MXR Sustain - NOT the nasty Dynacomp!

mictester · Post by **mictester** » 27 Mar 2012, 19:55

The MXR Dynacomp (and the almost legendary "Ross" variant of it and all the hundreds of boutique clones) is actually a pretty poor design. The fundamental problem is the OTA (Operational Transconductance Amplifier). In the Dynacomp, it's a CA3080. The "mojo" version of this IC (in its little round can) should show you that it's a very old-fashioned IC and has been around for almost as long as the 741 (and you wouldn't put quality audio through one of those, either!).

The rectifier circuit is probably the best part of the Dynacomp design, but doesn't have the facility for much adjustment (despite all the Boutique tweaks), and because the output audio is usually extracted from the emitter of the phase splitter transistor, you can get some amount of halfwave clipping (by the diode in the rectifier). A more sensible place to extract the audio is from the output of the OTA through a buffer transistor, but the damned OTA is still in the audio path, adding its own distortion and noise. Some manufacturers improved things by using better OTAs - the BA662, BA6110, LM13700 were all used - but the same mistakes were made: extracting the audio from the phase splitter transistor (trying to partially use it as a buffer as well), and using the same old rectifier circuit....

One team at MXR tried to do something different. They looked at the feedback around an op-amp. Negative feedback has a number of good side-effects (apart from just setting the gain) - as you increase negative feedback, amplifier bandwidth improves, distortion is cancelled, and noise diminishes.... So what happens if you vary the negative feedback around a reasonable quality op-amp? That's what I do with the Really Cheap Compressor, and you've seen how people (mostly) really like the way it sounds.

The MXR Sustain takes a slightly different approach and uses a 4558 as the audio path with just a little bit of gain (which is tamed by the output level pot). The level out of the op-amp is detected by a transistor rectifier (a bit more complicated than the Dynacomp one, but it works better and doesn't introduce distortion). The rectifier output is used to adjust the gain of a 3080 OTA, but the OTA is in the negative feedback loop of the audio path op-amp.... Sustain nirvana!

The range of compression is smaller than the Dynacomp, but is perfectly ample. it actually acts more as a limiter than a compressor. It has very low distortion and noise, and doesn't eat batteries.... Why build a Dynacomp?

PCB to follow shortly!

briggs · Post by **briggs** » 27 Mar 2012, 23:07

Interesting one, it's not a circuit I've looked at before and certainly not a circuit structure I've seen for a compressor before! I've just found a heap of old ca3080's in an old organ, I think I know what at least one of them is going in

roseblood11 · Post by **roseblood11** » 28 Mar 2012, 14:37

Where's the veroboard expert who makes a layout for a 1590B enclosure? Harald?!?

stonefreed · Post by **stonefreed** » 28 Mar 2012, 14:59

This is great mictester !! Is there going to a pcb design as well ?

I defenitely want to try this one.
I 've always thought of the dynacomp as fairly good comp but the distortion had started to bother me a bit.
I remember that I sold my script version and bought a Yamaha which was a little better , distortionwise , but not the sustain.
Then years later I built my own dynacompclone which I'm still using now and then.

Alf

stonefreed · Post by **stonefreed** » 28 Mar 2012, 15:01

By the way , is there something to be gained by using a better opamp like a TL0.. or a Burr brown instead of the 4558 ?

roseblood11 · Post by **roseblood11** » 28 Mar 2012, 16:54

Not really much, as it doesn't amplify much. Maybe use a lm833

Dirk_Hendrik · Post by **Dirk_Hendrik** » 28 Mar 2012, 19:16

mictester wrote: Why build a Dynacomp?

Because a Dynacomp contains more that 15 components already and therefore is in the category "very difficult". This circuit is in the category "insane" from boohteekers perspective.

Do you know how much time it will take to tune the sound of that side chain by ear? Let alone to hand select all those caps in the rectifier circuit for sonic capabilities.

(And no Mic, I do not subscribe to your statements on caps. But I do when it comes to circuits containing non-audio signals)

juanro · Post by **juanro** » 28 Mar 2012, 19:50

True, from the "swap&listen" perspective, the posibilities grow exponentially with the number of components, right? Ear-tune one of these would take YEARS, for BillFinneganChrist sake!

merlinb · Post by **merlinb** » 28 Mar 2012, 20:09

Dirk_Hendrik wrote:This circuit is in the category "insane" from boohteekers perspective.
Do you know how much time it will take to tune the sound of that side chain by ear? Let alone to hand select all those caps in the rectifier circuit for sonic capabilities.

That's either really good dry humour, or scary sincerity...

uncleboko · Post by **uncleboko** » 28 Mar 2012, 20:44

merlinb wrote:
Dirk_Hendrik wrote:This circuit is in the category "insane" from boohteekers perspective.
Do you know how much time it will take to tune the sound of that side chain by ear? Let alone to hand select all those caps in the rectifier circuit for sonic capabilities.
That's either really good dry humour, or scary sincerity...

or as Hank Wangford would say - "insincere sincerity"!

mictester · Post by **mictester** » 28 Mar 2012, 21:51

OK- here's a PCB. Not perfect, but I made two of them and they worked. Take the output from the wiper of the 50k pot.

: MXR_Sustain_Tracks.png (21.25 KiB) Viewed 5042 times

mictester · Post by **mictester** » 28 Mar 2012, 21:53

Another Mod I did - used two NE5534s (instead of the dual op-amp) - the PCB layout gets much easier! Noise was improved.

PokeyPete · Post by **PokeyPete** » 29 Mar 2012, 06:27

mictester wrote:Another Mod I did - used two NE5534s (instead of the dual op-amp) - the PCB layout gets much easier! Noise was improved.

Do you think an NE5532 on your board above would give similar noise results?

mictester · Post by **mictester** » 29 Mar 2012, 10:29

PokeyPete wrote:
mictester wrote:Another Mod I did - used two NE5534s (instead of the dual op-amp) - the PCB layout gets much easier! Noise was improved.
Do you think an NE5532 on your board above would give similar noise results?

Yes, but I found that if the gain stage was separated from the biasing stage (I used a TL071 for the biasing), I got lowered noise.

I'm going to put up a modified PCB (later today or early tomorrow) that uses separate ICs for the gain and for the biasing. A further useful modification I found was to put a unity gain buffer stage after the Level pot, to isolate the controls from external loading - it shouldn't matter, but I've done it anyway. The gain and buffer op-amps I've used are the NE5532 and the LM833, with virtually identical results (< 1mV change in the noise floor - it could have been down to specific ICs).

I built another one yesterday, using half of an LM13700 for the OTA. The noise and distortion figures were improved still further!

I'm not entirely convinced by the sidechain circuit. The principle's OK, but it seems quite a lot of transistors to do something that should be fairly simple. I'm considering ways of doing the same thing using op-amps (probably just a dual), so eventually I'll probably end up with a 13700, 5532 and an 074. The circuit principle will stay the same, but I'll get much better results from it!

I used to do a dual ratio compressor stage based on this kind of topology, but it was designed as a rackmount unit, not a stompbox. A lot of musicians used it as a "set and forget" first stage in an effects loop. In retrospect, given the results I'm getting from this relatively simple circuit, my rackmounted effort was grossly over-complicated!

stonefreed · Post by **stonefreed** » 29 Mar 2012, 14:44

"Yes, but I found that if the gain stage was separated from the biasing stage (I used a TL071 for the biasing), I got lowered noise."

That's what I wanted to know and thanks for the suggestion of another OTA.

Gonna try it. I'm very curious. And rasing the supply voltage would that be a benefit or do you have to change certain things ?

mictester · Post by **mictester** » 29 Mar 2012, 21:33

stonefreed wrote:"Yes, but I found that if the gain stage was separated from the biasing stage (I used a TL071 for the biasing), I got lowered noise."

That's what I wanted to know and thanks for the suggestion of another OTA.

Gonna try it. I'm very curious. And rasing the supply voltage would that be a benefit or do you have to change certain things ?

You'll get bigger headroom, and you'll have to be certain to use electrolytic capacitors rated for the voltage you're going to use. Just for the sake of interest, I powered one at 12V and measured the results - there wasn't any significant change from 9V, except that the noise floor was very slightly lowered, but not enough to be worthwhile.

cpm · Post by **cpm** » 29 Mar 2012, 21:52

but...
if the OTA is noisy by itself, it would still inject the noise in the feedback path, doesn't it?

mictester · Post by **mictester** » 29 Mar 2012, 22:50

The noise contribution from the OTA is small - most of the noise comes from the op-amps.

The advantage of this topology is that the OTA isn't directly in the audio path, so the distortion (predominantly non-linearity distortion) that it introduces is minimised. The Dynacomp (and its brothers) puts the OTA straight in the audio path. Some people like the distortion that an OTA introduces, but (as far as I'm concerned) the object of this kind of compressing sustainer is to provide sustain without distortion...

The more I experiment with this transistor rectifier, the more I think I can do better with a couple of op-amps! So, back to the breadboard......

cpm · Post by **cpm** » 29 Mar 2012, 23:11

is the noise from the OTA is smaller than the "in-path" topology because of it going for less gain?

What about a OTA in the path, but settled at near-unity gain, and amplifying only the sidechain input. That would be the same noise as this ota-in-the-feedback, but one less opamp in the main path.

merlinb · Post by **merlinb** » 30 Mar 2012, 08:18

cpm wrote: What about a OTA in the path, but settled at near-unity gain, and amplifying only the sidechain input. That would be the same noise as this ota-in-the-feedback, but one less opamp in the main path.

You would need lots of make-up gain if you did that, which would amplify the OTA's noise, so you're back where you started.

OTA noise is proportional to its control current (gain). By putting it in a feedback loop, the OTA gain is lowest when there is no audio input, so it produces its lowest noise level. This is the exact opposite of when you have in the Dynacomp, where the OTA gain is highest when there is no audio input.