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

[jonny.reckless]

That's a great idea, thanks. I will add makeup gain. For this I'll have to switch from a NE5534 to an NE5532 or maybe even LM833 for the main audio signal path op amp.

The switch to inverting was motivated by the desire to have the gate attenuate the signal by at least 20dB. The original non-inverting design does not allow for gains less than unity. I also wanted to put a JFET front end on the unit, as I have found that something like a J112 input stage generally sounds sweeter and more articulated (more high frequencies, better transients) than the JFET of a TL072 input pin, especially with single coil pickups. I don't know why this is, but if you A/B it, you can clearly hear a difference.

It's still a forward compressor with a single VCA. I parallel up the 2 halves of the LM13700 for a 3dB improvement in noise figure. The gate sets the gain reduction to maximum.

I am working on the next generation of this idea, with full control and 8 knobs: threshold, attack, release, gate threshold, gate attack, gate release, ratio and volume.

R3, R4, C2, C5 form an emphasis filter, same idea as in the original. Roughly 6dB boost at HF, first order. I moved the corner frequency down a bit. This was done purely for personal taste, and was tweaked until it sounded right on my Squier tele. C5 = 22n also sounds nice, as does 4n7.

[PRR]

> The original non-inverting design does not allow for gains less than unity.

I have not looked at the design in a while.

But the OTA can have gain. So the op-amp loop can be driven below unity gain.

I think.

[jonny.reckless]

> The original non-inverting design does not allow for gains less than unity.
But the OTA can have gain. So the op-amp loop can be driven below unity gain.
I think.

Yes, you are right.  I was neglecting the fact that the OTA could have a transconductance higher than the loop gain, and thus allow for overall gain less than unity. Thanks for the correction.

[jonny.reckless]

I did quite a bit of work on this over the weekend. I fixed the noise gate, added hysteresis to the threshold, and generally cleaned up the performance. I also added a negative voltage converter and made the unit split power supply, which simplified things a bit. It's getting further away from the original engineer's thumb but I like the flexibility.

I'll post more when I have chance to make some recordings. This is where I have gotten to with the schematic. It's hacked together on the bench at this stage, so quite noisy, but I'll fix that with a PCB. The LMC6482 should actually be TL072 - they can't handle 17V supply.

[samhay]

> The original non-inverting design does not allow for gains less than unity.

I have not looked at the design in a while.

But the OTA can have gain. So the op-amp loop can be driven below unity gain.

I think.

Good point. I have some recollection this won't be the case in feed back designs, but as the ET is feed forward it is possible, although I suspect you would need a hot signal and the ratio set quite low.
Perhaps someone with an ET can confirm?

[PRR]

> some recollection this won't be the case in feed back designs

We all know that the Non-Inverting operational amplifier connection only goes down to unity gain.

Why? Because we use a resistor divider to set loss which is the inverse of the gain we want. But resistors alone can't make "loss less than unity" (i.e. gain). So the minimum gain is unity.

Putting gain in the feedback divider is really strange. Amplifiers are inconsistent, resistors are consistent. Who would do that? In fact it does get done frequently enough, even if not in stomp-land.

For gain of 10, we use 10:1 resistor divider. This could just as well be some (lame) "amplifier" with "gain" of 10:1 (a loss).

For gain of 1, we use 1:1 resistor divider (a strap). (Except my sim objects "floating pin" so I threw it a negligible bone.) This could just as well be some (lame) "amplifier" with "gain" of 10:1 (a loss).

For gain of 1, we use 1:10 resistor divider. Ah, that won't happen with positive resistances. β can not be greater than 1. But it could be an amplifier with gain of 1:10. Now the non-inverting connection gives gain of 0.1.

See below.

That's just DC/Audio gain. You also seem blocked on how this would work in a dynamic limiter. But any overall gain/loss can be managed by inserting gain/loss in the right places. Yes, Blesser's paper shows that as you go to extremes, feedforward and feedback have different problems. We rarely hammer on audio so hard that this is insurmountable.

[jonny.reckless]

I'm doing a PCB layout now. It's quite a challenge fitting it all on a 100mm x 70mm PCB. I think I've got something, it just needs a little polish.

[rockhorst]

Hats of to all of you, esp. Jonny, for all the great work in this thread. I read through all 27 pages yesterday, it underlines what a great community this is (I should visit it more often).

[samhay]

>That's just DC/Audio gain. You also seem blocked on how this would work in a dynamic limiter...

Thanks Paul. I seem to have mis-remembered something regarding the limits of feedback automatic gain control. Will have to put out the reading material again.
Simulation says that the stock ET can do a gain of < 1 when the signal input is hot enough (~ 1V peak with moderate-high ratio settings). Conversion to a feedback  topology can give similar behaviour.

Happy to stand corrected.

[digi2t]

Maybe a dumb request, but since I've found some contradictions on the net, here goes...

Would it be possible to indicate the lug 1 for the pots on the schematic? I could never get my head past volume controls where that's concerned. Just seems that the conventions for this are all over the map.

[rankot]

I usually put numbers on pot lugs, or CW/W/CCW.

[digi2t]

I usually put numbers on pot lugs, or CW/W/CCW.

I usually just indicate "1" on my schematics. KISS principal. Like I said, volume or gain aren't a problem, I've gotten my head around those, but sometimes tone or other functions throw me completely off. I can never figure out "are they bridging 1&2, or 2&3?". "Are all the lug 1's on the left? On the bottom? On top?" Aghhhh! 

And then, generally without failure, I build it only to realize that no.... I guessed wrong.

[merlinb]

I usually just indicate "1" on my schematics. KISS principal.

But '1' doesn't mean anything. I have no idea which pin '1' is on a pot; it's not standardised (I think only ICs have a standard pin numbering pattern). Depending on who you buy your pots from, their datasheet could label any of the pins '1'. The footprint in you PCB design program could have still different numbering!
But CW/CCW is completely unambiguous.

[digi2t]

I usually just indicate "1" on my schematics. KISS principal.

But '1' doesn't mean anything. I have no idea which pin '1' is on a pot; it's not standardised (I think only ICs have a standard pin numbering pattern). Depending on who you buy your pots from, their datasheet could label any of the pins '1'. The footprint in you PCB design program could have still different numbering!
But CW/CCW is completely unambiguous.

Or, the square pad on a PCB...



Not meaning to be a shmuck or anything... but in the general scheme of things (without getting into the "how many angels on a head of a pin" bullshit), pretty damn unambiguous to me.

[Govmnt_Lacky]

Not meaning to be a shmuck or anything... but in the general scheme of things (without getting into the "how many angels on a head of a pin" bullshit), pretty damn unambiguous to me.

Agreed! For as long as I have been building or dealing with potentiometers, the CCW-most lug has always been Lug 1.

[merlinb]

Not meaning to be a shmuck or anything... but in the general scheme of things pretty damn unambiguous to me.

How about now?


By using CW/CCW you don't have to look anything up on google, you can always figure it out with the part in hand.

[digi2t]

Not meaning to be a shmuck or anything... but in the general scheme of things pretty damn unambiguous to me.

How about now?

Dude... really? Tell you what. When a DIY project gets to that point, we can settle that dust then and there. In the meantime, try keeping with the program. The subject today is apples. Not oranges.

[merlinb]

Dude... really?

Dude you've never used a trimpot? Or a Bourns pot? All pots are apples.

[samhay]

While I wouldn't have to think too hard about what lug 1 of a pot looks like (and I tend to use that notation on schematics) there is a further fly in the ointment:

[digi2t]

You know... I thought I was asking a fairly simple, maybe even mundane question. Generally, I've always been a true believer in the "there are no stupid questions" mantra, and as such, try to be true to that when I'm feeling competent enough to answer someones question.

In this case, not only does it seem that I'm going home without an answer to my question, but with Merlin's magic wand shoved up my a$$ as a parting gift.

If anyone wishes to illuminate me insofar as my question is concerned, I would be greatly appreciative. I can even do the "CW or CCW" thing. Just please don't bore me with the semantics of which lug is "LUG 1" on a 15 lug multi wafer pot.

While I wouldn't have to think too hard about what lug 1 of a pot looks like (and I tend to use that notation on schematics) there is a further fly in the ointment:

How about 1A, 2A, 3A and 1B, 2B, 3B? Sort of like what's used on rotary switches? (Now I duck to avoid the incoming flurry from the experts as to why that makes no sense).