OTAs: What, why and what for?

[KarenColumbo]

I don't get them. I "know" what an op-amp does. I think. I absolutely don't know what an OTA does. And why I would need one, if I have an op-amp. Wikipedia don't help at all. All those schematics I looked at don't tell me their secrets. (I think the problem is that I still need to get to grips with AMPERAGE vs. VOLTAGE. I can't seem to "translate" this so my non-technical brain can compute it)
Can somebody explain this to me?
I'm stupid.

[Rob Strand]

Can somebody explain this to me?
I'm stupid

I seriously doubt that.

The reason we know opamps is because they are familiar.    They have familiar patterns so when we look at a circuit with a whole heap of junk around an opamp we have an idea how to break it down.

OTAs don't come up much so we aren't familiar with them.  When we see a whole heap of junk around an OTA it's not so easy.   To tell you the truth I don't know a lot of OTA patterns.  I've got a away of breaking it down but it's a very mechanical process.

The first aspect is the current output:

If you look at a simple one transistor BJT amplifier with no emitter resistor you will know immediately what it does.   Guess what, the transistor itself can be viewed as voltage in and current out.    When that current flows into the collector resistor the current gets converted to voltage.  From an AC point of view vout = - Rc * ic  where vout is the ac output voltage, ic is the ac collector current and Rc is the collector resistor.  The AC current goes into the collector and that's why you the minus sign.  (FYI: The instantaneous collector voltage is  Vc = Vcc - Ic * Rc and you can see that when Ic increases the collector voltage Vc drops and that's like a negative output swing.)    The main point here is not the calculations but the fact the output current is converted to a voltage by putting a resistor on the output.

If you look at the OTA on the dynacomp  the output drives the 150k resistor.   That's doing the same thing as the BJT amplifier.  The following buffer makes sure that resistor doesn't get loaded down.

The "gain" of an OTA is the output current divided by the input voltage.   That type of ratio is called transconductance.    If you multiply the transconductance by the resistor load you get the voltage gain.   A single transistor does in fact have a transconductance; see wikipedia.   The T in OTA stands for transconductance.

One difference between an opamp and an OTA is the opamp has a very large gain and we rely on feedback to make the gain a more useable figure.     The OTA is a relative low gain.    We may use the OTA with feedback or without feedback.   In the case of a dynacomp the OTA is used with without feedback in the normal sense of feeding back the AC signal.  (What's seen as feedback is the control signal, this is  DC and a completely different feedback to ac feedback.)

The gain of an OTA is programmable by the control pin, IABC.   The more current down the control pin the higher the transconductance.    That's how the dynacomp works as a compressor.

With things like phasers it's a bit tricker.   The load is no longer a resistor, it's capacitor.  And to make things more difficult a cap not connected to ground.   The OTA gain is controlled by the sweep.   It acts like an RC filter where the R is being adjusted by the OTA current.   To see how that works needs circuit analysis and maths.   I doubt anyone can "see" how it works without formally analyzing it in some way.  [I actually posted a long winded mathematical breakdown of this a while back.]

The OTA has a maximum input voltage otherwise they overload, sort of like clipping some diodes.   The voltage is around 10mV to 25mV for low distortion but there's a few tricks to get that up higher to 200mV.   If you look at the dynacomp you can see a divider on the input.  All that does is drop the level down to prevent overload.  It also reduces the gain, since it's a divider.

The last thing is the input is differential like an opamp.  The input voltage appears between the + and - inputs.

That's the basic idea.    You might be able find the old OTA article by Ray Marston on-line.   It might give you a few more tips.   There's a few example of oscillators etc.



Here's one version,
pt1
https://www.nutsvolts.com/uploads/magazine_downloads/11/April%202003%20Ray%20Marston%20-%20Understanding%20and%20Using%20OTA%20Op-Amps.pdf
pt2
http://www.nutsvolts.com/uploads/magazine_downloads/11/May%202003%20Ray%20Marston%20-%20Understanding%20And%20Using%20OTA%20OP-Amps.pdf

[KarenColumbo]

Can somebody explain this to me?
I'm stupid

I seriously doubt that.

The reason we know opamps is because they are familiar.    They have familiar patterns so when we look at a circuit with a whole heap of junk around an opamp we have an idea how to break it down.

OTAs don't come up much so we aren't familiar with them.  When we see a whole heap of junk around an OTA it's not so easy.   To tell you the truth I don't know a lot of OTA patterns.  I've got a away of breaking it down but it's a very mechanical process.

The first aspect is the current output:

If you look at a simple one transistor BJT amplifier with no emitter resistor you will know immediately what it does.   Guess what, the transistor itself can be viewed as voltage in and current out.    When that current flows into the collector resistor the current gets converted to voltage.  From an AC point of view vout = - Rc * ic  where vout is the ac output voltage, ic is the ac collector current and Rc is the collector resistor.  The AC current goes into the collector and that's why you the minus sign.  (FYI: The instantaneous collector voltage is  Vc = Vcc - Ic * Rc and you can see that when Ic increases the collector voltage Vc drops and that's like a negative output swing.)    The main point here is not the calculations but the fact the output current is converted to a voltage by putting a resistor on the output.

If you look at the OTA on the dynacomp  the output drives the 150k resistor.   That's doing the same thing as the BJT amplifier.  The following buffer makes sure that resistor doesn't get loaded down.

The "gain" of an OTA is the output current divided by the input voltage.   That type of ratio is called transconductance.    If you multiply the transconductance by the resistor load you get the voltage gain.   A single transistor does in fact have a transconductance; see wikipedia.   The T in OTA stands for transconductance.

One difference between an opamp and an OTA is the opamp has a very large gain and we rely on feedback to make the gain a more useable figure.     The OTA is a relative low gain.    We may use the OTA with feedback or without feedback.   In the case of a dynacomp the OTA is used with feedback in the normal sense of feeding back the AC signal.  (What's seen as feedback is the control signal, this is  DC and a completely different feedback to ac feedback.)

The gain of an OTA is programmable by the control pin, IABC.   The more current down the control pin the higher the transconductance.    That's how the dynacomp works as a compressor.

With things like phasers it's a bit tricker.   The load is no longer a resistor, it's capacitor.  And to make things more difficult a cap not connected to ground.   The OTA gain is controlled by the sweep.   It acts like an RC filter where the R is being adjusted by the OTA current.   To see how that works needs circuit analysis and maths.   I doubt anyone can "see" how it works without formally analyzing it in some way.  [I actually posted a long winded mathematical breakdown of this a while back.]

The OTA has a maximum input voltage otherwise they overload, sort of like clipping some diodes.   The voltage is around 10mV to 25mV for low distortion but there's a few tricks to get that up higher to 200mV.   If you look at the dynacomp you can see a divider on the input.  All that does is drop the level down to prevent overload.  It also reduces the gain, since it's a divider.

The last thing is the input is differential like an opamp.  The input voltage appears between the + and - inputs.

That's the basic idea.    You might be able find the old OTA article by Ray Marston on-line.   It might give you a few more tips.   There's a few example of oscillators etc.



Here's one version,
pt1
https://www.nutsvolts.com/uploads/magazine_downloads/11/April%202003%20Ray%20Marston%20-%20Understanding%20and%20Using%20OTA%20Op-Amps.pdf
pt2
http://www.nutsvolts.com/uploads/magazine_downloads/11/May%202003%20Ray%20Marston%20-%20Understanding%20And%20Using%20OTA%20OP-Amps.pdf

Whoa. Thanks a heap for this!!!! I go meditate on my breadboard!

[anotherjim]

I think the Iabc control is probably the easiest part to understand, or at least figure out what it's for.

The part I find difficult is the choosing of resistor values. I think you want the input bias currents to match and then the signal modulates one of the input currents depending on if you want inverting or non-inverting output. This creates a difference voltage between + and -inputs. Iabc then determines the magnitude of the output current, which then needs to be converted to a voltage.

Most explanation is dual supply, but ground becomes Vref and -Vs becomes ground for single supply.

The trimmer setting the input bias currents are usually 2 equal resistors connected to vref. Those resistors values are scaled to suit supply voltage and fix bias current. I don't know what the ideal bias current is. Some designs look very different, but they all I think let the current of a driven input get modulated by input signal. Have a look for the EDP Wasp synth filter for a very different looking, but same principle scheme.

Rin is larger than the bias resistors because its easy to overdrive the input. I don't know how to choose it.

RL is turning the output current into voltage. I don't know how you size that.

The middle input is a feed for "linearizing diodes". The CA3080 doesn't have this. It is often left disconnected. I don't know how you would choose the feed resistor for it.

So I just steal designs 
At least I think I know enough to go bug hunting in them!

[Rob Strand]

n the case of a dynacomp the OTA is used with without feedback in the normal sense of feeding back the AC signal

I fixed this stuff up.

[deadastronaut]

excellent timing.. i was looking at an lm13700 phaser earlier today...

[ElectricDruid]

In simple terms, an OTA is just like an op-amp, except that the gain is controlled by a current (the Iabc input).

Other than that, the overall idea is exactly the same as an op-amp. The other differences are minor:

1) OTA's have a pathetic input level before distortion, so you have to scale everything down through a voltage divider before you feed it in.
2) OTA's have a current output. Stick an I-to-V stage on the end of it if you don't have one later in the circuit already and off you go.

You can build all the same structures with OTAs that you can with ordinary op-amps - inverting, non-inv, mixers, all-pass stages, whatever. The thing is that it doesn't make sense to do that *unless* there's some benefit to be gained by having the variable gain feature. Hence OTAs get used instead for circuits where voltage-control (ok, voltage turned into current-control) is crucial, like VCAs, VCFs, VCOs, etc etc.

One of the few applications I remember seeing for an OTA that *didn't* use the variable gain was one of the Moog Filter schematics, which uses it as a differential amp, and takes advantage of the extremely high input impedance and common-mode noise reduction. Doing the same thing with ordinary op-amps needs a "instrumentation" differential amp design with three op-amps (which appears in a different Moog filter schematic, iirc).

[Mark Hammer]

One of the great things about OTA-based phasers is that the individual stages can be easily configured to other types of filters, simply by moving one end of the cap in each stage around.

The venerable SSM2040 chip was essentially four OTAs on a chip, and used for many "classic" 4-pole lowpass filters, but also various phaser projects showing up in magazines.  The late John Blacet realized that the easy reconfiguration of the SSM2040 stages to form other filter types could make for an interesting module, giving rise to his Phasefilter ( https://hammer.ampage.org/files/Device1-6.PDF ).  Some years later, while looking at a schematic for the EHX Small Stone and black Ross Phaser (a Small Stone derivative), I was reminded of the 2040 and realized those phasers could be easily modded to do the same thing as the Phasefilter.  And by gum, it worked.  I gave one I had made to Joey Landreth, when he passed through town in Spring 2019.  He had some nice things to say when he tried it out, but I doubt it'll see the light of day on any recordings or gigs.

One of the both good and bad aspects of OTAs is that they can easily clip (CA3080s most of all).  Indeed, Craig Anderton uses one in the voltage-controlled distortion module he designed ( https://hammer.ampage.org/files/Device1-8.PDF ).  The other good/bad feature is that they respond instantaneously.  If the current source driving it is an envelope-follower, it had better be a damn good one or else every last drop of envelope ripple WILL be audible.  In that respect, LDRs/vactrols can often be preferable as control elements because their response is a little sluggish.

[iainpunk]

i had a huge feeling of confusion about OTA's and the American symbol really threw me off (especially the double circle current source and its odd placing on the gain stage output, while its actually closer to the input.), but after having a look at the INTERNAL schematic of the ca3080 the thing revealed its secret... Iabc is just the bias current of the long tailed pair, thus dictating the gain. the rest are just fancy current mirrors.

cheers

[11-90-an]

Here's a thread about discrete OTAs, i learned a lot from here...

https://www.diystompboxes.com/smfforum/index.php?topic=125100.msg1189419#msg1189419

[teemuk]

One of the both good and bad aspects of OTAs is that they can easily clip (CA3080s most of all).

Yes. As previously stated, their linear input signal range is very small, just few millivolts. Past that they overdrive and clip.

But since OTA is basically a differential amp - and more than often operated open loop - they don't brickwall but have very graceful and soft transition to clipping. They sort of "gain compress" smoothly. There are many distortion effects and amps that exploit this soft clipping characteristic.

In linear applications such non-linearity is - of course - a severe drawback so some of the OTA's feature "current compensation diodes" in input that sort of pre-distort the signal in opposite manner than the inherent gain compression. Two distortions thus cancel themselves. (Think about expanding the signal before compressing it). The diodes can linearize the OTA a great deal, but as a tradeoff the OTA looses the soft clipping characteristic. In many chips the current compensation diodes can be toggled on or off depending on what voltage you reference them to.

[DrAlx]

With things like phasers it's a bit tricker.   The load is no longer a resistor, it's capacitor.  And to make things more difficult a cap not connected to ground.   The OTA gain is controlled by the sweep.   It acts like an RC filter where the R is being adjusted by the OTA current.   To see how that works needs circuit analysis and maths.   I doubt anyone can "see" how it works without formally analyzing it in some way.  [I actually posted a long winded mathematical breakdown of this a while back.]

Rob do you have a link for what you posted?  I tried to work this out myself a while back but I didn't get a nice expression, or at least one that didn't make me think I made a mistake somewhere.  I got this odd transfer function

H(s)  = ( s C Reff - 1 - Reff /Rload ) / (s C Reff  + 1)

with the variable resistance of the all pass (Reff) given by

    Reff = (2r + R) / ( g * r )

where r and R are the "scale down" resistors at the OTA input and g is the transconductance.
So I get something that is only an approximation to an all-pass filter and requires the load resistor on the Darlington emitter (Rload) to be much larger than Reff. 
Using TriVibe circuit as an example, minimum transconductance for LM13700 corresponds to (1/g) of about 150 ohms,
and R/r could be something like 23, so Reff could be as large 3k5 which is very close to the 4k7 buffer resistor used on the Darlington.
So I think I must have made a mistake somewhere.

[ElectricDruid]

The diodes can linearize the OTA a great deal, but as a tradeoff the OTA looses the soft clipping characteristic. In many chips the current compensation diodes can be toggled on or off depending on what voltage you reference them to.

This would make a great "soft / hard clipping" switch. I bet it's been done somewhere.

[KarenColumbo]

As the universe would have it I just discovered 3 LM13700s in a box where all those have-been breadboarded ICs end up sooner or later. And 20 more are on the way from all over Europe. Gotte clear some breadboard space now and get to work!

[Rob Strand]

This would make a great "soft / hard clipping" switch. I bet it's been done somewhere.

IIRC, the Peavey Standard MKIII (the 260D IC version from the late 70's) used an OTA as a "soft" clipper.
They then fed the rectified input signal to the OTA control, essentially as an expander to *increase* the output level
and offset the effect of the clipping.

It was an interesting idea but didn't seem to achieve much in practice.

As far as I rememeber MKIII Bass amp had a similar preamp but didn't have that part of the circuit.

I don't think the Peavey Standard MKIII schematic is around for the IC version.

The fact Peavey had their own part numbers made that part of the circuit quite mysterious.   Finally it clicked it was a CA3094.

[Frank_NH]

Of course, there's the Lab Series amps which also used an OTA for distortion.  Bob Moog had a patent on this idea, which makes for some interesting reading:

http://www.muzique.com/misc/patent47.pdf

I own a Lab Series L5 amp and the OTA distortion works really well in that amp.

[PRR]

...Peavey had their own part numbers made that part of the circuit quite mysterious.   Finally it clicked it was a CA3094.

Some of those funny-number parts in Peavey were Approved Vendor, Approved Lot, or Tested. Not special parts, just the junk and rejects sorted-out.

[teemuk]

I recall there was OTA distortion at least in one gen. Of Peavey VT series amps. OTA overdrive is also "trademark circuit" of Pearce, and thus appears in Lab Series amps, all three generations of Pearce amps and in 1990's A.R.T. amps, preamps and multieffects.

Interesting OTA-related distortion concepts are also found from
- Traynor Dynagain series: No overdriven OTAs. Distortion is achieved with generic asymmetric Zener  clipping diodes in feedback loop, but the distortion stage's gain is also dynamically controlled with OTA. Gain decreases with sustained overdrive.
- Fender Rumble V3 series: No overdriven OTAs but generic VCA stages configured to distort. OTA's gain is controlled by feeding full wave rectified input signal to Iabc input. As result, gain decreases when input amplitude increases, which results to logarithmic transfer curve. (Soft clipping due to instantaneous gain compression). Transfer curves of different stages are shaped by varying gain ratio and (a)symmetry of the rectified signal. (Soft clipping) power amp's OTA also has additional input for the archetypal Fender DeltaComp side chain.

[Rob Strand]

I recall there was OTA distortion at least in one gen. Of Peavey VT series amps.

https://www.thetubestore.com/lib/thetubestore/schematics/Peavey/Peavey-Mace-Deuce-VT-Schematic.pdf

That output circuit looks extremely close to the Peavey Standard MKIII.    I can't vouch for every component value but it's the same idea.   The rest of the preamp is similar as well.  Except no phaser, and one channel was a Baxandall + parametric mid.  That part was close to the MKIII Bass amp.

[KarenColumbo]

I sure am willing to try a OTA distortion/overdrive But I'd feel a bit violated - those 13700s sure don't come as cheap as your average op-amp.