Re: normalization on dst/idst

[Paul Kienzle]

On Dec 16, 2006, at 6:12 AM, Joan Picanyol i Puig wrote:

> * Paul Kienzle <address@hidden> [20061206 03:09]:
> > Anybody have an idea what the normalization I should use for discrete
> > sine transform (dst/idst) functions?
>
> I believe it is a matter of convention, but sqrt(2/N) sounds 
> reasonable.
>
> > I added them to octave-forge, and would like it to be compatible
>
> to?

I thought there might be a version somewhere in Matlab.
Indeed I found them in the pde toolbox, which is documented
as putting the normalization in the idst function.  I'll
go with that convention, and ignore the fact that the
dct/idct functions spread the scaling between them.

- Paul