TD convolution versus FD multiplication

Hi all.

I am trying to convolute two time doamin signals. singal 1 is imported to matalb as a FD signal while i am creating signal2 in matlab so i do create it in TD.

I tried:

(1) convert signal1 to TD and perform direct convolution with signal2

(2)Keep signal1 in FD ,convert signal 2 to FD, and perform multipication. then convert the results back to TD

I need the final results to be in TD.

However, Both reaults dont agree. Can someone Please help me to figure out the reason?

Answer 1

[Matt J]

Remember, linear convolution is not the same as cyclic convolution. Make sure you do appropriate zero padding, e.g.,

x=rand(1,5);
y=rand(1,5);

conv(x,y)

ans = 1×9

0.2999 1.0313 1.7011 2.3099 2.6181 2.2161 1.5177 0.9398 0.2859

ifft( fft(x,9).*fft(y,9)  ,'symmetric')

ans = 1×9

0.2999 1.0313 1.7011 2.3099 2.6181 2.2161 1.5177 0.9398 0.2859

[mzm]

Thank you. Do you mind having a look at the script attached with the question and let me know what's wrong?

[Matt J]

I don't see any zero-padding being done in your code.

[mzm]

How to choose the zero padding size? I mean how many zeros should I add?

[Matt J]

it should be length(x)+length(y)-1