Basic operations using DSP Processor

Unlike other experiments which were performed on Scilab or with the help of C programming language, this experiment was somewhat hardware based where we used the actual DSP processor. The DSP processor we used to implement the basic instructions like arithemetic, logical and shift operations. The results were observed by checking the change in the value of the register before and after the execution of the program. The DSP processors are capable of processing real time signal by storing them in long arrays and this is what makes them different from other normal proccesors. 

FIR Filter Design : Frequency Sampling Method

Out of the two method which can be used to design FIR filter, here we have used the second method i.e  Frequency Sampling Method. It is the most simplest and direct technique to design an FIR filter for getting the desired frequency response. In this method we simply sample the desired frequency response and perform an IDFT in order to obtain the corresponding FIR.
In this experiment a Scilab code was written to perform the operations.

FIR Filter design : Windowing Method

The method to design an FIR filter are different from that of IIR filter. There are two methods to design an FIR filter : 1)Windowing Method 2)Frequency Sampling Method. In this design of FIR filter we are using windowing method.
There are many window functions available. Depending upon the requirements a particular window function is selected. The two main criteria for deciding the window function are 1) Attenuation in stop band 2) Transition width

With the help of Scilab a code was written to design an FIR filter using Hanning window. The attenuation in stopband was chosen accordingly and hence a low pass filter was designed.

Chebyshev Filter Design

Chebyshev filter design is used to design an IIR filter. In this type of filter design there are ripples present in the magnitude response of the filter unlike butterworth filter design. There are four types of chebyshev filter design depending upon where the magnitude response of the filter. Here we have designed Chebyshev-I filter. In Chebyshev-I filter design, the ripples are present in the passband and monotonic in stopband.
A Low pass filter and High pass filter was designed using Chebyshev-I filter design. For the design the values of attenuation in stop band, attenuation in pass band, stop band frequency, pass band frequency and sampling frequency are the basic requirement. These values are to be given as input values. At the end, to get the final transfer function BLT method is used.

Analog and Digital Butterworth Filter

This session involved analyzing analog and digital butterworth filter. We used Scilab for creating a code for filter design. Then we entered the input parameters in the console of Scilab. Finally we obtained the output which is the magnitude response. We have compared the attenuation values and order of the filter. Implementation on Scilab is very helpful as we learned the code while debugging it.

Overlap Add and Overlap Save Method

Overlap Add and Overlap Save Method are two methods which are helpful for finding discrete convolution of a very long input signal.

In Overlap Add Method, we use linear convolution and in Overlap Save Method we use circular convolution. F.F.T algorithm is used for computing the convolution.

Overlap Add and Overlap Save Method are known as block processing techniques because the input sequence is divided into different blocks and each block is processed separately.
These methods are suitable for processing of real time. 

Fast Fourier Transform

Fast Fourier Transform is implemented where fast response is required. F.F.T is fast implementation of D.F.T. In F.F.T then number of computations are reduced as a result it is computationally faster than D.F.T.

In this experiment, we computed the F.F.T of signals having N=4 and N=8.
Radix-2 algorithm is used for computing F.F.T of an input signal i.e the input signal must 2^N points. To achieve this zero padding of the input signal is done. The input and output sequence in radix-2 F.F.T algorithm is in bit reverse manner.

We also calculated the number of computations required to calculate the F.F.T of a given signal and thus verified that F.F.T is computationally faster than D.F.T.