Control Systems 2

Lecture 13: Digital controller design

Roy Smith

2018-5-22 13.1

Digital control system design

Sampled-data closed-loop

y(k) y(t) u(t) u(k) r(k)

G(s) ZOH Kd (z) +
T
−

GZOH (z) equivalence

GZOH (z)

y(k) r(k)
G(s) ZOH Kd (z) +
T
−

2018-5-22 13.2
Zero-order hold equivalence — transfer function

y(k) y(t) u(t) u(k)

G(s) ZOH
T

| {z }
GZOH (z)

(
1 k = 0,
Input: u(k) = , u(t) = step(t) − step(t − T ).
0 k 6= 0
  G(s)
Output: y(s) = 1 − e−T s .
s

We now sample this, and take the Z-transform,

  
−T s G(s)
GZOH (z) = Z 1−e
s
 
−1 G(s)
= (1 − z )Z .
s

2018-5-22 13.3

Zero-order hold equivalence — state space

y(k) y(t) u(t) u(k)

G(s) ZOH
T

| {z }
GZOH (z)

Integrating Φ(t) over a single sample period (kT to kT + T ):

Z kT +T
AT
x(kT + T ) = e x(kT ) + eA(kT +T −τ ) Bu(τ ) dτ,
kT

 Z T

  AT Aη
A B ZOH-equivalence  e e B dη 
 0 
C D =⇒
C D

2018-5-22 13.4
Zero-order hold equivalence — frequency domain
(2 − s)
Example: G(s) = , T = 0.6
(2s + 1)(s + 2)

0.1 1 π/T 10
log ω
1 (rad/sec)

GZOH (ejω )

0.1
G(jω)

G(jω)e−jωT /2
Magnitude
0.01

1 π/T 10
log ω
0 (rad/sec)

−90

G(jω)
−180
GZOH (ejω )

Phase (deg.) G(jω)e−jωT /2

2018-5-22 13.5

Digital control system design

Sampled-data closed-loop

y(k) y(t) u(t) u(k) r(k)

G(s) ZOH Kd (z) +
T
−

K(s) approximation
≈ K(s)

y(k) r(k)
G(s) ZOH Kd (z) +
T
−

2018-5-22 13.6
Design approaches

Continuous-time
design
G(s) K(s)

ZOH-equivalence Approximation
of G(s) and of K(s) with
sample/hold Kd (z)

GZOH (z) Kd (z)

Discrete-time
design

2018-5-22 13.7

Design approaches

Continuous-time
design
G(s) K(s)

Sampled-data
design
ZOH-equivalence Approximation
of G(s) and of K(s) with
sample/hold Kd (z)

GZOH (z) Kd (z)

Discrete-time
design

2018-5-22 13.8
Design by approximation

1. Design a continuous-time controller, K(s)

I Verify stability, performance and bandwidth
I Verify margins and robustness

2. Select a sample-rate, T

3. Find Kd (z) approximating K(s)

4. Calculate the ZOH-equivalent GZOH (z)

5. Check the stability of the GZOH (z), Kd (z) loop

6. Simulate Kd (z) with G(s) (including sample/hold).

I Verify simulated performance
I Examine intersample behaviour

2018-5-22 13.9

Controller approximation

Approach: approximating the integrators

y(t) x(t) y(k) x(k)

1/s ≈ F (z)

If F (z) ≈ 1/s, then, s ≈ F −1 (z), =⇒ Kd (z) = K(s) |s=F −1 (z)

1
I F (z)
s
sx(s) x(s) zx(z) x(z)

A B A B
y(s) u(s) y(z) u(z)
C D C D

2018-5-22 13.10
Integration

x(t)
y(t) x(t)
1/s
x(kT +T )
Z t
y(t) = y(0) + x(τ ) dτ,
0
x(kT )

y(kT +T ) − y(kT )
Time
[sec.]
kT kT +T

The signal, y(t), over a single T second sample period is,

Z kT +T
y(kT + T ) = y(kT ) + x(τ ) dτ.
kT

2018-5-22 13.11

Trapezoidal approximation
ŷ(kT + T ) = ŷ(kT ) + T x(kT ) + (x(kT + T ) − x(kT ))T /2.

Taking z-transforms,
T
z ŷ(z) = ŷ(z) + T x(z) + (z − 1)x(z),
2

Approximation: x(t)

ŷ(z) T z+1
= F (z) = . x(kT +T )
x(z) 2 z−1
So the substitution is,
2 z−1 x(kT )
s ←− .
T z+1
ŷ(kT +T ) − ŷ(kT )
Time
[sec.]
kT kT +T

This is known as a bilinear (or Tustin) transform.

2018-5-22 13.12
Frequency mapping
Pole locations under bilinear transform:
bilinear
{ s | real(s) < 0 } { z | |z| < 1 }
−→

K(s) stable ⇐⇒ Kd (z) stable.

imag imag
2π/T 1

2 (z−1)
s= T (z+1)
π/T

−→ real
real
−1 1

−π/T

−1
−2π/T

2018-5-22 13.13

Bilinear frequency distortion

Ω : discrete-frequencies: ejΩT , Ω ∈ (−π, π].

Frequency mapping:

Continuous frequencies, ω to discrete frequencies, Ω.

2 z−1
Substitute s = jω and z = ejΩT into s = :
T z+1
−jΩT


2 1−e 2 j sin(ΩT /2) 2
jω = −jΩT
= = j tan(ΩT /2).
T (1 + e ) T cos(ΩT /2) T

Frequency distortion:
2
Ω= tan−1 (ωT /2)
T

2018-5-22 13.14
Bilinear frequency distortion

2
Ω= tan−1 (ωT /2)
T

Discrete frequency
Ω [rad/sample]
π

Bilinear distortion

Continuous frequency
ω [rad/sec]

−3π/T −2π/T −π/T π/T 2π/T 3π/T

−π
Ω = ωT

The Ω = ωT line is the sampling mapping.

2018-5-22 13.15

Prewarping

α(z − 1)
s= , α ∈ (0, π/T ), maps {real{s} < 0} to {|z| < 1}.
(z + 1)

Modifying the frequency distortion

Select a frequency ωpw .


Solve for α such that K(jωpw ) = Kd ejωpw T .

The “prewarped” transform makes K(jω) = Kd (ejωT ) at ω = 0 and ω = ωpw .

α(ejωpw T − 1)
jω0 = = jα tan(ωpw T /2),
(ejωpw T + 1)
ωpw
which implies that: α= .
tan(ωpw T /2)

2018-5-22 13.16
Prewarping

2
Frequency distortion (bilinear:) Ω = tan−1 (ωT /2).
T

2
Frequency distortion (with prewarping): Ω = tan−1 (ω/α)
T

Discrete frequency
Ω [rad/sample]
π Prewarping distortion

Bilinear distortion

Continuous frequency
ωpw ω [rad/sec]

−3π/T −2π/T −π/T π/T 2π/T 3π/T

−π
Ω = ωT

2018-5-22 13.17

Controller approximations
Bilinear approximation: Kbl (ejω ), Prewarped Tustin approximation: Kpw (ejω )
10 Magnitude
Kbl (ejω )
10 ωpw K(jω)
1 log ω (rad/sec)

Kpw (ejω )
0.1

0.01

90

Kbl (ejω )
10
0 log ω (rad/sec)
ωpw

Kpw (ejω )
−90 K(jω)

Phase (deg.)
−180

2018-5-22 13.18
Choosing a prewarping frequency

The prewarping frequency must be in the range: 0 < ωpw < π/T .
I α = 2/T (standard bilinear) corresponds to ωpw = 0.

Possible options for ωpw depend on the problem:

I The cross-over frequency (helps preserve the phase margin);
I The frequency of a critical notch;
I The frequency of a critical oscillatory mode.

The best choice depends on the most important features in your control design.

Remember: K(s) stable implies Kd (z) stable.

But you must check that (1 + GZOH (z)Kd (z))−1 is stable!

2018-5-22 13.19

Example

Plant model:

5(1 − s/zrhp ) (s2 + 2ζηωm s + η 2 ωm

2
) 1
G(s) =
(1 + τ s) (s + 2ζωm s + ωm ) η 2
2 2

where τ = 0.5, zrhp = 70, ωm = 20, ζ = 0.05, and η = 1.2.

IMC desgn
Tideal (s) = 3rd order Butterworth filter with bandwidth: 25 [rad./sec.]

Q(s) = Tideal (s)G−1

mp (s)

K(s) = (I − Q(s)G(s))−1 Q(s).

2018-5-22 13.20
Loopshapes
100 Magnitude

L(jω)
10
K(jω)
G(jω) 1 10 100
1 log ω (rad/sec)

0.1

0.01

0.001

90

1 10 100
0 log ω (rad/sec)
K(jω)

−90
L(jω)
G(jω)
−180

Phase (deg.)
−270

2018-5-22 13.21

Sensitivity and complementary sensitivity

Magnitude
10

2
S(jω)
10 100
1 log ω (rad/sec)

T (jω)

0.1

2018-5-22 13.22
Step response
Output

Amplitude
1.4

1.2

1.0

0.8

0.6

0.4 y(t)

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.23

Step response

Actuation
Amplitude
1.4

1.2
u(t)
1.0

0.8

0.6

0.4

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.24
Bilinear/Trapezoidal/Tustin transform

Bilinear transform
π
Nyquist frequency: 100 radians/second =⇒ T = .
100


Kbl (z) = K(s) 2 z−1
s= T z+1

Discrete-time analysis
GZOH (z)

y(k) r(k)
G(s) ZOH Kd (z) +
T
−

2018-5-22 13.25

Loopshapes: bilinearly transformed controller

100 Magnitude

10
1 10 100
1 log ω (rad/sec)
Kd (ejω )
G(jω)
0.1
GZOH (ejω )

0.01 Ld (ejω )
K(jω)
0.001 L(jω)
90

1 10 100
0 log ω (rad/sec)

−90
G(jω)
−180
GZOH (ejω ) K(jω)
Phase (deg.)
−270 Ld (ejω ) Kd (ejω )
L(jω)

2018-5-22 13.26
Sensitivity and complementary sensitivity: bilinearly transformed controller

Magnitude
10

2
S(jω)

1 log ω (rad/sec)
10 100
Sd (ejω )

Td (ejω )
T (jω)

0.1

2018-5-22 13.27

Step response: bilinearly transformed controller

Output

Amplitude
1.4

y(k)
1.2

1.0

0.8

0.6 y(t)

0.4

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.28
Step response: bilinearly transformed controller

Actuation
Amplitude
1.4

1.2
u(t)
1.0

0.8
u(k)

0.6

0.4

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.29

Prewarped Tustin transform

Prewarped Tustin transform

π
Nyquist frequency: 100 radians/second =⇒ T = .
100
Select the prewarping frequency at ωpw = ωm (20 radians/sec.).


Kpw (z) = K(s) z−1
s=α z+1

where,
ωpw
α= .
tan(ωpw T /2)

2018-5-22 13.30
Loopshapes: prewarped Tustin controller
100 Magnitude

10
1 10 100
1 log ω (rad/sec)
Kd (ejω )
G(jω)
0.1
GZOH (ejω )

0.01 Ld (ejω )
K(jω)
0.001 L(jω)
90

1 10 100
0 log ω (rad/sec)

−90
G(jω)
−180
GZOH (ejω ) K(jω)
Phase (deg.)
−270 Ld (ejω ) Kd (ejω )
L(jω)

2018-5-22 13.31

Sensitivity and complementary sensitivity: prewarped Tustin controller

Magnitude
10

2
S(jω)

1 log ω (rad/sec)
10 Sd (ejω ) 100

Td (ejω )
T (jω)

0.1

2018-5-22 13.32
Step response: prewarped Tustin controller
Output

Amplitude
1.4

y(k)
1.2

1.0

0.8

0.6 y(t)

0.4

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.33

Step response: prewarped Tustin controller

Actuation
Amplitude
1.4

1.2
u(t)
1.0

0.8
u(k)

0.6

0.4

0.2

0 time (sec)
0.5 1.0 1.5 2.0
−0.2

2018-5-22 13.34
Sample rate selection

Sample rate selection is critical to digital control design.

Main considerations
I Sampling/ZOH will (approximately) introduce a delay of T /2 seconds.

I Anti-aliasing filters will need to be designed and these will also introduce
phase lag.

I The system runs “open-loop” between samples.

I Very fast sampling can introduce additional noise.

I Very fast sampling makes all of the poles appear close to 1. The controller
design can become numerically sensitive.

2018-5-22 13.35

Designing for digital implementation

Sampled-data implementation
d(t)
Gd (s)

r(k)
y(t) u(t) u(k)
+ G(s) ZOH Kd (z)

n(t) ym (t) ỹm (t)

+ Fa (s) ỹm (k)
T

Continuous-time design
d(t)
Gd (s)

r(t)
y(t) u(t)
+ G(s) e−sT /2 K(s)

n(t) ym (t) ỹm (t)

+ Fa (s)

2018-5-22 13.36
Sensitivity function

We want a similar discrete sensitivity function up to the frequency where

|Ŝd (jω)| returns to 1.

Magnitude
5
Ŝd1 (jω)
1 10
log ω
1 (rad/sec)
π/T1
ωB

S(jω)
0.1

0.01

Ŝd (s) = (I + Fa (s)G(s)e−sT1 /2 K(s))−1 Approximate discrete sensitivity

2018-5-22 13.37

Sensitivity function

In this example, for ω > 20 rad./sec.,

|1 − Ŝd (jω)|  1 =⇒ π/T = 20 is about the minimum.

Magnitude
5
Ŝd1 (jω)
10 20
ω
1 (rad/sec)
π/T1
ωB

S(jω)
0.1

0.01

Ŝd (s) = (I + Fa (s)G(s)e−sT1 /2 K(s))−1

2018-5-22 13.38
Loop-shaping interpretation
For ω up to where |Fa (jω)G(jω)K(jω)| <  and remains very small,
   
we want Fa (jω)G(jω)K(jω)e−jωT1 /2 ≈ Gd ejωT1 Kd ejωT1 .
(Gd (z) is the ZOH-equivalent of Fa (s)G(s))

imag

|S(jω)| > 1

−1 1 real

−j
L̂d1 (jω) L(jω)

2018-5-22 13.39

Discrete-time sensitivity function

What is the actual discrete sensitivity function,

1
Sd (s) = for π/T1 = 20.9 [rad./sec.]?
(1 + Gd (ejω ) Kd (ejω ))

Magnitude
5
Sd1 (jω) Ŝd1 (jω)
1 10
log ω
1 (rad/sec)
π/T1
ωB

S(jω)
0.1

0.01

Kd (z) is a prewarped Tustin approximation with ωpw = ωB .

2018-5-22 13.40
Discrete-time loopshape: Ld1 (s) = Gd (s)Kd (s), T1 = 0.15 seconds

Gd (s) is a ZOH-equivalent transform of G(s).

Kd (z) is a prewarped Tustin approximation with ωpw = ωB .

imag

|S(jω)| > 1

−1 1 real

Ld1 (jω)

−j
L̂d1 (jω) L(jω)

2018-5-22 13.41

Choosing a slower sample rate.

Choosing T2 = 0.5 seconds =⇒ π/T2 = 6.28 rad./sec.

Magnitude
50
Ŝd2 (jω)
10
Sd2 (jω)
S(jω)
1 10
log ω
1 (rad/sec)
π/T2
ωB

0.1

0.01

Kd (z) is a prewarped Tustin approximation with ωpw = ωB .

2018-5-22 13.42
Choosing a slower sample rate.
Choosing T2 = 0.5 seconds =⇒ π/T2 = 6.28 rad./sec.

imag

L̂d2 (jω)

−1 1 real
Ld2 (jω)

−j
L(jω)

2018-5-22 13.43

Fast sampling

Fast sampling period: Tf .

Control appropriate (slower) sampling period: Ts

(typically Ts = M Tf for integer M > 1).

ỹs (k) ỹ(t) y(t)

Fas (z) Faf (s)
Ts Tf

Digital Digital Fast Anti-aliasing

downsampling filter sampler filter

2018-5-22 13.44