From: Jeff Goode <jeffg7_at_gmail.com>
Date: Fri, 10 Jul 2009 20:04:38 -0400

Hi,

I finally have a functional limiter DSP feature. As noted in FlySpray,
it does work, but it needs tweaking. It's currently set up to reduce
gain in .6 dB steps, which in my opinion is too coarse. Finer steps are
certainly possible, at the expense of compute time for each sample. This
could be mitigated by replacing the current linear peak finding
algorithm with some kind of sort-type algorithm, halving and halving
again until arriving at the correct figure. The worst case for arriving
at the correct figure with maximum dB steps (0.0375 dB) is around 9
iterations, vs the current 20. This precision would come at the expense
of RAM because it involves much larger tables. I'm sure there's a happy
medium in there somewhere.

Please check it out and provide feedback at your convenience. It's
definitely not ready for committing, but there's a solid framework there
to work with.

Thanks,
Jeff
Received on 2009-07-11