1
$\begingroup$

I have a real world system I am analyzing consisting of actual mechanical components that oscillate by rotating back and forth in a fixed axle (kind of like those finger fidget spinners but my system oscillates back and forth in its rotations within an angle bandwidth instead of endlessly rotating in the same direction).

I would like to obtain the frequencies at which these components are doing their back and forth oscillations.

Which kinds of frequency values is most reasonable to request from my DSP package when doing my FFT? Only positive absolute values? All absolute values? Only positive complex values? All complex values?

My gut is telling me that I might want to stay away from anything dealing with complex values, since my oscillations are from a real-world system and might therefore only be a real-valued function (?), but to be honest, I don't know if his is a fair assessment (I am dipping my feet into DSP for my present project so not a pro in the field).

If there's a justification for picking one over the others, I am interested in hearing it. Thanks for any recommendations anyone can give.

EDIT: What I am feeding to the DFT is my angle vs. time data and I'm interested in identifyig the frequencies which are found in the oscillation rotation behavior.

$\endgroup$
4
  • $\begingroup$ As a brief note, "imaginary" is an unfortunate misnomer and in no way implies "not real". Whether to take abs or not depends on whether phase information matters. $\endgroup$ Dec 6, 2022 at 19:41
  • $\begingroup$ What are you actually measuring, as feedstock for your DFT? Please edit your question with this information. It sounds like what you really want is a power spectral density (PSD) plot -- these use DFTs under the hood, but you would either use a packaged PSD routine, or build one from a recipe. $\endgroup$
    – TimWescott
    Dec 6, 2022 at 19:47
  • $\begingroup$ @TimWescott - I went ahead and edited my response. Here it is also: What I am feeding to the DFT is my angle values vs. time data and I'm interested in identifyig the frequencies which are found in the oscillation rotation behavior. I hope this clarifies. Thanks. $\endgroup$ Dec 6, 2022 at 23:49
  • $\begingroup$ @OverLordGoldDragon - Thanks for clarifying. I really don't know if phase information matters for my system but what I am trying to do is feed my DFT my angle vs. time data and I'm interested in identifying the frequencies which are found in the oscillation rotation behavior. I hope this clarifies on my end. $\endgroup$ Dec 7, 2022 at 0:19

1 Answer 1

0
$\begingroup$

What you're really looking for is a power spectral density plot, not a raw FFT result. You'll want to search on the term to find out how to make one. The difference is that an FFT is a true mathematical transform that preserves all sorts of information that is utterly irrelevant in your context, while a PSD plot gives you a best estimate of the power of various spectral components in your signal -- which is relevant in your context.

A power spectral density will give you all positive numbers, from 0Hz to whatever upper frequency you choose. For your application (real-valued signals with no specific time reference) this is the relevant information you're interested in.

$\endgroup$
1
  • $\begingroup$ Sounds good, I'll take a look at power spectral density then. Thanks. $\endgroup$ Dec 15, 2022 at 22:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.