0
$\begingroup$

Given a sampling frequency Fs lets say we plot the magnitude of the fft of a temporal signal $x$, for different frequencies above the Nyquist frequency, to show the effect of aliasing:

Fs=10;
t=0:1/Fs:1-(1/Fs);
f=6;%7,8,9, ... to see aliasing starting a 6 since 6 is just above the Nyquist freq
x=sin(2*pi*f*t);
freq_range=Fs*((0:length(x)-1)/length(x));
plot(freq_range,abs(fft(x)));

Can someone help to formulate mathematically to which lower frequencies a given freqency f > Fs/2 will be aliased to?

Experimentally, we can see that (for integer frequencies, dont know if it really makes sense to use fractionnal ones?):

6 -> 4

7 -> 3

8 -> 2

9 -> 1

10 -> 0 <- should be zero? but the spectrum goes crazy and there are peaks everywhere.

Why? Then one would think that this would be periodic i.e. f=11 -> f_aliased=1, etc. back to zero in decreasing order, then again some strange things happens at 15 then it goes on cyclically etc. What is the relation?

Also, w.r.t. spectral resolution ($\Delta f = F_s/N$, let $N$ be the nb of samples, i.e. length(x)), one could ask if the next frequency abouve the Nyquist is in fact $f_{Nyq}+\Delta f = (F_s/2)+\Delta f$ ? But in this example the spectral resolution $\Delta f$ = 1 right?

$\endgroup$

1 Answer 1

1
$\begingroup$
  1. If you do spectral analysis is typically better to use WAY more points then 10 in the time domain
  2. Look at your y axis and make sure your graphs are properly scaled. $f=10$ aliases indeed back to $f=0$. All you see is numerical noise and your y-scale is $10^{-15}$. This would be different if you choose a a different phase, cosine instead of sine.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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