Frequency Response and Bode Plots

Advantages of root locus design approach:

 Good indicator of transient response.
 Explicitly shows location of closedloop poles.
 Tradeos are clear.
Disadvantages of root locus design approach:
 Requires the transfer function of the plant to be known.
 Dicult to infer all performance values
.  Hard to extract steady state response (for sinusoidal inputs).
Frequency response methods can be used to supplement root locus:
 Can infer performance and stability from same plot.
 Can use measured data when no model is available.
 Design process is independent of system order (number of poles).
 Time delays are handled correctly.
 Graphical techniques (analysis/synthesis) are quite simple.
To design a controller for more complicated reference inputs than simple
step function, we need a method to approximately determine the output for
any arbitrary input. Another reason such a method is required is to analyze
the eect of disturbance and noise on the output, since the controller has to be
designed so that their eect is small.
This lecture is concerned with the question: given a transfer function and an
arbitrary input signal that may not be a step, how do we quickly obtain quali-
tative information about the output without doing complicated calculations?

1
 where A is called the amplitude and ω is called the frequency of oscilla-
tion (whose unit is rad/sec) and φ is called the phase angle. We dene the
frequency response of a system as the steady-state response of the system to a
sinusoidal input. When a system is subject to a sinusoidal input of frequency ω,
the response of the system is also sinusoidal at the same frequency. The output
sinusoid diers from the input sinusoid only in amplitude and phase angle.
The frequency response of a system G(s) is simply the transfer function's val-
ues on the imaginary axis: G(jω) for ω ∈ [0, ∞). G(jω) is a complex function
of the real variable ω. The real functions |G(jω)| and ∠G(jω) are called the
magnitude response and phase response, respectively, of the transfer func-
tion G(s). The sinusoidal response result is important because it shows that a
sinusoidal input essentially pass through a stable system without any
change in frequency, only the amplitude and phase changes. Moreover,
the amplitude changes by a factor that is precisely equal to the the system's
magnitude response at the frequency of the input, and the phase shifts by an
amount exactly equal to the phase response of the system at the frequency of
the input.

2
 20
Example: Let G (s) = s2 +2s+8 and u (t) = sin (2t), what is y (t) at steady
state?

y (t) ≈ 1 × |G (j2)| sin (2t + ∠G (j2))

20 20 5
G (j2) = 2 = = = 3.53∠ − 0.7854
(j2) + 2 × j2 + 8 4 + j4 1+j
Therefore,

y (t) = 3.53 sin (2t − 0.78)

20
Example: Let G (s) = s2 +2s+8 and u (t) = sin (10t), what is y (t) at steady
state?

y (t) ≈ 1 × |G (j10)| sin (10t + ∠G (j10))

20
G (j10) = 2 = 0.21∠ − 2.93
(j10) + 2 × j10 + 8
Therefore,

y (t) = 0.21 sin (10t − 2.93)

3
Interesting Notes:

• Any signal can be decomposed into a sum of sinusoids, approximately.

• For a linear dynamic system, if the input consists of a sum of signals, the
output is the sum of outputs in response to each component of the input.

• The frequency response of a system - whose transfer function is not known

- can be determined experimentally. The system is driven with a sinusoidal
input at some known frequency, say ω1 , and a known amplitude. From
the amplitude and phase of the steady state sinusoidal response, the gain
and phase of the system at that frequency are determined. This process is
repeated at various frequencies ωi ,i = 1, ..., N . At the end, we have a list
of the gain and phase at those frequencies |G(jωi )|, ∠G(jωi ). A transfer
function G(s) is now tted to this data, so that the frequency response
of the system matches with the data at the frequencies frequencies ωi
,i = 1, ..., N . The MATLAB command invf reqs is useful in the nal
step. Obtaining models of dynamic systems from experimental data is the
subject of an entire eld of study, called system identication.

• The word gain is common in the study of control systems. It can be

confusing at rst since the same word means dierent things in dierent
contexts. Let us summarize all the instances of gain we have encountered
so far.

1. The gain in zero-pole-gain representation of a transfer function: G (s) =

k(s−z1 )(s−z2 )...
(s−p1 )(s−p2 )...
2. P gain, I gain, and D gain: These occurs in the context of PID
controllers. The PID control command is given by the formula:u (t) =
KP e + KI t0 e (θ) dθ + KD ė (t). The design variables KP
R
is called the
P-gain, KI , the I-gain and KD , the D-gain.

3. DC gain (also known as low frequency gain): The DC gain of a

transfer function G(s) is the real number G(0).
4. System gain, or gain of a transfer function: The gain of a transfer
function at frequency x. For a transfer function G(s), this refers to
the magnitude response |G(jx)| at the frequency x.
• Can we talk about one transfer function G1 being larger or smaller than
another transfer function G2 ? Since G1 (s) and G2 (s) are complex num-
bers, it is not possible to compare them directly. However, we can compare
the magnitudes of their frequency responses. For a particular frequency
ω0 , the numbers |G1 (jω0 )| and |G2 (jω0 )| are real numbers (in fact they
are non-negative) so they can be compared as being larger or smaller. And
such comparisons are useful since the system gains|G1 (jω0 )| and |G2 (jω0 )|
are exactly the factors by which the amplitude of a sinusoidal input with
frequency ω0 will be multiplied by at the output. So a plant that has
a higher gain than another plant (at the same frequency) will produce a
larger output for the same sinusoidal input at that frequency.

4
 Graphical representation of frequency response
The magnitude and phase response can be plotted against ω. The plot of
the magnitude response|G(jω)| vs. ω is the slice of the surface plot of |G(s)|
along the imaginary axis. Similarly, the plot of the phase response ∠G(jω) vs.
ω is the slice of the surface plot of ∠G(s) along the imaginary axis.

Evaluating the equation Y (s) = G(s)U (s) at s = jω leads to Y (jω) =

G(jω)U (jω). By taking the magnitude of both sides, and the phase of both
side, we get the following two relations:

|Y (jω)| = |G(jω)||U (jω)|

∠Y (jω) = ∠G(jω) + ∠U (jω)

The transfer function acts on the inputs frequency by frequency, by modify-
ing the amplitude and phase.

5
 100
Consider the system is G (s) = s2 +2s+100 and the input signal u(t) is such
that



 0.1 ω < 10

2 10 ≤ ω ≤ 20
|U (jω)| =


 0.1 20 ≤ ω ≤ 60
0 ω > 60

(
π
− 0.01ω ω ≤ 60
∠U (jω) = 2
0 ω > 60

• Notice that at every frequency ω0 , the output's Fourier transform mag-

nitude |Y (jω0 )| is increased or decreased over that of the input, |U (jω0 )|,
depending on whether the system gain at that frequency, |G(jω0 )|, is
larger than 1 or smaller than 1.

• The output has high energy in frequency near 10 since the plant has high
gain at that frequency. That is, sinusoidal components of the input at
frequency near 10 rad/s will be amplied at the output.

• Sinusoidal components of the input at frequency near 40 rad/s would be

suppressed at the output since the plant has a gain much smaller than 1
near 40 rad/s.

• Notice also that the input is band limited to [10, 20] rad/s.

• Its energy in that band is 22 × (20 − 10) = 40 while its energy in the rest
of the frequencies is 0.01 × (10 + 40) = 5 × 10−3 , which is negligible
2

compared to 40.

6
 Bode plot

The Bode plot of G(s) is a plot of magnitude response and phase response
as a function of frequency ω. It consists of the following two plots, where the
magnitude is expressed in dB and phase in degrees.
Let consider a typical transfer function can be given as

(s + a) (s + b)
H (s) = K
sn (s + c) (s + d) (s + e)
where a, b, c, d, e, K are real positive numbers. Let s = jω , the following expres-
sion can be obtained as

(jω + a) (jω + b)
H (jω) = K n
(jω) (jω + c) (jω + d) (jω + e)
The normalized transfer function H (jω) can be found as

jω
+ 1 b jω jω jω





a a b +1 ab a +1 b +1
H (jω) = K n
=K
jω
+ 1 d jω jω cde (jω)n jω jω




jω

(jω) c c d +1 e e +1 c +1 d +1 e +1

The Bode Form of the transfer function H (jω) can be written as

jω

jω

ab a +1 b +1
H (jω) = K
cde (jω)n jω

jω
jω

c +1 d +1 e +1
Magnitude Plot Rule

• Curve breaks up by 20 dB/dec at a, b

• Curve breaks down by 20 dB/dec at c, d, e
Bode Phase Plot Rule

• Curve breaks up by 45 degrees one decade before a, b and breaks down by

45 degrees one decade after a, b
• Curve breaks down by 45 degrees one decade before c, d, e and breaks up
by 45 degrees one decade after c, d, e
Initial Magnitude Plot
Case 1: n=0
Calculate H (jω) |ω=0 and then convert to dB i.e., 20 log (H (jω)) |ω=0
Case 2: n>0
Kab

n1
• Calculate the ω -intercept, i.e., ω − intercept = cde .

• Calculate the initial slope, i.e., initial − slope = −20n dB/dec

• Calculate the magnitude slope, i.e., f inal − magnitude − slope = −20.RD

dB/dec

7
Example
s+a
Calculate the initial angle and nal angle of G (s) = s+c

Initial − Angle = (Angle(jω + a) − Angle(jω + b)) |ω=0 = 0 − 0 = 0

F inal − Angle = (Angle(jω + a) − Angle(jω + b)) |ω=∞ = 90 − 90 = 0

8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31