Re: multiband xover

On Fri, Feb 08, 2013 at 11:52:06PM +0000, Steve Harris wrote:
> On 8 Feb 2013, at 21:38, Patrick Shirkey wrote:
> > On Sat, February 9, 2013 8:00 am, Steve Harris wrote:
> >> The existing algorithm uses the FFT data, which makes a poor x-over - I
> >> can't remember exactly why, I think there might be ringing in the freq
> >> domain or something?
> >> 
> > 
> > Do you think that might have been fixed with the switch to the
> > Kaiser-Bessel window?
> 
> It could yes, but I'm too rusty on exactly what that would do.

I just tested 0.97.14 (Debian testing package) with a spectrum analyser.
With noise input, soloing the compressor bands shows fairly wide, periodic,
out-of band peaks about 90dB below the signal level. That's probably inaudible
in a compressor, but might be a concern in a PA crossover.

There's also some low level aliasing when compressing a constant-level
sine wave. This is worst with short attack and release times, and depends
on the amount of gain reduction. Raising the threshold well above the
signal made it disappear, as did bypassing the band under test. At the worst
settings I tried it was about 100dB below the signal level. This was with a
9kHz sine wave, 48kHz fs. By contrast, SC4 mono running as a plugin in
ardour2 has a totally clean output with the same test signal.

John

------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013 
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb