Skip to content
Related Articles

Related Articles

Design IIR Bandpass Chebyshev Type-1 Filter using Scipy – Python

Improve Article
Save Article
  • Last Updated : 10 Nov, 2021
Improve Article
Save Article

IIR stands for Infinite Impulse Response, It is one of the striking features of many linear-time invariant systems that are distinguished by having an impulse response h(t)/h(n) which does not become zero after some point but instead continues infinitely.

What is IIR Chebyshev Filter?

IIR Chebyshev is a filter that is linear-time invariant filter just like the Butterworth however, it has a steeper roll off compared to the Butterworth Filter. Chebyshev Filter is further classified as Chebyshev Type-I and Chebyshev Type-II according to the parameters such as pass band ripple and stop ripple.

How is Chebyshev Filter different from Butterworth?

Chebyshev Filter has a steeper roll-off compared to the Butterworth Filter.

What is Chebyshev Type-I Filter?

Chebyshev Type-I minimizes the absolute difference between the ideal and actual frequency response over the entire passband by incorporating an equal ripple in the passband.

The specifications are as follows:  

  • Pass band frequency: 1400-2100 Hz
  • Stop band frequency: 1050-24500 Hz
  • Pass band ripple: 0.4dB
  • Stop band attenuation: 50 dB
  • Sampling frequency: 7 kHz

We will plot the magnitude, phase, impulse, step response of the filter.

Step-by-step Approach:

Step 1: Importing all the necessary libraries.

Python3




import numpy as np
import scipy.signal as signal
import matplotlib.pyplot as plt


Step 2:Defining the user defined functions such as mfreqz() and impz().

Python3




def mfreqz(b, a, Fs):
   
    # Compute frequency response of the filter
    # using signal.freqz function
    wz, hz = signal.freqz(b, a)
 
    # Calculate Magnitude from hz in dB
    Mag = 20*np.log10(abs(hz))
 
    # Calculate phase angle in degree from hz
    Phase = np.unwrap(np.arctan2(np.imag(hz), np.real(hz)))*(180/np.pi)
     
    # Calculate frequency in Hz from wz
    Freq = wz*Fs/(2*np.pi)
     
    # Plot filter magnitude and phase responses using subplot.
    fig = plt.figure(figsize=(10, 6))
 
    # Plot Magnitude response
    sub1 = plt.subplot(2, 1, 1)
    sub1.plot(Freq, Mag, 'r', linewidth=2)
    sub1.axis([1, Fs/2, -100, 5])
    sub1.set_title('Magnitude Response', fontsize=20)
    sub1.set_xlabel('Frequency [Hz]', fontsize=20)
    sub1.set_ylabel('Magnitude [dB]', fontsize=20)
    sub1.grid()
 
    # Plot phase angle
    sub2 = plt.subplot(2, 1, 2)
    sub2.plot(Freq, Phase, 'g', linewidth=2)
    sub2.set_ylabel('Phase (degree)', fontsize=20)
    sub2.set_xlabel(r'Frequency (Hz)', fontsize=20)
    sub2.set_title(r'Phase response', fontsize=20)
    sub2.grid()
 
    plt.subplots_adjust(hspace=0.5)
    fig.tight_layout()
    plt.show()
 
# Define impz(b,a) to calculate impulse response
# and step response of a system
# input: b= an array containing numerator coefficients,
#a= an array containing denominator coefficients
def impz(b, a):
   
    # Define the impulse sequence of length 60
    impulse = np.repeat(0., 60)
    impulse[0] = 1.
    x = np.arange(0, 60)
 
    # Compute the impulse response
    response = signal.lfilter(b, a, impulse)
 
    # Plot filter impulse and step response:
    fig = plt.figure(figsize=(10, 6))
    plt.subplot(211)
    plt.stem(x, response, 'm', use_line_collection=True)
    plt.ylabel('Amplitude', fontsize=15)
    plt.xlabel(r'n (samples)', fontsize=15)
    plt.title(r'Impulse response', fontsize=15)
 
    plt.subplot(212)
    step = np.cumsum(response)
     
    # Compute step response of the system
    plt.stem(x, step, 'g', use_line_collection=True)
    plt.ylabel('Amplitude', fontsize=15)
    plt.xlabel(r'n (samples)', fontsize=15)
    plt.title(r'Step response', fontsize=15)
    plt.subplots_adjust(hspace=0.5)
 
    fig.tight_layout()
    plt.show()


Step 3:Define variables with the given specifications of the filter.

Python3




# Given specification
Fs = 7000  # Sampling frequency in Hz
fp = np.array([1400, 2100])  # Pass band frequency in Hz
fs = np.array([1050, 2450])  # Stop band frequency in Hz
Ap = 0.4  # Pass band ripple in dB
As = 50  # stop band attenuation in dB


Step 4: Computing the cut-off frequency

Python3




# Compute pass band and stop band edge frequencies
wp = fp/(Fs/2# Normalized passband edge frequencies w.r.t. Nyquist rate
ws = fs/(Fs/2# Normalized stopband edge frequencies


Step 5: Compute cut-off frequency & order

Python3




# Compute order of the  Chebyshev type-1 filter using signal.cheb1ord
N, wc = signal.cheb1ord(wp, ws, Ap, As)
 
# Print the order of the filter and cutoff frequencies
print('Order of the filter=', N)
print('Cut-off frequency=', wc)


Output:

Step 6: Compute the filter co-efficient

Python




# Design digital Chebyshev type-1 filter
# using signal.cheby1 function
z, p = signal.cheby1(N, Ap, wc, 'bandpass')
 
# Print numerator and denomerator coefficients
# of the filter
print('Numerator Coefficients:', z)
print('Denominator Coefficients:', p)


Output:

Step 7: Plotting the Magnitude & Phase Response

Python3




# Call mfreqz to plot the magnitude and phase response
mfreqz(z, p, Fs)


Output:

Step 8: Plotting the Impulse and Step Response

Python3




# Call impz function to plot impulse
# and step response of the filter
impz(z,p)


 
 

Output:

 

 

Full Code:

 

Python3




# import required library
import numpy as np
import scipy.signal as signal
import matplotlib.pyplot as plt
 
 
def mfreqz(b, a, Fs):
   
    # Compute frequency response of the filter
    # using signal.freqz function
    wz, hz = signal.freqz(b, a)
 
    # Calculate Magnitude from hz in dB
    Mag = 20*np.log10(abs(hz))
 
    # Calculate phase angle in degree from hz
    Phase = np.unwrap(np.arctan2(np.imag(hz), np.real(hz)))*(180/np.pi)
     
    # Calculate frequency in Hz from wz
    Freq = wz*Fs/(2*np.pi)
     
    # Plot filter magnitude and phase responses using subplot.
    fig = plt.figure(figsize=(10, 6))
 
    # Plot Magnitude response
    sub1 = plt.subplot(2, 1, 1)
    sub1.plot(Freq, Mag, 'r', linewidth=2)
    sub1.axis([1, Fs/2, -100, 5])
    sub1.set_title('Magnitude Response', fontsize=20)
    sub1.set_xlabel('Frequency [Hz]', fontsize=20)
    sub1.set_ylabel('Magnitude [dB]', fontsize=20)
    sub1.grid()
 
    # Plot phase angle
    sub2 = plt.subplot(2, 1, 2)
    sub2.plot(Freq, Phase, 'g', linewidth=2)
    sub2.set_ylabel('Phase (degree)', fontsize=20)
    sub2.set_xlabel(r'Frequency (Hz)', fontsize=20)
    sub2.set_title(r'Phase response', fontsize=20)
    sub2.grid()
 
    plt.subplots_adjust(hspace=0.5)
    fig.tight_layout()
    plt.show()
 
# Define impz(b,a) to calculate impulse response
# and step response of a system
# input: b= an array containing numerator coefficients,
# a= an array containing denominator coefficients
def impz(b, a):
   
    # Define the impulse sequence of length 60
    impulse = np.repeat(0., 60)
    impulse[0] = 1.
    x = np.arange(0, 60)
 
    # Compute the impulse response
    response = signal.lfilter(b, a, impulse)
 
    # Plot filter impulse and step response:
    fig = plt.figure(figsize=(10, 6))
    plt.subplot(211)
    plt.stem(x, response, 'm', use_line_collection=True)
    plt.ylabel('Amplitude', fontsize=15)
    plt.xlabel(r'n (samples)', fontsize=15)
    plt.title(r'Impulse response', fontsize=15)
 
    plt.subplot(212)
    step = np.cumsum(response)  # Compute step response of the system
    plt.stem(x, step, 'g', use_line_collection=True)
    plt.ylabel('Amplitude', fontsize=15)
    plt.xlabel(r'n (samples)', fontsize=15)
    plt.title(r'Step response', fontsize=15)
    plt.subplots_adjust(hspace=0.5)
 
    fig.tight_layout()
    plt.show()
 
 
# Given specification
Fs = 7000  # Sampling frequency in Hz
fp = np.array([1400, 2100])  # Pass band frequency in Hz
fs = np.array([1050, 2450])  # Stop band frequency in Hz
Ap = 0.4  # Pass band ripple in dB
As = 50  # stop band attenuation in dB
 
 
# Compute pass band and stop band edge frequencies
wp = fp/(Fs/2# Normalized passband edge frequencies w.r.t. Nyquist rate
ws = fs/(Fs/2# Normalized stopband edge frequencies
 
# Compute order of the  Chebyshev type-1
# filter using signal.cheb1ord
N, wc = signal.cheb1ord(wp, ws, Ap, As)
 
# Print the order of the filter and cutoff frequencies
print('Order of the filter=', N)
print('Cut-off frequency=', wc)
 
# Design digital Chebyshev type-1 filter using
# signal.cheby1 function
z, p = signal.cheby1(N, Ap, wc, 'bandpass')
 
 
# Print numerator and denomerator coefficients of the filter
print('Numerator Coefficients:', z)
print('Denominator Coefficients:', p)
 
# Call mfreqz to plot the magnitude and phase response
mfreqz(z, p, Fs)
 
# Call impz function to plot impulse and
# step response of the filter
impz(z, p)


Output:


My Personal Notes arrow_drop_up
Related Articles

Start Your Coding Journey Now!