Re: DynamicsCompressor and emphasizing high frequencies

Chris Wilson writes:

> Thanks, that's good background.  It seems from this, though, that
> either 1) users should do this themselves, or 2) we should provide a
> control parameter on the emphasis, possibly multiple controls (after
> all, the "default" on the LA2A is no emphasis).  My inclination is to
> remove the emphasis, but that will require changing all current
> implementations.  Alternately, we need to carefully specify what
> emphasis is being applied, so it can be undone (by
> pre-pre-de-emphasizing, and post-post-emphasizing, which seems a
> bit...odd.)
>
> This is causing problems in multiband compression (which we use in
> Chrome OS for speaker tuning), as well as other side effects.
>
> Any comments on removing this emphasis?

The zero/pole de-emphasis and attempted inverse that WebKit uses
and Gecko copied would have some interesting effects on phase, so
I can see the advantages in allowing for different filters.

I'd prefer that any added filter control was a separate node from
DynamicsCompressorNode because it would give more flexibility.
Pre-emphasis may be useful for purposes other than compression.

If the first order filter shapes are important for those that want
the pre-emphasis, then it may be necessary to add API (to
BiquadFilterNode perhaps) to emulate them.  I expect zero/pole
filters to be faster than convolution.

Received on Thursday, 9 January 2014 23:12:59 UTC