THE ENGINEER'S THUMB... At last, a better compressor!

[Gandalf_Sr]

Thanks for the comments.

I am still learning to play but yesterday I had the 2 pedals that I've made so far connected in series - a Germanium fuzz pedal followed by the Engineer's Thumb (ET) compressor and I had a kind of epiphany as to why the compressor is useful.  The fuzz pedal is kind of quiet until you turn up the gain and then it's silly loud which means that I was always messing with the Amp's volume control; putting the ET compressor after the fuzz pedal cured all that, I can mess with the fuzz to get the sound I like and the output volume of the compressor stays pretty much the same - cool! 

[merlinb]

Part of what I'm missing is how it's supposed to work - I think it's supposed to keep the output at a constant(ish) amplitude regardless of the input level.

the attack pot is set fully clockwise (max time of about 30 mS

That's what attack means; the time it takes for the comp/limit to react to a change. So your screenshot is exactly right; it clamps down within about 30ms.

[rankot]

Guys, I've built this on a PCB I designed and I have a problem - it is distorting the sound all the way whatever I do. So I even tried to remove LM13700, thinking it could be false, just to discover that even without it I have this kind of noise on 4.5V rail (scope division is 0.1V):



This is how I did the power section - similar to the issue 4 schematic, but I have also added those two 100n capacitors. Could they be the cause of this oscillation? And I used NE5532 instead of TL072 in this position.



Also, I measure 9.2V DC at power jack, but only 4.2V at pins 2 and 3 of this IC. It shall be 4.6V, if I get it correct.

Ideas?

[Gandalf_Sr]

The 4.5V Rail is delivered via R19, I think it's meant to be 1k Ohm, not 10k?

[Rob Strand]

The 4.5V Rail is delivered via R19, I think it's meant to be 1k Ohm, not 10k?

Temporarily short R19.  Check for on pin 3 and on the opamp output.

[PRR]

....I have also added those two 100n capacitors.....

Why?? And why haven't you cut them out yet?

Hanging caps on opamp outputs is begging for oscillations. Which also hints why half is not half (when opamp is oscillating it usually isn't symmetric).

I also don't see why LED and 10k.

[Rob Strand]

I also don't see why LED and 10k.

IIRC, it was Merlin's trick to detect when the current source was driven hard.  It's used as the comp indicator.
There was a segment in the zillion of posts about it, maybe you were involved  .

[rankot]

....I have also added those two 100n capacitors.....

Why?? And why haven't you cut them out yet?

Hanging caps on opamp outputs is begging for oscillations. Which also hints why half is not half (when opamp is oscillating it usually isn't symmetric).

I also don't see why LED and 10k.

That explains a lot - I used them to further stabilize rails, but it seems that I made the opposite. LED and 10k are there as a compression indicator. I have shorted them (my first step when noticed problems), but it didn't solve a thing. I will remove those 100n caps right now and report what happens.

[rankot]

PRR was right (well, when doesn't he?), so removing those 100n caps fixed oscillations.

Now it works, but still not as expected - it either farts or not produce any sound at all. Just when I touch Threshold CCW pin with my finger, the sounds goes through, but after few seconds farts again. I will have to measure all the voltages, but what I find suspicious is the fact that I have 8.8V on V+ and only 3.6V on V_ref, which shall sit on 4.4V. But I don't have a clue why.

[rankot]

The 4.5V Rail is delivered via R19, I think it's meant to be 1k Ohm, not 10k?

In issue 4 schematic that I have, it is 10k, and voltage divider is formed using 100k resistors.

[rankot]

10uF capacitor across voltage divider was faulty, so that's why V_ref was lower. I removed it, put the good one and it seems to be working at least. But now I have another problem - I really don't hear it compressing, it seems that it is just passing the signal with some faint distortion.

[rutgerv]

Dear all,

I've been playing around with a variation to the engineer's thumb on my breadboard, using an SSI2164 VCA chip instead of the LM13700 OTA. By using a second section of this chip, I've been able to linearize the control law (normally logarithmic), such that it parallels the linear control law of the original OTA. I've used the original side chain of the ET, but obviously the control signal that comes out of it needs different scaling. The current version on my breadboard is working, but not quite right. In order to get it just right, I was trying to understand how the original ET actually implements its threshold. I understand the ET has a soft knee characteristic, so the threshold is more or less a continuous area, but don't understand which part of the circuit is actually responsible for creating this soft knee. Is it a by-product of the way the filter in the envelope detector works? Or is the soft knee happening even later (inside the current drive circuit for the OTA perhaps?)?

Can anyone enlighten me?

Rutger

[Gandalf_Sr]

10uF capacitor across voltage divider was faulty, so that's why V_ref was lower. I removed it, put the good one and it seems to be working at least. But now I have another problem - I really don't hear it compressing, it seems that it is just passing the signal with some faint distortion.

That was what I thought too (see my earlier post).  Then I put the compressor after a Germanium fuzz pedal that I have made that gives crazy high output when activated which means that I normally have to crank down the volume. However, with the compressor added after the fuzz box, the signal levels were tamed which is kind of the point of the compressor I think.

[merlinb]

I was trying to understand how the original ET actually implements its threshold.

The threshold is a by-product of the offset characteristics of the side chain opamp. In theory the circuit should start compressing with infinitely small signals, but in practice the precision rectifier just doesn't rectify signals smaller than about 10mVpk; it pretty much ignores them. That's why you can't see any obvious threshold-detection mechanism in the schematic; it's sort of a happy accident.

[rankot]

Now I removed everything I had extra and reverted my ET to issue 4 stock version. Even tried to use LM13700 from another batch, and still the same: sound is distorted when playing low stings or chords and also the tone is different - like there is less bass. I really can't figure this out. I have another ET that I built few years ago and it simply works as expected - tone is not altered, bright switch does what it shall do, all the pots (even Threshold) do their job. And it compress like heaven. This new one simply f*cks the things up.



H
E
L
P
!

[PRR]

...I was trying to understand how the original ET actually implements its threshold...

Analytically, the 1Meg bypassing the transistor sets 4uA of Iabc current at idle, so "nothing happens" until transistor and 1K resistor get 4mV of DC from the rectifier.

As Merlin says, 4mV is also the ballpark of slack in simple active rectifiers. I can well believe there's no strong action for some mV more.

The SSI2164 is kin to dBx and THAT Corp VCAs so a study of THAT's extensive literature may be enlightening (altho a different path than Merlin's). (The kinship is loose but as you know these are log not linear parts which makes a real difference.)

[Rob Strand]

Now I removed everything I had extra and reverted my ET to issue 4 stock version. Even tried to use LM13700 from another batch, and still the same: sound is distorted when playing low stings or chords and also the tone is different - like there is less bass. I really can't figure this out. I have another ET that I built few years ago and it simply works as expected - tone is not altered, bright switch does what it shall do, all the pots (even Threshold) do their job. And it compress like heaven. This new one simply f*cks the things up.

It's a good strategy to start from the basics.

I'd be lifting R10 and driving C5 from a pot wired from 0V to 4.5V.  As you adjust the pot you should measure the voltage between C5 and 4.5V, the check the same voltage appears across R12,  then confirm the current "V across 1K"/1k  = 2 * "V across R7" / 220 = 2 * "V across R9" / 220.    Adjusting the pot should control the level.  It will be maximum when the C5 voltage is near 4.5V and minimum when the C5 voltage is near 0V (maybe limit the lowest voltage to about 1V).   That will at least confirm the is gain is being controlled.      If the added pot is set to high gain and the ratio pot is set to max then you will get very high gain, probably enough to cause clipping.   You should be able to just what is clipping due to high gain settings and distortion even low gain settings.

After that I'd repeat the test but instead wire the added pot to IC "B"'s + input.

If you get that far perhaps the problem is dynamic in that there's something wrong with the filtering C4 and C5.

The weird symptom is the loss of bass.  Which can only come from C3 at the input and C3 at the output.    There's two C3's  so if you have 10nF for the output cap it's going to cut the bass for sure.

[merlinb]

This new one simply f*cks the things up.



H
E
L
P
!

Troubleshooting of personal builds should be in a separate thread.

[rankot]

Troubleshooting of personal builds should be in a separate thread.

OK, sorry.

I'm moving my troubleshooting here: https://www.diystompboxes.com/smfforum/index.php?topic=105610.msg1203445#msg1203445

[Eb7+9]

I was trying to understand how the original ET actually implements its threshold. I understand the ET has a soft knee characteristic, so the threshold is more or less a continuous area, but don't understand which part of the circuit is actually responsible for creating this soft knee.

like I pointed out above this is a very loose adaptation of the NE570/571 "soft knee" AGC architecture, except it uses a non-inverting op-amp gain stage (same current based NFB mechanism) ...

here's Walt Jung's reference paper on the subject:

https://www.yumpu.com/en/document/read/24667882/gain-control-ic-for-audio-signal-processing-walt-jung

similar to the NE570/571 architecture the ET exhibits near-zero/fixed threshold behavior ... actually the NE570/571 chip provides ideal rectification down to zero, and does so way more simply - so, itself providing a purely zero-threshold dynamic response ...

your attempt to use a precision trimmed OTA is definitely a step in the right direction, but that alone doesn't explain why so many people, including yourself, are still having a hard time making this build work on demand, every time ... or even why there's a fourth version with two OTAs in there (??) ... btw, using two OTAs makes things even worse from a build success point of view, never mind that it also increases the relative noise floor by SQRT(2) at idle via the fudging bias resistor R13 (1M) ... ie., the AGC/NFB mechanism is always on, even with no signal ... as a result, a relatively noisy circuit that gets worse with higher ratio settings 

it's obvious wherein the hang lies, but it's not my dog ... by studying Walt's paper carefully and reflecting on one comment I made above maybe someone will spot the crucial "duh" element in this saga ...

---

btw, for those interested in seeing a very good positive threshold active feedback style Limiter check out Steve W. Rabe's one-knob approach, and the eq'd side-chain variant I built from it for bass use ...

http://www.lynx.net/~jc/SWR-StyleLimiter.html