Chapter 15

Frequency Response of Linear

Systems

In the previous chapters we were concerned about the response of a linear system to
inputs like step and ramp signals. What about periodic inputs? How does a linear system
respond to, say, sinusoidal input? Does this response reveal important characteristics of
the system? Can these be used to design controllers? This is the subject matter of the
present chapter.

15.1 Response to Sinusoidal Input

Consider a system where G(s) is either a closed-loop or an open-loop system.

Figure 15.1: Input-output of a system

Frequency response analysis asks the following question:

For

r(t) = A sin ωt

170
Lecture Notes on Control Systems/D. Ghose/2012 171

what is y(t) ? In other words, we want to know what’s the response to a sinusoidal
input.
For a linear system driven by a sine wave, the output is also a sine wave with the same
frequency as the input but with a different amplitude and phase. To prove this assertion,
let us consider the frequency response which is related to a transfer function G(s) as
follows: Let,

r(t) = A sin ωt

Then
Aω
R(s) =
s2 + ω 2 m
(s − zk ) Aω
Y (s) = G(s)R(s) = K k=1
n
j=1 (s − pj ) s + ω
2 2

For simplicity, assume that G(s) has no repeated poles. Then, using partial fraction
expansion,
k1 kn B1 s + B2
Y (s) = + ··· + + 2
s − p1 s − pn s + ω2
On taking inverse Laplace transform, we get,

y(t) = k
1
ep1 t + ...

+ kn epnt + B sin(ωt + φ)
  
transient response forced response

Assume G(s) stable so that all the poles are negative, i.e., all pi < 0.

epj t → 0, as t → ∞

So, in the steady state,

yss (t) = B sin(ωt + φ)

So,
B1 s + B2
Yss (s) =
s2 + ω 2

To find B1 and B2 :

(s2 + ω 2 )Y (s)|s→jω = (B1 s + B2 )|s→jω

Aω
= (s2 + ω 2 )G(s) |s→jω
s2 + ω2
jωB1 + B2 = AωG(jω)
Lecture Notes on Control Systems/D. Ghose/2012 172

Now G(jω) is a complex number. So,

G(jω) = |G(jω)|ejθ(ω)
= |G(jω)| cos θ(ω) + j|G(jω)| sin θ(ω)
= X(ω) + jY (ω) (say)

with
 
−1 Y (ω)
θ(ω) = tan
X(ω)

|G(jω)| = X(ω)2 + Y (ω)2

So,

jωB1 + B2 = AωG(jω)
= AωX(ω) + jAωY (ω)

From which,

ωB1 = AωY (ω) ⇒ B1 = AY (ω)

B2 = AωX(ω)

Then,
B1 s + B2 s ω
Yss (s) = 2 2
= A[Y (ω) 2 2
+ X(ω) 2 ]
s +ω s +ω s + ω2
yss (t) = A[Y (ω) cos ωt + X(ω)sinωt]
= A[|G(jω)|sinθ(ω) cos ωt + |G(jω)| cos θ(ω)sinωt]
= A|G(jω)|sin(ωt + θ(ω))

where,

Y (ω)
θ(ω) = tan−1 =  G(jω)
X(ω)

Summarizing the above results:

For Y (s) = G(s)R(s)

r(t) = A sin ωt ⇒ yss (t) = B sin(ωt + θ)

B
= |G(jω)|, θ =  G(jω)
A
So, given G(s) as the system transfer function, we evaluate G(s) along the imaginary
axis to get the frequency response.
Lecture Notes on Control Systems/D. Ghose/2012 173

15.2 System Modelling

Frequency response may be used to describe the system model. Let g(t) be the system
impulse response. Then,

∞
G(s) = L[g(t)] = g(t)e−st dt
0

Substituting s = jω,

∞
∞
G(jω) = g(t)e−jωt dt = g(t)e−jωt dt = F[g(t)]
0 −∞   
  
since g(t) = 0 for t < 0 Fourier transform

So, now we have 3 different ways of modelling a system.

1. Time domain: Through impulse response g(t).

2. s-domain: Through Laplace transform L(.).


(s − zk )
G(s) = 
(s − pj )

3. Frequency domain: Through Fourier transform F(.).

G(jω) and  G(jω)

The system characteristics described by each kind of model is given in the table below.

Characteristics time domain s-domain frequency domain

Speed

Rise time Tr ωn bandwith

Settling time Ts ζωn –
Accuracy

Steady state error FVT: lims→0 sY (s) DC gain

Overshoot Mp ζ phase margin
Stability

Response to initial gain margin > 0

condition dies down Re(pi ) < 0 phase margin > 0
Lecture Notes on Control Systems/D. Ghose/2012 174

15.3 Frequency Response Plots

Plots of the frequency response of a system come in several forms. However, note that
each is a plot of the open-loop frequency response, but it gives information about closed
loop characteristics. Consider,

Figure 15.2: A closed loop system

F (s)
The open-loop transfer function E(s) = H(s)P (s)K(s) = G(s). There are three ways in
which we can plot the frequency response. These are:

1. Bode plots : |G(jω)| vs. ω

 G(jω) vs. ω

2. Nyquist plots : Im[G(jω)] vs. Re[G(jω)]

3. Nichols plots : |G(jω)| vs.  G(jω)

Each plot gives the same information but with different interpretation.

15.3.1 Bode Plot

Consider the system below.

Figure 15.3: A simple closed loop system

It has open loop transfer from G(s)K(s) with the decomposition,

(s + z1 )...(s + zm )
G(s)K(s) = K̄
(s + p1 )...(s + pn )
Lecture Notes on Control Systems/D. Ghose/2012 175

This is also called the root locus form.

(jω + z1 )..(jω + zm )
G(jω)K(jω) = K̄
(jω + +p1 )..(jω + pn )
   
jω
z1 z2 · · · zm 1 + z1
· · · 1 + zjω
m
= K̄  
p1 p2 · · · pn 1 + jωp1
) · · · (1 + jω
pn

z1 ..zm rz1 ejθz1 · · · rzm ejθzm
= K̄
p1 ..pn rp1 rjθp1 · · · rpn ejθpn
rz ejθz1 · · · rzm ejθzm
= K 1 jθp
rp1 e 1 · · · rpn ejθpn

(this is often called the Bode form). In the above,


z1 ..zm
K = K̄
p1 ..pn

rz1 ...rzm j[(θz1 +···+θzm )−(θp1 +···+θpn )]

G(jω)K(jω) = K e
rp1 ...rpn
rz1 · · · rzm
|G(jω)K(jω)| = K
rp1 ...rpn
 G(jω)K(jω) = (θz + · · · + θz ) − (θp + · · · + θp )
1 m 1 n

To make plotting easier, the gain |G(jω)K(jω)| is plotted in decibels.

dB(x) = 20 log10 (x)

In the following, for simplicity, we will drop the base ’10’ from the expressions involving
logarithms. So,

rzk rpj
⎡ ⎤
     
m
  jω  n  
jω ⎥
⎢ 
20 log |GK(jω)| = 20 ⎣log K + log 1 + − log 1 + ⎦
  zk pj
k=1 j=1
m  n 
 jω  jω
 GK(jω) =  1+ −  1+
k=1 zk j=1 pj

Note (for quick plots)

1
1 ≡ 0 dB 0.1 ≡ −20 dB 10 ≡ 20 dB 2
≡ -6 dB
√
2 ≡ 6 dB √1 ≡ −3 dB 2 ≡ 3dB
2

Summary
Lecture Notes on Control Systems/D. Ghose/2012 176

Figure 15.4: A system with input and output

For the system given above

r(t) = A sin ωt ⇒ yss (t) = B sin(ωt + θ)

B
= |G(jω)|, θ =  G(jω)
A

1. Bode Plot: |G(jω)| vs. ω

 G(jω) vs. ω

2. Nyquist Plot: Im[G(jω)] vs. Re[G(jω)] (amplitude plot)

3. Nichols Plot: |G(jω)| vs.  G(jω) (phase plot)