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

[antonis]

@ Mark: As for Relase control, I used a CCW 1M linear pot in series with a 220k resistor..
As for Attack control, I've made a "mark" on knob's travel for my own prefered spot.. 

[Mark Abbott]

Thanks for your reply.

How pronounced is the effect of the release control?

Regards

Mark

[antonis]

How pronounced is the effect of the release control?

To be honest, not much..
(at least, not as much as I expected..)

It could be replaced by a selector switch with 3 - 4 resistors ranging from 220k to 1.2M, with indermediate values..

[Mark Abbott]

I currently have the 1M resistor as the release control. I think it hangs on a bit too long.

[antonis]

Why don't you try a 560k, say..??

[midwayfair]

I did have success with a 47k substitute for the 10k going from #3 on the LM13700 to the op amp output. I got just the boost I was looking for.

That accomplishes the same thing, yes. Reducing the 220R is a lower noise, lower-offset solution than increasing the 10k, but not by much!

Huh. I would have thought the 13700 would still drive you down to unity no matter what you did in the feedback loop. Am I wrong about that? I've always sort of accepted that the max output of a sufficiently large signal in this design is 1K dividing the volume pot.

[merlinb]

To be honest, not much..
(at least, not as much as I expected..)
It could be replaced by a selector switch with 3 - 4 resistors ranging from 220k to 1.2M, with indermediate values..

Yeah I was a bit overgenerous with the release time. In the latest version I suggest a 100k resistor combined with a 470k or 500k pot.

[merlinb]

Huh. I would have thought the 13700 would still drive you down to unity no matter what you did in the feedback loop. Am I wrong about that?

If you reduce the 220R then the opamp has to produce more output signal in order to get the same signal fed back to its inverting input, compared with the stock 220R. Hence more gain. This is a feedforward compressor remember, so modding the feedback path does not alter the side chain (e.g. the current being fed into the OTA's control pin).

[rankot]

Merlin, have you ever though about using THAT 2181 instead of LM13700 in Engineer's Thumb? I would like to try this on my own, but my knowledge is still too poor for such an adventure.

[PRR]

> though about using THAT 2181 instead of LM13700 in Engineer's Thumb?

It wouldn't be "using", it would be a complete re-design. The '3080/'13700 OTAs are linear current control; the THAT parts are log voltage control. The side-chain design would be quite different. The OTAs have a low input voltage overload, which dominates good design; the THAT parts are current input and overload is simple scaling.

No doubt Merlin could design around the THAT chips. For that matter, THAT publishes comprehensive "how to use our stuff" papers you can just use verbatim. I seem to recall they published some stuff for low voltage (instead of the +/-15V world assumed in most THAT docs). Look at THAT 4315.

AH!! See 9V-Powered Effects Pedal Designs. Pedal Design Three: 4316-based Simple Limiter is one THAT chip and one dual opamp. I suspect it wants an input buffer to avoid loading a guitar. Design Four has such a thing.

The straw is that the THAT chips are harder to get and higher price than the OTAs. People balk at Small Bear's price for known-good '3080 ($5). Merlin may have swung to '13x00 because they sell under $2. The 2181 go $5-$10. Which I think is cheap, but this is a cheap crowd.

[merlinb]

[EBK]

Then there's the issue (an inconvenience, mostly) of the QSOP footprint for in-production THAT chips.

[mlp-mx6]

Then there's the issue (an inconvenience, mostly) of the QSOP footprint for in-production THAT chips.

Not all of them.  For example, the 4301 is still available in DIP-20 form.
http://www.mouser.com/ProductDetail/THAT/4301P20-U/?qs=sGAEpiMZZMvxTCYhU%252bW9mSHGcoR5HnTi6ZnSaFFEg3U%3d

[mlp-mx6]

Huh. I would have thought the 13700 would still drive you down to unity no matter what you did in the feedback loop. Am I wrong about that?

If you reduce the 220R then the opamp has to produce more output signal in order to get the same signal fed back to its inverting input, compared with the stock 220R. Hence more gain. This is a feedforward compressor remember, so modding the feedback path does not alter the side chain (e.g. the current being fed into the OTA's control pin).

Just a quick note of confirmation.  My ET had only ever barely reached unity volume.  I changed the 220R to 100R and the available level is now MUCH higher.  Perfect. Thanks, Merlin.

[jonny.reckless]

It's a really nice circuit and sounds great! Thanks. I made a few small mods to my build.

I paralleled up the 2 OTAs for a modest improvement in SNR. I just added a 1k resistor in series with each bias control input to make sure the two halves shared the bias current equally. I also replaced the main signal path op amp "A" with an NE5534 using a 68pF compensation cap (it needed to be this big for unconditional stability on my build). This reduces the overall output noise by about 8dB compared with the stock design.

[stan-thomas]

I paralleled up the 2 OTAs for a modest improvement in SNR. I just added a 1k resistor in series with each bias control input to make sure the two halves shared the bias current equally. I also replaced the main signal path op amp "A" with an NE5534 using a 68pF compensation cap (it needed to be this big for unconditional stability on my build). This reduces the overall output noise by about 8dB compared with the stock design.

Lower noise? Awesome. Please could you share the schematic.

[rankot]

Perhaps something like this? I have tried to modify schematics according to jonny.reckless notes, hope I understood him well. Take a look at this PDF:
https://drive.google.com/open?id=0B1wLPwWafEnXTkRKMl93UVZkVms

[POTL]

Hello.
A couple of questions are interesting.
1) Why is there such a large 10M resistor at the input, is it protection from the 3PDT cheat instead of which it is possible to install more common 1M 2M2 or for this scheme necessarily the value of 10M?
2) Why are LEDs on the input?
3) Transistor BC327, is it necessary or can it be replaced by another common PNP transistor like 2N3906 or 2N5087 / BC560?
4) The volume control has a non-standard value of 10K instead of the classic 100K
How important is the value of this potentiometer?

[merlinb]

Hello.
A couple of questions are interesting.
1) Why is there such a large 10M resistor at the input, is it protection from the 3PDT cheat instead of which it is possible to install more common 1M 2M2 or for this scheme necessarily the value of 10M?

It is the anti-pop bleed resistor. You can use 1Meg if you prefer, or 2M2 on each side of the cap if you want to maintain ~1Meg input impedance.

2) Why are LEDs on the input?

They're pre-clippers. They're needed because the circuit does not clip pleasantly by itself.
More explanation here: http://valvewizard.co.uk/engineersthumb2.html

3) Transistor BC327, is it necessary or can it be replaced by another common PNP transistor like 2N3906 or 2N5087 / BC560?

Yes any general purpose transistor should work, see the webpage.

4) The volume control has a non-standard value of 10K instead of the classic 100K
How important is the value of this potentiometer?

10k is the standard for opamp circuits (actually 5k may be more more common now that opamps have got so good). But 100k will 'work', though output impedance will be much worse of course, so personally I would never recommended a 100k output pot.

[PRR]

> Transistor BC327

"Any PNP". The specs are nearly unimportant. I can't think of any PNP that won't work. >4V, >4mA, hFE>1, fT>0.1KHz.

You do NEED to know the pin-out. If you get the Base wrong it can't work. I suspect C and E could be swapped and 9 of 10 times it would "work", maybe less-good.