[vorbis-dev] More mdct questions

[Segher Boessenkool]

Marshall Eubanks wrote:

> There is no standard for normalization of fourier transforms.
> I myself prefer 2/sqrt(n) on both the transform and its inverse, so that these
> are the same (except for signs). An FFT is 1/n, spread between the
> transform and its inverse. Confusingly, some also throw in 2 PI's

You get the PI's if you have a continous transform (non-sampled, by
integration; this one is called Fourier transform). You get no PI's if
you have a discrete tranform (sampled in both time and frequency domain,
by summation; this one is called _discrete_ Fourier transform (DFT), for which
the _fast_ Fourier transform (FFT) is an efficient means of computation).

> in the normalization. If you actually care about units in the transform
> domain, this is a real pain. For these purposes, it doesn't matter.

Yep. Only the combined forward/backward transform need to get the normalization
correct.

Dagdag,

Segher

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.