[ANN] Repro-1

[Urs]

Polivoks filter is 12 dB LP/BP and the most f**ked up filter ever put in hardware. I'd be happy to model it, but I currently dont know how.

What's the problem, too computationally expensive or not possible to solve with current approach you guys use in Diva/RePro?

Well, the problem is that we'd need to do a rather thorough model of the K140УД12 OpAmp, maybe the uA776 instead. We can't just measure a tanh-ish non-linearity and use slap it onto a standard filter equation system. I'm not even sure if the distortion fulfills the criteria required for guaranteed convergence of a Newton solver...

We could try a black-box approach instead and model the behaviour for a set of conditions, but then we might not cover its behaviour for the cases we haven't taken into account. Not sure if that's how I'd go about it...

[Urs]

Hello Urs & co.,

I know that the original Pro-1 did not have a sustain pedal input - but I can't imagine being able to play a vst without being able to sustain input via CC64 (ie. sustain pedal on typical midi controllers). Can there be an option somewhere (or modulation control somewhere) to respect CC64? Currently, it doesn't seem like RePro-1 does this.

Thanks,
Kevin L

Yeah, seeing that we're already implementing some sort of MIDI-CV interfacing, why not...?

[EvilDragon]

Yes sustain pedal support is a must!

[Urs]

As for "esoteric" stuff: I took this as "stuff that bloats the feature list but hasn't got anything to do with Pro-Ones". Hence no trigger modes that didn't exist, no alternative routings, added modules whatsoever.

Modeling part differences may sound esoteric, but for me it isn't. It's something that's absolutely real between several Pro-Ones and I think a good emulation of this thing should cover some of the bandwidth out there. It doesn't need to model double triggering keys, but it should at least cover the range of character among differently calibrated/serviced/aged synths.

Hell, we might even add a "Youtube Demo" mode which uses hot mixer input, reverb and delay for a larger-than-life sound

[david.beholder]

Well, the problem is that we'd need to do a rather thorough model of the K140УД12 OpAmp, maybe the uA776 instead. We can't just measure a tanh-ish non-linearity and use slap it onto a standard filter equation system. I'm not even sure if the distortion fulfills the criteria required for guaranteed convergence of a Newton solver...

We could try a black-box approach instead and model the behaviour for a set of conditions, but then we might not cover its behaviour for the cases we haven't taken into account. Not sure if that's how I'd go about it...

Urs please pay attention to previous messages

1. Since all soviet electroics is cloned кр140уд8 series is clone of µA740HC, µA740PC
http://www.volt.ac/A740HC.html <-- but i see only one 50 pF capacitor on block diagram
2. They are still produced and sold in Russia
3. I can dig more info in Russian

[aaron aardvark]

Urs,
The "Youtube Demo" mode sounds like a good idea to me. I typically like a big sound, and often the Pro-One sounds a bit thin compared to say Zebra.

[niektb]

@Aaron aardvark:
Don't forget to put a small detune (~0.10) to one of the oscillator. It gives that bit of a thicker unison sound.

[urosh]

1. Since all soviet electroics is cloned кр140уд8 series is clone of µA740HC, µA740PC
http://www.volt.ac/A740HC.html <-- but i see only one 50 pF capacitor on block diagram
2. They are still produced and sold in Russia
3. I can dig more info in Russian

KR140UD8 is not relevant, it's part of expo current source. Topology of that rather primitive but interesting expo current source is more relevant but not critical. Critical part is KR140UD12, which seems to be clone of uA776 (I've found some, rather new technical documents in Russian that claim that, but it is not certain). Do you know of any source for KR140UD12 or even better KR140UD1208?

[david.beholder]

Critical part is KR140UD12, which seems to be clone of uA776 (I've found some, rather new technical documents in Russian that claim that, but it is not certain). Do you know of any source for KR140UD12 or even better KR140UD1208?

You're right. I thought they are all КР140УД8. So 1208 is µ A776HC or µ A776PC clone.

http://www.chipdip.ru/product/kr140ud1208/ over here but this is russian store

[urosh]

Well, the problem is that we'd need to do a rather thorough model of the K140УД12 OpAmp, maybe the uA776 instead. We can't just measure a tanh-ish non-linearity and use slap it onto a standard filter equation system. I'm not even sure if the distortion fulfills the criteria required for guaranteed convergence of a Newton solver...

We could try a black-box approach instead and model the behaviour for a set of conditions, but then we might not cover its behaviour for the cases we haven't taken into account. Not sure if that's how I'd go about it...

So, I've done some analysis, and it doesn't look promising. I've assumed that uA776 is device to emulate, since internals of 140UD12 are unknown (specs are close enough to assume they are very close siblings).
Input stage is behaving nicely. Local feedback is taking care of biasing, so it behaves pretty much like OTA (only, since there are 4 junctions in total across input, it can take twice as much input compared to traditional diff pair). Simple diff-to-single ended configuration compared to OTA, but kept at constant output voltage, so all in all, you could replace whole input stage with voltage input current output block with single tanh nonlinearity.
Voltage amplification stage and output stage are different story. First, they rebias with change of control current, unlike situation in typical VCF (where only transconductance stage changes). They wont bias properly (when control current changes) without global feedback. So, as far as I can tell, you can't assume bias point of transistors, based on control current, and derive simple(simpler) nonlinear equations with constants that represent control value / cutoff. At best, it seems you need to use (low frequency) Ebers-Moll model for 5 transistors, to model accurately behavior at higher input levels. And this will end up with bunch of variables locked into system of implicit equations, so it might get really really ugly for solving (I once ended up with system of 4 equations for exotic first order filter, and gave up when deriving Jacobian for multivariable Newton method).
Further more, I've red comments that Polivoks VCF selfoscillates, so additional pole per amp is needed. I can't tell for sure, if it's pole from first stage, it might be doable, if it's output stage, well, I guess that's really bad, you would have to use Ebers-Moll with junction capacitances. Also, it's really picky about devices used. I've simulated ti with 2N3904 and 2N3906, and it wont meet the specs (voltage offset, and some other stuff). And it bleeds CV like crazy. And simulation behavior is quite crazy, especially at low control current, especially influence of output stage.
I guess this is really difficult, if not most difficult VCF to accurately emulate.

[knowix]

Was Cytomic modelling the Polyvoks filter, or the Wasp? I can't remember.

[urosh]

http://www.chipdip.ru/product/kr140ud1208/ over here but this is russian store

Thank you very much.

[david.beholder]

http://www.chipdip.ru/product/kr140ud1208/ over here but this is russian store

Thank you very much.

Noprob. Call if you need more help, especially if you in SF Bay Area.

[EvilDragon]

Was Cytomic modelling the Polyvoks filter, or the Wasp? I can't remember.

WASP.

[aaron aardvark]

@Aaron aardvark:
Don't forget to put a small detune (~0.10) to one of the oscillator. It gives that bit of a thicker unison sound.

Yes, I have owned a real Pro-One since 1981 (mine currently could use some service though), and I almost always did what you say (though I used a larger detune amount). Though surprisingly, I recently came up with a real Pro-One bass synth patch that sounds best/fattest (to me) with both oscillators tuned at 0.00.