[music-dsp] Adaptive (auto) release

Citizen Chunk citizenchunk at nyc.rr.com
Sun Jan 18 14:39:00 EST 2004


Hi James! thank you for the in depth response. there are a lot of great 
ideas there, a lot to think about. that's what i love about this stuff: 
there are so many possibilities to be explored, if you have the time, 
patience and drive.

anyway, thank you. i'll let you know what i come up with.

- Citizen Chunk

On Jan 18, 2004, at 12:06 PM, James Chandler Jr wrote:

> Hi Chunk
>
> Thats an interesting idea, controlling release by the ratio between 
> Peak and
> RMS. In case you didn't know, the ratio between Peak and RMS is 
> sometimes called
> Crest Factor. A web search on Crest Factor might turn up random 
> nuggets of
> useful info.
>
> I've never used RMS detection in a compressor, can't offer practical 
> advice.
>
> Some analog compressors of the past, implemented program-adaptive 
> release by
> running several attack-release envelopes in series (parallel might 
> work too,
> depending on the design). Run two or more envelope detectors, with 
> progressively
> longer attack-release times.
>
> The compressor gain stage is modulated by whichever detector has the 
> highest
> instantaneous level.
>
> A short transient would only affect the first detector, because the 
> slower
> detector(s) will not charge up fast enough to exceed threshold.
>
> Longer transients charge detectors "down the line" to above-threshold 
> levels,
> causing longer release.
>
> Sustained loud passages charge the long-release detector at the end of 
> the line,
> giving long-term smoothing on full mixes. One typically does not want 
> transients
> like kick and hat, to pump the level of the entire mix (jerking around 
> the level
> of sustained parts like organ, synth pad, or lead vocal).
>
> I've used multiple cascaded detectors on analog compressors and 
> digital plugins,
> seems to work pretty well. A crude fine-grained example--
>
> Detector    Attack    Release
> 1           0 mS      10 mS
> 2           10 mS     20 mS
> 3           20 mS     40 mS
> 4           40 mS     80 mS
> 5           80 mS     160 mS
> ..... etc .....
>
> Whatever works well in practice for what you want to do.
>
> The above configuration is probably best suited for a peak limiter, or 
> an
> "automatic" compressor which attempts to always do the right thing 
> with only
> user threshold and ratio controls, not offering user-adjustable 
> Attack-Release
> controls.
>
> You could try lots of fine-grained detectors, or just two or three 
> wider spaced
> detectors.
>
> A compressor might work better with fewer, wider spaced detectors. The 
> first
> detector might contain the attack control, and then the release 
> control might
> affect a second or third detector.
>
> Multiple first-order IIR detectors are pretty simple, so this can be 
> coded to
> run pretty efficiently.
>
> =============
>
> Another strategy I've used (in Peak Limiter rather than Compressor)-- 
> just one
> or two simple first-order IIR detectors (multiple series IIR detectors 
> can
> better smooth the envelope, minimizing intermodulation distortion).
>
> But... If the detector is above threshold, the program gradually 
> increases the
> release value (up to some maximum, perhaps 500 mS). For instance, you 
> might code
> it so that release would arrive at maximum if the signal stayed above 
> threshold
> for 100 mS, 500 mS, or whatever experimentally works for you on 
> real-world
> signals.
>
> If the detector is below threshold, the release value is gradually 
> decreased
> (down to some minimum value, maybe 10 mS).
>
> This scheme tends to "seek" whatever release is appropriate for the 
> music. If
> there are very few over-theshold transients, the release stays short. 
> As the
> number of over-threshold transients increase, the release gets 
> progressively
> longer.
>
> ==============
>
> However, your Peak vs RMS idea is very good. It might be potentially 
> better than
> the aforementioned methods.
>
> James Chandler Jr
>
>
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book 
> reviews, dsp links
> http://shoko.calarts.edu/musicdsp
> http://ceait.calarts.edu/mailman/listinfo/music-dsp
>




More information about the music-dsp mailing list