MATLAB: Pwelch and accuracy/number of points

downsamplefrequencypwelchresample

Hello, I have a signal sampled at 200Hz and when I perform the Pwelch method the returning frequency vector is till 100Hz as expected. My interest, though, is only at the frequencies 0-2Hz. Using xlim([0 2]) i can see the frequency content of interest but i want to use more points or maybe downsample my signal in order to get better accuracy at this frequencies. What do you propose?

Best Answer

If you only need to get more points in pwelch plot, I think increasing the number of FFT points in pwelch should do the trick.

Related Solutions

MATLAB: How to convert the units of the y-axis into miliseconds^2/Hz in power spectral density analysis of an ECG signal

The y-axis on a PSD estimate will not be in units of the time squared, it will be in amplitude squared per Hz, so for an ECG signal I assume that will be in millivolts, not milliseconds. The recorded values of the waveform are not in time.

Since you are using dspdata.psd, why not just use spectrum.periodogram?

    psdest = psd(spectrum.periodogram,ecg,'NFFT',length(ecg),'Fs',fs);
    plot(psdest)

If you want to plot without using dB (that is preferred by the way)

   plot(psdest.Frequencies,psdest.Data)

MATLAB: Setting pwelch frequency resolution

Refer to the documentation of “pwelch” function for the calculation of different parameters to be used to achieve the desired frequency resolution. If you want to learn more about how the “pwelch” function is used in a practical application, refer this example here.

The following code snippet will help you in using the “pwelch” function in your application.

% Reset the random number generator for reproducible results.  
rng default 
  
fs = 100;  % Assuming the sampling frequency is 100 Hz
t = 0:1/fs:60-1/fs; % Assuming that the signal is of 60 Seconds length
x = cos(2*pi*0.01*t) + randn(size(t)); % Create a test signal in which some randomized AWGN noise 
% is added to a sinusoidal signal of 0.01 Hz frequency 
% Use Welch's power spectral density estimation function to estimate the PSD.
  
% Calculation of the input parameters for "pwelch" function to get the required frequency resolution
  
% Resolution Required = 0.01 Hz  
resolutonReqd = 0.01;
% Calculate number of FFT points (NFFT) required for the required resolution
% fs/NFFT = resolutionReqd
NFFT = fs / resolutonReqd;
% Set window size for averaging to 200 
window = 200; % This makes number of FFT averages to 6000 samples/ 200 = 30
% Set overlap to 0 
overlap = 0;
% Now use pwelch with above parameters.
[pxx,f] = pwelch(x,window,overlap,NFFT,fs); 
% Now plot the PSD
plot(f,10*log10(pxx))
xlabel('Frequency (Hz)')
ylabel('PSD (dB/Hz)')

Best Answer

Related Solutions

MATLAB: How to convert the units of the y-axis into miliseconds^2/Hz in power spectral density analysis of an ECG signal

MATLAB: Setting pwelch frequency resolution

Related Question