In papers or textbooks, I have found several different definitions of PSD. I think I understand the meaning of PSD and all these papers or textbooks agrees that PSD and Fourier Series share a proportional relationship. But still there are different formulas from P(f)=1/T(〖F(f)〗^2), P(f)=〖∆t〗^2/T(〖F(f)〗^2), P(f)=1/4∆f(〖F(f)〗^2), P(f)=(〖F(f)〗^2) or others. From what I learn, I think on the point of conception of power(or energy) and Parseval's theorem, discrete PSD should be described as simply P(f)=1/T(〖F(f)〗^2). So why is there so many different formulas? Are they wrong or do I misunderstand something?


closed as unclear what you're asking by Peter K. Dec 19 '18 at 15:42

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ Related: this question $\endgroup$ – Matt L. Dec 19 '18 at 8:17
  • $\begingroup$ Your post isn't clear. Refer to formatting help. $\endgroup$ – BlackMath Dec 19 '18 at 8:20
  • $\begingroup$ They differ by a constant so yes there are many formulas but the pad is applied to many physical quantities. There is a tendency to treat power in an abstract way in many texts but would you expect hires power and watts having the same constant used in a formula? $\endgroup$ – Stanley Pawlukiewicz Dec 19 '18 at 13:29

Power Spectrum Density (PSD) represents the spectral distribution of total power of a given WSS random signal. Conventionally, it's defined as the Fourier transform of the Auto-Correlation function (sequence) of the WSS random process as:

$$ P_x(\omega) = \mathcal{ F } \{ r_{xx}[m] \} $$

where the auto-correlation sequence (ACS) $r_{xx}[m]$ of the WSS $x[n]$ is defined by:

$$ r_{xx}[m] = E\{ x[n]x^*[n-m] \} $$

These are theoretical entities, you may not compute them in practice where only a single or a few realizations of the random process are available. In such a case, the PSD is estimated and there are different estimation techniques, one of the most common and simplest is known as the periodogram estimate of PSD and given by

$$ \hat{P}_{x}(\omega) = \frac{1}{N} |X(\omega)|^2 $$

where $X(\omega)$ is the discrete-time Fourier transform of the observed signal $x[n]$ of length $N$.


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