Side Chain Response

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

hagtech

Well-known member
Joined
Oct 21, 2005
Messages
49
Location
Hawaii
I've been looking at various circuits for doing the attack and release envelope follower for a side chain. My first stab was sort of a current pump architecture, in that it would pump up or down using constant current sources. Saw something like that in a THAT app note. I was using an LM13700, which makes the circuit much simpler. Anyway, the operation didn't seem right to me. The pump action was linear. That is, the rising or falling slopes were constant and independent of signal amplitude.

So I thought maybe an RC type circuit would be more musical. The slopes would be exponential and independent of amplitude. A 5ms attack would be just that. I made a quick circuit to test an odd approach to this, came up with this response:

sidechain.jpg


Not tweaked, but you get the idea. Set for pretty fast release. Playing music through this, seems to me to respond awfully fast, a lot of "pumping" action. Not sure if this is good or not.

Has anyone tried both approaches? Wondering which is more musical and transparent.

jh
 
Hi Jim,

The linear mirror makes sense for the That parts because of their exponential scaling.

What gain control element are you using? The control law would likely define what sort of A/R response curve would be appropriate.

There are a couple of compressors that have switchable release types. On is the Groove Tube Glory Comp. Another interesting one is made by Crane Song.
 
> Wondering which is more musical and transparent.

Yeah, that's been the #3 question for 80 years. (#1 and #2 are "does it stop overload?" and "is it LOUD?", not always in that order...)

The original approach is dual time-constant. R-C based, but looking at the control law of tube grids, it is closer to dB/Second linear than exponential. A 2R2C scheme has advantages in musicality, and also reduces driver current demands for similar peak-catch speed at the same maximum DC resistance.

There are also true linear gain-cells like your 13700, and exponential cells like DBX-heritage VCAs. These sound different when used with R-C filtering. DBX also favored their RMS detector and an "exponential capacitor" to handle the exponential control law.

In the 1970s, chip-amps allowed very complex sidechain processing. Program dependent time-constant, gain hold, some really wild things were tried. 1980s radio-wars led to digitally controlled processing and now we just drop audio in a DSP and whack away at it with algorithms to make the loudest possible on-air signal.

Despite all that, plain old dual-constant RC is still popular in tracking and mastering.

Another beloved scheme is a naked photo-resistor, which acts as many-many time constants due to the photo-chemical actions of a complex brew. My thought is that your ear learns the time-constant of a simple R-C limiter, the 2R2C "auto" setting on some limiters is harder to learn, and photoresistors never get boring.

> musical and transparent

Time-constant may have more to do with it than implementation. For mild control of small overload on good music, use 1mS-10mS attack and 500mS-2S release. (No, there isn't a real standard for measuring this stuff, use your ears.) Speech can stand a longer attack (let short loud sounds clip) and a shorter release (bring up the soft quick consonants). Percussive tracks will just "duck" with long release times.

> using an LM13700

Handy for experimenting with linear control. But the maximum dynamic range is 70dB-80dB. If used in the in-out signal path of a limiter, your output dynamic range is reduced by the maximum gain-reduction you will allow, which leads to <60dB dynamic range. You can also put a 13700 in the feedback loop of an opamp, keeping it Off when not reducing gain, and get plenty of S/N for loudspeaker amp protection (probably over 100dB, but I have not really pondered it.) This was Crown's usual scheme in the late 1980s power amps.
 
What gain control element are you using?

Good point. Was thinking of Vactrol. I can see how this is part of the equation. So maybe linear pump up/down not so bad.

original approach is dual time-constant. R-C based

You mean separate attack and releas times? Or two attack and two release times, mixed somehow? Isn't there also some sidechains split into two frequency bands?

true linear gain-cells like your 13700

Oh, sorry. I meant this chip was in the side chain. I would never let one of these touch an audio signal!

jh
 
[quote author="PRR"]...My thought is that your ear learns the time-constant of a simple R-C limiter, the 2R2C "auto" setting on some limiters is harder to learn, and photoresistors never get boring.[/quote]
Mark well this wisdom.

Keith
 

Latest posts

Back
Top