Scribd
Upload
23 views7 pages

Chebyshev Filter Design Methods

Uploaded by

Untaught Stuff
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
23 views7 pages

Chebyshev Filter Design Methods

Uploaded by

Untaught Stuff
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

ECL352 – DIGITAL SIGNAL PROCESSING

1. Design a Chebyshev low pass filter with the specifications αp = 1 dB ripple in the
passband 0 ≤ ω ≤ 0.2π, αs = 15 dB ripple in the stopband 0.3π ≤ ω ≤ π, using (a),
bilinear transformation, (b). Impulse invariance.

Given data αp = 1 dB ; ωp = 0.2π, αs = 15 dB; ωs = 0.3π


Prewarped frequencies are given by

2 ωp 0.2π
Ωp  tan  2tan  0.65
T 2 2
2 ωs 0.3π
π (2k  1)π
1 s  tan
Ω  2tan  1.02
0.1α
2 φk  22 2N k  1, 2,3,4
10 p
1
cosh T
10 0.1αs 1
N  3.01 φ1  112.5  ; φ2  157.5  ; φ3  202.5  ;
Ω
cosh -1 s
ΩP φ4  247.5  ;
Let us take N  4
s1  a cos φ1  jbsinφ1   0.0907  j0.639
ε  (10 0.1α P
 1) 0.5
 0.508 s 2  a cos φ2  jbsinφ2   0.2189  j0.2647
1 2
με  1 ε  4.17 s 3  a cos φ3  jbsinφ3   0.2189  j0.2647
s 4  a cos φ4  jbsinφ4   0.0907  j0.639
The poles of a Chebyshev filter
 μ 1/N  μ 1/N 
a  ΩP    0.237
 2 
μ  μ
1/N 1/N

b  ΩP    0.6918
 2 
The denominator of H(s) =[(s+0.0907)2 +(0.639)2]
=(s2+0.5306s+0.3516)
As N is even, the numerator of H(s) =(0.3516) /√1+ε2

=0.28

The transfer function H(s) = 0.28/(s2+0.5306s+0.3516)


H(z) = H(s) | 2 1  z 1 
s  
T 1  z 1 
0.001836(1  z 1 )4

(1  1.499z 1  0.8482z  2 )( 1  1.5548z 1  0.6493z  2 )

Impulse Invariance Method:


Given data αp = 1 dB ; ωp = 0.2π, αs = 15 dB; ωs = 0.3π

10 p  1
0.1α

cosh 1
10 0.1αs  1
N  3.2
-1 Ω s
cosh
ΩP
Let us take N  4
π (2k  1)π
φk   k  1, 2,3,4 ε  (10 0.1αP  1)0.5  0.508
2 2N
φ1  112.5  ; φ2  157.5  ; φ3  202.5  ; μ  ε 1  1  ε  2  4.17
φ4  247.5  ;
The poles of a Chebyshev filter
s1  a cos φ1  jbsinφ1   0.0876  j0.619  μ 1/N  μ 1/N 
a  ΩP    0.229
s 2  a cos φ2  jbsinφ2   0.2115  j0.2564  2 
s 3  a cos φ3  jbsinφ3   0.2115  j0.2564 μ  μ
1/N 1/N

b  ΩP    0.67
s 4  a cos φ4  jbsinφ4   0.0876  j0.619  2 

The denominator of H(s) =[(s+0.0876)2 +(0.619)2] [(s+0.2115)2 +(0.2564)2]


=(s2+0.175s+0.391) (s2+0.423s+0.11)
As N is even, the numerator of H(s) =(0.391) (0.11)/√1+ε2

=0.03834

The transfer function H(s) = 0.03834 / [(s2+0.175s+0.391) (s2+0.423s+0.11)]


A A
H(s)  s  ( 0.0876  j0.619)  s  ( 0.0876  j0.619)
B B
 
s  ( 0.2115  j0.2564) s  ( 0.2115  j0.2564)

Using Impulse invariant transform

 0.083  0.0245z1 0.083  0.0238z1


H(z)  
1  1.49z1  0.839z2 1  1.56z1  0.655z2

2. Design an ideal high-pass filter with a frequency response


1 for π/4  ω  π
H d (e jω )  
0 for | ω |  π/4

Evaluate the values of h(n) for N = 11 and compute the transfer function of the
realizable filter H(z).
The desired frequency response is shown in Fig.1
We know

Figure1: the ideal frequency response of highpass filter of example 1


-
Truncating to 11 samples, we have
for
= 0 otherwise
For n=0

=
From the given frequency response we can find that α =0. Therefore, the
filter coefficients are symmetrical about n = 0 satisfying the condition

For n=1

= - 0.225
the transfer function of the filter is given by

= 0.75 +
The transfer function of the realizable filter is

= 0.0045 – 0.075 - 0.159 -0.225 + 0. 75 -0.225 -


0.159 -0.075 -0.045

3. Repeat the problem 2 using Hanning Window.


Hanning window

= 0 otherwise
For N = 11

= 0 otherwise

The filter coefficients can be obtained as below

The filter coefficients using Hanning window are

= 0 otherwise

The transfer function of the filter is given by

The transfer function of the realizable filter is

1. For the given specifications, design an analog Butterworth filter.


0.9 ≤ H(jΩ) ≤ 1 0 ≤ Ω ≤ 0.2π
H(ejω) ≤ 0.2 0.4π ≤ω≤ π
1 1
From the data, we find Ωp = 0.2π ; Ωs = 0.4π ;  0.9 ; and  0.2
1  2 1  2
From which we obtain
ε = 0.484 and λ = 4.898

 
log  
N     3.34
Ω
log s
ΩP
Let us take N  4
For N = 4, the transfer function of normalized Butterworth filter is
1
H(s) =
( s  0.76537 s  1)(s 2  1.8477 s  1)
2

p
We know Ωc = 0.1 p
 0.24
(10  1)1 / 2 N
H(s) for Ωc= 0.24π can be obtained by substituting s → s/ 0.24π in H(s)

1 1
H(s) = x
 s  2
 s    s  2
 s  
   0.76537   1    1.8477   1
 0.24   0.24    0.24   0.24  

0.323
=
( s  0.577 s  0.0576 2 )(s 2  1.393s  0.0576 2 )
2

2. An analog filter has transfer function H(s)=10/s2+7s+10. Design a digital filter


equivalent to this using impulse invariant method.

3. Using the Bilinear transformation, design a highpass filter, monotonoic in


passband with cutoff frequency 1000 Hz and down 10 dB at 350 Hz. The
sampling frequency is 5000Hz.
Prewarping the digital frequencies we have

Therefore we take N =1. The 1st order Butterworth filter for

Common questions

Powered by AI

Prewarping compensates for the non-linear frequency mapping inherent in the z-domain when using methods like bilinear transformation. This ensures that designated critical frequencies, such as cutoff or passband edges in the analog domain, linearly map onto corresponding digital domain frequencies. Without prewarping, frequency components might not exactingly maintain their positions post-conversion, altering filter characteristics such as bandwidth or cutoff. Prewarping aligns analog and digital frequency characteristics, resolving discrepancies and aiding precise digital filter implementation .

Poles in a Chebyshev filter directly impact the slope of the magnitude response, defining ripple behavior in passbands and stopbands. They are calculated based on tolerance specifications (e.g., ripple size defined by ε) using trigonometric identities involving hyperbolic and arccosine functions to reflect edge frequencies and ripple characteristics. Specifically, poles are critical in shaping response profiles due to their positions calculated from frequency-specific informations like Ωp and Ωs, configured by N, the filter order, and Chebyshev's inequality principles .

A Butterworth filter design starts by identifying passband edge Ωp and stopband edge Ωs frequencies, and the respective passband and stopband ripple specifications. These determine the minimum filter order N using the formula: 1/cosh(N) > (1/ε) and λ, being found based on ripple limits. The resultant Butterworth transfer function is characterized by a flat passband, defined using the formula H(s) = 1/[1 + (s/Ωc)^(2N)], where Ωc is the cutoff frequency. Adjustments for digital realization often involve bilinear transformations or prewarping techniques .

The design involves defining the passband and stopband edge frequencies, ωp = 0.2π and ωs = 0.3π, with corresponding ripples αp = 1 dB and αs = 15 dB. Using the bilinear transformation, prewarped analog frequencies are calculated. The magnitude of prewarping is determined to adjust the response accordingly. The Chebyshev polynomial degree (N) and parameters like ε are determined from the ratio of Ωp and Ωs using predefined equations. The transfer function, H(s), is specifically constructed based on these parameters and expressed in s-domain, where H(s) = 0.28/(s² + 0.5306s + 0.3516). This transfer function is then transformed to the z-domain using the bilinear transformation .

The impulse invariance method retains the sampled impulse response of the continuous-time filter, which can lead to aliasing unless the original signal is band-limited. This involves directly mapping the roots of the analog transfer function to the digital domain. Prewarping is still applied, with specific computations as given (Ωp and Ωs), but poles and zeros are mapped considering the aliasing effects contrary to the bilinear transformation, which avoids aliasing by frequency warping. For instance, using H(s) with prewarped frequencies, a transfer function in terms of z is derived, resulting in differences in digital filter characteristics when compared to bilinear transformation results .

The design starts with specifying the desired frequency response. For an ideal high-pass filter, response values are 1 for frequencies π/4 ≤ ω ≤ π and 0 otherwise. A Hanning window is applied to smooth out truncations at the ends of the filter response. The specified window length (N=11) is used to calculate filter coefficients, aiming for a symmetric and smooth transition. The transfer function in z-domain, H(z), is determined by these coefficients, ensuring it aligns closely with the truncated response profile .

Starting with an analog transfer function like H(s) = 10/(s² + 7s + 10), the impulse invariant method translates this directly to the digital domain by sampling the impulse response. This method maps poles to the complex z-plane, respecting the original mechanics of the s-plane structure. Calculations integrate sampling frequency effects, potentially leading to aliasing in non-bandlimited signals. The analog response is thus carefully reconstructed in the digital sample domain, maintaining original dynamics except where alias effects disrupt, which requires cautious oversight .

Designing a digital highpass filter via bilinear transformation includes setting a criterion for cutoff frequency and requisite attenuation rate. Prewarping is executed on key frequencies to compensate for non-linearities of frequency warping in digital representation. For a monotonic passband, the Butterworth filter is often used due to its flat response ideal for unaltered frequency ranges. The cutoff frequency (e.g., 1000 Hz) is established, and filter order (N) is derived based on gain reduction at a specified frequency (e.g., 10 dB down at 350 Hz). Final transformations align analog and digital specifications accurately .

The impulse invariance method is preferable when precise retention of a specific analog filter's transient characteristics in a sampled system is crucial, especially when the analog filter is a well-designed prototype with minimal aliasing risks (e.g., pre-sampled or bandlimited input). Unlike bilinear transformation, which warps frequency but retains a linear phase, impulse invariance method focuses on maintaining the signal's time-domain integrity by directly mapping the analog impulse response. This is appropriate for band-pass or band-stop filters where frequency linearity is not disrupted by the sampling process .

Window functions, such as Hanning, mitigate the side effects of abrupt truncations when creating finite impulse responses (FIR) from ideal infinite ones. They compensate for edge discontinuities by smoothing transitions, effectively reducing spectral leakage and sidelobes that arise from sharp frequency cuts. Window shapes (like Hanning or Hamming) determine the trade-off between main lobe width (resolution) and side lobe level (ripple), optimizing filter frequency response depending on the desired application. Hence, windowed designs offer more controlled spectral attributes .

You might also like