Trying to understand the envelope of the Ross/MXR compressors.

[dirk]

I've searched a long time, but can't find the answers. Or get opposing answers, witch is even more confusing.

As I understand it: The attack and release times are determined by the charging and discharging of the 10uF capacitor between the 2 transistors Q3, Q4 and ground. Is this correct?
The 150k resistor mod, is that the attack or is it release? Or maybe both?
And how can you make separate attack and release controls? (for use in a home studio environment)

[brett]

Hi

The attack and release times are determined by the charging and discharging of the 10uF capacitor. Is this correct?
The 150k resistor mod, is that the attack or is it release? Or maybe both?
And how can you make separate attack and release controls? (for use in a home studio environment)

Yes, the 10uF cap holds the "control" charge.
I think the 150k affects attack and release (because it affects how the charge on the 10uF cap drives Q5).
Simplest attack and release controls would be to vary the 10uF up and down.  e.g. 3.3, 10 or 33 uF. 
cheers

[dirk]

I tried different capacitors, but that only makes very small changes to the sound.

I hoped that, because the 150k resistor is connected to the 9V, it only affected the charging of the capacitor. So influencing only the attack (or decay, I don't know).

Going further onto this road of thought, I hoped to control the discharge of the capacitor by putting a pot between ground and the positive end of the 10uF capacitor. I have tried a 1M pot, but that gave only very small changes to the sound.

What I want is an attack time of about +- 1-40 msec and a decay of +- 30 msec - 1 sec.

[freddd]

Not too hi-jack but; Another question - how is the 3080 used in this circuit? What sets the gain? I've never understood the CA3080!

[brett]

Hi
The 3080 is a transconductance op-amp.  The voltage on its control pin sets the gain (which is very handy for a compressor).  You can see in the schematic that everything "flows" to Q5, which, when turned "on" connects the control pin to the power supply, via a 27k resistor and the "sensitivity" control.

I tried different capacitors, but that only makes very small changes to the sound.

If I'm reading the schematic correctly, for a rising envelope amplitude, the cap discharging, and feedback via Q3 and Q4, control the rate at which Q5 turns off and the gain goes down.  Normally, a cap might give rise to a decay roughly equal to its time constant, but maybe the feedback is reducing the effect?  Try 0.1 and 220uF caps and see if you get fast and slow responses ?  

There's also interaction with that 150k resistor.  If Q3 and Q4 are off (no signal), the cap can't fill any faster than with a half life of 150k x 10uF = 1.5 seconds. (Jack's schematic says 1000mS, which seems consistent).  So if you want to shorten up the decay time, that resistor will need to be reduced.  For 30 mS you might need it to be about 3.3k.  But a better solution would probably be a 0.22 uF cap and keep the 150k resistor.  Or am I missing something?

cheers

[Mark Hammer]

I suppose if *I* was trying to make money from 19 year-old boys, I'd probably prefer to call a control "attack", rather than "recovery", because it sounds sexier, but that doesn't make the ensuing miscommunication any less frustrating.

What happens when you play a note into a 3080-based compressor like the Dynacomp?  Normally, the 3080 is set to some maximum gain (this is responsible for the tendency towards noisiness during the silent passages).  When a sudden transient occurs, and is detected by the envelope follower, gain is essentially subtracted.  How much gain is subtracted depends partly on the total series resistance fed to pin 5 of the 3080.  When that resistance is high (Compression control set to 7:00 position) there is little influence over gain from the current coming in on pin 5.  Of course what comes in via the pot and 27k resistor is essentially the result of two sources, all summed at the node where the 150k and 10uf components and Q3/Q4 collectors converge.  One source is what is coming in from the two complementary halves of the signal envelope, via Q3 and Q4, and the other is the supply voltage (V+) via the 150k resistor.  The 10uf cap provides some smoothing or lag.  Think of it like smudging charcoal or oil pastels to provide what seems like a smoother transition between changes.

During silent passages, the 10uf cap is recharged via the 150k resistor to its capacity.  When it is "at capacity", the  3080 gain is reset to maximum.  How long it takes to reset the gain to maximum in the absence of any input signal depends on the value of that 150k resistance.  If you make it smaller, the recharging takes less time.  Stock, the recharging takes about a second.  Jack has termed this "decay" in his diagram, but it is more properly described as "gain recovery".  Yes, the recovery shows up in how long the simulated sustain appears to be in effect (which is why people like to think of it as "decay", but the simulated sustain is a product of a recovery process like the way your eyes "recover" from exposure to bright light in the dark, and much like that recovery it will depend on what happened before and how long ago it happened.  "Decay" suggests a property that is independent of what is being played.  "Recovery" implies a resumption depending on what is being recovered from.

Okay.  Enough fancy footwork.  What's going on here?  The question you have to ask yourself is what the audio implications are for recovering from profound events.  In our case a profound event is a strongly picked note with a huge peak.  If that 10uf cap is discharged by a sudden large transient and it takes a full second for the cap to be completely recharged again via the 150k resistor, then the gain applied to any notes immediately following will be drastically reduced.  So much so that note N+1 will not have the same perceptible onset characteristics as note N that caused the gain to be suddenly reduced.  Of course, if you waited a reasonable portion of that 1-second period and then played a note, the gain would have returned enough that you would hear that second note at a reasonable volume and would also hear its gain being reduced in response to that particular transient.  But if the notes come too close together, notes N+1, N+2, etc will seem to be suppressed by the discharging of the cap caused by the first note.  It is the suppression of subsequent notes by the first one that resulted in the control being labelled "attack".  The fact that it only impacts on subsequent notes and note every note (including the first one) is why it is NOT an "attack" control.

But this is largely semantics for some folks.  What are the implications of altering the 150k resistance?  As the 150k resistor is reduced in value, this permits the 10uf cap to recharge or "recover" faster, because there is a more efficient path between V+ and the + side of the cap.  When the cap charges up faster, it returns to maximum gain faster.  You still get the same dynamic reduction effect since gain IS reduced in response to transients, but the slow and gradual resumption of gain, that creates the illusion of sustain by slowly increasing gain as the signal itself dies out, is replaced with a faster resumption of gain.  Faster recovery means that notes N, N+1, N+2, etc don't have to be so far apart in time for you to hear the "attack" portion of each of those notes when picked.

Because the change in recovery is something that only affects time, then the timing of your playing will determine whether that change is heard or not.  And this has been a continuing source of confusion amongst modders/builders.  They expect the effect to be instantly audible, and it isn't, because it will always depends on the speed and strength of your picking.  Overall, though, faster recovery time will mean that the dynamic reduction will seem to be imposed on notes individually, rather than as a cluster.  This will give a different feel to passages.

[dirk]

Thanks Mark for the clarification. I wish the world had more people like you.

Could the real attack be adjusted by replacing the 2 1M resistors with a dual 1M (or more) pot?

[Mark Hammer]

Thanks for the vote of confidence.

My gut tells me that the attack time will be altered by the value of the (currently) 10uf cap, since it is the time for that to be "counter-charged" by the envelope signal that provides the short delay for the gain reduction to take place.

Should you tamper with it?  That wouldn't be my first inclination.  Largely because in this envelope follower design, the same cap accounts for BOTH the attack, decay, and recovery-time properties (this would be collectively refered to as the "averaging" or "time constant" cap).  Monkey around with one aspect and you affect the others, possibly in ways that are difficult to compensate for.

While I have the attention of sidechain lovers, here is a hypothetical question.  When one has a simple half-wave rectifier like that on the Dr. Q and so many other devices, the small resistance (generally <1k) in series with the diode, and just before a cap to ground, determines the charge-up time of the cap, and consequently the "attack" time.  One can reduce that resistance to quicken attack, or increase it to slow attack down.  Unfortunately, because it is a current-limiting resistor, adding resistance to slow attack time also reduces apparent sensitivity, such that one generally needs to increase sensitivity of the envelope follower if you want to have the same response but slower.

The fact that 1-1.5k will all but obliterate the envelope-sensitivity, that doesn't leave much opportunity for audibly slow envelope sweeps; one generally has a choice of fast and not quite so fast.  So, I'm wondering, is it possible to use some sort of high current buffer to make use of much larger attack resistors (e.g., 5-10k) possible?  In other words, in between the gain stage-plus-rectifying diode, and the RC combination that averages out the envelope and provides the time characteristics, could one stick, say, a paralleled pair of NE5532 sections so as to make the circuit more immune to the current-limiting properties of the series resistor?

[Freddy205]

Sorry for being a goof but can anyone explain just how Q5 is reacting here? It's got me confused!

[Mark Hammer]

There is certainly more to it electronically than I understand or can explain, but in a functional sense at least, Q5 sums the two parallel rectifier paths and isolates/bufferes the Iabc pin in the 3080 from the rest of the circuit.

[R.G.]

The two parallel rectifier paths are summed by the 10uF cap.

As you noted Mark, the "attack" time is how fast an incoming transient can reduce gain. In this instance, that is how fast can Q3 and Q4 pull charge out of the 10uF cap. The recovery/delay/decay time is governed by the 150K/10uF. Q5 is only there to buffer the 10uF cap from the current needs of the Iabc terminal. Q5's base current reflects an error term, making the recovery time longer than it would otherwise be, and the ultimate gain less than it would otherwise be.

Slowing down the speed with which Q3/Q4 can discharge the 10uF is a real attack time control. Right now, it's set up to be as fast as possible. Joining the Q3/Q4 collectors and feeding that through a pot to the junction of the 150K/10uF would add the ability to slow down how fast the gain can reduce, although that's not all that useful a thing to do, as not reducing gain as fast as possible would either produce a starting "peep" of uncompressed sound or a squawk of distortion before the gain came down.

The sustain control, as you note only reduces the biggest possible gain.

[dirk]

The two parallel rectifier paths are summed by the 10uF cap.

As you noted Mark, the "attack" time is how fast an incoming transient can reduce gain. In this instance, that is how fast can Q3 and Q4 pull charge out of the 10uF cap. The recovery/delay/decay time is governed by the 150K/10uF. Q5 is only there to buffer the 10uF cap from the current needs of the Iabc terminal. Q5's base current reflects an error term, making the recovery time longer than it would otherwise be, and the ultimate gain less than it would otherwise be.

Slowing down the speed with which Q3/Q4 can discharge the 10uF is a real attack time control. Right now, it's set up to be as fast as possible. Joining the Q3/Q4 collectors and feeding that through a pot to the junction of the 150K/10uF would add the ability to slow down how fast the gain can reduce, although that's not all that useful a thing to do, as not reducing gain as fast as possible would either produce a starting "peep" of uncompressed sound or a squawk of distortion before the gain came down.

The sustain control, as you note only reduces the biggest possible gain.

Thanks!

For drumsounds that lack punch, it's very usefull to clip the attack part. This gives bite to the sound, a fast release then gives more body to the sound (breathing). It's a very common technique for electronic music.
But I agree its not very usefull for guitar/bass.

[ayayay!]

WOW that was really good info.  I just learned more from Mark & R.G. in the past 5 minutes on compressors than all the previous articles I've read combined.  I've suddenly got very into compressors, and especially the Dyna Comp, which was somewhat mysterious to me.  The Boss CS-3 seems simpler to work on (read:hack) but the DC, especially the use of the 3080, always had me a bit confused.

Thanks guys!  That helped me so much...

[Mark Hammer]

Electronic Musician had a decent article on compressors and compression in the Feb 2001 issue ( http://emusician.com/mag/emusic_big_squeeze/wall.html?return=http://emusician.com/mag/emusic_big_squeeze/index.html ).

[Le québécois]

The sustain control, as you note only reduces the biggest possible gain.

I know, It's a very old post but I think it's a good place for that question since it is directly related. What set the biggest possible gain of the CA3080? Only the 150K resistor that connect Vb to pin 6 of the IC?
From what I have scavenged here and there, this 150K will be see as a load for pin6 and will determine the voltage associated to the output current of the OTA.
Meanwhile, is it only a highpass filter (15k//0.1uF to ground) that is connected to pin 3 of the CA3080 or it also play a role in determining the maximal gain of the OTA?

My goal is to understand how to increase that maximum gain in an attempt to achieve higher loudness out of this fx.

many thanks 

[matt239]

The two parallel rectifier paths are summed by the 10uF cap.

As you noted Mark, the "attack" time is how fast an incoming transient can reduce gain. In this instance, that is how fast can Q3 and Q4 pull charge out of the 10uF cap. The recovery/delay/decay time is governed by the 150K/10uF. Q5 is only there to buffer the 10uF cap from the current needs of the Iabc terminal. Q5's base current reflects an error term, making the recovery time longer than it would otherwise be, and the ultimate gain less than it would otherwise be.

Slowing down the speed with which Q3/Q4 can discharge the 10uF is a real attack time control. Right now, it's set up to be as fast as possible. Joining the Q3/Q4 collectors and feeding that through a pot to the junction of the 150K/10uF would add the ability to slow down how fast the gain can reduce, although that's not all that useful a thing to do, as not reducing gain as fast as possible would either produce a starting "peep" of uncompressed sound or a squawk of distortion before the gain came down.

The sustain control, as you note only reduces the biggest possible gain.

Doesn't the sensitivity control adjust the MINIMUM gain; or "the maximum amount of gain REDUCTION?" Or am I misunderstanding?
If this is so, what DOES set the maximum gain of the 3080?
I'd actually like to reduce the gain a little, since the output control is just an output level trim, and not a gain control.
It seems to me there's more gain than we need, and with it noise..

For a "real" Attack control we could just lift the collector leg of Q3 & Q4 and re-connect them through a small resistor? (A bit awkward physically no doubt..)
1k perhaps?
Probably would want to still modify the release or "recovery" time also.

[PRR]

A sweet mod for the forum would be "dust". The older a post is, make it darker and low-contrast. There are reasons to dredge-up old threads; but also useful to know when a question is stale.

> Doesn't the sensitivity control adjust the MINIMUM gain

No.

Gain of '3080 is control current times load resistor times a constant.

Load resistor is the 150K+0.001uFd at '3080 output.

In this plan, control current comes from Q5 which simply buffers the voltage at 10uFd/150K node. At idle, this node is way high. '3080 control port is low (about 0.6V). So we have about 8 volts, through Sens. pot, which sets '3080 current, and thus '3080 gain.

This is low-price 1980s technology. It lingers because its "flaws" suit some guitar sounds. If you are mashing drums, there are MUCH better plans, and now not very expensive.

Merlin's "Engineer's Thumb" is DIY, the same complexity, and a much better approach to the Dyna's limitations.

[matt239]

A sweet mod for the forum would be "dust". The older a post is, make it darker and low-contrast. There are reasons to dredge-up old threads; but also useful to know when a question is stale.

> Doesn't the sensitivity control adjust the MINIMUM gain

No.

Gain of '3080 is control current times load resistor times a constant.

Load resistor is the 150K+0.001uFd at '3080 output.

In this plan, control current comes from Q5 which simply buffers the voltage at 10uFd/150K node. At idle, this node is way high. '3080 control port is low (about 0.6V). So we have about 8 volts, through Sens. pot, which sets '3080 current, and thus '3080 gain.

This is low-price 1980s technology. It lingers because its "flaws" suit some guitar sounds. If you are mashing drums, there are MUCH better plans, and now not very expensive.

Merlin's "Engineer's Thumb" is DIY, the same complexity, and a much better approach to the Dyna's limitations.

Thanks. I always appreciate your help.
I agree 100% there are MUCH better designs.
I'm just trying to understand this out of; raw curiosity, and I happen to have one laying around that I'd like to make work better..
As I was saying in another thread, I may just build something else in the box!