Take the 2-minute tour ×
Signal Processing Stack Exchange is a question and answer site for practitioners of the art and science of signal, image and video processing. It's 100% free, no registration required.

I am basing my question on this post here, because I would like additional details on it, as I have not had any success in re-creating it.

I would like to simply learn, in no uncertain terms, how to do upsampling via the FFT. There seem to be a lot of details regarding if the signal is even or odd, where exactly to zero-pad, how much exactly, etc, and I am not getting those correct numbers for some reason.

Question: So let us say I have an even or odd length signal, of length $N$. I would like to upsample it to, say, $2N$, using FFTs.

How is this done exactly?

What I tried:

I tried following the post and others on the web but there seems to be very little consistency. It basically amounts to "zero-padding the frequency domain and then IFFTing" but I do not get proper results. In my case, I have a signal of length $100$, and I would like to upsample it to length $200$. So I do the following command:

ifft(fftshift([zeros(1,50) fftshift(fft(signal, 100)) zeros(1,50)]));

This however gives me complex data back, and not a signal that is simply upsampled by a factor of $2$. I am not sure where the problem is.

Thank you.

share|improve this question
    
In short, you have been misinterpreting how the spectrum is stored. Thanks for the answer, Peter K. –  user7358 Mar 10 at 19:37
    
@user7358 Ok, but why the downvote? –  TheGrapeBeyond Mar 10 at 20:27
add comment

1 Answer 1

up vote 7 down vote accepted

One trick, for even-length signals, is what to do with the "middle" sample. Here, I've split it half and half between each side of the FFT.

The other trick is to ensure that you have the right amplitudes in the resampled signal. Here's it's a factor of 2.

Try this in scilab:

x = rand(1,100,'normal');
X = fft(x);
XX = 2*[X(1:50) X(51)/2 zeros(1,99) X(51)/2 X(52:100)];
xx = ifft(XX);

clf;
plot([0:199]/200,xx)
plot([0:199]/200,xx,'go')
plot([0:99]/100,x,'r.')

which appears to generate the right thing. The red dots are the original samples, and the green circles (and blue connecting lines) are the resampled data.

enter image description here


UPDATE

For the case of an odd number of samples, there are fewer fiddly bits (no splitting the last coefficient):

x2 = rand(1,101,'normal');
X2 = fft(x2);
XX2 = 2*[X2(1:51) zeros(1,101)  X2(52:101)];
xx2 = ifft(XX2);

clf;
plot([0:201]/202,xx2)
plot([0:201]/202,xx2,'gx')
plot([0:100]/101,x2,'r.')
share|improve this answer
    
Thanks Peter. May you also please place an addition to your answer for the odd-cases for completion, based on the same example? Much appreciated! –  TheGrapeBeyond Mar 10 at 14:02
    
@TheGrapeBeyond:Done! –  Peter K. Mar 10 at 15:21
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.