Your mission, should you choose to accept it: Opto-FETs in the audio path

[midwayfair]

... without distortion.

Following mine and Jazznoise's discussion in another thread.

For those who are unaware of the issue: If you stick and otpo-FET like the H11F1 in the audio path of a circuit, you'll get distortion above about 100mA.

I know there are ways to alleviate this. Ray Ring did it in his opto-Squeezer by shunting overvoltage away from the chip:
http://circuitsaladdotcom.files.wordpress.com/2013/03/simple-opto-comp1.gif
and
http://circuitsalad.com/2012/10/17/opto-fet-issue-solved-with-negative-feedback/

CultureJam tried it in his Shoot the Moon tremolo (it's like a shoot the moon) with bad results. I've tried it in several designs with similar failures.

I'm intensely interested in getting these chips to work in tremolo designs. But I would think shunting voltage around the chip would almost assuredly result in very little effect. Cutting signal would result in less signal (and possible a redesign would be in order).

So ... who has ideas for some ways to prevent distortion in these chips?

[alparent]

Sorry that I cannot (as always) offer nothing more then questions.

Why are you so intensely interested in getting these chips to work in tremolo designs?
What would be the advantages?

Thanks for educating me!

[Govmnt_Lacky]

I am probably talking out of my rear end on this but...

Have you tried some sort of buffer set-up before the opto-FET and then a recovery stage after?

[R.G.]

For those who are unaware of the issue: If you stick and otpo-FET like the H11F1 in the audio path of a circuit, you'll get distortion above about 100mA.

Why would you ever want 100ma of signal current?

[garcho]

Typo?

I've read about these in the synth world but never used them. Those peeps just attenuate the signal, but those are synth signals, much bigger. What about attenuation or the diode solution do you want to avoid? They also say they find the distortion pleasing; obviously a moot point in a trem.

[Jazznoise]

I think Midway meant 100mV

A recovery stage is fine, and still leaves the huge question of headroom. If we've to divide our 1V signal down 10 times and bring it back up, that's at least 14dB of noise not including any circuitry we add. And our headroom is still pretty bad . Paralleling stages could be done, but we're only going to get a 3dB improvement for one and a 6db for 4.

I posted one idea I had earlier about using a sort of active feedback stage that would only introduce noise when attenuating, but I've no idea how well it would work in real life:




Another idea is to use a balanced system with two H11F1's on identical signals of inverse polarity and sum them via a balanced amp afterwards. Distortion products would largely cancel, but the signal driving the LED would also have to be inverted on the inverted channel.

[midwayfair]

I think Midway meant 100mV

Yes, thanks

alparent, these chips have several advantages:
-Exceptionally high dark resistance and low light resistance; and low light memory
-Cheaper than vactrols
-Smaller than vactrols
-Easily socketted, and less easily broken
-SMD packages are available, meaning they'll be obtainable for many years.

[jonasx26]

How about sticking the fet between the 'input-R' and inverting input of a inverting op amp. Shouldn't see much voltage across it, right?

[merlinb]

How about sticking the fet between the 'input-R' and inverting input of a inverting op amp. Shouldn't see much voltage across it, right?

+1.

This is a problem with any FET in the signal path, not just optos. Even with 6dB local feedback around the FET, they cannot normally stand more than about 1Vp-p across them before significant distortion. Place the FET in a virtual earth so it sees a reduced signal voltage.

[Jdansti]

Maybe you've thought of this, or maybe it defeats the purpose, but how about keeping the opto-FET out of the audio path and use it to drive or switch something else? 

[Jazznoise]

How about sticking the fet between the 'input-R' and inverting input of a inverting op amp. Shouldn't see much voltage across it, right?

That's much smarter than my idea! Might give this a go once the academic year finished.

For those reading, no, I still haven't gotten my Harmonic Percolator working! 

Jdansti, a problem with that is we want linear resistance across a broad range and with a very small recovery time. 0.5 mS is as slow as I'd like to go if I was building a ring modulator or doing crazy filter FM stuff. On that particular note, I've always been curious about the idea of a Phase Modulation based synth pedal for guitar. But I feel that with most guitar tones, you'd need to be driving it with a sine tone to get anything other than jibberish! But even in situations such as audible filter sweeping, phase shifting or amplitude modulation the even-ness of the change will make or break the effect.

We've all done a filter sweep on an 8 bit rotary encoder, right?

[midwayfair]

How about sticking the fet between the 'input-R' and inverting input of a inverting op amp. Shouldn't see much voltage across it, right?

Can you show an example? I'm having a hard time imagining it (my knowledge of op amps is pitiful -- I'm a transistor guy)

[Jazznoise]

I assume Jonas meant something like this?

The idea being that with half the signal on either side that the output would be 0. The two inputs of an op amp are compared to each other for an output. If they're exactly the same, the output is 0. In the standard guitar pedal we'd set one of the inputs to 4.5 volts so that the signal would be biased around half our power supply.



Just to elaborate on this, at max a 1V signal would be on this at which point a 0.5V would be on either side. Now the majority of the signal itself would be gone, but the distortion products would not. Much simpler layout and lower part count, though.

[jonasx26]

I assume Jonas meant something like this?

The idea being that with half the signal on either side that the output would be 0. The two inputs of an op amp are compared to each other for an output. If they're exactly the same, the output is 0. In the standard guitar pedal we'd set one of the inputs to 4.5 volts so that the signal would be biased around half our power supply.

No. Take your bog-standard inverting op amp amplifier. Then stick the fet in series, between the input resistor and inverting input.
That is: IN > R > FET > Neg. Input. Then add another resistor from op amp output to negative input. (Everything DC-coupled, centered around Vdd-Vss/2.)

The signal will "pass through" the fet "as a current", while the op amp 'virtual earth'-effect means the voltage over the fet will be kept very small.
I've seen this technique used for reducing distortion in analog switches (makes non-linear Ron vs Vsignal matter less)
Works very well for switches being either hard on or hard off. Haven't tried anything in between. Haven't seen any circuits using this approach either.
Probably not a coincidence. Probably very difficult to get good results.That is, difficult beyond fiddly scaling and offsets of CV..

[EATyourGuitar]

cool audio makes a V2162 that is a clone of a ssm2162. I know its not as cheap but it is available in SMT.

[samhay]

I assume Jonas meant something like this?

The idea being that with half the signal on either side that the output would be 0. The two inputs of an op amp are compared to each other for an output. If they're exactly the same, the output is 0. In the standard guitar pedal we'd set one of the inputs to 4.5 volts so that the signal would be biased around half our power supply.

No. Take your bog-standard inverting op amp amplifier. Then stick the fet in series, between the input resistor and inverting input.
That is: IN > R > FET > Neg. Input. Then add another resistor from op amp output to negative input. (Everything DC-coupled, centered around Vdd-Vss/2.)

The signal will "pass through" the fet "as a current", while the op amp 'virtual earth'-effect means the voltage over the fet will be kept very small.
I've seen this technique used for reducing distortion in analog switches (makes non-linear Ron vs Vsignal matter less)
Works very well for switches being either hard on or hard off. Haven't tried anything in between. Haven't seen any circuits using this approach either.
Probably not a coincidence. Probably very difficult to get good results.That is, difficult beyond fiddly scaling and offsets of CV..

Starting thinking about the 'challenge' last night. Couldn't find any opto-FETs in my stash, but have some BJT phototransistors like the 4n35. Jonas' approach is what I came up with too. When the light is on, the gain ->1, when off gain -> 0. Only at the simulation stage thus far. On the plus side, you can do ring modulation, on the minus, looks like you might get some partial half wave rectification of the signal through the optocoupler.

[Lurco]

How about sticking the fet between the 'input-R' and inverting input of a inverting op amp. Shouldn't see much voltage across it, right?

Can you show an example? I'm having a hard time imagining it (my knowledge of op amps is pitiful -- I'm a transistor guy)

Look at the 5th visible picture here: http://www.diystompboxes.com/smfforum/index.php?topic=39145.msg278531#msg278531 ("Variable Attenuator")

[slacker]

Works very well for switches being either hard on or hard off. Haven't tried anything in between. Haven't seen any circuits using this approach either.
Probably not a coincidence. Probably very difficult to get good results.That is, difficult beyond fiddly scaling and offsets of CV..

I think this only really works for switches, my understanding is the the input resistor and the FET will act like a voltage divider, with the signal at one end and ground at the other. When the FET is on (switch closed) its resistance is tiny compared to the input resistor so there's very little voltage across it. When it's off (switch open) it's resistance is huge compared to the opamp feedback resistor so the opamp gives a large negative gain muting the signal. The FET's resistance is also huge compared to the input resistor so most of the voltage drop will be across it, as there's no output from the opamp any distortion this causes is irrelevant.
If you use it as a variable resistance then you could still get too large a voltage drop across it, say you had it set so the FET's resistance was equal to the input resistor, you'd get half the signal voltage across it.

[merlinb]

You should read this, if you haven't already. See fig. 12 especially.
http://www.datasheetcatalog.org/datasheet/vishay/70598.pdf

[Jazznoise]

Not too dig this up, but is Fig. 12 not the same idea as using it as an input attenuator?

Would the distortion products cancel for some reason?