Automatic Control (1)
CH 10 Frequency Responses
Fall 2022
By Liang-kuang Chen
Dept. of Mechanical Engineering
National Taiwan University of Science and
Technology
1
Road Map
• Frequency response
• Bode plots
– Asymptotes approximation
– Examples
• Nyquist plot
• Nyquist Stability Criterion
• Stability margins
2
Frequency Response
• Concept: what you get is what you sowed (linearity)
– A periodic input will generate a periodic output
– What is the input-output relation?
– Magnitude and phase changed
• The derivation
– Sinusoid represented by phasors (AC CKT?)
• Implication
– Steady state relationship between the input and output
of the system w.r.t. a sinusoidal input
– Including both magnitude and
phase change
– cf: The phasor representation of
signals and systems
3
0.5
.8s+1 Scope
Sine Wave Transfer Fcn
Refer to the doc!
Q: Positive phase
or negative?
Frequency Response
• Check pdf file, results: G(jw) = G(s)|s→jw
– Magnitude frequency response M(w) and phase
frequency response f(w)
– Analytic representations of magnitude and phase?
• Plotting: M(w) in dB vs. log w
– dB: defined as 20log|G(jw)|
– Need the semilog paper
– If (|G(jw)|) vs. log w used: full-log paper needed
– Samples followed
5
Sample Semilog Paper
100
80
60
Units
40 ?
20
-20
-40
-60
-80
-100
-2 -1 0 1 2
10 10 10 10 10
6
How to Plot Frequency Response?
• Given G(s), compute G(jw) for each w of interest
• Mark the corresponding magnitude and phase vs
frequency on the semilog paper point by point
– Bode plot
• Polar graph also popular
– Resulting Nyquist Plot
– Frequency information not
presented
• Can this be expedited?
– Asymptotes method
7
Asymptotic Approximations
• Asymptotic lines of Bode plots
• Why appropriate?
𝐾 ( 𝑠+ 𝑧 1 ) ⋯ ( 𝑠+ 𝑧 𝑚 )
𝐺 (𝑠 )= 𝑡
– Log turns all multiplications into
𝑠 ( 𝑠 +𝑝 1) ⋯( 𝑠+ 𝑝𝑛 ) summations:
– Deal with them one by one
– Break frequency and maximum
errors
𝐾 |( 𝑠+𝑧1 )| ⋯|( 𝑠+𝑧 𝑚 )|
|𝐺(𝑠)|= 𝑡
|𝑠 ||(𝑠+𝑝1 )|⋯|(𝑠+𝑝𝑛 )| 8
Cntd. 𝐾 ( 𝑗 𝜔+𝑧 1 ) ⋯ ( 𝑗 𝜔+ 𝑧 𝑚 )
𝐺 ( 𝑗 𝜔)= 𝑡
( 𝑗 𝜔 ) ( 𝑗 𝜔+𝑝 1 )⋯ ( 𝑗 𝜔 +𝑝 𝑛)
Consider LHP zeros and poles for now, i.e. p and z >0
when 0
M 20log G j 20log K 20log j z1
20log j 20log j p1
t A straight line of slope
-20t dB/dec
if t 0, all terms are finite
if t 0, note log(0) -
M : some finite constant
M 20log j 20t log
t
G j 0
1
t corresponds to system type, G j t
j
t
under unity feedback 2
M 1e j1 M 2e j2 M 1M 2 j 1 1 3 4 5
Note : j3 j 4 j 5
e
M 3e M 4e M 5e M 3M 4 M 5 9
Cntd.
K j z1 j zm
G ( j )
j ( j p1 )( j pm )
t
when
M 20log G j 20log K 20log j z1
20log j 20log j p1
t
rd
M 20log j 20rd log
1 A straight line of slope
G j rd
j
rd
2 -20rd dB/dec
If some zeros and poles in RHP, non-minimum phase systems
10
Consider a mode of the
form (s+a)
G(jw)=(jw+a)
If w→0, G(jw)→a+0j
|G(jw)|=a, M=20loga
<G(jw)≈0
If w→∞, G(jw)→wj
|G(jw)|=w, M=20logw
<G(jw)≈p/2
11
11/12/2024 12
Normalize and Scale
• Common illustration techniques
• Standard plots resulted after normalization
– At least for low order systems
• For example, for a first order system, the Bode plots
always look like this (next 2 slides)
13
14
15
16
EX 10.2
17
11/12/2024 18
Second Order Mode
• Asymptotic lines: from limits as before
• Exact form analysis: let s=jw
– The standard form:
– Correction term on M at wn: -20log2z (normalized, let w=
wn)
• Normalized results shown next
– Resonant peak: dependent on z
• Almost negligible for z>0.6
– Resonant frequency reduced when damping increased
19
Second Order Systems
20
Example 10.3
23
24
10.3 Nyquist Criterion
• Nyquist Plot: G(jw) on the complex plane
• Relating closed loop stability to open loop
frequency response and pole location
• Concept: closed loop poles (roots) occur when
1+G(s)H(s)=0, i.e., G(s)H(s)=-1
• Conformal mapping: a complex function on s,
F(s), is actually a conformal mapping from s-
plane to F-plane
• Mark the discussion on page 561
25
Figure 10.21: Conformal Mapping
Mapping contour A through function F(s) to contour B
Analytic required?
26
Encircling of
zero/pole in
contour A Encirclement of the origin in
leads to a full contour B
360deg change in
phase
27
Figure 10.23
Vector representation of mapping
Q: Is the direction correct?
28
Sum Up the Mapping
• Poles inside the contour in the domain
– Reverse direction encirclement of the origin in the
range
• Zeros inside the contour in the domain
– Same direction encirclement of the origin in the
range
• These two may cancel each other
– Consequently, N=P-Z
29
Nyquist Stability Criterion
• Let F(s) be 1+G(s)H(s)
– Root locus equation?
– Zeros of F(s): roots (i.e., closed-loop poles)
– Poles of F(s): open loop poles
– Z=P-N with N: counter-clockwise encirclements of
the origin of 1+G(s)H(s)
• Finally move F(s) to G(s)H(s) and extend the
contour to the entire CRHP
30
Figure 10.25
Mapping
examples:
a. contour does
not enclose
closed-loop
poles;
b. contour does
enclose
closed-loop poles
31
10.4 Sketching the Nyquist Diagram
• EX 10.4
32
Figure 10.27
Vector evaluation of
the Nyquist diagram
for Example 10.4:
a. vectors on contour
at low frequency;
b. vectors on contour
around infinity;
c. Nyquist diagram
33
Detouring around open-loop poles:
Why: since the mapped function would become infinity
detour right is more common, to avoid encirclement of poles
34
Figure 10.29
a. Contour for Example 10.5;
b. Nyquist diagram for Example 10.5
35
10.5 Stability via Nyquist Diagram
• Scaling with K: the critical point becomes -1/K
• Objective: design the system so that the
Nyquist diagram has the correct number of
critical point encirclement
– Very often: this is simplified to shape the Nyquist
plot to avoid the critical point
– Analogy to root locus: find the range of K that
ensures the open-loop magnitude is less than
unity at the frequency where phase angle is 180.
36
Figure 10.30
Demonstrating Nyquist stability:
a. system; b. contour;
c. Nyquist diagram
37
Figure 10.31
a. Contour for Example 10.6;
b. Nyquist diagram
38
Figure 10.32
a. Contour and root locus of system that
is stable for small gain and unstable for large gain;
b. Nyquist diagram
39
Figure 10.34
a. Portion of contour to be mapped for Example 10.7;
b. Nyquist diagram of mapping of positive imaginary
axis
40
10.6 Gain Margin and Phase Margin
• Gain margin: the change in open loop gain,
expressed in dB, required at 180° of phase
shift to make the closed loop system unstable.
– How to compute: via the Routh Table
• Phase margin: The change in open loop phase
shift required at unity gain to make the closed
loop system unstable
– How to compute: Solve for the frequency at which
gain is 0dB (wg), then find the phase angle at wg.
41
Figure 10.35
Nyquist diagram
showing gain
and phase
margins
42
10.7 Stability Margins via Bode Plots
43
10.8 Relation between Closed-loop Transient
and Closed-loop Frequency Response
• Consider the standard 2nd order system
• Resonant peak is a function of z, and is therefore
related to %OS
• Furthermore, resonant frequency wp = wnsqrt(1-2z2)
• More involved derivation gives the bandwidth
44
Figure 10.38
Second-order
closed-loop
system
45
Selected Problems
• Review Questions: 2, 4, 6, 7, 8, 9, 10, 12, 14,
16, 17, 19, 20
• P: 1, 3, 4, 6, 7, 22, 26
46
