How to use fft exatly with data acquired by oscilloscope

Hi all i need to use the fft on an array of 5000000 values with fs=1e7 and plot it exactly. The shape is good but i'm very confused about the normalization for the fft. Someone can help me with a good code for the plot.

thx

Alessandro

Answer 1

[Star Strider]

It is all clearly and concisely explained in the text and code between the first two figures of the fft documentation.

[Alessandro Luciani]

thx for answering. I'm actually using this code but it doesn't satisfy parseval theorem, at least in my case

energy_a = sum(Signal.*conj(Signal)*dt) energy_b = sum(Y.*conj(Y) * 1/(fs*m))

where

Y = fft(Signal,NFFT)*ts m=length(Signal)

[Star Strider]

Using this definition and not normalising the amplitude of the fft, I get the same result for the energy of the signal and the fft of the signal:

s = randi(10, 100, 1);
Fs = fft(s);

Es = s'*s;
EFs = mean(abs(Fs).^2);

Use that approach and see if you’re still having problems.

[Alessandro Luciani]

with your code i understood the problem thx so much. Th fact is that i evalueted the fft with a window i.e Y = fft(Array,NFFT)/m NFFT = 2^nextpow2(length(array));

in this way Parseval is said. Using Y = fft(Array);

Parseval is satisfied.

Thx :)

[Star Strider]

My pleasure.

The window will filter the signals and remove much of the high-frequency energy.