I need to calculate power spectral density of a signal in MATLAB.

what I have to do is to calculate the noise in a signal and see how it depends on the frequency spectrum. I am trying to calculate PSD of a signal but everytime, I get an error saying "vectors must be the same lengths". I am not able to find a solution.

Here is command which I'm using. please let me know if there is some problem with the command.

I imported the excel file as vectors.

x=time; %time vector in excel file

y=current; %current vector in excel file

plot(x,y); % plot current-time graph

xlabel('time(s)');

ylabel('current(A)');

Fs=50; %sampling frequency

T=1/Fs; %sampling time

L=length(y); %length of the signal

L1= x/T;

t=(0:L-1)*T; time vector

%%%calculate the mean current

Iavg=mean(L);

%calculate the FFT

NFFT=2^nextpow2(L); %next power of 2 from length y

% calculate and plot PSD of conditioning mode

Nc=abs(fft((L-Iavg),NFFT)).^2./NFFT;% Nc is power spectral density

fc=Fs/2*linspace(0,1,NFFT/2+1);%frequency

plot(fc,Nc);

Answer 1

[Wayne King]

You are calculating your frequency vector for a one-sided PSD estimate, but then you try to plot the PSD estimate for the entire period (two-sided) against that.

Also, you do not want to find the PSD of L, that is just a scalar which is the length of the signal, y. So fft((L-Iavg))is definitely not what you want, because that is just 0!!!

 Nc=abs(fft((y-mean(y)),NFFT)).^2./NFFT;
 Nc = Nc(1:513);
 Nc(2:end-1) = 2*Nc(2:end-1);
 fc=Fs/2*linspace(0,1,NFFT/2+1);%frequency
 plot(fc,Nc)

Also, you probably want to plot it in dB

   plot(fc,10*log10(Nc))

[gaurav Nanda]

Thanks a lot for the answer. It still doesn't seem to work and it says "index exceeds matrix dimensions".

Actually this is the first time I am using Matlab and I have been to told to characterize the noise of my signal to find out low frequency noise 1/f, thermal noise and how do they depend on the concentration of Analyte, types of electrode used for measurement etc.

Also, I did not understand why you have used "Nc=Nc(1:513)"

Is there any way to perform curve fitting of this PSD?

Answer 2

[gaurav Nanda]

Thanks a lot for the answer. It still doesn't seem to work and it says "index exceeds matrix dimensions".

Actually this is the first time I am using Matlab and I have been to told to characterize the noise of my signal to find out low frequency noise 1/f, thermal noise and how do they depend on the concentration of Analyte, types of electrode used for measurement etc.

Also, I did not understand why you have used "Nc=Nc(1:513)" Is there any way to perform curve fitting of this PSD?

Answer 3

[Wayne King]

because you computed the DFT of a vector to length 1024, only 513 points are "positive" frequencies.

Why not just use spectrum.periodogram if you have the Signal Processing Toolbox?

 Fs = 1e3;
 t = 0:0.001:1-0.001;
 x = cos(2*pi*100*t)+randn(size(t));
 plot(psd(spectrum.periodogram,x,'Fs',1000,'NFFT',length(x)));