Está en la página 1de 9
Lecture 2: z-transform and I/O models
Shift-operator
• Shift operator
• I/O models
• Direct sampling
• z-transform
• Poles and zeros
• Selection of sampling interval
• Frequency response of sampled-data systems
• Lyapunov theory for discrete-time systems
Forward shift operator
qf (k) = f(k + 1)
Backward shift (delay) operator
q −1 f(k) = f(k − 1)
The range of the shift operator is double inﬁnite sequences
Compare with the differential operator p =
d
dt
Shift-operator calculus
Reciprocal polynomials
y(k + na) + a 1 y(k + na − 1) + ⋅ ⋅ ⋅ + a na y(k)
y(k + na) + a 1 y(k + na − 1) + ⋅ ⋅ ⋅ + a na y(k)
= b 0 u(k + nb) + ⋅ ⋅ ⋅ +
b nb u(k)
= b 0 u(k + nb) + ⋅ ⋅ ⋅ +
b nb u(k)
where na ≥ nb. Using the shift operator gives
can be written as
y(k) + a 1 y(k − 1) + ⋅ ⋅ ⋅ + a na y(k − na)
(q na + a 1 q na−1 + ⋅ ⋅ ⋅ + a na )y(k) =
Introduce the polynomials
(b 0 q nb + ⋅ ⋅ ⋅ + b nb )u(k)
= b 0 u(k − d) + ⋅ ⋅ ⋅ + b nb u(k −
d − nb)
A(z)
= z na + a 1 z na−1 + ⋅ ⋅ ⋅ + a na
B(z) = b 0 z nb + b 1 z nb−1 + ⋅ ⋅ ⋅ + b nb
the difference equation can be written as
Pole excess d = na − nb
Reciprocal polynomial
A ∗ (z) = 1 + a 1 z + ⋅ ⋅ ⋅ + a na z na = z na A(z −1 )
The system description in the backward shift operator
A(q)y(k) = B(q)u(k)
B(q)
A ∗ (q −1 )y(k) = B ∗ (q −1 )u(k − d)
y(k) =
u(k)
−1 )
A(q)
y(k) =
B ∗ (q
A ∗ (q
u(k − d)
−1 )
Pulse-transfer function operator
SISO systems
State-space system
B(q)
H(q) = C(qI − Φ) −1 Γ + D =
A(q)
x(k + 1) = qx(k) = Φx(k) + Γu(k)
If no common factors
Use the shift operator
deg A = n
(qI − Φ)x(k) = Γu(k)
A(q) = det[qI − Φ]
and
Eliminate x(k)
y(k) + a 1 y(k − 1) + ⋅ ⋅ ⋅ + a n y(k − n)
y(k) = Cx(k) + Du(k) = C(qI − Φ) −1 Γ + D u(k)
= b 0 u(k) + ⋅ ⋅ ⋅ + b n u(k − n)
Pulse-transfer operator
where a i are the coefﬁcients of the characteristic polynomial of
H(q) = C(qI − Φ) −1 Γ +
D
Φ.
In the backward-shift operator
H ∗ (q −1 ) = C(I − q −1 Φ) −1 q −1 Γ + D = H(q)
Poles, zeros, and system order
Example – Double integrator with delay
B(q)
h = 1 and τ = 0.5 gives
H(q) = C(qI − Φ) −1 Γ + D =
A(q)
1
1
0.375
0.125
Φ =
 
 
Γ 1 =
 
 
Γ 0 =
 
 
Poles: A(q) = 0
0
1
0.5
0.5
Zeros: B(q) = 0
Then
System order: deg A(q)
H(q) = C(qI − Φ) −1 (Γ 0 + Γ 1 q −1 )
q − 1
1
Important to use the forward shift operator for poles/zeros,
0
q − 1
system order, and stability.
0.125 + 0.375q −1
0.5 + 0.5q −1
=   1
0 
(q − 1) 2
The backward shift operator is suited for causality considera-
=
=
tions.
0.125(q 2 + 6q + 1)
q(q 2 − 2q + 1)
0.125(q −1 + 6q −2 + q −3 )
1 − 2q −1 + q −2
Order: 3
Poles: 0, 1, and 1
Zeros: −3 ± √ 8
How to get H(q) from G(s)?
z-transform
Use Table 2.1
Deﬁnition of z-transform
Consider the discrete-time signal { f(kh) : k = 0, 1,
. .
.}.
Zero-order hold sampling of a continuous-time system, G(s).
b 1 q n−1
+ b 2 q n−2 +
⋅ ⋅ ⋅ + b n
Z( f(kh)) = F(z) =
f(kh)z −k
H(q) =
q n +
a 1 q n−1 + ⋅ ⋅ ⋅
+ a n
k=0
The inverse transform is given by
G(s) H(q)
1
f(kh) =
2π i F(z)z k−1 dz
1
h
s
q−1
1
h 2 (q+1)
2(q−1) 2
where the contour of integration encloses all singularities of
s 2
F(z). Maps a semi-inﬁnite time sequence into a function of a
a
1−exp(−ah)
complex variable
s+a
q−exp(−ah)
Example
Properties of z-transform
Let y(kh) = kh for k ≥ 0. Then
1.
Deﬁnition.
Y(z) =
0 + hz −1 + 2hz −2
+
F(z) =
f(kh)z −k
k=0
=
h(z −1 + 2z −2 + ⋅ ⋅ ⋅
2.
Time shift.
hz
=
(z − 1) 2
Zq −n f = z −n F
Z{q n f } = z n (F − F 1 )
n−1
where F 1 (z) =
f(jh)z −j
j=0
• Similarities with Laplace transform
3.
Initial value theorem.
• Common in applied mathematics
4.
Final-value theorem.
• How the theory of sampled-data systems started
5.
Convolution.
k
Z( f ∗ ) = Z
f(n) (k − n) = (Z f)(Z )
n=0
Pulse-transfer function
Why both q and z?
x(k + 1) = Φx(k) + Γu(k)
• Could be sufﬁcient with only the shift operator q
y(k) = Cx(k) + Du(k)
• Many books contain the z-transform
Take the z-transform of both sides
• Must be aware of the difﬁculties with z-transform
z
z −k x(k) − x(0) =
Φz −k x(k) +
Γz −k u(k)
• Remember q operator and z complex variable
k=0
k=0
k=0
Hence
z(X (z) − x(0)) = ΦX (z) + ΓU(z)
X (z) = (zI − Φ) −1 (zx(0) + ΓU(z))
Y(z) = C(zI − Φ) −1 zx(0) + (C(zI − Φ) −1 Γ + D)U(z)
Pulse-transfer function
H(z) = C(zI − Φ) −1 Γ + D
A warning
Calculation of H(z) given G(s) using z-transform
tables
!!!Use the z-transform tables correctly!!!!
{u(kh)}
u(t)
y(t)
{ y(kh)}
Zero-order
G(s)
hold
f(kh)
Lf(t)
Z f(kh)
H(z)
δ (k) (pulse)
1
z
1
k ≥ 0 (step) 1
s
z − 1
1.
Determine the step response of the system with the
1
hz
kh
s 2
(z − 1) 2
transfer function G(s).
1
1
h 2 z(z + 1)
2 (kh) 2
2.
Determine the corresponding z-transform of the step
s 3
2(z − 1) 3
response using the table.
T
z
e −kh/T
1 + sT
3.
Divide by the z-transform of the step function.
1
z − e −h/T
z(1 − e −h/T )
1 − e −kh/T
s(1 + sT)
(z − 1)(z − e −h/T )
G(s)
˜
Y(s) =
→ Y = Z(L −1 Y)
s
Warning. Notice that Z f in the table does not give the zero-
˜
→ H(z) = (1 − z −1 ) Y(z)
order-hold sampling of a system with the transfer function Lf.
Double integrator – Sampling using table
Formula for H(z)
Transfer function G(s) = 1/s 2
The following formula can be derived:
Introduce the step
2π i γ +i∞
e sh
z − 1 1
G(s)
1
H(z) =
ds
Y(s) =
z
z − e sh s
γ −i∞
s 3
If G(s) goes to zero at least as fast as s −1 for a large s and
Use the table
has distinct poles (none at the origin)
h 2 z(z + 1)
˜
Y = Z(L −1 Y) =
2(z − 1) 3
1
H(z) =
G(s)
Get the pulse transfer function
e sh Res e sh − 1
z −
s
s=s i
h 2 (z + 1)
˜
H(z) = (1 − z −1 ) Y(z) =
where s i are the poles of G(s)
2(z − 1) 2
Multiple poles inﬂuence the calculations of the residues.
Modiﬁed z-transform
Interpretation of poles and zeros
Can be used to determine intersample behavior
Poles:
Deﬁnition: Modiﬁed z-transform
A pole z = a is associated with the time function z(k) = a k
A pole z = a is an eigenvalue of Φ
F(z, ˜ m) =
z −k f(kh − h + mh),
0 ≤ m ≤ 1
Zeros:
k=0
The inverse transform is given by
A zero z = a implies that the transmission of the input
u(k) = a k is blocked by the system
1
f(nh − h + mh) =
F(z, ˜ m)z n−1 dz
• A zero is related to how inputs and outputs are coupled to
2π i Γ
the states
Γ encloses all singularities of the integrand
Transformation of poles λ i (Φ) = e λ i (A)h
New evidence of alias problem
s
z
z = e sh
ω N
Several points in the s-plane is mapped into the same point in
the z-plane.
The map is not bijective
−ω N
ω N
3 π / h
p 2
x
π / h
p
−ω N
p 1
x
x
S 0
p 1
x
x
− π / h
ω N
p 2
x
−3π / h
−ω N
Sampling of a second order system
Transformation of zeros
ω
+ 2ζ ω 0 s + ω
2
More difﬁcult than poles
0
2
s 2
In general, more sampled zeros than continuous
0
For short sampling periods z i e s i h
Poles of the discrete-time system are given by the mapping
For large
s then G(s) s −d
where d = deg A(s) − deg B(s)
1
ζ= 0
The r = d − 1 sampling zeros go to the zeros of the polynomi-
als Z d
ζ = 0.2
d Z d
ζ = 0.4
0.5
1
1
ζ = 0.6
ζ = 0.8
2
z + 1
3
z 2 +
4z + 1
0
ζ = 1.0
4
z 3 + 11z 2 + 11z + 1
−1
−0.5
0
0.5
1
5
z 4 + 26z 3 + 66z 2 + 26z + 1
Real axis
Imaginary axis
Systems with unstable inverse
Selection of sampling period
Continuous-time system is nonminimum phase if it has right
(a)
1
1
half-plane zeros or time delays.
0
Number of samples per
−1
0
0
5
0
5
A discrete-time system is in many books deﬁned to be non-
rise time
(b)
1
1
minimum phase if it has zeros outside the unit disc
0
T r
N r =
4 − 10
−1
0
h
0
5
0
5
We will use the following notation:
(c)
1
1
The rise times of the
0
Deﬁnition – Unstable inverse
signals are T r = 1.
−1
0
A discrete-time system has an unstable inverse if it has zeros
a)
N r = 1, b) N r = 2,
0
5
0
5
(d)
1
outside the unit disc
c)
N r = 4, d) N r = 8
1
0
?
−1
0
Nonminimum phase Unstable inverse
0
5
0
5
Time
Time
Second order system
Pole-zero variation with h
1
G(s) =
T r
(s + 1)(s 2 + s + 1)
N r =
4 − 10
h
Corresponds to (for
(a)
(b)
dominating modes)
1
1
1
ω 0 h 0.2 − 0.6
0
0
0
5
0
5
h=0+
h=0+
h=0+
(c)
(d)
ζ = 0.5, ω 0 = 1.83 gives
0
1
1
T r = 1;
a) h = 0.125 (ω 0 h =
0
0
0
5
0
5
−1
0.23)
Time
Time
b)
h = 0.25 (ω 0 h = 0.46)
c)
h = 0.5 (ω 0 h = 0.92)
−4
−3
−2
−1
0
1
Real axis
d)
h = 1.0 (ω 0 h = 1.83)
h = 0.0001, 0.2, 0.5, 1, 2, and 3
Imaginary axis
Nyquist and Bode diagrams
Example
Nyquist curve: H(e iωh )
0
for ωh ∈ [0,π ], i.e. up to ω N
−0.5
• Periodic
1
G(s) =
• Interpretation
s 2 + 1.4s + 1
−1
−0.5
0
0.5
1
Real axis
• Higher order harmonics
Zero-order hold sampling
1
• Discuss more in connection with Chapter 7
h = 0.4
0.01
0.066z + 0.055
H(z) =
0.1
1
10
z 2 − 1.450z + 0.571
Continuous-time (dashed),
0
discrete-time (full)
−180
0.1
1
10
A. M. Lyapunov
Lyapunov theory
1857–1918
Consider the system
x(k + 1) = f(x(k)),
f(0) = 0
Monotonic convergence x(k + 1) < x(k) a too strong
condition for stability
Find other "norm", a Lyapunov function V(x)
V(x) is
continuous in x and V(0) = 0
• V(x) is positive deﬁnite
∆V(x) = V( f(x)) − V(x) is negative deﬁnite
• V(x) → ∞, x → ∞
Existence of Lyapunov function implies asymptotic stability for
the solution x = 0
Phase
Gain
Imaginary axis
Geometric interpretation
Linear system
x(k + 1) = f(x(k)),
f(0) = 0
x(k + 1)
= Φx(k)
V(x(k + 1))
x 2
x (k + 1)
V(x) = x T Px
P > 0
∆V(x) = V(Φx) − V(x) = x T Φ T PΦx − x T Px
= x T Φ T PΦ − P x = −x T Qx
x(k)
V is a Lyapunov function iff there exists a P > 0 that satisﬁes
the Lyapunov equation
Φ T PΦ − P = −Q
Q > 0
x 1
V(x(k))
Example
Summary
0.4
0
1
0
• Piecewise constant input and periodic sampling gives time-
Φ =
 
 
Q =
 
 
−0.4 0.6
0
1
invariant discrete-time system
2
• Solution of the system equation, λ(Φ)
• Shift operator q and pulse transfer operator
• z-transform and pulse transfer function
• Be careful with z-transform tables
0
• Poles, zeros, and system order
• Selection of sampling period
T r
N r =
4 − 10
h
−2
−2
0
2
ω 0 h 0.2 − 0.6
State x 1
• Frequency function
State x 2