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

[Cliff Schecht]

Vref itself is coming from the 1/2 supply divider and U2B is buffering this because it is used in a lot of places. If any of the 4.5V reference circuits draw current from the resistor divider it can cause the reference voltage to move around. The buffer provides a low impedance output which keep the reference voltage at 4.5V pretty much regardless of current draw (within reason).

[Mark Hammer]

Hats off, my friend.  Hats off.

[Jdansti]

Very nice!  This is probably one of those "duh!" questions, but here goes:

I don't see IC2/U2 on the Vero layout. Is the Vero an equivalent circuit to the schematic and operates the same without IC2?

[Jdansti]

Very nice!  This is probably one of those "duh!" questions, but here goes:

I don't see IC2/U2 on the Vero layout. Is the Vero an equivalent circuit to the schematic and operates the same without IC2?

EDIT:
I think I just had the "duh!" moment. The TL074 takes the place of the 2 TL072's. Correct?

[merlinb]

Why not cut down on op amp count and use the OTA for the envelope follower or expo current source? A FET could buffer the 4.5V reference just as well as the op amp circuit and you could maybe shrink the overall footprint of the design a bit.

There's also the two on-board darlington buffers that you could still put to some use.

Believe me, I tried to use both OTAs and darlingtons in many ways as I developed the circuit. But in the end I could not get the 'good behaviour' from an OTA rectifier or current source as I could from opamp versions, and whatever I did with the darlingtons they always ended up redundant, one way or another.

It certainly lends itself to a multi-band or stereo application. This small version is only the start! Hopefully some adventurous people will find the basic topology useful for their own compressor experiments!

EDIT:
I think I just had the "duh!" moment. The TL074 takes the place of the 2 TL072's. Correct?

Yep that's right. I wanted to keep the parts designators the same in the PCB and vero, but a quad opamp made for a smaller vero.

[Jdansti]

[/quote]
Yep that's right. I wanted to keep the parts designators the same in the PCB and vero, but a quad opamp made for a smaller vero.
[/quote]

Thanks!

[Mike Burgundy]

This small version is only the start! Hopefully some adventurous people will find the basic topology useful for their own compressor experiments!

Ooh, I know a glove in the face when I taste it! I just wish I had the time. I'm not even thinking about pretending I could come up with something like this on my own ;P
Still haven't gotten around to building my last 2 projects beyond breadboard, and the first of those has been brewing for almost a year.... Any takers?

(edit: gotta learn to type slower, or more accurate)

[Keppy]

Nice! I've been looking forward to this one. Thank you!

[bonaventura]

understood. thanks cliff.

[mr_deadmaxxx]

 

[CaptainTuna]

Great! Just great!
I am definitely gonna build this one next week.
Do you suggest any modification to it for using it with bass?

[merlinb]

Great! Just great!
I am definitely gonna build this one next week.
Do you suggest any modification to it for using it with bass?

It will work fine with bass right off the bat. You could increase the 100pF cap to 220pF if you like, which will reduce noise, since you don't need as much bandwidth.

[frank_p]

Could someone do a little walk-trough of how the electronics works ?   ...please.  That would be great !

[PRR]

> a little walk-trough of how the electronics works?

Very little walk:

This is all old-hat design. (What is new is getting it all to work nice under 9V supply.) You can find the general plan (over-complicated) in Crown and Peavey and other power amps.



Signal goes in at U1a "+" and out at U1a "out".

Assume super-small signal, no limiting.

Assume U3a is essentially dead. (There is a teeny trickle through 1Meg to 4.5V.)

U1a gain is roughly 1Meg/10K (actually the 1Meg is varable and there's a 1K and a +1 in the formula). "Treble mod" brings in another 10K for double the treble gain.

Now put signal in. U1b is standard half-wave low-offset diode referenced to 4.5V. Negative signal peaks are captured on 1uFd cap. The peak voltage is forced by U2a onto a 1K resistor between a transistor and 4.5V. Transistor current varies from zero to 3mA.

This is why we need U2b, to hold "4.5V" steady despite 3mA thunks in transistor current.

Transistor current "wakes up" OTA U3a. It passes U1a output signal back to U1a input, in shunt with the 1Meg pot. This turns-down U1a gain, and thus the signal.

This is "backward" from the basic plans on OTA datasheets. Usually they run signal forward through the OTA, idle it at maximum current/gain, and reduce current/gain for limiting action. That works, good-enuff for many purposes, but isn't optimum. The high-current OTA has significant hiss, in part because the input is so very sensitive that most high-level signals must be attenuated before the OTA, and low-level signals necessarily get the same pre-attenuation. Also the turned-down OTA can be starved for headroom. In the backward connection, for super-small signals the OTA can be essentially "off", adding no hiss, and on strong signals the OTA current is rising so headroom is good. When limiting, input level to the OTA is essentially constant, and the designer can select this level on the OTA's THD plot so that distortion is small.

"Backward" connection would seem to need an additional op-amp. However in practice we always need to buffer the low impedance of the OTA attenuator (10K+220r) so it isn't more parts than the "basic" connection.

This design is feed-forward. It is easier to make a feedback limiter hold a constant level, but overshoot must be considered. A feedforward limiter needs very accurate rectifier and VCA to produce accurate limiting. It "can be shown" that both types can limit equvalently well, disregarding practical implementation details. In this case the OTA is a "perfect" VCA over more than the required range, and the rectifier is plenty accurate for the useful range of guitar signals.

[frank_p]

It's great to have a small explanation in the same topic than the schematic is. Thanks Paul and Merlin.

[Mike Burgundy]

This small version is only the start! Hopefully some adventurous people will find the basic topology useful for their own compressor experiments!

Ooh, I know a glove in the face when I taste it! I just wish I had the time.

I'm not picking up any gloves just yet, but if anyone's interested and wants to tinker, I just cobbled up a simple 3-band Linkwitz/Riley crossover that looks quite okay and can be used for multibanding. Just the schem simmed.
Yes, 3-band, will leave you with yet another unused half chip, I know... I just feel 2 bands won't quite cut it, and 4 is getting too complicated by far as far as knobs and board space is concerned. It's quite easy to extend this to 4-band, if needed. 4 bands might have the need for a 3-pole filter, which only needs an extra cap and resistor per opamp. Edit: actually that might not be as easy as it seems... 2-pole it is, I guess ;P

Lemme know, I'll put it up. Edit2: it's up.

[merlinb]

Looks like my web page has run out of bandwidth, so I've added the files to photo bucket too. PCB, layout, schematic and everything is here:

http://s81.photobucket.com/albums/j207/merlinblencowe/Engineers%20Thumb%20Compressor/

[Cliff Schecht]

Just knocked out a little SMD board for this compressor. The one thing I'm really interested in figuring out is how this compressor will compare to my beloved Orange Squeezer and the Dyna/Ross comps I'm building up as well (neither of which I've really ever used either). I've been using my OS for many years and enjoy the flaws that this circuit has. Makes for a very musical compressor, at least when running a guitar into an amp. I'm wondering if this compressor (the Engineers Thumb) will be too good and not the whacked out squishiness that the OS does so well. I'm curious to hear Merlin's thoughts on this, especially being that he is a EE as well and certainly knows the theory side of things. Perhaps this is why he chose the exponential current source over the simpler linear one for the OTA bias circuit..

I'm hoping to populate this board tonight. In my haste to finish the board quickly I of course forgot to mirror the transfer so I'll be flipping the SMD op amp and OTA over to salvage this board. For a board using 10 mil traces everywhere, this transfer came out too clean to just waste it!

[PRR]

> why he chose the exponential current source over the simpler linear

Thre's no exponential current source here.

[midwayfair]

Just finished wiring this up after waiting a bit for the 13700 chip. I'm a bit of a comp affectionado, so it was nice to try out a new "easy" design. I hope it's okay to post my thoughts here:

Overall sound:
It sounds beautiful as a boost on certain milder compression settings.
In the stock setting, the most extreme compression settings lost some highs. The treble boost mod cures this, but I can see how players who use picks (I'm strictly fingers) might find this desirable.
The bass response is a little smeared at the most extreme comp settings.

How does the compression "feel"?:

The lowest ratio resistance provides no available boost, but it is extremely even limiting and feels very natural. The highs are fairly attenuated, so, again, I found the treble mod to be a perfect cure for my complaints here.

The next 100K of resistance was all very usable in every "normal" testing situation I messed with - it sounded right and reacted right at all areas of the sweep, but with varying levels of overall compression, high end, and boost available.

The compression is very natural compared to a Dynacomp (despite certain people's assumptions elsewhere), though I must admit that the range of the ratio knob for the purposes of compression can be fairly limited with any given guitar. Using my tele w/ broadcasters, anything above 250K was virtually indistinguishable from the settings below it except that I got more pumping like a mis-biased Orange Squeezer, and the release was a little too long when switching between slow soft notes and fast runs (meaning that the fast runs would keep a quiet part immediately following from sounding correctly); I tried a strat and found that the 100KA pot would suffice (but the low-input pickups did like the very top of the comp knob)! I  don't have a super high output guitar to test, so I ran a boost in front and I could see how, say, active pickups would like to have the 1M pot. It still had some transients and didn't distort (which is very nice), but there were some noise floor issues.

Mods?:
I built it almost stock - I used a 500KA pot for the ratio because that was more than sufficient with every guitar I own and gave me the most usable sweep. A 250KA would have been ideal for my purposes, but I didn't have one. After testing the attack mod (which I didn't like -- but then, I don't typically like that mod on anything) and the release mod (which I did like for more transparency, but the release is already very short), the treble mod was the only one I really found essential, as it cures the only problem I had with the highest compression setting.

Overall:

This walks the line between hard "effect" compressor boxes like the Orange Squeezer and Dynacomp and really transparent comps like the Flatline. It would certainly please people who want more squish than the optical comps but don't necessarily want to know if it's working.

I'd recommend this project to people looking for another "effect" comp, but maybe not people looking for absolute transparency.

Is is a "better compressor"? I think it's certainly better than the Ross for my tastes. I've never really liked the Ross.

One last thought:
It's tough on perf, and the etch is complicated with a lot of jumpers. Maybe a group buy for double-side fabbed boards is in order?

Anyway, it's going in a pretty green box as soon as I get to the drill press tomorrow.