--- Steve Harris <[EMAIL PROTECTED]> wrote: > On Fri, May 17, 2002 at 12:52:27 +0100, mike rawes > wrote: > > > How efficient are these techniques? > > > > At the moment, not very: > > That's pretty much what I do, except I dont use a > tree, I use a large > vector (so there are no duplicates) and a lookup > table.
I don't have any duplicates either: each successive table has a decreasing number of harmonics and less samples per period - although the smallest tables needed oversampling to remove noise at very high frequencies. I was thinking of using a lookup table too, using some multiple of (long)frequency as key. My understanding of cubic splines is fairly poor, so I don't know how to improve this (yet). I've used the code from the musicdsp archives. Applying algrebraic methods to reduce the number of multiplies is the best I can do right now. How do you do your speed tests? Mike > > With gcc, linear interp it takes about 20 PIII > cycles per sample, cubic > takes about 50. > > I should use a faster cubic algorithm though. > > - Steve __________________________________________________ Do You Yahoo!? Everything you'll ever need on one web page from News and Sport to Email and Music Charts http://uk.my.yahoo.com
