Frequency response using Bode plot presentation

• Learning Objectives
• to understand how the magnitude and phase points are calculated from a transfer function
representation
• to be able to enter appropriate data in order to produce a computer-based Bode plot
• to divide the frequency range into low, middle and high frequency ranges.
• to find out the basic components that make up any simple transfer function
• to be able to sketch a Bode plot of a simple transfer function
• to be able to sketch a Bode plot of a cascaded system
• to understand how adding simple components changes the shape of the Bode plot
The Bode plot

The Bode plot comprises two plots

(i) a Magnitude plot: The gain of the system versus frequency

(ii) a Phase plot: The phase shift induced by the system versus frequency

Frequency axis: semi-log scale;

Magnitude axis: usually in dB, that is, 20 log10( gain) versus log10ω.
Phase axis: has unit of degrees.

Gain and Phase calculation without a computer

• Given a model of the process, G(s),
• Substitute for s=jω
• Choose a desired frequency range ω Low to ω High
• Calculate |G(jω)|, and the phase, ∠G(jω),
Example
0.5
G(s) =
s+0.5

Substitute for s=jω,

0.5
G(jω)=
jω +0.5
Simplify
0.5 (0.5 - jω) 0.52 0.5 ω
G(jω)= 2 2 = 2 2 - j
(ω +0.5 ) (ω +0.5 ) (ω2 +0.52)
Calculate the gain and phase

0.5
| G(jω)| = Real2 + Imag2 = ,
ω2 +0.52
∠ G(jω) = tan-1 (Imag/Real)= tan-1 (-ω/0.5)

For ω =0.2 rad/s and ω =1.8 rad/s.

0.5 -1 -0.2
|G(j0.2)| = = 0.928 and ∠ G(j ω ) = tan ( )= -0.385 rads= -21.8o
2
0.2 +0.5 2 0.5
0.5 -1 -1.8
|G(j1.8)| = = 0.268 and ∠ G(j ω ) = tan ( )= -1.300 rads= -74.5o
2
1.8 +0.5 2 0.5
Using computers to form a frequency response plot

6 6
enter G(s) = (= 2 )
(4s+1)(s+1) 4s + 5s+1
s=tf(‘s’);
g=6/(4*s^2+5*s+1)
simply use
bode(g)

bode(g,{wmin,wmax})
bode(g,{0.1,10})
[mag,phase,w] = bode(g)

rltool
What are low, middle and high frequencies?

LF range: frequencies near zero. Gain is often ‘constant’, or constantly decreasing.

MF range: gain/phase change; this range determines the closed loop stability.
HF range: slope (cut-off rate) of gain is constant. This range determines the performance of
the closed loop system to high frequency disturbances such as noise.
Motivation for sketching Bode plots
Bode plots are fairly easy to sketch ‘on the back of an envelope’. Complex systems could be
represented easily. We still use Bode plots for control system design.

Transfer function components

We can write any transfer function as a product of
• A constant gain
• integrators
• first order terms,
• 2nd order underdamped (complex poles)

20 s(3s + 1) 1 1
G1( s ) =  20, , 3s+1, 2
(10 s + 1)(s 2 + 0.2 s + 1) 10s+1 s +0.2s+1
6( s 2 + 10 s + 100) 1 1
G2 ( s ) =  600, , ,(0.01s2 + 0.1s +1)
s 2 (50 s 2 + 15s + 1) 10s +1 5s + 1
Magnitude and phase of transfer function components

Logarithm rules
Multiplication of a and b: log ab = log a + log b
Division of a and b: log a/b = log a – log b
Powers of a: log ab = b log a

Not that the magnitude on a Bode plot is written in terms of 20log10 |G(jω)|.
Since we are effectively multiplying the terms together in the transfer function, it turns
out that we will be adding them as logarithmic values.

Calculation of phase from complex variables

(i) for a complex variable x = a + bj, the phase is ∠x = tan-1 b/a,
a+bj
(ii) for, x= , ∠x = ∠num - ∠ den, ∠x = tan-1(b/a) – tan-1(d/c)
c+dj
(iii) for x= (a+bj)*(c+dj), ∠x = ∠ 1st term + ∠ 2nd term, ∠ x = tan-1 (b/a) + tan-1(d/c)
Magnitude and phase of constant gain K

System gain K (Magnitude)  20 log10K

System gain K (Phase)  0o

Magnitude and phase of s-terms

1 1
for G(s) = , replace s = jω, G(jω) =
s jω
1 1
Gain: |G(jω)| = = , |G(jω)|dB = 20log10(1/ω) = -20 log10(ω)
|jω| ω
Phase: ∠G(jω)=∠(1/jω) = -90o.

for G(s) = s, replace s = jω, G(jω) = jω

Gain: |G(jω)| = | jω|= ω, |G(jω)|dB = 20log10(ω) = 20 log10(ω)
Phase: ∠G(jω)=∠(jω) = 90o.

Magnitude of sp -term
Let G(jω) = (jω)p.
| G(jω)|dB = 20log10ωp = p * 20log10(ω)
∠G(j(jω)p = p*90o
Introducing a Sketching Table
Draw the Bode plot for the following transfer function
6
G(s) =
s
1
Solution terms are: 6,
s

Sketching Table (magnitude)

Term Low Freq. Mid-freq. High-freq. r

K= 6 15.6dB 15.6dB 0 dB/decade

1/s line at -20 dB/ decade at 1 rad/s 0dB -20 dB/decade
(1 rad/s, 0 dB)
TOTAL line at -20 dB/ decade at 1 rad/s, 15.6 dB 0- 20 = -20
(1 rad/s,15.6dB) dB/decade

Sketching Table (phase)

Term LF MF HF
o
K= 6 0 0o 0o
1/s -90o -90o -90o
TOTAL -90o -90o -90o
Magnitude and phase of first order terms
1
G1(s) = and G2(s) = (τ s+1) .
(τs+1)
1
We will examine only G1(s) = , and apply the negation to the results to find the
(τs+1)
corresponding results for G2(s).

The following table shows the magnitude and magnitude(dB) calculations for G1(s)
TF 1
G1(s) =
τs + 1
s = jω 1
G1(jω) =
jωτ + 1
Gain G(jω) 1 1
|G1(jω)| = =
|jτ ω +1| τ 2ω 2 + 1
Gain in dB |G1(jω)|dB = 20 log10 (τ2ω2+1)-0.5
Phase ∠G1(jω) = ∠ (1/(1 + jτ ω))
= - tan-1 (τω)
Low frequency and high frequency asymptotes.
Bode Diagrams
Gain in dB |G1(jω)|dB = 20 log10 (τ 2ω 2+1)-0.5
LF ω≈0 |G1(jω)|dB →20log10(1) = 0dB 20
10
Low Freq.
Asymptote
horizontal line at 0dB 0 High Freq.

Phase (deg); Magnitude (dB)

Asymptote
MF ω=1/τ |G1(jω)|dB → 20 log10(1+1)-0.5
-10
-20

= -3dB -30

-40
! =1/!

point (1/τ,−3dB) 0
-10

HF (ω>>1) |G1(jω)|dB ≈ 20log10(τ2ω2)-0.5 -20

-30
-40
= -20 log10 τ -20 log10 ω -50
-60
-70
Line with slope of : -80
-90

–20dB/decade -100
10
-1 0
10
1
10 10
2

Pass through (1/τ, 0 dB) Frequency (rad/sec)

∠G(jω) ∠G1(jω) = - tan-1 (τω)
LF ω≈0 ∠G1(jω)≈ - tan-1(0) ≈0
The phase at low frequency
is approximately zero.
MF, ω=1/τ ∠G1(jω) = - tan-1(1) = - 45o
The phase at the corner
frequency is -45o .
HF(ω>>1) ∠G1(jω) = - tan-1(∞) = -90o
The phase at high frequency
equals –90o.
Lead and lag terms
1
The terms and (τs+1) appear commonly in transfer functions.
τs +1
1
the term introduces –90o of phase lag.
τs +1
the term (τs+1) introduces +90o of phase change,

Procedure to Sketch Bode Plot

Step 1: Identify all the tf components and order them in increasing corner frequencies.
Step 2: Choose the frequency range of interest, i.e. (0.1 to 100 rad/sec)
Step 3: Complete a Sketching Table for the range of frequencies chosen.
Step 4: Find the low frequency asymptote from the Sketching Table.
The value of the asymptote will depend on whether there are any ‘s-terms’ in the
transfer function.
(i) If there are no s-terms, then the low frequency asymptote will be a horizontal line
with d.c.gain of 20log10K, where K is the transfer function gain.
(ii) If there are s-terms, then the low frequency asymptote is formed by the logged
sum of the slope of these lines, which pass nominally
through 0dB at 1 rad/s and are then raised by 20 log10K.
Step 5: If there is more than one corner frequency, put them in order of numerically
increasing frequency values. Find the (first) corner frequency, ω c = 1 / τ. Find the
corresponding point on the low frequency asymptote.
Step 6:Draw the high frequency asymptote. This is a straight line of slope -20 dB/decade for
lag terms starting from the point ( ωc , 20 log10 K ) , and a slope of + 20 dB/decade for lead
terms.
Step 7:Find the next corner frequency and the plot will then decrease (or increase) at an extra
–20dB/decade or + 20dB/decade for lag and lead terms repsectively.
Step 8: Plot the phase response using the phase shift values at different frequency ranges.

0.1 Elementary examples

We illustrate the use of the Sketching Tables by examining several examples which are very
common in process control systems, all of which involve a combination of the three basic
terms , gain K, s-terms, and first order terms.
Example 1: Bode plot of gain and first order lag
Problem
Sketch the Bode plot for the following transfer function
20
G(s) =
4s+1
Solution
Firstly we list the transfer function component terms:
1
20,
4s + 1
We identify the corner frequencies , as these will usually occur in our mid-frequency range.
In the above transfer function, there is only one corner frequency and it is at ωc = ¼ = 0.25
rad/s. We can then complete the Sketching Table.
 Sketching Table (magnitude)
Compon Low Freq. Mid-freq. High-freq. roll-
ent asymptote Value(s) off
0.25 rad/s
K= 20 20log10(20) = 26.02dB 0 dB/decade
26.02dB
1 0 dB -3 dB -20 dB/decade
4s+1
TOTAL 26.02 dB 23.02 dB -20dB/decade

Sketching Table (phase)

Compon Low Freq. value Mid-freq. High-freq. value
ent ω < 0.025 rad/s Values: ω > 2.5 rad/s
0.25 rad/s
o
K= 20 0 0o 0o
1 0o tan –1 -90o
4s+1 4ω = −45ο
TOTAL 0o -45o -90o

We make two remarks about the phase table. Firstly, we note again that the constant gain K
has no effect on the phase at all and secondly, that we consider ‘low frequency’ to be a
decade below the (lowest) corner frequency and ‘high frequency’ to be a decade above the
(highest) corner frequency.
We can then sketch the Bode plot using the information in the Tables.
Figure 0.1 shows the low and high frequency asymptotes as well as the magnitude and phase
at the corner frequency. We have used MATLAB to plot the Bode plot in dashed lines to
illustrate how good the Sketching approximation is. (In MATLAB we used the command :
bode(g,{0.01,10}).

Bode Diagrams

30

X
20
low freq. asympt. high freq. asympt.
Phase (deg); Magnitude (dB)

-20 dB/dec
10

0 23.02dB
ωc = 0.25 rad/s
-10

0
-10
-20
-30
low freq. ~ 0o -45o
-40
X
-50
-60
high freq. ~ –90o
-70
-80
-90
-100
-2 -1 0 1
10 10 10 10

Frequency (rad/sec)
Figure 0.1 Bode plot of 20/(4s+1).
Example 2 : Finding a simple transfer function from a Bode plot
Just as we have used a transfer function description to sketch a Bode plot, we can use a Bode
plot to help us find out the system transfer function. This is particularly useful when we have
some experimental results from a process that we have taken by injecting sinusoids at
different frequencies, and we wish to find a transfer function for the system.
Problem
An engineer tries to find the time constant of an RC network by applying unit
magnitude sine waves of different frequencies and zero phase at the input of the RC circuit.
The engineer measures the magnitude and phase of the output voltage, y(t), across the
capacitor and sets up the table shown below:

The frequency response of the RC network.

ω (rad/s) 0.01 0.0 0.1 0.5 1.0 5.0 10. 60.0 100.
5 0 0
|Y(jω)|dB 0.00 - - -8.6 -14 -28 -34 -49 -54
0.3 0.97
∠Y(jω) -3 -14 -26 -68 -79 -87 -88 -89 -90
a. Draw the Bode plot for the system
b. Draw the low frequency and high frequency asymptotes.
c. Find the time constant of the system.
We assume the input is a sine wave of unit amplitude and zero phase.
Solution:
a.
∠Y ( jω ) = ∠G( jω ) + ∠U ( jω )
U ( jω ) = 1 and ∠U( jω ) = 0
Thus, the system has a frequency response G(jω) similar to the voltage signal across the
capacitor Y(jω). To draw the Bode plot, transfer the data at each frequency onto the semilog
graph paper, Figure 0.2.
MATLAB commands
w=[0.01 0.05 0.1 0.5 1.0 5.0 10.0 60.0 100.0];
mag=[0 -0.3 -0.97 -8.6 -14 -28 -34 -49 -54];
ph=[-3 -14 -26 -68 -79 -87 -88 -89 -90];
subplot(2,1,1);semilogx(w,mag,+,)
subplot(2,1,2);semilogx(w,ph,+,)
 Magnitude(dB)

Low freq. asympt.

20 constant of 0dB
0 + +
+ + High freq. asympt.
-20 -20 dB./decade
+
-40
+
ωc = 0.2 rad/s
+
-60 -2 +
-1 0 1 2
10 10 10 10 10
Phase(degrees)
0+
-20 +
+
-40
low freq. phase high freq. phase
-60 ~0o -90o
+
-80 +
+ + + +
-100 -2 -1 0 1 2
10 10 10 10 10
Frequency (rad/s)

Figure 0.2 Data plotted on semilog paper

b. The RC network is a first order system. The features on the Bode plot which verify this are:
(i) roll-off rate
Find the slope of the gain response at high frequencies by selecting two
frequencies which are one decade apart, for example, ω=1 and ω=10. Subtracting
the gain at ω=1 (-14 dB) from the gain at ω=10 (-34 dB) , gives the slope as -20
dB/decade as the roll-off rate. This is the roll-off rate for a first order system.
(ii) Phase change
 The phase change is from 0o to –90o implying a first order lag term.
(iii) The low frequency asymptote is a constant horizontal line which implies there are
no ‘s-terms’ in the transfer function and which therefore gives us the gain value.
The horizontal gain asymptote is at 0 dB = 20 log10 K dB (equivalent to a gain K= 1).
c. The corner frequency is at 0.2 rad /sec from the Bode plot. The time constant is given by:

τ =1/0.2=5.0 seconds.
The transfer function of the RC circuit is therefore
1
G(s) =
5s+1
Example 3: Combination of s-term and first order lag
Problem
Sketch the Bode plot for the following transfer function.
10
G(s) =
s(s + 3)
Solution
To list the components, we need the transfer function in ‘unity-constant- coefficient form’.
We can rewrite the transfer function by dividing the numerator and denominator by 3.
3.33
G(s) =
s(0.33s +1)
The components are now:
1 1
3.33, ,
s 0.33s + 1
The corner frequency of the first order lag is ωc = 3 rad/s. We can now complete the
sketching table.

Sketching Table (magnitude)

Compon Low Freq. Mid- High-freq. roll-
ent asymptote freq.Value(s off
)
3 rad/s
K= 3.33 20log10(3.33) = 10.45dB 0 dB/decade
10.45dB
1 Line decreasing at Line -20 dB/decade
s –20dB/dec decreasing
Passes 1 rad/s at at
0dB –20dB/dec
1 0 dB -3 dB -20 dB/decade
0.33s+1
TOTAL low freq. asympt. 7.45 dB -40 dB/decade
Line at
–20dB/dec
passes 1 rad/s at
10.45 dB

Sketching Table (phase)

Compone Low Freq. value Mid-freq. High-freq. value
nt ω < 0.3 rad/s Values: ω > 30 rad/s
3 rad/s
K= 3.33 0o 0o 0o
1 -90o -90o -90o
s
1 0o -tan –1 -90o
4s+1 4ω = −45ο
TOTAL -90o -135o -180o

We can then sketch the Bode plot using the information in the tables. Once again we can use
MATLAB to check how close our sketching approximation is to the actual frequency
response plot. We have used MATLAB to plot the Bode plot using the command
bode(g,{0.1,100}), Figure 0.3.
 Bode Diagrams

40

20
0 high freq. asympt.
Phase (deg); Magnitude (dB)

low freq. asympt. -40 dB/decade

-20
-20 dB/decade
-40
-60
ωc = 3 rad/s
-80

-80
Phase at ωc = -135o
-100

-120 low freq. phase

high freq. phase
-90o
-140 -180o
-160

-180
-1 0 1 2
10 10 10 10

Frequency (rad/sec)
10
Figure 0.3 Bode plot of G(s) =
s(s + 3)

Example 4: Two first order lag terms

Problem
Consider the following second order transfer function. Choose an appropriate frequency
range and sketch the Bode plot.
10
G(s) = 2
0.66s + 2.33s + 1
Solution
We can check to find out if the system is under or over-damped. Recall that for an
overdamped system
(i) the damping rati , ζ, must be greater than 1
(ii) the second order transfer function has two real poles, that is, the transfer function can be
split into
K
G(s) =
(τ1s + 1)(τ2s + 1)
Check the value of b2 – 4ac in the original: 2.332 – 4*0.66 = 2.7889 > 0. The transfer
function represents a second order overdamped system and can be written as
10
G(s) =
(2s + 1)(0.33s + 1)
The corner frequencies of the two first order components are:
1
: corner frequency: ωc1 = 1/τ = 0.5 rad/s
2s + 1
1
: corner frequency: ωc2 = 1/τ = 3.0 rad/s
0.33s + 1

The transfer function components are then given as:

1 1
10, ,
2s + 1 0.33s + 1
We can now complete the sketching table.
Sketching Table (magnitude)
Compo Low Mid- Mid- High-freq.
nent Freq. freq.Value( freq.Value(s roll-off
asympto s) )
te (i) 0.5 rad/s (ii) 3 rad/s
K= 10 20log10( 20.0 dB 20.0 dB 0 dB/decade
10) =
20.0 dB
1 0 dB -3 dB line -20
2s+1 decreases at dB/decade
–20
dB/decade
from
intersect with
low freq.
asympt.
1 0 dB 0 dB -3 dB -20
0.33s+1 dB/decade
TOTAL low 20 – 3 = 17 line -40
freq. dB decreases at dB/decade
asympt.: –20 dB/
horizont decade from
al line at intersect
20 dB with low
freq.
asympt.,
then dec. at
– 40
dB/decade

Sketching Table (phase)

Compo Low Freq. Mid-freq. Mid-freq. High-freq.
nent value Value Value value
ω < 0.05 0.5 rad/s 3 rad/s ω > 30 rad/s
rad/s
K= 10 0o 0o 0o 0o
1 0o - 45o - tan-1 -90o
2s+1 (2ω) = -81o
1 0o - tan- -45o -90o
1
0.33s+1 (0.33ω) = -
9o
TOTAL 0o -54o -126o -180o

The Bode plot is shown in Figure 0.4. We notice the high frequency asymptote in the
magnitude plot is –40 dB/decade consistent with our 2nd order system. Likewise, the phase
decreases to –180o.
Bode Diagrams

20 -20 dB/dec.
0
Phase (deg); Magnitude (dB)

-20
-40 dB/dec.
-40
0.5 rad/s
3 rad/s
-60

-50

-100
-126o
-54o
-150

-200
-1 0 1 2
10 10 10 10

Frequency (rad/sec)
10
Figure 0.4 Bode plot of
(2s + 1)(0.33s + 1)
Summary: By using a combination of all constant K, s-terms, and first order lead and
lags, we can sketch all but second order underdamped features.
0.2 Second order underdamped system
A second order underdamped system will have a transfer function of the form:
1
G( s) =
1 2 2ζ
( 2 )s + ( )s + 1
ωn ωn
where ω n is the natural frequency and ζ is the damping ratio of the system. We remember that
(i) The damping ratio, ζ, will be less than 1.
(ii) The roots of the denominator (poles of the transfer function) will be complex
(iii) The step response of the system will show an overshoot depending on the value of the
damping.
The magnitude and phase plot of a second order underdamped system is shown in Figure 0.5.
We note the following remarks.
• The underdamped second order frequency response has a low frequency asymptote at 0
dB.
• It has a peak in the mid-frequency range.
• (The peak does not occur at the natural frequency, but at the damped natural frequency, ωd.
This can be found by differentiating |G(jω)| with respect to ω. If ζ is small (<0.4), then ωn
and ωd are similar in value.)
1
• The value of gain at the natural frequency ωn is or 20 log10(1/ 2ζ)
2ζ
• The peak is dependent on the damping ratio of the system: as the damping decreases, the
peak value of the magnitude plot is increased.
• The high frequency asymptote meets the low frequency asymptote on the 0 dB axis at the
natural frequency
• The natural frequency represents the corner frequency for the under damped second order
system.
• The phase changes over approximately two decades; from one decade below the corner
frequency to one decade above the corner frequency.
Gain dB Low
20 Frequency
asymptote High
Frequency
0
asymptote
Corner
-20 frequency

-40
-1 0 1
10 10 10
Phase deg Frequency (rad/sec)
0

-90

-180
-1 0 1
10 10 10
Frequency (rad/sec)
Figure 0.5 Bode plot of underdamped second order system
We now look at the transfer function analysis to justify our remarks.
Analysis
Magnitude of transfer function, |G(jω)|:
 1 1
G ( jω ) = =
1 2 ζ ω ω
( 2 )( jω)2 + ( )( jω) + 1 ( )2 j 2 + 2ζ( ) j + 1
ωn ωn ωn ωn
1
=
ω ω
(1 − ( )2 ) + j 2ζ( )
ωn ωn
The gain of the complex function G(jω) is the gain of its numerator divided by the gain of its
denominator:
1 1
G( jω ) = =
ω 2 ω ω 2 2 ω
(1 − ( ) ) + j 2ζ (1 − ( ) ) + 4ζ 2 ( ) 2
ωn ωn ωn ωn

Convert the gain into decibels:

0.5
⎡ 2
⎤
ω 2 2 2⎛ ω ⎞
G ( jω ) dB = 20log10 (1) − 20log10 ⎢(1 − ( ) ) + 4ζ ⎜⎜ ⎟⎟ ⎥
⎢⎣ ωn ⎝ ω n ⎠ ⎥⎦

0.5
⎡ 2
⎤
ω 2 2 2⎛ ω ⎞
= −20log10 ⎢(1 − ( ) ) + 4ζ ⎜⎜ ⎟⎟ ⎥
⎢⎣ ωn ⎝ ω n ⎠ ⎥⎦

Examine the general feature of the gain response over the different frequency ranges:
Gain in dB: |G(jω)|dB
|G(jω)|dB
 0.5
⎡ 2
⎤
ω 2 2 2⎛ ω ⎞
= −20log10 ⎢(1 − ( ) ) + 4ζ ⎜⎜ ⎟⎟ ⎥
⎢⎣ ωn ⎝ ω n ⎠ ⎥⎦
Low |G(jω)|dB →20log10(1) = 0dB
Frequency
Range ω≈0 horizontal line at 0dB
Mid- |G(jω)|dB → -20 log10(1-1 + 4ζ2)0.5
Frequency = -20log10(2ζ) = 20 log10
Range (1/2ζ)
ω=ωn If ζ<0.5, log10(2ζ) >0 and there will be
a peak in the magnitude above the 0
dB line
The gain at the natural frequency is
inversely proportional to the damping
ratio, ζ. As the damping decreases and
the system become more oscillatory,
the gain at natural frequency increases.
High
Frequency ω ω
|G(jω)|dB ≈ -20log10(( )4 + 4ζ2 ( )2
Range ωn= ωn
)0.5
(ω>>1) since ω4 >> ω2
ω
|G(jω)|dB ≈ -20log10(( )2)
ωn
= -40 log10(ω) –40 log10 ωn
Line with slope of :
–40dB/decade
(found by differentiating above w.r.t
log10 ω.)
 Line intersects the horizontal
0 dB line at ω = ωn .

The Phase Response of second order underdamped systems.

We can similarly find the phase of the transfer function, G(jω), as a function of frequency
over different frequency ranges. Using the transfer function G(jω):

1 1
∠G ( jω ) = ∠ =∠
1 2ζ ω 2 2 ω
2
()( jω)2 + ( )( jω) + 1 ( ) j + 2ζ ( ) j + 1
ωn ωn ωn ωn
1
=∠
ω 2 ω
(1 − ( ) ) + j 2ζ( )
ωn ωn
ω 2 ω
∠G(jω) = ∠1 - ∠( 1-( ) + 2ζ( ) j)
ωn ωn
ω ω
2ζ( ) 2ζ( )
ωn ωn
= 0 -tan-1( ) = -tan-1( )
ω 2 ω 2
1-( ) 1-( )
ωn ωn
Phase ∠G(jω) ω
2ζ( )
ω n
∠G(jω) = -tan-1( )
ω 2
1-( )
ωn
Low ∠G(jω) →-tan (0) = 0o
-1

Frequency
Range ω≈0 asymptote at 0o
Mid- ∠G(jω) → -tan-1(∞)
Frequency = -90o
Value Mid –frequency phase value is –90o.
ω=ωn
High ω
Frequency 2ζ( )
-1 ωn
Range ∠G(jω)≈ -tan ( )
ω 2
(ω>>1) 1-( )
ωn
2
since ω >> ω
∠G(jω) ≈ -tan-1(0)
= -180o
Phase is decreasing towards high
frequency. Asymptote at
∠G(jω) = -180o
0.3 Effect on gain and phase plots of increasing the damping ratio
Figure 0.6 shows how the gain and the phase plots of the underdamped second order system
change for different values of damping. The gain plot shows a large peak for low damping,
this corresponds to a higher overshoot on a step response plot.

Figure 0.6 Changes in Bode plot for different damping factors.

Sketching Procedure for simple 2nd order underdamped system

Step 1: Select the desired frequency range and mark it on the log scale of the semilog graph
paper.
Step 2: Select the desired range of the gain in dB and mark it on the linear scale of the log
paper.
Step 3: The low frequency asymptote is given by the 0 dB line (unless combined with other
gains or s-terms)
Step 4: Find the corner frequency as ωc = ωn. Draw a line of slope -40 dB/decade from the
point (ωn, 0 dB). This line is the high frequency asymptote.
Step 5: Find the gain at the corner frequency, ω n , using |G(j ω n )|=1/(2ζ)dB.
Step 6: Starting at the low frequency asymptote, draw a curve passing through the gain at ω n
and finish at the high frequency asymptote.
Step 7: Plot the phase response using the phase values of approx 0o at a decade down from the
corner frequency, -180o a decade up from ωc, -90O at ωc. If it is necessary to have a
more accurate approximation, use the expression for the phase, ∠G(jω), to fill in extra
points.

Problem: Identification of second order transfer function from Bode plot

An engineer has measured the frequency response of a position d.c. servo system by applying
a sine wave of amplitude of 2 volts and zero phase to the position reference point of the
system. The amplitude and the phase of the position output sine wave has been measured and
is given in Table 0.1.

The frequency response of the DC servo system

ω(rad/sec) 0.1 0.4 0.8 1 3 4 10
Magnitude(volts) 20 24. 45. 50.2 1.01 0.63 0.10
2 0
Output Phase -7 - - -89.0 - - -
(degree) 12. 49. 172. 173. 177.
2 3 3 1 7

Table 0.1 Gain and phase info for identification problem

a. Draw the Bode plot of the system.
b. Draw the low frequency and high frequency asymptotes.
c. Assuming the system is second order, find the d.c. gain, natural frequency and the
damping ratio of the system.
Solution:
a. The gain of the system is

output amplitude at ω
System gain =
input amplitude at ω
We should note that the input is not, in this case, a signal of magnitude 1, but a signal of
magnitude 2. The system gain can be calculated using the equation and then converted to dB.
The results are shown in the following table.
ω(rad/sec) 0.1 0.4 0.8 1 3 4 10
System Gain, 10 12. 22. 25. 0.50 0.31 0.05
|G(jω)| 1 5 1
|G(jω)|dB 20 21. 27. 28. 0.08 - -19.9
6 0 0 6 4.01

b. Plot the points as shown in Figure 0.7. The low frequency asymptote is the horizontal line
of height 20 dB. The high frequency asymptote is drawn at a slope of –40 dB/decade and
meets the low frequency asymptote at 1 rad/s. This is the corner frequency of the second
order system.
Magnitude (dB)
Low and high
40 frequency asymptotes

20

-20 -1 0 1
10 10 10
Phase (degrees)
0
-30
-60 -90o at ωn
-90
-120
-150
-180 -1 0 1
10 10 10

Frequency (rad/s)
Figure 0.7 Magnitude and phase points of 2nd order system plotted.
c. System gain: The system gain at low frequency is 20 dB. The gain K = 10.
Natural frequency ωn: The corner frequency is at ωn = 1 rad/s. For a second order system,
the phase at ω n is -90o.
Damping ratio, ζ: Calculate the damping ratio, ζ using the value of the gain at ω n . (Hint:
K
|G(jωn)| =
2ζ
Gain dB at ω n = sum of all logged gain components:
= 20log10(K) + 20 log10(1/2ζ)
28 dB = 20log10 K – 20 log10(2ζ)
20 log10(2ζ) = 20 dB – 28 dB = -8 dB
2ζ = 10^(-8/20) = 0.3981
ζ = 0.199 ∼ 0.2
nd
Therefore the 2 order system is given by
10
G(s) = 2
s + 0.4 s + 1

0.4 Further examples using MATLAB plots

Complex systems often comprise a number of subsystems in cascade. An example of this is
shown in Figure 0.8. The input signal is U(s), the output is Y(s) and Q(s) is an intermediate
variable. For example, a power system may have a steam turbine (subsystem 1) driving an
electric generator (subsystem 2). A second example is a d.c. servo system which has an
amplifier in cascade with a motor. Each subsystem may be represented by a first or second or
higher order system. We have approached the technique of plotting frequency responses by
considering that all systems are combinations of a basic few subsystems. Therefore when it
comes to plotting cascade systems, we really only need to extend the number of the smaller
subsystems to account for all the factors in the cascaded system.

U(s) Q(s) Y(s)

G1(s) G2(s)

Figure 0.8 Cascade system

Consider the two cascade subsystems G1(s) and G2(s). Define Gp(s) as the transfer function
of U(s) to Y(s) :
Y(s) = Gp(s)U(s)= G2(s)G1(s)U(s)
Gp(s) = G2(s)G1(s)
Magnitude:
|Gp(jω)| = |G2(jω)| |G1(jω)|
|Gp(jω)|dB = 20 log10(|G2(jω)| |G1(jω)|)
= 20 log10 |G2(jω)| + 20 log10|G1(jω)|
Phase Relationship:
∠Gp(jω) = ∠ G2(jω)G1(jω) = ∠ G2(jω) + ∠G1(jω)
This shows us that the magnitude of the combined transfer function is just the sum of
logarithmic magnitudes of the individual ones. Similarly, the combined phase is the sum of
the individual phases. We now just need to consider all the transfer function components in
G2(s) and G1(s) separately and combine them using our Sketching Table for adding
logarithmic magnitude and adding phase.
 We will look at three examples of combinations of our transfer function components.
We will plot the responses using MATLAB and will not produce Sketching Tables for the
following, although it would be easy to use them to combine the different components.

0.4.1 Problem: The effect of a adding a pole at the origin

Consider the following transfer functions, G1(s). Determine the effect on the Bode plot of
adding the term G2(s) = 1/s in cascade with G1(s).
4
G1(s) = 2
s +1.6s +4
Solution
The combined system is given by
4
Gp(s) = 2
s(s +1.6s +4)
We will use our method for examining the two plots. Firstly we note that the second order
term is not in ‘unity constant coefficient form’. By dividing the numerator and denominator
by 4 we find:
1
Gp(s) = 2
s(0.25s +0.4s +1)
We now check to see if the second order term can be formed from two first order
components: check b2 – 4ac: This value is less than zero which implies that the second order
term would give us complex roots and represents an underdamped system. We note that the
corner frequency ωc = ωn is given from the second order component as 2 rad/s.
 We can see from the magnitude plot in Figure 0.9 that the second order plot has a
constant d.c.gain, whilst the integrator adds a high gain at low frequency. The integrator also
increases the order of the system and therefore the roll-off rate is increased at high frequency.
In the mid frequency range, for this example, the integrator causes the gain to decrease
slightly at the natural frequency. The gain crossover becomes lower, moving from above 2
rad/s to between 1 and 2 rad/s.
For the phase plot, the integrator term adds –90o of phase at all frequencies, hence
lowering the second order plot entirely. This causes the phase to cross –180o and the phase
crossover frequency becomes ωpco = 2 rad/s. We can write down the the gain and phase
margins of the original and altered systems.
G1(s) : GM none
PM > 60o
Gp(s) : GM about 5dB
PM about 45o
 Movement of ωgco
40
20
Second order
0
Phase (deg); Magnitude (dB)

-20
-40
-60 Second order plus
-80 integrator
-100

0
-30
-60 Second order
-90
-120
-150
-180 Second order plus
-210 integrator ωpco
-240
-270
-1 0 1 2
10 10 10 10

Frequency (rad/sec)
Figure 0.9 Effect of adding an integrator to the system

0.4.2 Problem: The effect of adding a real pole

Let G1(s) be the following transfer function
1
G1(s) = 2
0.25s +0.4s +1
Consider the effect on the Bode plot of G1(s) by adding a first order lag, G2(s).
10
G2(s) =
2s+1
Solution
We note that the second order term is the same as in the previous example and therefore
represents a complex underdamped system. The combined system is given by
10
Gp(s) =
(2s + 1)(0.25s2+0.4s +1)
1
where we note that we have added a gain of 10 and a lag term .
2s + 1
We can see from Figure 0.10, that the gain has raised the magnitude plot by 20 dB at
low frequencies. The mid frequency values are 17 dB at 0.5 rad/s and 9.9 dB at 2 rad/s. The
roll-off has been increased from –40 dB/decade to –60 dB/decade by the addition of the lag
term. The lag term would have reduced the gain crossover but the addition of the gain term
has masked this effect and we see an increase in the gain crossover frequency from 2.5 to 3
rad/s due to the higher d.c.gain.
For the phase plot, the gain term makes no change to the phase plot. The first order lag
term has its corner frequency before the corner frequency of the 2nd order term. Therefore, we
see that the phase starts to decrease according to the first order term before the 2nd order term
has an effect. The overall phase lag has now become –270 o with a phase crossover
frequency at 2.2 rad/s.
The gain and phase margins of the original and altered systems are
G1(s) : GM none
PM > 60o
Gp(s) : GM about –9dB
PM about –33o
Since the GM and PM are negative, this tells us that the closed loop system would be
unstable.

Bode Diagrams

30 ωgco
20
10
Phase (deg); Magnitude (dB)

0
-10 second order
-20
-30
second order plus lag and gain
-40

0
-30
-60 ωpco
-90
-120
second order
-150
-180
-210
-240 second order plus lag and gain
-270
-300
-2 -1 0 1
10 10 10 10

Frequency (rad/sec)
Figure 0.10 Bode plot of 10/[(2s + 1) (0.25s2 + 0.4s + 1)]

0.4.3 Problem: The effect of adding a real zero

Let G1(s) be the following transfer function
1
G1(s) = 2
0.25s +0.4s +1
Consider the effect on the Bode plot of G1(s) by adding a first order lead term, G2(s).
 G2(s) = (2s+1)
We make a comment here that, in practice, pure lead terms are not implemented since they
form an improper transfer function. However, they do occur in combination with other
components.
Solution
We note that ωn is given from the second order component as 2 rad/s and the corner
frequency from the first order lead is 0.5 rad/s.
Although the low frequency asymptote has remained the same for both the cases with
and without the lead term, the lead term has introduced increased gain in the mid frequency
range, Figure 0.11. This has caused the gain crossover frequency to increase from 2.5 to 9
rad/s. The roll-off rate is not as high due to the introduction of the lead term.
The lead term has its corner frequency before the corner frequency of the 2nd order term.
Therefore we see that the phase starts to increase before the decreasing phase effect of the
underdamped second order system takes effect. The overall phase lag has now become –90o.
In some design cases we wish less phase lag in our systems and we introduce phase lead
through phase lead compensators. (Chapter 16).
Since the phase response does not cross –180o for both plots, there is no gain margin
before or after the lead term is added. However, the increase in ωgco causes the PM to
decrease.
 Bode Diagrams

20

0
Phase (deg); Magnitude (dB)

-20
second order
-40

-60 second order plus lead

-80

50

-50
second order
-100

-150 second order plus lead

-200
-2 -1 0 1 2
10 10 10 10 10

Frequency (rad/sec)
Figure 0.11 Bode plot of 10(2s + 1) /(0.25s2 + 0.4s + 1)

0.5 Magnitude plots of closed loop and sensitivity transfer functions

From Chapter 8, we remember that the closed loop transfer function GCL(s) and the sensitivity
transfer function S(s) are closely linked. We examine this in more detail and then look at the
magnitude plots for the two transfer functions.
The Sensitivity Function
We have produced designs for our system which satisfy specifications related to GCL(s).
However, we have not considered how our disturbance rejection problem fits into the
frequency domain. We will find out in the following that the disturbance rejection depends
on the Sensitivity function, which is the transfer function from the disturbance to the output.
Consider the unity feedback system in Figure 0.12.
D(s)
R(s) E(s) U(s) + Y(s)
+
K(s) G(s)
+

Figure 0.12The unity feedback system

We can write the output Y(s) in terms of the reference signal R(s) and the disturbance D(s) as
follows:
GOL(s) 1
Y(s) = R(s) + D(s)
1 + GOL(s) 1 + GOL(s)
K(s)G(s) 1
= R(s) + D(s)
1 + K(s)G(s) 1 + K(s)G(s)
= GCL(s) R(s) + S(s) D(s)
We can see that the closed loop response is a linear combination of the responses due to the
reference R(s) and the disturbance D(s). Because of the importance of the two transfer
functions, S(s) and GCL(s), we give them the special names of Sensitivity and
Complementary sensitivity functions.
1
Sensitivity Function: S(s) =
1 + K(s)G(s)
 K(s)G(s)
Complementary Sensitivity Function: T(s)=
1 + K(s)G(s)
For the standard unity feedback system that we use in this book, the complementary
sensitivity function, T(s), and the closed loop transfer function GCL(s) are the same transfer
function. We have also referred to these as the reference tracking transfer function. We note
that
GCL(s) + S(s) = 1
Thus, GCL(s) complements S(s) and hence the name complementary sensitivity function. This
relationship shows an important constraint in any control design problem. The relationship
between the reference and the output can be written as
Y(jω) = GCL(jω)R(jω)
where we would like the output Y to behave like the reference R. To achieve this, we should
design the controller so that GCL(jω) =1 over all frequencies. The complementarity between
GCL and S would then ensure that S(jω) = 0 and perfect disturbance rejection would follow.
Unfortunately this perfection is not possible for real processes and controllers. What can be
achieved is that GCL(jω) ~ 1 and S(jω) ~ 0 in the low frequency region and in the high
frequency region we obtain GCL(jω) ~ 0 and S(jω) ~ 1. Such designs will be satisfactory for
low frequency reference and disturbance signals. But we will have to use the controller to
shape the mid-frequency ranges for the stability margins and use controller roll-off to deal
with the high frequency measurement noise going round the loop.
This becomes clearer by examining the magnitude plots for the two transfer functions.
Example
We use the open loop transfer function given by
 6
GOL(s) =
s(s+1)(s+3)
(i) Closed loop magnitude plot:
Y(s) = GCL(s) R(s)
The following MATLAB code will produce the magnitude plot of GCL(s).
MATLAB code
gol = 6/(s*(s+1)*(s+3);
gcl=gol/(1+gol);
w = logspace(-1,2,300);
[mag,ph] = bode(gcl,w);
magdb = 20*log10(mag);
semilogx(w,magdb(:))
Figure 0.13 shows the magnitude plot. We can examine the plot in the low, middle and high
frequency ranges:
low frequency range: We note that we have a gain of 1 at low frequencies. Since Y(s) =
GCL(s) R(s), a gain of unity at low frequencies indicates that the closed
loop system will track low frequency reference signals accurately.
mid-frequency range: This section of the plot is usually shaped by the controller to give
the required performance in terms of, for example, acceptable
bandwidth ( speed of response) and overshoot. We note that the peak
on the closed loop magnitude plot is usually given the notation Mp and
occurs at the frequency ωp. The value of Mp can be evaluated for
second order systems to be
 1
Mp =
2ζ( 1-ζ2)
ωp = ωn 1-2ζ2
and can be used as a design specification for systems which exhibit
primarily second order characteristics.
high frequency range: Usually a degree of roll-off is required to remove noise or high
frequency disturbances which may affect the system.
Closed loop
magnitude, dB
20
Closed loop peak
magnitude, Mp
10

0
Gain of 1 (=0dB)
-10

-20

-30
High frequency
-40 roll-off

-50 Frequency point,

ωp
-60

-70

-80
-1 0 1 2
10 10 10 10
Frequency, rad/s

Figure 0.13 Magnitude plot of closed loop transfer function

(i) Sensitivity magnitude plot:

Y(s) = S(s) d(s)
We recall that the sensitivity plot gives us the information on how the disturbances affect the
output of the system.The following MATLAB code will produce the magnitude plot of S(s).
 MATLAB code
gol = 6/(s*(s+1)*(s+3);
sens=1/(1+gol);
w = logspace(-1,2,300);
[mag2,ph2] = bode(sens,w);
magdb2 = 20*log10(mag2);
semilogx(w,magdb2(:))
Figure 0.14 shows the magnitude plot of the sensitivity function. Once again, we can examine
the plot in the low, middle and high frequency ranges:
low frequency range: There is very low gain at low frequency. Hence any low
frequency disturbances (usually undesirable) will be attenuated by the
system.
mid-frequency range: The mid-frequency range shows a peak. The magnitude and
position of this peak on the frequency axis can be determined by tuning
the controller. For example, it may be important not to amplify signals
above 1 rad/s. In our plot, we would have to redesign the controller
(that is , change the open loop transfer function) to meet this
specification.
high frequency range: We find that the sensitivity function is usually unity at high
frequencies, (and this corresponds to the transfer function GCL(s) taking
near zero values at high frequency.
Magnitude of
Sensivity function, dB
15

10 Area of amplification:
Frequency range 0.9 to 7 rad/s
5

Unity gain value at

-5 high frequencies

-10

-15

-20

Attenuation at low
-25 frequencies

-30
-1 0 1 2
10 10 10 Frequency, rad/s 10

Figure 0.14 Magnitude plot of Sensitivity transf er function

What we have learnt

 how to use MATLAB to plot a Bode plot
 how to enter different frequency ranges into the MATLAB Bode command
 that there are 4 basic components which can be combined together to form a Bode plot:
 system gain K,
s-terms (sp),
first order lag/lead terms (τs+1),
1
second order underdamped terms
1 2 2ζ
(s + s +1)
ωn2 ωn
 how to form a Sketching Table which identifies key values of combinations of transfer
function components
 how to recognise low and high frequency asymptotes and find the corner frequencies for
systems.
 to identify the alterations to Bode plots that some simple cascade transfer functions
introduce.