Re: FFT - Defining Frequency Axis

[SpaceEngineer]

Hello Everyone,

I'm trying to analyse a time-varying signal for helicopter main rotor pitch rod using FFT in Matlab.

I have written the code and understand the concept of FFT too but the specification of the frequency axis does not seem to work too well as I'm getting a lot of frequency leakage for my signal.

If anyone can please take a look at the following code and make suggestions, I would be grateful, particularly for defining the frequency axis in Hz.

%Define Variables
x = 24;
N = 2^x;
f_datalog = 48;
fs = f_datalog*6;

X = fft(MRPR1,N); %Perform N-Point FFT on the data
Y = abs(X); %Determine the magnitude
f = ((0:N-1)*(fs/N)); %Specify the frequency range

plot(f,Y),axis([0 60 0 3.5*10^5])
title('Magnitude Versus Frequency for MRPR1 D8C1');
xlabel('Frequency (Hz)');
ylabel('Magnitude');
%grid on


Thanks heaps,
SpaceEngineer

 

[GregLocock]

I doubt any problems with leakage are associated with errors in the frequency axism which by eye looks Ok to within a factor of 2.

Hanning. There's a lot more to signal analysis than 'understanding ffts'.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

see:

http://en.wikipedia.org/wiki/Window_function

TTFN

FAQ731-376

 

[hacksaw]

if you are attempting to characterize the freq content of a time signal, especially if it is noisy, the segment of data may have a dc level that needs removing. you'll get cleaner results, windowing can help as well a averaging multiple data sets. In some cases when the spectrum is particularly noisy wavelet transforms are helpful

 

[SomptingGuy]

Always start with a trivial synthesised signal, it's the only way. When you're happy that the magnitude vs frequency plot is what you expect, then introduce real data.

- Steve