How can I get proper length of FFT

I'm trying to transforming an already know signal from time domain to frequency domain using fftfunction in MATLAB.

And I got different values of amplitude whenever I change the value(nfft) of length of signal.

I'm therefore wondering if there is a rule to choose the length of signal when calculating FFT.

From I understand the amplitude of signal should be same when it is on the time domain and frequency doamin.

am I wrong?

I have an acoustic signal got from a sensor and need to analyze it particularly SPL

I want to transform it into frequency domain using fftfunction in MATLAB. And I don't know how can I choose the length of signal.

Could somebody please help?

This is my code below

[y,Fs] = audioread('AA.wav')

%% Implementation FFT
t=linspace(0,length(y)/Fs,length(y));
nfft=512;
f=linspace(0,Fs,nfft);
g=abs(fft(y,nfft));
F2=f(1:nfft/2)
G2=g(1:nfft/2)

%% ploting FFT
figure(1)
plot(F2,G2)
xlabel('Frequency')
ylabel('Amplitude')
legend('FFT for SPL');

[David Goodmanson]

Hello Byeongoh,

How does the size of nfft compare to the length of y? Is it always less?

[Adam]

This thread might help. You need to scale the FFT for correct results.

[Byeongoh Min]

y is very huge as compare to nfft almost 800,000

[Star Strider]

From the documentation, the input can be:

• Data Types: double | single | int8 | int16 | int32 | uint8 | uint16 | uint32 | logical

The output will then be:

• If X is of type single, then fft natively computes in single precision, and Y is also of type single. Otherwise, Y is returned as type double.

Answer 1

[Star Strider]

To have the correct amplitudes, scale the fft by dividing it by the length of the original signal. The ‘nfft’ argument (usually greater than the length of the original signal) will increase the frequency resolution, so some of the amplitudes may be slightly different. If you use a one-sided fft, multiply the resulting amplitudes by 2.

x = your_signal;
t = your_time_vector;
Fs = 1/mean(diff(t));                                   % Sampling Frequency
Fn = Fs/2;                                              % Nyquist Frequency
N = length(x);
NFFT = 2^nextpow2(N);
FTx = fft(x, NFFT) / N;
Fv = linepace(0, 1, fix(N/2)+1)*Fn;                     % Frequency Vector
Iv = 1:numel(Fv);                                       % Index Vector

figure
plot(Fv, abs(FTx(Iv))*2)
grid

[昊 王]

> From I understand the amplitude of signal should be same when it is on the time domain and frequency doamin

I also think so, So can you explain why this sentence doesn't apply here?

Not from the perspective of matlab code, but the perspective of signal and system.