Talk:Sigma approximation

Hi.

I've just edited the mathematical expression to explicitly show that the sigma approximation can also be used in the form shown when one has a series of any arbitrary period T. I didn't think this was obviously the case from the article as it stood (where the series implicitly has a period of 2 pi). [The proof that this is the case is a straightforward extension of the proof for a period of 2 pi.]

However I appreciate this reduces the compactness of the expression somewhat, so please discuss if you think this modification detracts from the article...

I've also added explicit explanation of the meaning of normalized sinc function, since many mathematically literate people without experience in communications/signal processing theory (like myself) are unlikely to have heard of it and may just assume there is little difference with the standard (unnormalized) sinc function --- and yes I know this is also covered in the wikipedia article on sinc functions.

___________________________

The value of m isn't defined. If it is the index of the last term included in the Fourier series, why not take the sum on n from 0 to m, and put m+1 in the Lanczos factor?

Hess88 (talk) 01:25, 27 December 2008 (UTC)[reply]

_________________________ The formula given here differs by a factor of two from that in MathWorld, which agrees with the first edition of FS Acton, "Numerical Methods That Work", p. 228. Also, I've always seen sinc(x) = sin(x)/x elsewhere. — Preceding unsigned comment added by 129.6.136.190 (talk) 21:04, 10 December 2012 (UTC)[reply]