1

I want to find the period of a signal. On searching on the net, I found that the power spectrum density (PSD) can give an estimate of period, by finding the highest peak in the spectrum. However, there are many methods to estimate PSD. Some of the methods include welch, bartlett and periodogram. I am confused on which one to use to estimate the PSD. I also found that periodogram has good frequency resolution but is a bad estimator of PSD. Which among the above mentioned methods provides a balance between frequency resolution and estimate of PSD?

Dave Tweed
  • 172,781
  • 17
  • 234
  • 402
Jyothi Jain
  • 31
  • 1
  • 4
  • Also, it's not clear what you want. If you want to know the period of the signal like it says in the title, why do you also want a good estimate of PSD? – The Photon Jul 12 '15 at 15:41
  • Note that there can be a peak in the spectrum even if the signal is in fact not periodic. – avl_sweden Jul 13 '15 at 07:35
  • @avl_sweden can u further elaborate your answer.. – Jyothi Jain Jul 14 '15 at 05:29
  • Well, for instance a white noise signal fed through a bandpass filter will have a well defined peak in the frequency spectrum. But it isn't a periodic signal. This may not matter in your application, I just thought I'd point it out. – avl_sweden Jul 14 '15 at 05:34
  • In your question you say you want to find the period of your signal. But for non-periodic signals there is no period . So there's an implicit assumption here. Is it a true assumption for your application? Is your signal periodic? – avl_sweden Jul 14 '15 at 05:37
  • Let me give a detailed explanation. I wanted to find the cut off frequency of noise.I managed to separate the noise from the signal.But for further analysis, I wanted to find the cut off frequency.Therefore,I took the psd and found the peak (I hope this method of finding cut off frequency is right).But there are many methods for estimating psd.Therefore,I wanted to know which estimate to use. – Jyothi Jain Jul 14 '15 at 05:43

1 Answers1

1

If you want to find the periodicity of a signal, just calculate the autocorrelation of it and look for the highest peak.

Since that topic is almost entirely DSP and math related and has little to do with electronic engineering I suggest that you visit the sister-site https://dsp.stackexchange.com/ for more information.

Nils Pipenbrinck
  • 5,137
  • 1
  • 21
  • 30