1.

Inverse Z-transform - Partial Fraction

Find the inverse Z-transform of

2z 2 + 2z
G(z) = 2
z + 2z − 3
G(z) 2z + 2
=
z (z + 3)(z − 1)
A B
= +
z+3 z−1
Multiply throughout by z + 3 and let z = −3 to get

2z + 2  −4
A=  = =1
z − 1 z=−3 −4
Digital Control 1 Kannan M. Moudgalya, Autumn 2007
2. Inverse Z-transform - Partial Fraction

G(z) A B
= +
z z+3 z−1
Multiply throughout by z − 1 and let z = 1 to get
4
B= =1
4
G(z) 1 1
= + |z| > 3
z z+3 z−1
z z
G(z) = + |z| > 3
z+3 z−1
↔ (−3)n1(n) + 1(n)
Digital Control 2 Kannan M. Moudgalya, Autumn 2007
3. Partial Fraction - Repeated Poles

N (z)
G(z) =
(z − α)pD1(z)
α not a root of N (z) and D1(z)
A1 A2 Ap
G(z) = + + ··· + + G1(z)
z−α (z − α)2 (z − α)p
G1(z) has poles corresponding to those of D1(z).
Multiply by (z − α)p
(z − α)pG(z) = A1(z − α)p−1 + A2(z − α)p−2 + · · ·
+ Ap−1(z − α) + Ap + G1(z)(z − α)p

Digital Control 3 Kannan M. Moudgalya, Autumn 2007

4. Partial Fraction - Repeated Poles

(z − α)pG(z) = A1(z − α)p−1 + A2(z − α)p−2 + · · ·

+ Ap−1(z − α) + Ap + G1(z)(z − α)p
Substituting z = α,
Ap = (z − α)pG(z)|z=α
Differentiate and let z = α:
d
Ap−1 = (z − α)pG(z)|z=α
dz
1 dp−1 p
Continuing, A1 = (z − α) G(z)|z=α
(p − 1)! dz p−1

Digital Control 4 Kannan M. Moudgalya, Autumn 2007

5. Repeated Poles - an Example

11z 2 − 15z + 6 A1 A2 B
G(z) = = + +
(z − 2)(z − 1)2 z−1 (z − 1)2 z−2
× (z − 2), let z = 2, to get B = 20. × (z − 1)2,
11z 2 − 15z + 6 (z − 1)2
= A1(z − 1) + A2 + B
z−2 z−2
With z = 1, get A2 = −2.
Differentiating with respect to z and with z = 1,

(z − 2)(22z − 15) − (11z − 15z + 6)  2
A1 =  = −9
(z − 2) 2
z=1
9 2 20
G(z) = − − +
z−1 (z − 1)2 z−2
Digital Control 5 Kannan M. Moudgalya, Autumn 2007
6. Important Result from Differentiation

Problem 4.9 in Text: Consider

X∞
z  −1
1(n)an ↔ = anz −n, az  < 1
z−a n=0

Differentiating with respect to a,

∞
X
z n−1 −n n−1 z
= na z , na 1(n) ↔
(z − a)2 n=0
(z − a)2
Substituting a = 1, can obtain the Z-transform of n:
z
n1(n) ↔
(z − 1)2
Through further differentiation, can derive Z-transforms of n2,
n3, etc.
Digital Control 6 Kannan M. Moudgalya, Autumn 2007
7. Repeated Poles - an Example

9 2 20
G(z) = − − +
z−1 (z − 1)2 z−2
9z 2z 20z
zG(z) = − − +
z−1 (z − 1)2
z−2
↔ (−9 − 2n + 20 × 2n)1(n)
Use shifting theorem:
G(z) ↔ (−9 − 2(n − 1) + 20 × 2n−1)1(n − 1)

Digital Control 7 Kannan M. Moudgalya, Autumn 2007

8. Properties of Cont. Time Sinusoidal Signals

ua(t) = A cos (Ωt + θ)

ua(t) = A cos (2πF t + θ), − ∞ < t < ∞
A: amplitude
Ω: frequency in rad/s
θ: phase in rad
F : frequency in cycles/s or Hertz
Ω = 2πF
1
Tp =
F
Digital Control 8 Kannan M. Moudgalya, Autumn 2007
9. Properties of Cont. Time Sinusoidal Signals

ua(t) = A cos (Ωt + θ)

ua(t) = A cos (2πF t + θ), − ∞ < t < ∞
With F fixed, periodic with period Tp
1
ua[t + Tp] = A cos (2πF (t + ) + θ)
F
= A cos (2π + 2πF t + θ)
= A cos (2πF t + θ) = ua[t]

Digital Control 9 Kannan M. Moudgalya, Autumn 2007

10. Properties of Cont. Time Sinusoidal Signals

Cont. signals with different frequencies are different

   
t 7t
u1 = cos 2π , u2 = cos 2π
8 8
1

0.8

0.6

0.4

0.2
coswt

−0.2

−0.4

−0.6

−0.8

−1
0 0.5 1 1.5 2 2.5 3 3.5 4
t

Different wave forms for different frequencies.

Can increase f all the way to ∞ or decrease to 0.
Digital Control 10 Kannan M. Moudgalya, Autumn 2007
11. Discrete Time Sinusoidal Signals

u(n) = A cos (wn + θ), −∞ < n < ∞

w = 2πf
n integer variable, sample number
A amplitude of the sinusoid
w frequency in radians per sample
θ phase in radians.
f normalized frequency, cycles/sample

Digital Control 11 Kannan M. Moudgalya, Autumn 2007

12. Discrete Time Sinusoids - Identical Signals

Discrete time sinusoids whose frequencies are sepa-

rated by integer multiple of 2π are identical, i.e.,
cos ((w0 + 2π)n + θ) = cos (w0n + θ), ∀n
• All sinusoidal sequences of the form,
uk (n) = A cos (wk n + θ),
wk = w0 + 2kπ, −π < w0 < π,
are indistinguishable or identical.
• Only sinusoids in −π < w0 < π are different
1 1
−π < w0 < π or − < f0 <
2 2
Digital Control 12 Kannan M. Moudgalya, Autumn 2007
13. Sampling a Cont. Time Signal - Preliminaries

Let analog signal ua[t] have a frequency of F Hz

ua[t] = A cos(2πF t + θ)
Uniform sampling rate (Ts s) or frequency (Fs Hz)
n
t = nTs =
Fs
u(n) = ua[nTs] − ∞ < n < ∞
= A cos (2πF Tsn + θ)
 
F
= A cos 2π n + θ
Fs
4
= A cos(2πf n + θ)
Digital Control 13 Kannan M. Moudgalya, Autumn 2007
14. Sampling a Cont. Time Signal - Preliminaries

In our standard notation,

u(n) = A cos (2πf n + θ)
It follows that
F Ω
f = w= = ΩTs
Fs Fs
Reason to call f normalized frequency, cycles/sample.
Apply the uniqueness condition for sampled signals
1 1
− <f < −π < w < π
2 2
Fs π
Fmax = Ωmax = 2πFmax = πFs =
2 Ts
Digital Control 14 Kannan M. Moudgalya, Autumn 2007
15. Properties of Discrete Time Sinusoids - Alias

• As w0 ↑, freq. of oscillation ↑, reaches maximum

at w0 = π
• What if w0 > π?

w1 = w0
w2 = 2π − w0
u1(n) = A cos w1n = A cos w0n
u2(n) = A cos w2n = A cos(2π − w0)n
= A cos w0n = u1(n)
w2 is an alias of w1
Digital Control 15 Kannan M. Moudgalya, Autumn 2007
16. Properties of Discrete Time Sinusoids - Alias

t 7t
u1[t] = cos (2π ), u2[t] = cos (2π ), Ts = 1
8 8
7n 1
u2(n) = cos (2π ) = cos 2π(1 − )n
8 8
2π 2πn
= cos (2π − )n = cos ( ) = u1(n)
8 8
1

0.8

0.6

0.4

0.2
coswt

−0.2

−0.4

−0.6

−0.8

−1
0 0.5 1 1.5 2 2.5 3 3.5 4
t

Digital Control 16 Kannan M. Moudgalya, Autumn 2007

17. Fourier Transform of Aperiodic Signals

Z ∞
X(F ) = x(t)e−j2πF tdt
Z−∞
∞
x(t) = X(F )ej2πF tdF
−∞
If we let radian frequency Ω = 2πF
Z ∞
1
x(t) = X[Ω]ejΩtdΩ,
2π −∞
Z ∞
X[Ω] = x(t)e−jΩtdt
−∞
x(t), X[Ω] are Fourier Transform Pair
Digital Control 17 Kannan M. Moudgalya, Autumn 2007
18. Frequency Response

Apply u(n) = ejwn to g(n) and obtain output y:

∞
X
y(n) = g(n) ∗ u(n) = g(k)u(n − k)
k=−∞
∞
X ∞
X
= g(k)ejw(n−k) = ejwn g(k)e−jwk
k=−∞ k=−∞

Define Discrete Time Fourier Transform


∞ ∞ 
4 X X 
−k 
G(ejw ) = g(k)e−jwk = g(k)z 
k=−∞ k=−∞ 
z=ejw
= G(z)|z=ejw
Provided the sequence converges absolutely
Digital Control 18 Kannan M. Moudgalya, Autumn 2007
19. Frequency Response

∞
X
y(n) = ejwn g(k)e−jwk = ejwnG(ejw )
k=−∞

Write in polar coordinates: G(ejw ) = |G(ejw )|ejϕ - ϕ is

phase angle:
y(n) = ejwn|G(ejw )|ejϕ = |G(ejwn)|ej(wn+ϕ)
1. Input is sinusoid ⇒ output also is a sinusoid with following
properties:
• Output amplitude gets multiplied by the magnitude of
G(ejw )
• Output sinusoid shifts by ϕ with respect to input
Digital Control 19 Kannan M. Moudgalya, Autumn 2007
 2. At ω where |G(ejw )| is large, the sinusoid gets amplified
and at ω where it is small, the sinusoid gets attenuated.
• The system with large gains at low frequencies and small
gains at high frequencies are called low pass filters
• Similarly high pass filters

Digital Control 20 Kannan M. Moudgalya, Autumn 2007

20. Discrete Fourier Transform

Define Discrete Time Fourier Transform


∞
X X∞ 
4 
−k 
G(ejw ) = g(k)e−jwk = g(k)z 
k=−∞ k=−∞ 
z=ejw
= G(z)|z=ejw
Provided, absolute convergence
∞
X ∞
X
|g(k)e−jwk | < ∞ ⇒ |g(k)| < ∞
k=−∞ k=−∞

For causal systems, BIBO stability. Required for DTFT.

Digital Control 21 Kannan M. Moudgalya, Autumn 2007

21. FT of Discrete Time Aperiodic Signals - Defi-
nition

∞
X
4
U (ejω ) = u(n)e−jωn
n=−∞
Z π
1
u(m) = U (ejω )ejωmdω
2π −π
Z 1/2
= U (f )ej2πf mdf
−1/2

Digital Control 22 Kannan M. Moudgalya, Autumn 2007

22. FT of a Moving Average Filter - Example

1
y(n) = [u(n + 1) + u(n) + u(n − 1)]
3
X∞
y(n) = g(k)u(n − k)
k=−∞
= g(−1)u(n + 1) + g(0)u(n) + g(1)u(n − 1)
1
g(−1) = g(0) = g(1) =
3
X∞
jw


G e = g(n)z −n|z=ejw
n=−∞
1 jw −jw

1
= e +1+e = (1 + 2 cos w)
3 3
Digital Control 23 Kannan M. Moudgalya, Autumn 2007
 23. FT of a Moving Average Filter - Example

jω

1 jw

1
G e = (1 + 2 cos w), |G e | = | (1 + 2 cos w)|
3
( 3
0 0 ≤ w < 2π3
Arg(G) =
π 2π
3
≤w<π
1 / / U p d a t e d ( 1 8 − 7 − 0 7 )

2 / / 5 . 3

4 w = 0 : 0 . 0 1 : %pi ;
5 subplot ( 2 , 1 , 1 ) ;
6 plot2d1 ( ” g l l ” ,w , abs(1+2∗ cos (w) ) / 3 , s t y l e = 2 ) ;
7 l a b e l ( ’ ’ , 4 , ’ ’ , ’ Magnitude ’ , 4 ) ;
8 subplot ( 2 , 1 , 2 ) ;
9 plot2d1 ( ” g l n ” ,w , phasemag(1+2∗ cos (w ) ) , s t y l e = 2 , r e c t =[0.
10 l a b e l ( ’ ’ , 4 , ’w ’ , ’ Phase ’ , 4 )
Digital Control 24 Kannan M. Moudgalya, Autumn 2007
24. FT of a Moving Average Filter - Example

jw

1
|G e | = | (1 + 2 cos w)|,
(3
0 0 ≤ w < 2π 3
Arg(G) = 2π
π 3 ≤w<π
0
10
Magnitude

−1
10

−2
10

−3
10
−2 −1 0 1
10 10 10 10

200

150
Phase

100

50

0
−2 −1 0 1
10 10 10 10
w
Digital Control 25 Kannan M. Moudgalya, Autumn 2007
25. Additional Properties of Fourier Transform

Symmetry of real and imaginary parts for real valued sequences

∞
X
jw


G e = g(n)e−jwn
n=−∞
X∞ ∞
X
= g(n) cos wn − j g(n) sin wn
n=−∞ n=−∞
X∞
−jw


G e = g(n)ejwn
n=−∞
X∞ ∞
X
= g(n) cos wn + j g(n) sin wn
n=−∞ n=−∞
 jw

  −jw


Re G e = Re G e
 jw

  −jw


Im G e = −Im G e
Digital Control 26 Kannan M. Moudgalya, Autumn 2007
26. Additional Properties of Fourier Transform

Recall
jw

∗ −jw


G e =G e
Symmetry of magnitude for real valued sequences
jw

 jw

∗ jw
1/2
|G e |= G e G e
 ∗ −jw


−jw 1/2
= G e G e
= |G(e−jw )|
This shows that the magnitude is an even function. Similarly,
 −jw

  jw


Arg G e = −Arg G e
⇒ Bode plots have to be drawn for w in [0, π] only.

Digital Control 27 Kannan M. Moudgalya, Autumn 2007

27. Sampling and Reconstruction

u(n) = ua(nTs), −∞ < n < ∞

Fast sampling:
ua (t) Ua(F )
1

Fs
F0 − Fs 2
t F
−B F0 B F0 + Fs
 
F
ua (nT ) = u(n) U Fs

Fs

t F
Ts − 32 Fs − F2s Fs
2
3
F
2 s
Digital Control 28 Kannan M. Moudgalya, Autumn 2007
28. Slow Sampling Results in Aliasing

ua(t) Ua(F )

t F
−B B

 
F
U Fs

Fs

t F
Ts − F2s Fs
2

 
F
U Fs

u(n) Fs

t F
Ts − F2s Fs
2

Digital Control 29 Kannan M. Moudgalya, Autumn 2007

29. What to Do When Aliasing Cannot be Avoided?
Ua(F )

 
1 F
Fs
U Fs

Ua(F + Fs ) Ua(F ) Ua(F − Fs )

Ua′ (F )

 
1 F
Fs
U′ Fs

Ua′ (F + Fs ) Ua(F ) Ua′ (F − Fs )

Digital Control 30 Kannan M. Moudgalya, Autumn 2007

30. Sampling Theorem

• Suppose highest frequency contained in an analog

signal ua(t) is Fmax = B.
• It is sampled at a rate Fs > 2Fmax = 2B.
• ua(t) can be recovered from its sample values:
n o
∞
X sin Tπ (t − nTs)
s
ua(t) = ua(nTs) π
n=−∞ T (t − nTs)
s

• If Fs = 2Fmax, Fs is denoted by FN , the

Nyquist rate.
• Not causal: check n > 0
Digital Control 31 Kannan M. Moudgalya, Autumn 2007
31. Rules for Sampling Rate Selection

• Minimum sampling rate = twice band width

• Shannon’s reconstruction cannot be implemented,
have to use ZOH
• Solution: sample faster
– Number of samples in rise time = 4 to 10
– Sample 10 to 30 times bandwidth
– Use 10 times Shannon’s sampling rate
– ωcTs = 0.15 to 0.5, where,
ωc = crossover frequency
Digital Control 32 Kannan M. Moudgalya, Autumn 2007