OTAs

[Chuck D. Bones]

Some of the pedals we build contain a special kind of opamp known as an OTA. OTA stands for Operational Transconductance Amplifier. Transconductance in this context means voltage in and current out. What makes an OTA particularly useful to us is the fact the we can use a control voltage to vary the gain. Technically, it's a control current called IABC that sets the gain. Back in the old days, the OTA in use was the CA3080. It's no longer in production and has been replaced with a similar, better dual OTA: the LM13700. The LM13700 is essentially a pair of CA3080s and a pair of emitter follower buffers. Here's what half of an LM13700 looks like inside:

We'll ignore Q12 & 13 for the moment. Q4 & Q5 are a differential pair connected to the inputs. Q1 & Q2 are a current mirror that takes the current flowing into the AMP BIAS INPUT (IABC) and makes the same current flow in Q2. Q4 & Q5 share Q2's current, but not necessarily equally, depending on the voltage difference between the +INPUT and the -INPUT. Q4's collector current is reflected thru two more current mirrors (Q6 & Q7 - Q8 & Q9) while Q5's collector current is reflected thru current mirror Q10 & Q11. What we get at the end of all of this is a current at the output that is proportional to the difference between the +INPUT and the -INPUT multiplied by IABC. Diff amps are non-linear, so D2 & D3 are there to linearize the diff amp. We usually don't use those diodes in pedals. We can get away with that by making the difference between the +INPUT and the -INPUT small, on the order of a 10mV or less.

Now about Q12 & Q13... because the OTA output is a current, the output impedance is by definition very high. We don't necessarily want a high output impedance, so a simple unity gain buffer is provided in the form of a Darlington emitter follower made up of Q12 & Q13. some circuits make use of the buffer and some don't.

So now we have an amplifier whose gain is controlled by IABC. Making IABC larger increases the current in all 10 transistors which increases the gain. But there's a catch. The noise in a transistor is proportional to the current flowing in that transistor. Higher current means more noise. IABC is kept deliberately low in the Maestro FSH-1 voltage-controlled filter and for good reason. When I breadboarded the FSH-1, I experimented with higher IABC and the noise floor became objectionable. The same thing happens in most OTA compressors. The EHX Black Finger and the Valve Wizard's Engineer's Thumb are the exceptions because of the clever way they use the OTA.

Rather than explain the benefits and inner workings of the Engineer's Thumb, I'll point you to the source:

http://www.valvewizard.co.uk/engineersthumb.html

 

[bowanderror]

Great timing on the write-up, Chuck! Just today Electronic Audio Experiments posted a discrete OTA board he'd made with 3 matched SMD transistor pairs (at least from what I can see):



Here's the info he posted:
This little friend is a very simple discrete OTA using matched transistor pairs. An OTA, or operational transconductance amplifier, produces a current proportional to the input signal and a bias current. In short, it’s an amplifier you can control with a voltage or signal. It’s no mistake why OTAs are the building blocks of so many synthesizer VCAs and filters, as well as a few classic guitar compressors, phasers, ring mods, and tremolos.
A few different integrated circuit OTAs are available on the market, but I never really understood how they worked. So this was my way of grappling with the technology. I’m pleasantly surprised with the results so far. Perhaps this circuit will find its way into a product at some point?
Is this a drop in replacement for a DIP8 OTA? - Same pinout as a 3080! Not the same function but it makes it easier to breadboard. The 3080 has push/pull output which def helps with normal load driving, this one is happiest going straight into an I to V converter.
Whats the noise like compared to the LM13700? - That’s a good question. If I have any LM13700s around I will run that test! Subjectively it’s about as noisy as any other OTA. As a VCA I’m using a very low impedance network to attenuate the input especially compared to, say, a Dynacomp circuit. My guess is PSRR & CMRR are both worse than the 13700, but the distortion profile is a lot more pleasant.

So the Bones of my question to you: What are the biggest challenges in making a discrete OTA? Is it worth doing, even if only in certain use cases?

I recall a post from Jim Patchell, going back almost 10 years, where he started on an open-source OTA using bog-standard 2N3904/2N3906 transistors, but I couldn't find a followup.

Thanks Chuck!

 

[cooder]

How timely indeed. Did this come about because I poked you with an objectionable design from somewhere?
In any case, I need an extra dose of brain synapses now...

 

[Chuck D. Bones]

Build an OTA? It's a fun experiment, but a lot of effort for not a lot of benefit. All of the transistors should be matched, not just the pairs, and they should all be at the same temp. Easily achieved when they're all on the same die. As one of my early mentors told me: "Why buy the cow when milk is so cheap?" He wasn't talking about electronics, but in this case, it's applicable.

Sure you can make distortion with an OTA if you like. If it's diff amp distortion you're after, look no further than the Keeley Fuzz Head, Dinosaural Tube Bender or Subdecay Variac. You don't need all of the other stuff in an OTA. With a discrete design, you can deliberately unbalance the diff amp for some even-order harmonics and use local feedback (emitter resistors) to smooth out the distortion a bit. Check out the attached PDF where Mr. Keeley explains how the Fuzz Head works and how to tweak it.

 

[Chuck D. Bones]

How timely indeed. Did this come about because I poked you with an objectionable design from somewhere?
In any case, I need an extra dose of brain synapses now...

Yeah, you know you did. The Ross-based compressors are an inferior design and I take no stock is such truck.

 

[fig]

I've set mine on fire in show of support of this movement....uh oh, the output cable was plugged into the....Oh crap my AMP!

 

[Chuck D. Bones]

I got a visual!

If you ask me how to improve a circuit I think is shitty, don't be surprised when I suggest tossing it in the trash.

 

[fig]

Lipstick on a pig eh?

 

[lrgaraujo]

Nice one, as always, Chuck, thanks!
I have a little question about preamp circuits that use OTAs for distortion (like the Peavey and the Lab L5). It looks to me, that the signal being fed to the diff amp is also being fed to the amp bias input. Should it make the distortion more dynamic this way?

 

[Chuck D. Bones]

Good question!

Looking at the Moog L5 Preamp, IC A110 (CA3094) is the distortion stage. Injecting some of the audio signal into the Iabc (amplifier bias current) pin on the OTA causes the OTA to generate even-order harmonics, especially 2nd harmonic. The signal fed to the Iabc pin is high-pass filtered to prevent muddiness. This method does not create compression, so the dynamics are preserved.

It seems odd to me that there is no DC feedback loop to stabilize IC10's operating point. One resistor and one cap would do it. An LM13700 could be used in place of the OOP CA3094.

The Aphex Aural Exciter (type B) does a similar thing, but with some additional trickery to modulate the gain of the OTA.

We can get odd-order harmonics by overdriving the input stage of the OTA, I've seen that in the Spaceman Aphelion. Again, no DC feedback which seems crazy to me.

Yet another circuit to put in the breadboard queue!

Which Peavey amp?

 

[lrgaraujo]

Thanks for the explanation, Chuck.

It's the Peavey Mace. I forgot to include the model, my bad. Apparently, they also use a CA3094 (marked 478 in peavey schematics)

 

[Chuck D. Bones]

I'm looking at the Peavey schematic and there are 478's (CA3094) all over that circuit like a rash. Two make up an LFO. Four make up an OTA phase shifter. Looks like two of them are used as expanders (opposite of compressors).

 

[lrgaraujo]

I'm looking at the Peavey schematic and there are 478's (CA3094) all over that circuit like a rash. Two make up an LFO. Four make up an OTA phase shifter. Looks like two of them are used as expanders (opposite of compressors).

Cool! Do you think, in this case, the expander could be contributing with some distortion? I've heard samples from the preamp and it looks to me as if there is some grit coming from the preamp

 

[Chuck D. Bones]

Sure, but it's because the CA3094's diff input can be overdriven. There is also a diode clipper in there (CR6 & CR7) on the Effects channel. It's also possible that some of the other opamps can be driven to saturation.

 

[lrgaraujo]

Good question!

Looking at the Moog L5 Preamp, IC A110 (CA3094) is the distortion stage. Injecting some of the audio signal into the Iabc (amplifier bias current) pin on the OTA causes the OTA to generate even-order harmonics, especially 2nd harmonic. The signal fed to the Iabc pin is high-pass filtered to prevent muddiness. This method does not create compression, so the dynamics are preserved.

Ressurecting this since my breadboard is empty after a long period, even though my question is not entirely related to OTAs. Is there a discrete approach that could give me a result similar to this 2nd harmonic behavior? I'm guessing it would be a really bare bones OTA of some kind

 

[Chuck D. Bones]

Anything that clips asymmetrically will generate 2nd harmonic. The tricky bit is limiting how much of the higher-order harmonics get generated. Ever wonder why there are so FET-based dirt pedals out there? It's because FETs have a parabolic transfer function. They are good at making low-order harmonics. Not as good as tubes, but better than most other methods when done right.

The short answer is a single JFET amplifier, biased slightly cold so that is goes into cutoff before going into saturation, will make only low-order harmonics when driven lightly.

 

[lrgaraujo]

The short answer is a single JFET amplifier, biased slightly cold so that is goes into cutoff before going into saturation, will make only low-order harmonics when driven lightly.

Great, thanks! I'll have something in the breadboard later today. What exactly are you calling a cold bias?

 

[Chuck D. Bones]

In this context, neutral bias is when the drain voltage is halfway between the source voltage and Vcc. Hot bias is when the drain voltage is below halfway. Cold bias is when the drain voltage is above halfway. Example: Vcc = 9V and Vs = 1V. The halfway point is 5V (the average of 1V & 9V). If the drain is above 5V, the bias is cold. You can try differnt drain voltages and listen to the difference. There will be sweet spot where you can play clean at lower guitar volume settings and as you turn the volume up, the distortion comes on gradually. This assumes that there is a bypass cap across the source resistor. Setting the bias gets a little more complicated if there is no bypass cap. In any case, start with the drain voltage above the halfway point and tweak the bias by ear.

 

[lrgaraujo]

Nice, thanks!