DIY OTA questions

[11-90-an]

After staring a lot and analyzing OTAs after jonny.reckless posted his PMC, (https://www.diystompboxes.com/smfforum/index.php?topic=124777.0) I came up with this:



The schem on the left is from the CA3080 datasheet and the one from the left is from the PMC. Both have some similarities, like the LTP at the input and the power supply.
However, jonny stated that his design "can only insert about 100mVpp signal in this configuration before distortion becomes a problem so you have to pad the guitar signal down." Which also seems to be the use of the voltage divider.

The portions on the pink parts is (from my limited reckoning) how the makers bias and derive an output from the LTP. On the PMC, it simply utilizes a wilson current mirror, but on the CA3080, it uses... a current mirror in a current mirror?

Now I know that these 2 have great differences and jonny.reckless design one as a "simpler alternative", but still my mind can't be put at rest...

So for my questions:

1. Would you still need the voltage divider network in the PMC if one were to use it as a seperate "module" of some sort?
2. Or would the builder already anticipate that they need to supply the PMC OTA about 100mVpp signal only before the dreaded distortion?
3. Is this the same case as the CA3080?

4. Can somebody explain to me how the pink highlighted section works?

and finally...
5. Could the PMC OTA replace the LM13700 in the Engineer's thumb?
6. Any mods to make this possible?

[jonny.reckless]

So for my questions:
1. Would you still need the voltage divider network in the PMC if one were to use it as a separate "module" of some sort?
2. Or would the builder already anticipate that they need to supply the PMC OTA about 100mVpp signal only before the dreaded distortion?
3. Is this the same case as the CA3080?
4. Can somebody explain to me how the pink highlighted section works?
5. Could the PMC OTA replace the LM13700 in the Engineer's thumb?
6. Any mods to make this possible?

1. It depends on the incoming signal level
2. Assuming they are familiar with the LM13700 or CA3080, I assume yes.
3. Yes
4. They are just current mirrors. To get full voltage swing out of the OTA you need a couple more current mirrors than I used, as the collector of the LTP can't swing below its emitter. I just dealt with limited compliance by setting the DC bias points to use the headroom available. 4Vpp output is more than enough for an electric guitar effect.
5. In theory, yes
6. Why bother? The output compliance of the PMC transconductance stage might work OK with the ET, but you would probably lose output swing, increase distortion, and need a bigger board. The ET uses the OTA in the feedback loop so its self noise is not so much of a problem. A discrete OTA might cause parasitic oscillations in the feedback loop of an opamp unless you are very careful with layout and loop compensation.

Hope that helps 
Jonny

[11-90-an]

Thanks for the input, jonny...

R.G. made this thread (https://www.diystompboxes.com/smfforum/index.php?topic=31109.0) about discrete OTAs and since I didn't see any schem so i made one...



It seems closer to the CA3080, but the biasing stage (pink part) is now 2 current mirrors, and the output uses a standard 2 transistor current mirror...

7. Does this ckt allow the LTP collector to "swing below its emitter"?
8. Does the current mirror at the output give a big difference in sound quality? (i.e. the 2 transistor current mirror in R.G.'s vs the wilson current mirror in the CA3080)
9. Why does one one of the pin's output have to go through another current mirror?

EDIT: added question 9

[11-90-an]

Searched the lm13700 schem:



It seems like they used wilson current mirrors everywhere..

[antonis]

It seems like they used wilson current mirrors everywhere..

It's a residual habit from the times when "T" wasn't included in "OTA"..

[Rob Strand]

It seems like they used wilson current mirrors everywhere..

One of the biggest problems with OTA is feedthrough.   That is, when the control current  (IABC) is changed it creates an output signal (due to IABC).   When you have something like a compressor or limiter, sharp changes in IABC during the attack will show up in the audio as ticks.   If you have a tremolo the ticks would be in sync with the LFO.

So imagine if the current mirrors are not balanced.   The current through from the diff-pair collectors through to the output gets out of balance and you get feedthrough.    The Wilson mirrors have a more predictable current "gain" so it's easier to get a good balance.

[Rob Strand]

Check out these notes,
http://www.openmusiclabs.com/files/otadist.pdf

[11-90-an]

What would happen if one where to do something like this?



10. What are the problems with this?

[Rob Strand]

What would happen if one where to do something like this?

I suspect the currents would go to zero.


The first thing to notice is the tail current in the differential pair halves into each emitter but the two outputs of the top mirrors add together again.

If we call Iin the input to the lower mirror and the output the sum of the two top mirrors then,

Iout = Iin * M * (IC/IE) * M

where M is the current gain of the mirror (= output current of mirror / input current of mirror)
Since IC / IE = alpha


Iout = Iin * M^2 * alpha

We know alpha < 1 and for those mirrors M < 1
So M^2 * alpha < 1

In simple language each stage has a lower output current than it's input.

That means the overall output "6" is less than the input "6".  If you feed one into the other it will spiral to zero.

I haven't considered the Early effect but at least to first order the current should go to zero.

[11-90-an]

R.G. made this thread (https://www.diystompboxes.com/smfforum/index.php?topic=31109.0) about discrete OTAs and since I didn't see any schem so i made one...

Mr. Reckless said before that his design couldn't "swing below the emitter...
How does one solve this problem? Does R.G.'s design do this?

My brain hurts...

[Rob Strand]

Mr. Reckless said before that his design couldn't "swing below the emitter...
How does one solve this problem? Does R.G.'s design do this?

My brain hurts...

It's actually not that hard but you need to understand a few smaller steps:

For any "linear" transistor circuit the collector cannot go below the emitter, typically a few 10's of mV above it.  That's just how transistors work.   That's where the "collector cannot swing below the emitter" comes from.

If the base of a transistor is at some DC voltage, the emitter is about 0.6V below the base voltage.

From those two points the collector cannot swing more than 0.6V below the base.  ie. the base voltage limits the collector swing.

Now Mr. Reckless's PMC design the output voltage comes from the buffer.   The buffer is connected to the collector of the differential pair.    So the output swing is limited by the differential pair collector voltage.   However, the (negative) swing on the differential pair collector is limited by the base voltage.   Notice how the base voltage on the PMC is set to a low value.   That tries to give largest (negative) swing on the differential pair.   The base voltage is chosen just low enough that the current source at the emitters can still operate.

Now, have a look at the LM13700 design.  It's all current mirrors.   The signal from one stage to the next are all currents.    The voltages are all pinned to one or two Vbe's off the supply rails.   There's no voltage swing problems inside the design, only current swing limits set by the differential pair current source.     The only time you see a voltage is at the output terminal.   Here we have transistors which can operate very close to the rails in either direction.

The way the PMC works and biased doesn't lose much swing.  It's about as good as you can do without throwing a lot of transistors at it to make a LM13700 type design.  Like Johnny said,  it works well and it's a fairly simple and economical design.

[jonny.reckless]

Check out these notes,
http://www.openmusiclabs.com/files/otadist.pdf

That open music labs paper is great, do you know who wrote it?
It should have been called "Everything you wanted to know about OTAs but were afraid to ask"

[Rob Strand]

That open music labs paper is great, do you know who wrote it?
It should have been called "Everything you wanted to know about OTAs but were afraid to ask"

Yes, it's sure packed with info. 

No idea who wrote it.   It couldn't get any more anomymous!  Nothing in the text, nothing in the PDF info, nothing on the site:  "My summer vacation, by guest."

http://www.openmusiclabs.com/2015/10/my-summer-vacation-by-guest/trackback/index.html

There's a few other long and similarly styled documents on other topics.

[11-90-an]

Wow... at first i just skimmed through the article just because I thought it was all overly complicated... then Mr. Reckless started praising the paper, saying that it was great for beginners... so I read the article again and I CAN FEEL MY BRAIN'S GEARS TURN. I should have read this sooner.... (i still can't understand the equations, but hey, that's a start..)

Anyway, I came up with this "improved" discrete OTA... i know that some parts have small usage, but if it helps, it helps...

The diode bypass part can be disregarded when your circuit doesn't use a diode bias...



Feedback welcome... 

It should work in the engineer's thumb right? (Are there any problems?)

[Rob Strand]

Feedback welcome... 

It should work in the engineer's thumb right? (Are there any problems?)

It looks like it should just slot straight into a LM13700 design. crap

Keep in mind, in practice transistor matching is significant factor in the performance of an OTA.  Some aspects might look good on paper but problems due to mismatches can mask any goodness.

[jonny.reckless]

Anyway, I came up with this "improved" discrete OTA.

It depends on what you mean by improved 
By adding an "improved" 4 transistor Wilson mirror in the collector circuit, and a 3 transistor Wilson mirror in the emitter circuit, you have reduced the compliance of the long tail pair by quite a bit, it can probably only swing about 4Vpp in total before hitting the limits either way, depending on where you bias the bases. This may or not be OK for your application, but you should at least be aware of the trade-offs when making design decisions.
In the engineer's thumb, the inputs and outputs of the LM13700 sit at approximately the same voltage, around 4.5VDC. This is going to be a problem with this type of OTA, since it can't swing its output below the inputs. You need to change the DC bias conditions around the op amp to make that work.

That 220n should actually be much bigger, maybe a 10uF multilayer ceramic, this is just a rough sketch to illustrate the idea, and is missing a few parts for clarity.

[Rob Strand]

In the engineer's thumb, the inputs and outputs of the LM13700 sit at approximately the same voltage, around 4.5VDC. This is going to be a problem with this type of OTA, since it can't swing its output below the inputs. You need to change the DC bias conditions around the op amp to make that work.

I'm glad someone is paying attention.

[11-90-an]

Thank you soo much Jonny and Rob! 

So if i keep the 3 transistor current mirror instead of the 4 transistor one in the collector circuit, that should help?

I read that a wilson current mirror controlling the emmiter circuit could help in some ways... guess not in this application...

Looks like I'll learn about matching transistors now...

[antonis]

(i still can't understand the equations, but hey, that's a start..)

No need to dive into natural logarithms integration/differentiation troubled waters..
Just focus on formulas calculation final results..
(but some understanding of Shockley's/Ebers-Moll/Early's equations application should be useful although not essential..) 

[11-90-an]

Wait. I just realized... what am I looking for when I match transistors? The Vbe and β right?

Saw this site: http://www.bestsoldering.com/transistor-matching/
Should the power supply really be +/-12...?
I only have one 9v battery for all my pedal needs...