Skip to content
Related Articles

Related Articles

Gibb’s Phenomenon Rectangular and Hamming Window Implementation
  • Last Updated : 16 Mar, 2021

What is Gibb’s Phenomenon?

Gibb’s Phenomenon in Digital Filter’s refer to the manner how a Fourier series of periodic functions behave near jump discontinuation, the partial sum of the Fourier series has large oscillations near the discontinuation, which might increase the maximum of the partial sum above that of the function itself. Thus, there is an overshoot as the terms are getting added of the Fourier series at the discontinuity.

In the signal processing terms, we can translate Gibb’s Phenomenon as the step response of the filter & the truncation of our signal via Fourier series can be termed as filtering out the signal.

We will be understanding this via example problem:

Specifications of the Filter



  1. Digital Lowpass FIR filter with 41 taps. 
  2. Cutoff frequency of the filter is pi/4

We will be implementing the solution using both the Rectangular & Hamming Window Technique.

What is Rectangular Window Technique?

In the Rectangular window, the Fourier Transform converges to sinc function, giving a rectangular shape at the edges thus the name Rectangular Window. The Rectangular is not widely used in the industry for filtering signals.

What is the Hamming Window Technique?

Hamming Window is a far more optimized approach to filter signals as it cuts off signal points on either side for us to see the more clear picture of the signal’s frequency spectrum. 

Step-by-step Approach:

Step 1: Importing all the necessary libraries.

Python3




##import required library
import numpy as np
import scipy.signal as signal
import matplotlib.pyplot as plt

Step 2: Define variables with the given specifications of the filter.

Python3




#Given specification
wc =np.pi/4        #Cutoff frequency in radian
N1=int(input())    #Given filter length
M=(N1-1)/2         #Half length of the filter

Step 3: Computations to calculate the magnitude, phase response to get Rectangular Window coefficients

Python3




N = 512    ## Choose DFT size
n = np.arange(-M,M)   
#Desired impulse response coefficients of the lowpass 
#filter with cutoff frequency wc
hd = wc / np.pi * np.sinc(wc * (n) / np.pi)
  
#Select the rectangular window coefficients
win =signal.boxcar(len(n))
  
#Perform multiplication of desired coefficients and 
#window coefficients in the time domain 
#insted of convolving in the frequency domain
h = hd * win # Modified filter coefficients
  
##Compute the frequency response of the modified filter coefficients
w,Hh = signal.freqz(h, 1, whole = True, worN = N)## get entire frequency domain
##Shift the FFT to center for plotting
wx = np.fft.fftfreq(len(w))

Step 4: Plotting the Truncated Impulse Response, Frequency Response, Frequency Response of the Filter  using The Rectangular Window Method

Python3




##Plotting of the results
fig,axs = plt.subplots(3, 1)
fig.set_size_inches((16, 16))
plt.subplots_adjust(hspace = 0.5)
  
#Plot the modified filter coefficients
ax = axs[0]
ax.stem(n + M, h, basefmt =  'b-', use_line_collection = 'True')
ax.set_xlabel("Sample number $n$", fontsize = 20)
ax.set_ylabel(" $h_n$", fontsize = 24)
ax.set_title('Truncated Impulse response $h_n$ of the Filter', fontsize = 20)
  
#Plot the frequency response of the filter in linear units
ax = axs[1]
ax.plot(w-np.pi, abs(np.fft.fftshift(Hh)))
ax.axis(xmax = np.pi/2, xmin = -np.pi/2)
ax.vlines([-wc,wc], 0, 1.2, color = 'g', lw = 2., linestyle = '--',)
ax.hlines(1, -np.pi, np.pi, color = 'g', lw = 2., linestyle = '--',)
ax.set_xlabel(r"Normalized frequency $\omega$",fontsize = 22)
ax.set_ylabel(r"$|H(\omega)| $",fontsize = 22)
ax.set_title('Frequency response of $h_n$ ', fontsize = 20)
  
#Plot the frequency response of the filter in dB
ax=axs[2]
ax.plot(w-np.pi, 20*np.log10(abs(np.fft.fftshift(Hh))))
ax.axis(ymin = -80,xmax = np.pi/2,xmin = -np.pi/2)
ax.vlines([-wc,wc], 10, -80, color = 'g', lw = 2., linestyle = '--',)
ax.hlines(0, -np.pi, np.pi, color = 'g', lw = 2., linestyle = '--',)
ax.set_xlabel(r"Normalized frequency $\omega$",fontsize = 22)
ax.set_ylabel(r"$20\log_{10}|H(\omega)| $",fontsize = 18)
ax.set_title('Frequency response of the Filter in dB', fontsize = 20)



Step 5: Computations to calculate the magnitude, phase response to get Hamming Window coefficients

Python3




#Desired impulse response coefficients of the lowpass filter with cutoff frequency wc
hd=wc/np.pi*np.sinc(wc*(n)/np.pi)### START CODE HERE ### (≈ 1 line of code)
  
#Select the rectangular window coefficients
win =signal.hamming(len(n))### START CODE HERE ### (≈ 1 line of code)
  
#Perform multiplication of desired coefficients and window
#coefficients in time domain 
#insted of convolving in frequency domain
h = hd*win### START CODE HERE ### (≈ 1 line of code ## Modified filter coefficients
  
##Compute the frequency response of the modified filter coefficients
w,Hh =signal.freqz(h,1,whole=True,worN=N)### START CODE HERE ### (≈ 1 line of code)      
## get entire frequency domain
  
##Shift the FFT to center for plotting
wx =np.fft.fftfreq(len(w))### START CODE HERE ### (≈ 1 line of code)   

Step 6: Plotting the Truncated Impulse Response, Frequency Response, Frequency Response of the Filter  using The Hamming Window Method

Python3




##Plotting of the results
fig,axs = plt.subplots(3,1)
fig.set_size_inches((10,10))
plt.subplots_adjust(hspace=0.5)
  
#Plot the modified filter coefficients
ax=axs[0]
ax.stem(n+M, h, basefmt='b-', use_line_collection='True')
ax.set_xlabel("Sample number $n$",fontsize=20)
ax.set_ylabel(" $h_n$",fontsize=24)
ax.set_title('Truncated Impulse response $h_n$ of the Filter', fontsize=20)
  
#Plot the frequency response of the filter in linear units
ax=axs[1]
ax.plot(w-np.pi, abs(np.fft.fftshift(Hh)))
ax.axis(xmax=np.pi/2, xmin=-np.pi/2)
ax.vlines([-wc,wc], 0, 1.2, color='g', lw=2., linestyle='--',)
ax.hlines(1, -np.pi, np.pi, color='g', lw=2., linestyle='--',)
ax.set_xlabel(r"Normalized frequency $\omega$",fontsize=22)
ax.set_ylabel(r"$|H(\omega)| $",fontsize=22)
ax.set_title('Frequency response of $h_n$ ', fontsize=20)
  
#Plot the frequency response of the filter in dB
ax=axs[2]
ax.plot(w-np.pi, 20*np.log10(abs(np.fft.fftshift(Hh))))
ax.axis(ymin=-80,xmax=np.pi/2,xmin=-np.pi/2)
ax.vlines([-wc,wc], 10, -80, color='g', lw=2., linestyle='--',)
ax.hlines(0, -np.pi, np.pi, color='g', lw=2., linestyle='--',)
ax.set_xlabel(r"Normalized frequency $\omega$",fontsize=22)
ax.set_ylabel(r"$20\log_{10}|H(\omega)| $",fontsize=18)
ax.set_title('Frequency response of the Filter in dB', fontsize=20)
  
fig.tight_layout()
plt.show()




 Attention geek! Strengthen your foundations with the Python Programming Foundation Course and learn the basics.  

To begin with, your interview preparations Enhance your Data Structures concepts with the Python DS Course. And to begin with your Machine Learning Journey, join the Machine Learning – Basic Level Course

My Personal Notes arrow_drop_up
Recommended Articles
Page :