0

Consider a discrete signal S sampled at point x={2,5,9,11}, whose respective values are S(x)={300,312,340,400}.

How would you calculate its power? If it was an equi-sampled signal, with sampling period T=1, I would just calculate sum(S^2(i))/N with N=length(S(i)).

Intuitively, I would calculate the power over a non equi-sampled sequence with something like:

sum(S^2(x(i)))*(N-1)/(N*(x(N)-x(1)))

In the example: (300^2+312^2+340^2+400^2)*3/(4*(11-2))

But I am not sure this is a general result. Is there a well-known formula?

vitil
  • 256
  • 1
  • 3
  • 13
  • 1
    the first problem is you need to know how much time each measurement is representative of.once you can do that you just need to make a weighted average of the squared amplitude. – Jasen Dec 05 '14 at 10:38
  • Thanks for the reply. I guess each measurement with index i, excluding the ones on the edges is representative for `(x(i+1)-x(i-1))/2`. So you're saying that I should calculate: `sum((x(i+1)-x(i-1))/2 * S^2(x(i)) / (x(N)-x(1)))` ? – vitil Dec 05 '14 at 10:45
  • 1
    yes, you may want th re-write you data duplicating the first and last results: `x={2,2,5,9,11,11} , S(x)={300,300,312,340,400,400}` that way you have a x[n-1] and x[n+1] for the forst and last measurements – Jasen Dec 05 '14 at 10:50

0 Answers0