DSPP lab

We were already knowing that Talele sir was going to teach DSPP and that was best motivation for the subject. This was not just because talele sir teaches best but also he has good sense of humour. He makes us understand key concepts by so many examples and jokes that each and every student is satisfied by his teaching. I am a diploma student and weak in maths, but even I enjoyed the subject. We all agree that subject are never boring or awesome it depends on the faculty what he/she makes it. Talele sir is mixture of both Einstein and zakir khan of S.P.I.T.

We had observed this experiment that was performed by the some of my classmates. In this experiment we used TMS320C5505 processor board, MATLAB to design a filter and obtain coefficients of the transfer function, TI Code Composite studio. This is used to filter out background noise. The input was taken from microphone and out was observed using headphone.

Our topic was to blink the LEDs with music. Basically, it was an Audio signal to Light Signal conversion.  Red Green Blue LED strip is blinked with the music rhythm using FFT.  Fast Fourier Transform is used to obtain the amplitude of various frequencies in the audio signal and these amplitudes are converted into Pulse Width Modulation and sent to the transistor circuit which blinks the RGB lights accordingly. Also giving  a review on one IEEE paper and one patent on the above topic. Watch the video  here . Code Link:  GitHub Link .

I performed this experiment using c programming language in which I implemented the code to perform fast Fourier transform by using the butterfly structure of solving a discrete time signal in which we decompose the n point signal into smaller signals for faster speed of calculation. Using the formula for complex and real addition and multiplication I verified that indeed lesser calculations are required. On discussion with Prof Talele, he asked me to implement the code to find the correlation coefficient of two input signals and take one more case with different values of the input signal. On the basis of values of Pearson’s Coefficient the degree of similarity of input signals was found out. The degree of similarity is approximately 1 for highly similar signals and approximately 0 for highly dissimilar signals. 

This experiment was the design of another digital filter. In this experiment too we had to design Low Pass as well as High Pass Filter. Basically, the formula to find the order of the filter changed w.r.t Butterworth and various other input parameters were added. In this digital filter, the magnitude response was equiripple in the passband and monotonic(smooth) in the stop band. The number of valleys and peaks of the ripples equalled the order of the filter. In this filter design instead of convolution, recursion is used.

Sir had explained to us the basic of all filters. Once we understood the range of frequencies allowed and rejected by lpf,hpf,apf,bpf,bsf we decide the transfer function. The position of poles and zeroes for each filter was told by Talele Sir. Accordingly, we took a transfer function and plotted the magnitude response in Scilab using the frmag function.we also plotted the pole-zero graph. we observed the effect on amplitude and bandwidth by varying the poles and zeroes

Sir had taught us two methods for performing the above experiment;overlap-add and overlap-save method.I went forward with the overlap-add method. The algorithm was clear in my mind, so it was easy to proceed with my code. Also, the output sequence is the circular convolution of input sequences and transfer function. So all I did was break the input sequences into smaller sequences and implement the circular convolution function which I had prepared earlier.H however, I had a few glitches since the length of the decomposed input sequences were to be altered. After a few modifications, the code was implemented successfully.

We had learned how to find the discrete Fourier transform of a signal in theory lectures. In class, we have practised it by the formula method and the matrix method as well. Thus we were aware of the procedure to find DFT. We broke the code into parts: taking the input, the function, displaying the output. Among these, the function was the most challenging part. The calculated results of DFT we displayed in the rectangular form. We manually found out the polar form and plotted the magnitude and phase spectrum on paper. We then appended the signal with zeroes and observed the results. We noted that the shape of the spectrum remains the same but frequency spacing decreases. Thus we noted that the number of points in the input signal affects various factors of the DFT signal such as accuracy, frequency spacing, resolution.  While executing the program we got a few errors, however by patiently checking the program step by step, we were able to debug them.

Sir had covered the concepts of circular and linear convolution in class. So when we went for the practical we were aware of the logic behind convolution. We wrote the codes to perform linear, circular convolution and to find linear convolution using circular convolution. First, we performed linear convolution and observed that for linear convolution the length of the output sequence is one less than the sum of the lengths of the two input sequences. After that, we executed circular convolution and we observed that the length of the output of the circular convolution is the maximum of the length of the two input sequences. We also observed that for casual inputs the output sequence is also causal. We observed aliasing effect for circular convolution. Then we amended our circular convolution code to execute linear convolution using circular convolution. This experiment strengthened our basics of convolution and it was a great learning experience and we took home something substantial....