Help me identify a circuit

[Mark Hammer]

Many people have a "bin of shame", with populated boards, PCB, vero, and/or perf, that either work, or still required trouble-shooting, and have lost their identity by sitting ignored for years.  I have several bins myself.  Today, I brought home a "bin of shame" from former forum member Peter Snow, who passed away 9-1/2 years ago.

Most of the boards are pad-per-hole, and have no indication of the circuit contained.  The only one that came close to identifiability is the only one in an actual enclosure, with "Tube Driver" barely legible in pencil on the chassis.  It "works", and produces a nice healthy pleasing overdrive, but neither of the pots (50k and 5k) actually does anything.  It uses a 13600 dual OTA (both halves) and a dual op-amp (TL082 was used).  The output is taken from the 13600, and there are there is ONE diode in the feedback loop of one of the op-amp stages.  I'd show you a picture, but as a pad-per-hole build I doubt the layout would yield any clues.

Being a Brit by birth, Peter was partial to "Uncle" Bob Penfold circuits, so it could be one of those, though not necessarily.  I'd like to get it working properly, meaning that both controls function.  Anybody have any clues as to what the original circuit might have been?  A schematic, or link to one, would be even better.  Hopefully the clues I've provided might spark a recollection.  It is unlikely the circuit is an original one, but you never know.

Thanks in advance.

[Mark Hammer]

Never mind.  Found it.  I was partly right.  British mag, but different author.

[PRR]

...neither of the pots (50k and 5k) actually does anything.

You do now see that it has to be battery or a *dedicated* (or true isolated) power source. Internally it makes +/-4.5V. If you share power and audio with other boxes, the power rails and biasing go totally lop-sided. It may fuzz all to heck, but not work as intended.

I'm not man enough to decode what it does. I'd suspect a rectifier maybe forcing heavy even-order distortion, but it could be something else.

[stringsthings]

I'm guessing it wouldn't be too hard to modify the circuit for "normal" ground and Vbias at 4.5V
like we're used to seeing.

Looking at the circuit more, maybe it would be difficult. 

[Rob Strand]

You do now see that it has to be battery or a *dedicated* (or true isolated) power source. Internally it makes +/-4.5V. If you share power and audio with other boxes, the power rails and biasing go totally lop-sided. It may fuzz all to heck, but not work as intended.

Actually it's OK.  It's got dumb-ass labels on the socket grounds!

[ElectricDruid]

Nice find, Mark. It's getting pretty hard to do something interesting in a distortion pedal, but that's a fairly unique approach: using OTAs to make a nonlinear amplifier. When you first mentioned them I thought it was going to be something based on their soft tanh input characteristic, but that's not it at  all. In fact, in some ways it's closer to synths that use Chebyshev polynomials for wave shaping.

T.

[MR COFFEE]

Gibson Lab Series solid-state guitar amplifiers used OTAs for overdrive\soft clipping several decades ago. The circuitry was designed by none other than Bob Moog.

The Gibson Lab series Guitar amp was B.B. King's favorite amp, at least as of about 10 years ago. The schematics are available on the Internet and are worth checking out if you want to see something quite different. He included something called a multifilter circuit in the Lab Series that has a had to describe effect.

mr coffee

[Aph]

Gibson Lab Series solid-state guitar amplifiers used OTAs for overdrive\soft clipping several decades ago. The circuitry was designed by none other than Bob Moog.

The Gibson Lab series Guitar amp was B.B. King's favorite amp, at least as of about 10 years ago. The schematics are available on the Internet and are worth checking out if you want to see something quite different. He included something called a multifilter circuit in the Lab Series that has a had to describe effect.

Yes, I used to have one. Liked it. But, being an impulsive type, got rid of it. Now contemplating this:

https://aionelectronics.com/project/l5-preamp/

[Mark Hammer]

Craig Anderton used CA3080 OTAs for the voltage-controlled distortion module in his AMS-100 project.  Of course an OTA is not an OTA is not an OTA.  The 3080 tends to distort with inputs ignals greater than 100mv.  http://hammer.ampage.org/files/Device1-8.PDF

[ElectricDruid]

Craig Anderton used CA3080 OTAs for the voltage-controlled distortion module in his AMS-100 project.  Of course an OTA is not an OTA is not an OTA.  The 3080 tends to distort with inputs ignals greater than 100mv.  http://hammer.ampage.org/files/Device1-8.PDF

Ok, tell me more! How is an OTA not an OTA?!? I mean, I've seen the schematics for the 3080 vs the 13700, so I realise there are some differences (mostly in the type and sophistication of the current sources used, iirc) but does this really make such a huge sonic difference?

Aside from the 13600/700 and the CA/LM3080, I can only think of the CA3280 and the Roland BA662. There's also the CA3094 and the (very rare) CA3060. Oh, and the NE5517, which is supposed to be the same as the 13700. So many 3080 projects have been converted to 13700 that I'd have to regard those as pretty much interchangeable.

Aside from TB303 fanatics, I've never heard anyone seriously arguing that some of these are significantly better/different than others, with the exception of the CA3280, which was (as far as I know) the only chip of its generation - but that seems to be mostly that it was lower noise than the (rather noisy) others.
But this is a fascinating topic and if there's more I can learn, I'm all ears!

Tom

[Rob Strand]

Ok, tell me more! How is an OTA not an OTA?!? I mean, I've seen the schematics for the 3080 vs the 13700, so I realise there are some differences (mostly in the type and sophistication of the current sources used, iirc) but does this really make such a huge sonic difference?

The LM13600 and LM13700 have the pre-distortion diodes which increase the maximum signal capability. IIRC the CA3280 also has diodes.  There's a bit of a blurb about it in the datasheet.   In the back of my mind there was an app note with more detail.

While the improvement is measurable, it's debatable how noticeable the difference is in practice.

[Mark Hammer]

JimPatchell has some interesting info here:  http://www.oldcrows.net/~patchell/SecretsOfOTAs.pdf

I didn't mean to imply OTAs were desgned in fundamentally different ways.  But they vary in their susceptibility to distortion.

[MR COFFEE]

Hi Rob, Mark, and all,
In my experience with the 3280 dual OTA (and NOT using any "pre-distortion" diodes available in several of the later generations of OTA offerings) the CA3280 had quite a bit better noise and bandwidth. It is truly a shame that CA3280 is the model that failed to survive in production to the present. If the circuitry was well-designed for low noise, the CA3280 was measurably - and audibly - quieter.

Despite the improved OTA designs, they couldn't keep up with the Allison, DBX (now THAT), CEM, and SSM VCAs, in terms of signal-to-noise, and yet the old-school 3080-based designs persisted, and flourished, anyway.

It remains quite amazing to me that the old 3080 designs still persist to this day for guitar compressors, even in "Boutique" designs, where component cost should be such a small fraction of the unit cost, when such superior parts are available - just look at the THAT Corp. VCAs.

mr coffee

[Rob Strand]

n my experience with the 3280 dual OTA (and NOT using any "pre-distortion" diodes available in several of the later generations of OTA offerings) the CA3280 had quite a bit better noise and bandwidth.

I found it interesting the CA3280 actually had a noise spec but the CA3080/LM3080 does not.   I was never certain if the CA3280 was a dual CA3080 with input diodes or not.  It's stretching my memory a bit.

On one hand, the shot noise from a Gilbert cell should simply be a function of the IABC current.   It should be the same for all devices.  *However* given these are small devices the transistors might have significant transistor Rbb'.  So you get an extra Johnson noise component, plus the effect of the shot noise through Rbb'.   Years ago I came up with the idea of paralleling OTAs to reduce noise.  I tested the idea on NE570/NE571/NE572 and it does work; these have a slightly different input structure to raw OTA's and probably benefit more.    There was a thread on it around Christmas.    The benefit of paralleling normal OTAs depends largely on the size of the Rbb's.   If Rbb' is already small then paralleling OTAs might not be so beneficial because short noise just depends on the total current.

Getting to the point, it may well be that the Rbb's of the LM3080 produce significant noise whereas the CA3280 might have Rbb's small enough to push their contribution into the shot noise level.    That would imply paralleling LM3080 would reduce noise.      Things like SSM/That stuff are likely to have low Rbb's and I'd say these would be as low noise as you could get, as shot noise is a theoretical limit.

This is where the diodes (and other schemes http://www.openmusiclabs.com/files/otadist.pdf) ) come into play.  If you can't reduce the noise at the low end you try to increase the drive level at the high end.  That way the dynamic range and signal to noise can be improved.

I did some work on mixers for Medical Ultrasound devices and the mixers pretty much had theoretical shot-noise.   You get the shot noise down by operating at higher currents (in effect High Iabc) but you *must* drive them from a low impedance.

( Oh, there's another trick to get larger signal handling, you add Re resistors on the diff-pair.  Unfortunately you can't do this to LM3080 etc.)

[Rob Strand]

I had a poke around with CA3280.  The input noise spec is 7 to 8 nV/rtHz for Iabc=500uA.
In order to get input referred noise to this level each input needs about 1300 ohms in series with it (or driving it).

One thing not clear from the datasheet is if the noise spec is for a specific test circuit or just the raw device.
For example if the test circuit has 1k on each input then that implies Rbb' for the OTA diff pair transistors is 300 ohms.  If the noise spec is for a low impedance test circuit it means means Rbb's is 1300 ohms.  That's quite high.  Even 300 ohms is significant.

For Rbb' = 1300, and Iabc = 500uA,  the input noise is 11dB above the theoretical shot noise.  So that means parallel devices will help.  It also means keeping the input impedances down helps keep the noise down.  When Iabc is small the difference becomes less as the shot noise will dominate the Rbb's noise.  Paralleling devices will then help reduce noise.

[Rob Strand]

For comparison, to match the noise in the LM13600 datasheet I have to put 5k series impedance on each input.
I can't believe there is 5k on each input inside the chip,  so I can only conclude the noise measurement assumes a test circuit with external series resistance.

This is probably the case for the CA3280 as well.

Unfortunately I can't see or work out what the test circuits are from the datasheet.

[ElectricDruid]

My contact at Alfa Rpar in Latvia tells me that they're working on a re-release of the CA3280 - the AS3280. These are remanufactured chips, not necessarily direct clones. On some of their previous chips they've made some upgrades - improving internal current sources by using a better design that wasn't around at the time, or that was considered too expensive for the extra handful of transistors it used or whatever. So they should be at least as good as the originals and probably better.

They're also doing a quad OTA based on the 13700, but without those stupid darlington buffers, to be called the AS13704. Four uncommitted OTAs in a chip sounds like a 4-pole filter to me - Roland synth style - but I can think of a ton of other things to do with it too - OTA phaser, anyone?

So OTAs are making a comeback too!

Tom