ECE

EE 8443
3512 – PatternContinuous
– Signals: Recognition
and Discrete

LECTURE 29: BODE PLOTS AND FILTERS

• Objectives:
First-Order System
Second-Order System
Poles and Zeros
Butterworth Filters
Chebyshev Filters
Frequency Transformations

• Resources:
MIT 6.003: Lecture 18
Wiki: Bode Plots
EC: Bode Plots
Wiki: Butterworth Filters
Wiki: Chebyshev Filters

URL:
First-Order System Revisited
• Recall the transfer function
for a 1st-order system:
1 t / 
1/ 1 h(t )  e u (t )
H (s)  Re{s}   
s 1/ 
• The frequency response of this
system is:
1/
H ( j ) 
j  1 / 
1/
H ( j )  s (t )  1  e  t /  u (t )
 2  (1 /  ) 2
 1  0

  1/ 2   (1 /  )
~ 1 /( )   (1 /  )


H ( j )   tan 1  
 1  0

   / 4   (1 /  )
  / 2   (1 /  )

EE 3512: Lecture 29, Slide 2
Bode Plot
• Asymptotic approximation to
the actual frequency response.
• Plotted on a log-log scale.
• Allows approximation of the
frequency response using
straight lines:
20 dB/decade
 1  0

H ( j )   1 / 2   (1 /  )
~ 1 /( )   (1 /  )

• Note that:
 1 
20 log10 H ( j )  20 log10  
   Changes by /2
 20 log10  
 H ( j 0 ) 
20 log10    20 log10 10 
 H ( j100 ) 
 20 dB/decade

EE 3512: Lecture 29, Slide 3

Second-Order System
• The transfer function of a
2nd-order system:
n2
H (s)  2
s  2 n s  n2
• The frequency response of this
system can be modeled as: 40 dB/decade
2 2
  2   2 
H ( j ) dB  20 log 1  2    
 n   n 
• When   n :
2
 2  Changes by 
H ( j ) dB  20 log  2 
 n 
 
 40 log  
 n 
 2 n 
H ( j )   tan 1  
  
 180

EE 3512: Lecture 29, Slide 4

Example: Poles and Zeroes
• Transfer function:    
4 j    1
1000( s  2) 1000( j  2)  2 
H (s)  H ( j )  
( s  10)( s  50) ( j  10)( j  50)        
 j    10  j    1
  10    50  
• The critical frequencies
are ω = 2 (zero), 10 (pole),
and 50 (pole).
• MATLAB (exact resp.):
w = logspace(-1,3,300);
s = j*w;
H = 1000*(s+2)./(s+10)./(s+50);
magdB = 20*log10(abs(H));
phase = angle(H)*180/pi;
• MATLAB (Bode):
num = [1000 2000];
den = conv([1 1o], [1 50]);
bode(num, den);
• Bode plots are useful as an
analytic tool.
EE 3512: Lecture 29, Slide 5
Example: Comparison of Exact and Bode Plots

EE 3512: Lecture 29, Slide 6

Causal Filters
• Recall that the ideal lowpass filter
is a noncausal filter and hence
unrealizable.
• The filter design problem then
becomes an optimization problem
in which we try to best meet the
user’s requirements with a filter that
can be implemented with the least number of components.
• This is equivalent to trying to minimize the number of coefficients in the
(Laplace transform) transfer function. It is also comparable to minimizing the
order of the filter.
• Typically the numerator order is less than or equal to the denominator order,
so, building on the concept of a Bode plot, we can relate the maximum amount
of attenuation desired to the order of the denominator.
• There are often no “perfect” solutions. Users must tradeoff design constraints
such as passband smoothness, stopband attenuation and linearity in phase.
• Computer-aided design programs, including MATLAB, are able to design
filters once the user adequately specifies the constraints. We will explore two
interesting analytic solutions: Butterworth and Chebyshev filters.
EE 3512: Lecture 29, Slide 7
Butterworth Filters
• Butterworth filters are maximally flat in the
passband. Their distinguishing characteristic is
that the poles are arranged on a semi-circle of
radius ωc in the left-half plane.
• The filter function is an all-pole filter that exploits
the properties of Butterworth polynomials.
• Examples:
n2
N  2 : H (s) 
s 2  2n s  n2
c3
N  3 : H (s)  3
s  2c s 2  2c2 s  c3
• Frequency Response:
1
H ( j ) 
1   / c 
2N

• Design strategy: select the cutoff

frequency and the amount of
stopband attenuation, then
compute the order of the filter.
EE 3512: Lecture 29, Slide 8
Chebyshev Filters
• Chebyshev filter, based on Chebyshev polynomials, allows ripple in the
passband to achieve greater attenuation. This implies a lower filter order at the
expense of smoothness of the frequency response in the passband.
• Prototype:
1
H ( j ) 
1   2TN2  / c 
where:
T0 ( x)  1
T1 ( x )  x
TN ( x)  2 xTN 1 ( x)  TN  2 ( x)
and  is chosen based
on the amount of passband
ripple that can be tolerated.
• Filter design theory is based
largely on the mathematical
properties of polynomials.
• An third type of filter, the
elliptical filter, attempts to
allow ripple in both bands.
EE 3512: Lecture 29, Slide 9
Frequency Transformations
• There are two general approaches to
filter design:
 Direct optimization of the desired
frequency response, typically
using software like MATLAB.
 Design a “normalized” lowpass
filter and then transform it to
another lowpass, highpass or
bandpass filter.
• The latter approach can be achieved
by using these simple frequency
transformations:
To Lowpass : s  s1 / 2
To Highpass : s  12 / s
s 2  12
To Bandpass : s  c
s2  1 

• This latter approach is popular

because it leverages our knowledge
of the properties of lowpass filters.
EE 3512: Lecture 29, Slide 10
Summary
• Analyzed the frequency response of a first and second-order system.
• Demonstrated how this can be approximated using Bode plots.
• Demonstrated that the Bode plot can be used to construct the frequency and
phase responses of complex systems.
• Introduced Butterworth and Chebyshev lowpass filters.
• Described how lowpass filters can be converted to bandpass and highpass
filters using a frequency transformation.

EE 3512: Lecture 29, Slide 11