yc (t ) = lim sYc ( s ) = 0.667
t → s →0

The same steady state for the continuous and the discrete
systems.
Observations
Compare the results: yc (t ) = 0.667(1 − e −6t ) Continuous
y (kT ) = 0.667 1 − (0.4562) k  Discrete

kT y(kT) yc(t)
0 0 0
0.1 0.363 0.300
0.2 0.528 0.466
If the sampling frequency
0.3 0.603 0.557 is increased (sampling
0.4 0.639 0.606 period is decreased), the
0.5 0.654 0.634 discrete time response will
0.6 0.661 0.648 approach the continuous
time case.
1.0 0.666 0.665
Example
Find y(t) for discrete system.

 R *( s )  1
Y ( s ) = G ( s ) X *( s ) X *( s ) =   R ( z ) =
1 + G *( s )  1 − z −1
 R *( s )   1 − e −Ts   4   R( z ) 
Y (s) = G (s)   =  s   s + 2  1 + G ( z ) 
 1 + G *( s )      z =eTs
 4  (1 − z −1 )   1 
=  −1  = Y ( s ) 1 + G ( z ) 
+ − +
1
 s ( s 2)  (1 z )(1 G ( z ))  z =eTs   z =eTs

2 2 1 z − 0.8187
Y1 ( s ) = − = = 1 − 0.363 z −1 − 0.165 z −2 −
s s+2 1 + G ( z ) z − 0.0465
−2 t  1  −Ts −2Ts
y1 (t ) = 2(1 − e ) 1 + G ( z )  = 1 − 0.363e − 0.165 y −
  z =eTs
y (t ) = −1
Y (s)(1 − 0.363e
1
−Ts
− 0.165e −2Ts − )
= 2(1 − e −2t )u (t ) − 0.363(2)(1 − e −2(t −T ) )u (t − T ) − 0.165(2)(1 − e −2( t − 2T ) )u (t − 2T ) −
Example
Continuous: yc (t ) = 0.667(1 − e −6t )

Discrete: y (kT ) = 0.667 1 − (0.4562) k 

Discrete: y (t ) = 2(1 − e −2t )u (t ) − 0.363(2)(1 − e −2( t −T ) )u (t − T )


− 0.165(2)(1 − e −2(t − 2T ) )u (t − 2T ) − ...
y (0) = 0 y(t)
y (0.1) = 0.363
y (0.2) = 0.528
y (0.3) = 0.603 Sampled-data system

The same values as for Continuous-time system


z-transform method;
but with this form, one
can find the system
output for any time t.
Example
Time constant of 1st order systems and the sampling interval
From the previous example, we had: Impulse Response
Yc ( s ) 4 0.667 K
= = =  hc (t ) = 0.667(e − t /0.1667 )
Rc ( s ) s + 6 0.1667 s + 1  s +1
The time constant for this 1st order system is: τ = 0.1667.

Rule of thumb (1st order systems):


set the sample rate (1/T) > 5/τ or T < τ/5

dc gain:
The dc gain of a stable system may be obtained
from the closed-loop transfer function.
Example
0.3625
G( z) = from before
z − 0.8187

A) For sampled systems, the final value to the step input is:
 G( z) 
dc gain = lim y ( nT ) = ( z −1)Y ( z )  = ( z −1) R( z ) 
n→ z =1 1+ G ( z )
  z =1

 G( z) z  zG ( z )
= ( z −1)  =
 1 + G ( z ) z − 1  z =1 1 + G ( z )
z =1
G (1) 2
= = = 0.667
1 + G (1) 1 + 2

Then simply evaluate the closed-loop transfer


function at z = 1 to determine the dc gain.
Example
B) For continuous systems, the final value to the step input is:

G (s)
dc gain = lim y (t ) = lim sY ( s) = lim s R( s)
t → s →0 s →0 1 + G ( s)
G (s) 1 G (0) 2
= lim s = = = 0.667
s →0 1 + G ( s ) s 1 + G (0) 1 + 2

We have the same steady state (dc gain) for both the
discrete and continuous systems. Either may be evaluated
using the final value theorem, or by simply evaluating the
closed loop transfer function at dc (s = 0 or z = 1).
Example
Calculate the step response for the following second order,
underdamped system
R(s) + X(s) X*(s) 1 − e−Ts X ( s) 1 Y(s)

T = 1.0 s s ( s + 1)

z −1  1  z − 1  z[(1 − 1 + e −1 ) z + (1 − e −1 − e −1 ] 
G( z) =  s 2 ( s + 1)  = z  −1 
z  T =1  ( z − 1) 2
( z − e ) 
0.368 z + 0.264
= 2
z − 1.368 z + 0.632

G( z) z (0.36 z + 0.204)
Y ( z) = R( z ) =
1 + G( z) ( z − 1)( z 2 − z + 0.632)
Example
z (0.368 z + 0.264)
Y ( z) = = 0.368 z −1 +1.00 z −2 +1.40 z −3
( z − 1) ( z 2 − z + 0.632)
+1.40 z −4 +1.15 z −5 + 0.90 z −6 + 0.80 z −7 + 0.87 z −8
+ 0.99 z −9 +1.08 z −10 +1.08 z −11 +1.00 z −12 + 0.98 z −13
+ 

The final value of y(nT)


 z (0.368 z + 0.264)  0.632
lim y (nT ) = lim ( z −1)Y ( z ) = lim ( z −1) = =1
2
n→ z →1 z →1  ( z − 1) ( z − z + 0.632)  0.632
Example
Continuous system equivalent (remove the sample and hold):
1
Y (s) G ( s) s ( s + 1) 1 n 2
= = = 2 = 2
R( s) 1 + G ( s) 1 + 1 s + s + 1 s + 2n s + n 2
s ( s + 1) n =1,  = 0.5
1 1
Y (s) = 2 R( s) = (for the unit - step input)
s + s +1 s ( s + s + 1)
2

y(t)

In this case the


continuous and discrete
systems do not match
well. Sampling rate is
too low.
Example
Rule of Thumb (2nd order systems):
τ = 1/ζωn sample @ T < τ/10
1 1  2
= = =2 sample at T = = 0.2 sec
n 0.5 10 10

Y (s) 2n
=
R ( s ) s 2 + 2n s + 2n

− / 1− 2
Percent Overshoot = e 100
Characteristic Equation
An important step is to define and analyze the characteristic
equation of the system. A typical expression of a transfer
function of a closed-loop control system is the following:
m
K  ( z − zi )
G( z) i =1
Gcl ( z ) = =
1 + GH ( z ) n
 ( z − pi )
i =1
where zi and pj are the zeros and poles of the closed-loop
system.
The output of the system can be written as
m
K  ( z − zi )
G( z) i =1
Y ( z ) = Gcl ( z ) R ( z ) = R( z ) = R( z )
1 + GH ( z ) n
 ( z − pi )
i =1
Characteristic Equation
The system response typically has two types of terms:
k1 z kn z
Y ( z) = + + + YR ( z )
z − p1 z − pn
Comes from R(z)
Comes from 1+GH(z)
The transient response terms (first n terms) contain the
poles of the system, which are independent of the input.
The second group depends on the forcing function (i.e., the
system input).
The equation 1+ GH ( z )
is called the characteristic equation of the system and
includes the poles of the transfer function.
−1  ki z 
Originates from the characteristic
Then  = ki ( pi ) equation, and gives the character
k

 z − pi  of the transient response.
Characteristic Equation
Consider

GR ( z ) R(z) is not separable from G(z)


Y ( z) =
1 + GH ( z ) → no explicit transfer function

For the situation where we can’t write the transfer


function explicitly (i.e., the input cannot be separated
and is trapped in a transform with other
(analog/continuous) elements of the transfer
function), the effect of the input on Y(z) is embedded
and cannot be isolated as before.
Mapping the s-Plane into the z-Plane

The s-plane poles map into the z-plane by the mapping


z = esT .
In the primary band of the s-plane, between ±ωs/2, the
poles of the characteristic equation map one-for-one to
the z-plane. These z-pole locations are functions of both
the sampling interval T, and the s-plane pole locations.

Using s = σ + jω and z = esT, we can map between s and z.


Mapping the s-Plane into the z-Plane
Mapping Function
s-Plane z-Plane
z = e sT = e( + j )T = e T e jT = e T e j (2 /s )
s-Plane z-Plane
j Im z
 D C ws / 2
B
B 1
A C D A
Re z
0
 F E 0
G
G
E −s / 2
F
Mapping the s-Plane into the z-Plane
z = e sT = e T e jT = e T e jT
magnitude phase

s-Plane z-Plane
jw Im z
r = e2T
r = e1T
0
s2 s1 s 1
Re z

Loci of constant  on both planes

Lines of constant  (vertical lines) in the s-plane map to


circles of constant radial distance, r = esT in the z-plane
Mapping the s-Plane into the z-Plane
z = e sT = e T e jT = e T e jT
magnitude phase

s-Plane z-Plane
jw
Im z
ws / 2 w 2T
w 2

w 1
w1T
0 s 1 Re z

Constant frequency ω loci in both planes

Lines of constant ω (horizontal lines) in the s-plane map to


lines of constant angle, ωT, (radial lines) in the z-plane.
Mapping the s-Plane into the z-Plane
s-Plane z-Plane
Im z
jw s
2 5
4
0
s 1 Re z
3
- jw s
2

The jω axis between ωs/2 and –ωs/2 maps to the unit circle.

The left-hand s-plane maps inside the unit circle.

The right-hand s-plane maps outside the unit circle.


Mapping the s-Plane into the z-Plane
Loci for constant damping ratio (ω/σ = tan β):
 
j T j T
z = e sT = e  T e j T = e T e 
= eT e 
= e  T e jT tan 
s-Plane z-Plane
jw Im z
ws / 2
w1
b Re z
1

s1 s
Radial line
Logarithmic
spiral
Mapping the s-Plane into the z-Plane
Mapping the s-Plane into the z-Plane
Consider the case in the figure with
s-plane poles at s =   j .
Such poles result in a system
transient-response term of the
form Aet cos(t + ) .
In the z-plane:
z = e sT = eT e  jT = eT (T ) = r ()
s = j

Each pair result in a transient-response term of the form


AekT cos(kT + ) = A(r ) k cos(k + )
Transient Response
A(r ) k cos(k + )
Example
Find the time constant and settling time for the first-order
system
G( z) 0.3625
T ( z) = =
1 + G ( z ) z − 0.4562
Closed - loop characteristic equaion : z − 0.4562

z1 = 0.4562 = e s1T = e0.1s1


ln(z1 ) ln(0.4562)
s1 = = = −7.848
T 0.1
1 • Time constant for the sampled-data
= = 0.127 s system (SDS).
s1
• The time constant for the continuous-
tss  4  0.54 s
time system was computed as τ=0.1667
before.
• Different sampling rate will lead to
different time constant of the SDS.
Mapping the s-Plane into the z-Plane
Mapping second-order systems
2n
G (s) = 2 s1,2 = − n  jn 1−  2
s + 2 n s + n2
Poles in the z-plane
z = e sT = e −nT (nT 1 −  2 ) = r ()
s = s1,2

−nT n T 1 −  2 = 
e =r

nT = − ln r 
=
−ln r
1−  2 

Solving for parameters


− ln r −T
r = e −T / 
1 1
= n = (ln r ) 2 + 2 = =
(ln r ) 2 + 2 T n ln r
Example
Find the characteristic equation roots, damping ratio,
natural frequency, and time constant for the second-order
system G( z) 0.368 z + 0.264
T ( z) = = , T = 1s
1+ G ( z ) 2
z − z + 0.632 Compare with:
Roots of the characteristic equation: 1. Continuous case
z 2 − z + 0.632 = ( z − 0.5 − j 0.618 )( z − 0.5 + j 0.618 ) = 0  = 0.50
n = 1 rad/s
z = 0.5  j 0.618 = 0.795(51.0o ) = 0.795(0.890 rad)
 = 1  n = 2 s
T
z=e (T ) = r (T ) = 0.795(0.890)
2. Faster sampling, T =
Computing the parameters: 0.1 s
− ln r − ln 0.795  = 0.475
= = = 0.250
(ln r ) 2 + 2 [(ln 0.795) 2 + (0.890) 2 ]1/2 n = 0.998 rad/s
1 1  = 2.11 s
n = (ln r ) 2 + 2 = [(ln 0.795) 2 + (0.890) 2 ]1/2 = 0.9191 rad/s
T 1 Faster sampling was
1 −T −1
= = = = 4.36 s needed.
n ln r ln 0.795
Choosing the Sampling Interval
Real and complex pole mapping:
s + 1/  → z − e − T / radius sinusoid
( s + 1/ ) 2 + 2 → z 2 − 2 ze −T / cos T + e −2T /
= ( z − z1 )( z − z1 ) 360
r = e−T /   =
Td T

Complex pole magnitude:


z1 = e −T / e jT = e −T /  (T ) = r 
z1 = e −T / e − jT = e −T /  (−T ) = r (−)
Samples per time Samples per period
constant

z1 z1 = e −2T /  r τ/T θ° Td / T
0.999 999.5 10° 36
−T / 
| z1 |= e 0.99 99.5 20° 18
0.95 19.5 30° 12
Sinusoidal period, Td: 0.9 9.5 45° 8
T = 1 fs Td = 1 f >5 0.8 4.48 60° 6
0.7 2.80 90° 4
2T 0.6 1.96 120° 3
 = T = (2f )T =
Td 0.4 1.09 180° 2

Td 2 360 0.2 0.62


= =
T  
Example
Suppose we have an open loop G(s)
4.25
G ( s) =
s ( s + 1)

Yielding the closed-loop T(s)


G ( s) 4.25 4.25
T ( s) = = 2 = 2
1 + G ( s ) s + s + 4.25 s + 2 4.25s + 4.25
1
n = 4.25  2.06 =  0.24
2(2.06)
1 1 t 2
t= = =2 sample@ T < = = 0.2 s
zw n 0.5 10 10
T < 0.2 seconds
Example (for T = 0.001)
4.25 N = 0 0 4.2500
G (s) = D = 1.0000 1.0000 4.2500
s ( s + 1)
G ( s) 4.25 Transfer function:
T (s) = = 2 4.25
1 + G ( s ) s + s + 4.25 --------------
s^2 + s + 4.25
poles at s = –0.5 ± j2 Transfer function:
2.124e-006 z + 2.124e-006
-------------------------
% MATLAB® z^2 - 1.999 z + 0.999
num=[0 0 4.25] Sampling time: 0.001
den=[1 1 4.25]
sysc=tf(num,den) ans =
sysz=c2d(sysc,0.001) 0.9995+ 0.0020i
pole(sysz) 0.9995- 0.0020i

poles are very near z = 1


Example (for T = 0.01)
% MATLAB® % MATLAB®
sysz=c2d(sysc,0.01) sysz=c2d(sysc,0.1)
pole(sysz) pole(sysz)
Transfer function: Transfer function:
0.0002118 z + 0.0002111 0.02049 z + 0.01981
----------------------- ----------------------
z^2 - 1.99 z + 0.99 z^2 - 1.865 z + 0.9048
Sampling time: 0.01 Sampling time: 0.1

ans = ans =
0.9948+ 0.0199i 0.9323+ 0.1890i
0.9948- 0.0199i 0.9323- 0.1890i

poles are moving away poles are moving


from z = 1 further away from z = 1
Example (for T = 0.2)
Transfer function:
%MATLAB ® 0.0785 z + 0.07341
sysz=c2d(sysc,0.2) ----------------------
pole(sysz) z^2 - 1.667 z + 0.8187
Sampling time: 0.2

For T = 0.2 (our upper ans =


0.8334+ 0.3524i
limit for sampling time) 0.8334- 0.3524i
we see a significant
deviation from the high poles are moving
fs sampling rate. further away from z = 1
Example (for T = 0.5)
Transfer function:
%MATLAB® 0.4154 z + 0.3496
sysz=c2d(sysc,0.500) -----------------------
pole(sysz) z^2 - 0.8416 z + 0.6065
Sampling time: 0.5

We see the trade-off between: ans =


sampling slowly and incurring 0.4208+ 0.6553i
0.4208- 0.6553i
significant error in mapping
the s-plane design to the z- Note, as T becomes larger,
plane implementation the placement of the z-plane
or roots is a poorer system
sampling fast and placing a high representation.
demand on the processor with INSTABILITY is possible!
possible failure to complete the
algorithmic tasks in T seconds.
Example (step responses) T (s) = 4.25
>> num=[4.25]; den=[1 1 4.25]; s 2 + s + 4.25
Ts=tf(num,den); >> Tz5=c2d(Ts,0.5); %T = 0.5
Tz01=c2d(Ts,0.01); %T = 0.01 Tz8=c2d(Ts,0.8); %T = 0.8
Tz1=c2d(Ts,0.1); %T = 0.1 Tz10=c2d(Ts,1); %T = 1
Tz2=c2d(Ts,0.2); %T = 0.2 figure(2), step(Tz5,Tz8,Tz10,Tz2,5)
figure(1), step(Tz2,Tz01,Tz1,5), grid grid
z-Plane Design Goals
1. Keep poles “close to” z = 1

2. s-plane real pole,

s = –a → T < 1/(5a) = τ/5


3. s-plane complex pair, s1,2 = − n  jn 1−  2

s = s2 + 2ζωns + ωn2 → T < 1/(10ζωn) ) = τ/10


4. Choose the most severe constraint above.
Provides a good starting point for the design.

5. Simulate (continuous and discrete) to evaluate the


sample time selection.
Steady-State Accuracy
Control designers are always interested in transient
behavior, such as overshoot, rise time, settling time, etc.
Another performance measure in control system design is
steady-state accuracy. Steady-state accuracy considers
the error signal e(k) = r(k) – y(k), which ideally is zero, and
in practice must be acceptably small.
To discuss steady-state accuracy consider the system
R E Y
+ T G(s)
+
-
m
K  ( z − zi )
G( z) = p
zi  1, zj 1
( z − 1) N ( z − z j )
The value of N determines the system type (0, I, II, III, ...).
Steady-State Accuracy
The system type (0, I, II, III, etc.) determines the steady-
state accuracy for various types of standard input signals.

TYPE STEADY-STATE ERROR VS INPUT


Step Ramp Parabolic
0 Finite Infinite Infinite
I 0 Finite Infinite
II 0 0 Finite
III 0 0 0
Steady-State Accuracy
m
K  ( z − zi )
R E Y G( z) = p
, zi  1, z j  1
+ G(s)
T ( z − 1) N ( z − z j )
+
- m
(1 − zi )
and let K dc = ( z − 1) G ( z ) |z =1 = K
N i =1
p
(1 − z j )
i =1

Kdc is the open-loop dc-gain, with the poles at z = 1 removed.


R( z )
E( z) =
1 + G( z)
In general, the final value of the error is (from the final value theorem):
 ( z − 1) R( z ) 
ess (kT ) = lim ( ( z − 1) E ( z ) ) = lim  
z →1 z →1
 1 + G ( z ) 
provided it exists*
• Sometimes for unstable systems,
finite ess may be calculated, but is
meaningless.
Steady-State Accuracy
m
STEP : K  ( z − zi )
G( z) = i =1
p
z ( z − 1) ( z − z j )
Let R( z ) = r (k ) = 1
N

z −1 i =1

 ( z − 1) R( z )  z 1
ess (k ) = lim   = lim =
z →1
 1 + G( z)  z →1 1 + G ( z ) 1 + lim G ( z )
z →1

Define the position - error constant : K p = lim G ( z )


z →1
m
1
ess (k ) = (1 − zi )
1+ K p K dc = ( z − 1) G ( z ) |z =1 = K
N i =1
p

Two cases : (1 − z j )


i =1

1 1 1
N = 0, K p = K dc , ess (k ) = = =
1 + lim G ( z ) 1 + K p 1 + K dc
z →1

N  1, K p → , ess (k ) → 0
Steady-State Accuracy
m
RAMP: K  ( z − zi )
Tz G( z) = i =1

R( z ) = r (k ) = k
p
Let ( z − 1) ( z − z j )
N
( z - 1) 2 i =1

 ( z − 1) R( z )  ( z − 1) T z
ess (k ) = lim  =
 z →1
lim
z →1
 1 + G ( z )  ( z − 1) 2
(1 + G ( z ))
T T 1
= lim = =
z →1 ( z − 1) + ( z − 1)G ( z ) lim( z − 1)G ( z ) K v
z →1

1
Velocity - error constant K v = lim ( ( z − 1)G ( z ) )
T z →1
Three cases : If N = 0, K v = 0, ess (k ) → 

K dc 1 T
N = 1, Kv = , ess (k ) = =
T K v K dc
N  2, Kv → , ess (k ) → 0
Steady-State Accuracy
PARABOLIC : m
K  ( z − zi )
T 2 z ( z + 1) k2 G( z) = i =1

R( z ) = r (k ) =
p
, ( z − 1) ( z − z j )
N
2( z − 1)3 2 i =1

 ( z − 1) R( z )  ( z − 1)T 2 z ( z + 1)
ess (k ) = lim   = lim
z →1
 1 + G ( z )  z →1 2( z − 1) 3 (1 + G ( z ))

T2 1
= =
lim ( ( z − 1) 2 G ( z ) ) K a
z →1

Acceleration-error constant: K a =
T
1
lim
2 z →1 ( ( z − 1) 2
G( z) )

Three cases:
N  1, K a = 0, ess (k ) → 
K dc 1 T2
N = 2, Ka = 2 , ess (k ) = =
T K a K dc
N  3, K a → , ess (k ) → 0
m
Steady-State Accuracy
( z − zi )
G( z) = K p We evaluate the steady-state error
( z − 1) ( z − z j )
coefficients from the open-loop G(z) in
N

K dc = ( z − 1) N G ( z ) |z =1 factored form.
m
m
(1 − zi ) K (1 − zi )
=K Type 0 : K p = p
= K dc
(1 − z j )
p
(1 − z j )
Kv = Ka = 0
K p = lim G ( z )
z →1  m 
 −
Type I : K p = lim G ( z ) = lim  →
K (1 zi )
z →1 
 ( z − 1) (1 − z ) 
1
lim ( ( z − 1)G ( z ) )
p
Kv = z →1

T z →1  j 

K a = 2 lim ( ( z − 1) 2 G ( z ) )  
1 m
 −
K v = lim  ( z − 1)  = K dc
1 K (1 z )
T z →1 i

T z →1  ( z − 1) 
p
(1 − z )
 T
1  j 
ess = (step)
1+ K p Ka = 0
1 Type II : K p = K v = 
ess = (ramp)
Kv  m 
 −
K a = 2 lim  ( z − 1)  = K dc
1 2 K (1 zi )
1
ess = (parabolic) T z →1  p  T 2
( z − 1) (1 − z j ) 
2
Ka 
Example
D(z) = 1
R E Y
T
1− e −Ts
 K  + + D(z) G(s)
G (s) =  s ( s + 1)  -
s  
z − 1  1  K ( z − 1)  z  ( e −T
+ T − 1) z + (1 − e −T
− Te −T
) 
G( z) = K   =
z  s ( s + 1) 
2
z ( z − 1) 2 ( z − e −T )
K (e −T + T − 1) z + (1 − e −T − Te −T ) 
= −T
Type I system
( z − 1)( z − e )

1 1
K p = lim G ( z ) =  For a step input ess = = =0
z →1 1+ K p 

1 K (e −T + T − 1) z + (1 − e −T − Te −T ) 
K v = lim ( z − 1)G ( z ) = lim −T
=K
z →1 T z →1 T (z − e )
1 1
For a ramp input ess = =
Kv K
Example
R T
E 1
_____ Y
+ + D(z) ZOH s+1

1− e −Ts
 1  D(z) = 1
G (s) =  ( s + 1) 
s   Design specification ess(ramp) <= 0.01
z −1  1  ( z − 1) z (1 − e −T ) (1 − e −T )
G( z) =  s ( s + 1)  = z ( z − 1)( z − e −T ) = ( z − e −T )
z  
1 1
Type 0 system  K v = 0  ess (ramp) = = =
Kv 0
To achieve ess (ramp)  0.01 → there must be a pole @ z = 1 Must be a
type I system
KI z (KI + KP ) z − KP
Introduce D( z ) = + KP = (PI compensator )
z −1 z −1
1 (K + KP )z − KP  (1 − e −T ) 
K v = lim ( z − 1) D( z )G ( z ) = lim( z − 1) I  ( z − e −T ) 
z →1 T z →1 T ( z − 1)  
(K + KP ) − KP  (1 − e −T )  K I
= I  (1 − e −T )  = T ( D( z )G ( z ) → Type I system )
T  
Example
Type I system, with a ramp input:
1 T
ess (ramp ) = =  0.01  K I  100T
Kv K I

KI > 100T will give the required design specification of


ess < 0.01
Provided that the system is stable at this gain.

Note: Steady-state error analysis is only valid if


the system is stable at the required gain.

You might also like