Why does my Engineer's Thumb sound more like a volume swell than a compressor?

[solderfumes]

I've got a 5-knob Engineer's Thumb on my breadboard right now, following the schematic posted by Valve Wizard here:
http://www.freestompboxes.org/viewtopic.php?f=13&t=17213&start=80
I substituted 1uF electrolytic caps for the 1uF caps in the rectifier stage, and am using a TL074 instead of two TL072s.

The circuit seems to be working, but it overshoots wildly.  When I start playing, it will squish the signal down to near 0, and then the signal comes back to full volume as the compressor releases.  The speed of the initial squishing changes with the attack knob, and the speed of the release changes with the release knob.  This happens in pretty much any setting when you play loud enough; you can change the level at which this occurs by fiddling with the threshold and the ratio, but it still happens.  Basically it sounds like a "duck-and-swell" of extreme proportions.  The end result is that it behaves more like an auto-volume-swell than a compressor.  I also adjusted R14 hoping to reduce this clamping, and it seems to have helped a tiny bit but not enough.

I saw in the thread I linked above that this circuit was at the centre of a pretty heated debate when it was first published, with a couple of people having problems that sounded similar to mine.  However I've also seen Jon Patton's demo of this circuit, and it works beautifully for him.  Has anyone else had this experience with the circuit?  I can post voltages if that would help, with the attack, threshold, and release set at their "default" values as in the regular 2-knob schematic.

[solderfumes]

Huh... weird, I can't get the link to work at all.  Anyway, the schematic is the same as posted at

http://www.valvewizard.co.uk/engineersthumb.html

but with the threshold mod and treble boost mods added, as well as replacing two of the resistors in the rectifier section: the one that controls attack is a 1M linear pot, and the one that controls release is a 1M linear pot in series with a 220k resistor.

The link I posted with the debate on the subject is titled "THE ENGINEER'S THUMB... At last, a better compressor!" on *that other pedal forum*.

[bluebunny]

The link I posted with the debate on the subject is titled "THE ENGINEER'S THUMB... At last, a better compressor!" on *that other pedal forum*.

FYI, there's an equivalent thread right here.   <link>

[Mark Hammer]

From the behaviour you describe, seems to me that there is too much control signal feeding the 13700, turning the gain down too much.  Does it behave this way even when set for minimal compression?

Perhaps one or more of the resistor values are off?

[solderfumes]

FYI, there's an equivalent thread right here.   <link>

Yeah that one probably has the correct schematic as well somewhere in its 14 pages.  That thread was, um, less heated  

From the behaviour you describe, seems to me that there is too much control signal feeding the 13700, turning the gain down too much.  Does it behave this way even when set for minimal compression?

Perhaps one or more of the resistor values are off?

It doesn't do it when the threshold or ratio are turned all the way down, no, but it does happen even when they're set to quite low settings.  Pretty much any setting with audible compression does it.

I will definitely double-check the resistor values, but I kinda doubt that's the problem.  I already double-checked the 1k emitter resistor connected to Q1 and it measured fine, and that seems to be the most important resistor in the sidechain (bearing in mind I used pots for the attack and release resistors).  I thought it might be because the BC327 I used for Q1 had a high-ish gain (around 248) so I tried trimming that resistor with a 10k, to some avail, but it's still not workable.

I've never used an OTA before.  I know correctly using the diode bias pin is supposed to enhance the linearity of the amplifier, but I'm guessing that's only with respect to the differential inputs and not the gain pin.  Could I tune this overshooting behaviour with that biasing?  Or is there somewhere else I can trim other than the emitter resistor of Q1?

[solderfumes]

Double-checked every resistor and they all checked out.  I also tried the other side of the LM13700 and it did the same thing.  Here are my voltages:

supply: 9.30

LM13700
pin1: 1.126V
pin2: 0
pin3: 4.58
pin4: 0
pin5: 4.59
pin6: 0
pin7: 0
pin8: 0

pin11: 9.31
all the rest are 0

TL074:

U1b (rectifier)
pin7 (out): 4.62
pin6 (-): 4.59
pin5 (+): 4.13

pin4: 9.31

U2a (current source)
pin3 (+): 4.19
pin2 (-): 4.59
pin1 (out): 4.13

U2b (power buffer)
pin8 (out): 4.59
pin9 (-) : 4.59
pin10 (+): 4.37

pin11: 0V

U1a (main driver)
pin12 (+): 4.13
pin13 (-): 4.59
pin14 (out): 4.23

Q1
B: 4.13
C: 1.126
E: 4.59

As an aside, I also tried rigging it up to feed backwards by feeding the rectifier chain with a tap from the output of U1A through a decoupling cap.  It still does the same thing in a different fashion: when I start playing, the volume is reduced to near 0, it releases back to full volume, *re-triggers itself*, and then the whole thing repeats.

[PRR]

> pin3: 4.58
> pin4: 0


Pin 4 = 0V is not correct. It has to be same-as pin 3 for the '13700 to work right.

As designed it "can't" be zero. It has 220 Ohms to 4.5V and 10K to a pin which should be near 4.5V (and is).

See if this is a mis-measurement or you have a solder-blob to ground.

[solderfumes]

Oops, haha.  Thanks, I'll double-check that later.  It's on breadboard so perhaps a resistor lug was touching another resistor lug.  I suspect, though, that it was just a mis-measurement.  Perhaps the power connection got nudged out of place while I was reaching in with the probe.

[solderfumes]

That did the trick!  It was missing a connection to 4.5v.  Thanks so much!

Initial reactions: wow this is a clean sounding compressor.

[bluebunny]

Initial reactions: wow this is a clean sounding compressor.

It is indeed.  I built the two-knob version and I think it's a very sweet compressor.  10/10 for Merlin!

[solderfumes]

As an aside, I think the pin numbering on PRR's diagram is off: pin 4 is the negative input and pin 3 is the positive.  Makes it even more inexcusable to have a 0V reading at pin 4!

I'm thinking I'm gonna try making it switchable so that you can choose between a feed-forward and feed-backward topology.  I haven't played around with the feed-backward setup since I got it properly working, but my general impression from what I've read is that feed-forward is perhaps more accurate, while feed-backward is a bit more musical.  Hopefully I can get a little bit more of that Ross/Dynacomp character out of the feed-backward setup.

[midwayfair]

As an aside, I think the pin numbering on PRR's diagram is off: pin 4 is the negative input and pin 3 is the positive.

It's inverting vs. noninverting, not negative vs. positive.

[solderfumes]

Right, thanks.