[music-dsp] Anti-Aliasing Filter With FFT/iFFT

[Daniel Werner]

vesa norilo wrote:

> Hi,
>
> You could do it like this but I guess you'd rather not...
>
> - Every time your sample playback frequency changes, you'll have to 
> reconstruct your anti-aliasing filter. What about when a note starts 
> mid-block? Part of the block requires one antialiasing filter, part of 
> it another. What about when there's a vibrato? This could mean you 
> have to reconstruct the filter and FFT it at all control rate frames. 
> Plus the problem of filtering different parts of your block with 
> different filters.

The plugin I'm writing is for DarkWave Studio which never changes the 
samplerate or block size mid-block, so this isn't a problem. Do you mean 
that I have to change the filter when the note changes? I thought you 
would only have to change the filter when the oversampling rate, sample 
rate or block size changes.

> - The filter orders where FFT starts to have a significant advantage 
> over FIR are probably very much overkill for sampler interpolation. 
> Commercial samplers have for some time used 4th to 8th order 
> resampling and for filter orders so small, FFT is going to be slower 
> than straightforward FIR.

I assume by order you mean the amount of oversampling, what about the 
number of zero crossings in the filter?

> For this application, you might want to stick with Lagrange 
> interpolation (which is equivalent to sinc with binomial window)

I'll consider Lagrange, I haven't read about it yet, but If you think 
it's good I'm sure it's worth a look, thanks.

> Vesa

- Daniel Werner
http://experimentalscene.com