From my understanding PSD tries to give a good estimation on the power each frequency attributes to the overall signal power.

If I am only interested in a frequency range $(\omega_0 \leq \omega \leq \omega_1)$, could I put the signal through a band pass filter for that same frequency range then calculate the signal power using this formula:

$$P_x = \frac{1}{N_1 - N_0 + 1}\sum_{n = N_0}^{n=N_1} \left|x(n)\right|^2 $$

Would this technique give me what I am looking for, (the power of the signal from $N_0$ to $N_1$ for the specific frequency range)?

How does this compare to PSD estimation using any of the popular techniques?

Pros, Cons?

Thank you!

  • $\begingroup$ i would say the answer is "yes, it should give you what you are looking for". Pros: simple and consistent. Cons: might be more expensive if $N_0 \ll N_1$ than an FFT. your BPF will likely need to be sharp on both left and right. $\endgroup$ Sep 7, 2016 at 2:54

1 Answer 1


You can switch to the frequency domain with regards to Parseval's theorem. I guess you can transform your problem via discrete-time fourier transformation.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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