Say, buddy, where'd you get your sidechain?

[Mark Hammer]

I was going to whip up some sort of frankenstein Orange Squeezer for a fund raiser at work, and it occurred to me that many of the designs I'm most familiar with tap the sidechain in compressors after the stage/point where the gain/signal reduction occurs.

For instance, the signal envelope in a Dynacomp circuit (and its many, many variations) is drawn from the 3080 OTA.  That signal envelope, in turn, controls the OTA, such that the envelope follower is not responding to the "true" input signal, but to the modified signal instead, including whatever artifacts have been imposed on it by imperfect envelope followers.  The same is true for the optoisolator-based setup in the Anderton EPFM compressor and its soulmate, the Whisper.  The Orange Squeezer does not reduce gain during compression but attenuates the signal passively, and applies a fixed gain to whatever the signal level happens to be at that moment.  The same downstream sidechain applies, though, because the rectifier uses the signal available after that gain stage.

So what happens if the sidechain (that portion of the circuit where the envelope-follower/rectifier is located) detects the input signal before it hits the level-adjusting element, be it OTA, LDR, or FET?  Is there a logical and/or aesthetic reason why the journey through the sidechain AFTER the modified signal, rather than before?  Or is this simply a case of reducing the size of the overall circuit by doing so?  Can anybody point to some circuits that start the sidechain BEFORE the input signal gets altered, or could they describe what might be different about the two approaches in terms of feel?

[Phorhas]

I think the LA2 compressor's side chain is taken right from the input

[Mark Hammer]

Well, I looked through what I had, since posting, and the only things I was able to clearly identify as tapping the sidechain before the control element were the What compressor (and Joe Cheap), Craig Anderton's NE570-based limiter, Johan Blomdahl's LA-light, and a couple of unsourced optical compressors.  I also ran into another pair of schematics I had downloaded from somewhere that showed the sidechain tapped before the gain adjustment, but then recommended breaking that link and redirecting the sidechain input to the output of the gain-reduction stage.

So, it happens, and it seems to happen in optical units more than anywhere, but I'm not sure why.

[bioroids]

Hi Mark!

I think of it this way: you are comparing the output amplitude against some predefined value, and using this info to control the amount of gain applied. This works out to maintain the output level at the specified fixed value.

I think this is an aproach derived from "control systems" theory (i.e. a closed loop with negative feedback to maintain some variable between specified limits), but it makes sense: if you want the output to be at a fixed level, why measure the input instead of the output ?

Does this make sense?

Luck!

Miguel

[R.G.]

It depends on what you're doing with the sidechain.

Studio and recording compressors are trying to actually compress the signal, in that they want to reduce the incoming dynamic range from, say 80db to 40db. There is no real knee where the output does not grow larger with larger input signals. It just grows more slowly. What suits this best is a sidechain that senses the input signal.

Musical instrument "compressors" are actually much more like hard limiters. The signal hits a threshold and never gets any larger for increasing input signals. For that kind of operation, it makes sense to detect the output level and make it be constant by using it in a feedback loop to the gain changing element.

I think that's really the biggest difference. We use "compressor" in a different sense than they do in  pro audio. What we call a compressor is more like what they'd call a hard limiter.

[StephenGiles]

I've never quite thought of it that way RG, but you're right.
Stephen

[lovekraft0]

I think most stompbox compressors are designed more as "sustainers" (if you will), and as such, the feedback sidechain makes more sense, sinxce it keeps a fairly constant output level above the threshold. A true compressor should simply make a large change in the input level smaller at the output, like a log amp in analog computing (but without mangling the waveform), so a feed-forward sidechain without any influence from the output level is ideal.

Which, now that I look at it, is pretty much what RG already said. 

[Elektrojänis]

Some material also says that opto compressors usually use feedback configuration because the opto elements (LDR's etc.) are not very linear and the feedback topology linearize the compression (to make the comression curve closer to ideal).

Hmmm... At least thats what this page says: http://www.geocities.com/m_natsume/what_compressor.html (check out the text on the green sidebar)

I think there was some more info about feedforward vs. feedback on some What Compressor / Joe Cheep / Weak Joe -page but I seem to be unable to find it now.

[Mark Hammer]

Studio and recording compressors are trying to actually compress the signal, in that they want to reduce the incoming dynamic range from, say 80db to 40db. There is no real knee where the output does not grow larger with larger input signals. It just grows more slowly. What suits this best is a sidechain that senses the input signal.

Musical instrument "compressors" are actually much more like hard limiters. The signal hits a threshold and never gets any larger for increasing input signals. For that kind of operation, it makes sense to detect the output level and make it be constant by using it in a feedback loop to the gain changing element.

I think that's really the biggest difference. We use "compressor" in a different sense than they do in  pro audio. What we call a compressor is more like what they'd call a hard limiter.

You hit a line drive, brother.  Makes absolutely perfect sense.  In which case, if one was attempting to make a relatively transparent "dynamics reduction box" for guitar, and NOT a "sustainer", you'd try to tap the input signal.  Hmmm, I wonder if it would end up being less hissy than your average tap-the-output device?

[puretube]

in studio-equipment,
limiter and compressor functions
IMHO are being described by getting
the CV pre or post VCA

[davebungo]

My MXR Limiter (the 4 knob mains powered unit) exhibits typical feed-forward behaviour with a ratio of 5:1 above threshold - I've never thought of it as sounding particularly like a "squeezer".

I know from experience that most studio/broadcast products are feed-forward but some are feed-back and they are generally regarded as more "musical" sounding.  The feed-forward type has, I believe, become more accepted as standard simply because it has a more "technical" (for want of a better expression) transfer function.

[davebungo]

Studio and recording compressors are trying to actually compress the signal, in that they want to reduce the incoming dynamic range from, say 80db to 40db. There is no real knee where the output does not grow larger with larger input signals. It just grows more slowly. What suits this best is a sidechain that senses the input signal.

Musical instrument "compressors" are actually much more like hard limiters. The signal hits a threshold and never gets any larger for increasing input signals. For that kind of operation, it makes sense to detect the output level and make it be constant by using it in a feedback loop to the gain changing element.
hat they'd call a hard limiter.

Nearly all (feed-forward) compressor/limiters have a ratio control and a threshold control (as well as attack/release etc).  By setting the ratio to 10:1 or greater this does act as a limiter with a very distinct knee.  In fact some units (DBX 160A for example) offer a hyper compression mode where the output level actually falls as the input increases beyond the threshold.  The limter mode is probably the most common use of compression in broadcast applications where the concern is not to overload the transmitter and to keep control of the audio level in general.

[R.G.]

It's certainly possible to do the job with signal pickoff either place.

I speculate that it winds up being simpler to build more limiter-oriented (and hence guitar oriented) compressors within the constraints of guitar pedals ( that is, 9V unipolar power, few and cheap components, not many controls, small pcb space, etc) with feed back chains. That being the case, Occam is gonna slice you. There's more flexibility in studio equipment, in terms of space, components, price, power, all of that.

I don't have all the facts, of course, just a suspicion based on years of perusing schematics. There may be one of Mother's rules hiding in there somewhere. Or maybe I'm just perceiving a pattern in the static. 

[davebungo]

You're right that for more simple guitar oriented units, it is always going to be cheaper and less complex to have a fixed ratio (as well as maybe fixed attack and recovery times to minimise controls) and if it's going to be useful it may as well be on the limiter end otherwise users won't think it really does anything useful at all.

[Mark Hammer]

So I think this is what I'm going to do when I get a chance.

1) Stick a unity gain op-amp buffer on the input of an Orange Squeezer (currently, the input goes directly to the cap/resistor feeding the gain stage).

2) Stick a unity gain buffer after the resistor/FET attenuator in place of the gain stage that's there now.

3) Use the gain-of-23 rectifier/sidechain (well, the plan is to have some variable gain), but have the op-amp's input switchable to either the output of the input buffer OR the buffer after the FET (i.e., choice of feed-forward or feed-back).

4) Stick a gain stage at the output with either fixed gain and an attenuator or variable gain.

Stick a bypassable bandpass filter just ahead of the rectifier gain-stage, and I'll bet that'd make a dandy little de-esser.  When you consider having choice of feed-forward and feed-back, variable rectifier gain, some control over attack and recovery time, variable output level, and tunable de-essing, that'd make a nice little control strip, whose noise level might be pleasingly low, and bandwidth be pleasingly high.