Our answer to modular polyphony: Supercritical Synthesizers

[cioaudio]

he product acronym is unfortunate, but the first post does describe it as a digitally controlled ocillator (singular). I'm hoping it comprises 16 digitally controlled oscillators.

I was really hoping for an answer from the manufacturers. The original posting is vague and makes some contradictory claims, like the "apologies to aliasing fans" thing, which if literally and absolutely true is not compatible with the temporal aliasing of traditional DCOs - so I took that as marketing hype rather than a literal technical description. Maybe they just mean the controller's clock speed is fast enough that it's not a problem. The way it's worded almost makes it sound like they are saying there's no aliasing specifically because the module is made in Finland, which is ridiculous - so what I get from it is that the wording is not careful and should not be taken too seriously. Your guess is as good as mine or Timmy's when we attempt to read between the lines of the advertisement, but without the specific technical answers that their engineer is supposed to be eager to give, we're just shooting in the dark.

The thing is, 16 analog integrators, 16 precise current drivers for them, and 16 reset circuits, plus the entire digital side, is a lot to pack into a 12HP 25mm-deep Eurorack module. As a rough guess I think they need something like three op amps per oscillator minimum, and that makes 48; 12 standard quad op amp packages. It's surely possible with a high-density multi-layer PCB, but I'd really like to see a good photo of that PCB. Even just stuff like doing the power distribution well, to prevent crosstalk among the oscillators and noise coming over from the digital side, is impressive when we're talking about that level of density on a board. If I'm understanding the ad copy for the expander right, it's adding a further 16 VCAs (so, what's that, four 2164 packages or similar, plus the driver and support circuitry for those, as well as all the other things the expander does?) in 8HP and that's impressive density, too.

16 DCOs only requires 16 integrators, 16 digital lines to reset them, a 16 channel DAC to provide the signal to be integrated and maybe a few other minor bits (an ARM MCU). These oscillators do NOT alias, they merely have discretisation of possible frequencies and with high resolution timers this temporal discretisation is not likely to be a problem. I know for a fact you can pack 12 such oscillators in a very small module so 16 is feasible. A DCO swaps the frequency error of a VCO for amplitude error and this effectively means you may end up with separate control of amplitude and frequency in order to be able to compensate the amplitude error. I'm not aware of any reason why you couldn't use this amplitude control as a VCA and it doesn't need any more parts because it just varies the value to be integrated.

But I agree that the initial post is a bit too vague and we're all just guessing so I too look forward to more details in the coming weeks. Really just wanted to point out that 16 real digitally controlled oscillators isn't perhaps as hard as you might think

[zeroseven]

Oh wow.

Will definitely be keeping an eye on this

[Sandrine]

These sound & look really great! Can't wait to see more as I love MIDI control and DCO's yummy!
The dual polarity for MIDI TRS is totally required, I did that on my Reflex LiveLoop Flash-8 expansion but then I ran into one that uses the sleeve and tip (I think the BSP is that) so now there needs to be 3 jumpers ugg
Most are tip-ring though thankfully.

[ckwjr]

loved the demo video. The module and expansion look fantastic.

[hzzzu]

Hey hey! Back to work after a busy launch weekend, thank you all for your replies. Timo will get back on the technical side of things shortly.

Meanwhile the prices in the EU are now public. 16 Voice Demon Core Oscillator: 490€, Demon Core Expander: 280€ incl. VAT. US prices will be similar and be announced soon.

The DCO and Expander will be distributed by Alex4.

[tisaalho]

Hi everybody, and great to hear your feedback on our first pair of modules!

I'm Timo, and I do the engineering, coding etc in Supercritical Synthesizers. The whole thing got started as my DIY synth electronics hobby started to get a bit out of hand, and here we are now

A lot of good questions! Let me take those one at a time, in separate posts to avoid one ultralong post to rule them all.

It is described as a DCO, thus the waveforms are in fact created using a ramp or triangle core analogue circuit, not via a DAC - see https://en.wikipedia.org/wiki/Digitally ... oscillator

This is pretty close! And indeed, the ambiguity with the product name acronym as Digitally Controlled Oscillator is intentional

Basically, it's an analog ramp generator with a digital reset driven from an MCU, just like a Juno DCO, except that our frequency resolution is between ten to one hundred times better than the Juno's (depending on which range setting on the Juno you're comparing to). However, we played a little trick with the reset circuit allowing us to combine much of the analog electronics from multiple oscillators into one, and that's why we can fit it into 12HP. There is a cost though: that means that there are no individual outputs. That would be a different module with different tradeoffs.

The core is then not quite a saw core or a triangle core, but something a bit more powerful... a Demon Core (so there, now there's at least a few options to choose from when interpreting the name)! You can also check out how the movie cameras in Pratchett's Moving Pictures work

So, from a sound quality / psychoacoustics perspective, the end result is:

1. no aliasing, since we round the period in clock ticks to an integer
2. because of the above, frequency resolution is quantized, however, the master clock is 72MHz, so the the pitch resolution is 0.05 cents at 2KHz, which is good enough to do things like very subtle pitch jitter etc if desired (and you hear that in the demos, at least in my own use the Core Stability control is usually a bit below halfway by default, and more if I want to sound like a dodgy vintage VCO poly)
3. No digital quantization, i.e. the ramp is true analog

So basically, with the Core Stability at zero, it sounds like DCO synths such as the Junos, but due to the better pitch resolution we can model the drift and jitter of a VCO, do subtle detuning etc etc.

[tisaalho]

Next topic, the amplitude envelopes:

... A DCO swaps the frequency error of a VCO for amplitude error and this effectively means you may end up with separate control of amplitude and frequency in order to be able to compensate the amplitude error. I'm not aware of any reason why you couldn't use this amplitude control as a VCA and it doesn't need any more parts because it just varies the value to be integrated...

This is more or less correct, there's no VCAs but rather the amplitude can be controlled per voice when generating the waveform, and that's where the envelopes are applied.

[hzzzu]

The core is then not quite a saw core or a triangle core, but something a bit more powerful... a Demon Core (so there, now there's at least a few options to choose from when interpreting the name)! You can also check out how the movie cameras in Pratchett's Moving Pictures work

Wrong Moving Pictures. This is the correct and only Moving Pictures.

[kay_k]

..
If it's not the only MIDI controlled thing in your rack, how would you set it up? Will there be MIDI through or something, or can you set it up with an Expert Sleepers FH-2 somehow?

TIA!

sorry for the off topic advertisement of my shit but there is an answer to this:
https://www.modulargrid.net/e/lpzw-modu ... -midi-thru

[tisaalho]

If it's not the only MIDI controlled thing in your rack, how would you set it up? Will there be MIDI through or something, or can you set it up with an Expert Sleepers FH-2 somehow?

The Expander has a MIDI thru on TRS, with polarity selectable via a pair of jumpers. That's exactly because we anticipated that you'll often want to feed the same MIDI to control other things too, and it would be annoying to split the MIDI externally.

If i'm understanding the voice allocating possibilities correctly:

-Over the midi In/thru jack, each of the 16 voices can be played independantly via midi.
-And with the expander attached, you have 5 V/oct inputs adressable over CV. so you could make any possible division of the 16 voices, over those 5 V/oct inputs? Or are those all the ones you've mentioned already (8x2, 5x3, 4x4, 16x1)?
-What happens when you control the voices over Midi and over V/oct inputs at the same time? is this possible?

Correct on the first point. The second and third are related, so let me clarify how that works:

1. The main module V/Oct input always shifts everything up or down (it's bipolar, +-5V range), whether or not the expander is connected
2. When the expander is not connected, or the expander is set to Stream -mode, the full number of voices as set by the current polyphony plays all the time.
3. When MIDI notes are played, they're allocated in round-robin schedule to the voices, as determined by the polyphony setting. If the expander is set to Gated -mode, MIDI notes gate the envelopes for each voice
4. The expander CV inputs always shift the pitch of voices 1 - 4
5. You can divide the 16 voices almost as you like, i.e. you can do for example (poly x stack) 7x2, or 3x4, or whatever. As a UI simplification though some combinations like 1x14 are skipped simply because there not much difference in the sound for 1x13 and 1x14 for example, and it gets tedious to click through all the combinations.
6. The possible stack depths are 1, 2, 3, 4, 5, 8, 12, 16, paired with any polyphony from the list below such that poly x depth doesn't exceed 16.
7. The possible polyphonies are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 16, and again paired with any stack depth from the above list such that the total does not exceed 16. There's a bit more choices for the polyphony, since sometimes you might want to have a very specific number of voices playing in the Stream -mode to play chords of specific sizes

So just a few use cases for this (Heikki might have a few more ideas, and I'm sure there will be tons more once people start using the modules!):

The main module alone: the 1V/Oct input works as usual in an oscillator

The main module + expander, using only MIDI, with envelopes: if none of the V/Oct CV's are connected, this works like an ordinary MIDI polysynth, with MIDI triggering the voices and their envelopes. This would also be your basis for a string synth -like paraphonic setup, where you throw in an external filter to process the sound further. You'll probably want to use the Gate / Trig Out on the expander to gate an external envelope for the filter.

The main module + expander, using only the expander CV's: Here you'd set the polyphony to 4, switch the expander to Stream -mode and plug in the four CV's on the expander. You then have four voices playing all the time, which you can articulate with an external VCA, filter etc. Again, the Gate / Trig Out will probably come in handy.

Chord memory: Set the expander to Stream -mode, play a single chord over MIDI, and you get chord memory with MIDI determining the chord and the main module CV shifting that chord up and down.

[tisaalho]

You can also check out how the movie cameras in Pratchett's Moving Pictures work

Wrong Moving Pictures. This is the correct and only Moving Pictures.

Hmm, how about reading Moving Pictures while listening to Moving Pictures, would that be ok?

[tisaalho]

Anyway, let's continue with the CV question:

So what happens if you have, say, polyphony of 5, and you play a Cmaj9 chord over MIDI. If none of the CV's are attached, then that is exactly what you get. Now, say you attach all four expander CV's, with say +1V, -1V, +2V and +3V (I'm picking easy numbers here, of course you'd usually use something other than full octaves). Then the first voice gets shifted up by an octave, next down by an octave, the third up by two octaves, the fourth up by three octaves, whereas the fifth one stays put (since there's no CV input for it). Which of the voices of the chord are the first, second etc? That would depend on the order you played the notes in, and the state of the voice allocator before playing the notes (which you can reset from the front panel by toggling the stream/gated -mode).

Then, if you further insert a CV, let's say -1V to keep things easy, to the main module V/oct input, what happens? That shifts everything down by one octave, i.e. you get the first voice is now at the same pitch it was before, the next is down by two octaves, the third up by one octave, the fourth up by two octaves, and the fifth one down by one octave. These are relative to what the MIDI notes would do without any CV's.

So, I hope these answers clear things up a little, and don't hesitate to ask more, we'll be happy to answer!

[mskala]

However, we played a little trick with the reset circuit allowing us to combine much of the analog electronics from multiple oscillators into one, and that's why we can fit it into 12HP. There is a cost though: that means that there are no individual outputs. That would be a different module with different tradeoffs.

The core is then not quite a saw core or a triangle core, but something a bit more powerful... a Demon Core (so there, now there's at least a few options to choose from when interpreting the name)! You can also check out how the movie cameras in Pratchett's Moving Pictures work :hihi:

If how it works is a secret you don't want to reveal, that is your prerogative. But the above is ad copy, not a technical description.

[cioaudio]

The main module + expander, using only the expander CV's: Here you'd set the polyphony to 4, switch the expander to Stream -mode and plug in the four CV's on the expander. You then have four voices playing all the time, which you can articulate with an external VCA, filter etc. Again, the Gate / Trig Out will probably come in handy.

This appears to be exactly what folk with ACL Sinfonions ordered will require for the chord section. Excellent.

Great to have you here.

[flo]

If it's not the only MIDI controlled thing in your rack, how would you set it up? Will there be MIDI through or something, or can you set it up with an Expert Sleepers FH-2 somehow?

The Expander has a MIDI thru on TRS, with polarity selectable via a pair of jumpers. That's exactly because we anticipated that you'll often want to feed the same MIDI to control other things too, and it would be annoying to split the MIDI externally.

I'd like to see MIDI any connection on the rear with a 3p KK header (on any module with MIDI, actually). Then everyone can wire that to whatever connector they like in whatever polarity is needed (which certainly wouldn't be a 3.5mm TRS jack in my case), on the front or the rear of the modular... No one does that though except Graham Hinton How does your module connect to the expander?

Cheers

[Dstepatl]

This looks interesting, I am looking forward to hearing more about it in the future.

[Carl Licroy]

The main module + expander, using only the expander CV's: Here you'd set the polyphony to 4, switch the expander to Stream -mode and plug in the four CV's on the expander. You then have four voices playing all the time, which you can articulate with an external VCA, filter etc. Again, the Gate / Trig Out will probably come in handy.

This appears to be exactly what folk with ACL Sinfonions ordered will require for the chord section. Excellent.

Great to have you here.

Same with Instruo Harmonaig

[mdoudoroff]

This appears to be exactly what folk with ACL Sinfonions ordered will require for the chord section. Excellent.

Same with Instruo Harmonaig

Main caveat is that it’s all mixed down to one output. I started a list of such high density oscillators here.

[chrisj]

So, from a sound quality / psychoacoustics perspective, the end result is:

1. no aliasing, since we round the period in clock ticks to an integer
2. because of the above, frequency resolution is quantized, however, the master clock is 72MHz, so the the pitch resolution is 0.05 cents at 2KHz, which is good enough to do things like very subtle pitch jitter etc if desired (and you hear that in the demos, at least in my own use the Core Stability control is usually a bit below halfway by default, and more if I want to sound like a dodgy vintage VCO poly)
3. No digital quantization, i.e. the ramp is true analog

So basically, with the Core Stability at zero, it sounds like DCO synths such as the Junos, but due to the better pitch resolution we can model the drift and jitter of a VCO, do subtle detuning etc etc.

Works for me. I'm quite fond of the analog rawness of my old Farfisa Commander (currently parked as I was overusing it) and that uses a top octave divider to produce oscillators, just in a fixed way rather than a configurable way.

I'd be interested to hear demos using bright raw-output chords in a high register: far from using this to find fault, if the pitch quantization gets sketchier but stays not-aliased you'll probably get a distinctive and desirable sound out of the lil' bugger. Already sounds nice to me

[JES]

So what happens if you have, say, polyphony of 5, and you play a Cmaj9 chord over MIDI.

Thanks for patiently answering so many questions. Let me re-ask one you may have missed. Any planned suppor for MPE? Or, how would the module handle MPE?

[tisaalho]

Thanks for patiently answering so many questions. Let me re-ask one you may have missed. Any planned suppor for MPE? Or, how would the module handle MPE?

Of course, sorry, I meant to answer but that one slipped my mind!

So the initial version of firmware will not have MPE support. At the moment, we don't have explicit plans to support that in a firmware update either, but on the other hand we don't want to exclude that.

What would you like to control with MPE? Per note pitch and volume are obvious targets, on the other hand Spread and Core Stability might give some interesting opportunities for expression

Hmm... maybe implementing MPE would be the perfect excuse for me to get myself a Linnstrument/Seaboard etc

[kay_k]

I'd like to see MIDI any connection on the rear with a 3p KK header (on any module with MIDI, actually). Then everyone can wire that to whatever connector they like in whatever polarity is needed (which certainly wouldn't be a 3.5mm TRS jack in my case), on the front or the rear of the modular... No one does that though except Graham Hinton How does your module connect to the expander?

Cheers

I'm putting MIDI inputs and outs/thru on the back of my modules (5V serial signal, not current loops) and I am planning an internal buffered clock & midi thru board too.
Also: Pamela's New Workout has a internal out and Mutable Instruments Grids has a MIDI input for clock (I can also say Grids works with my modules Out ports)

In the moment there is no standard for something like this, so single pins with jumper cables is the most probable future for such a thing. I personalty don't like that people are putting such things on the CV/GATE lines of the standard connector cos it's going to confuse people.

Sorry for being off topic!

[dubonaire]

In Ringo's defence, I've got a PhD and write about the history of technology and didn't get the reference either. But I just looked it up, so thanks everyone.

This is thread drift but, anyhow...

I find the Slotin criticality incident one of the most fascinating events in the history of 20th century technology. Not least because Los Alamos was a bunch of gifted scientists creating weapons of mass destruction that they didn't even fully understand. It's also an amazing case study about human behaviour and safety. It blows my mind that Slotin was levering the plutonium sphere with a clumsy screwdriver. Given the time it occurred, I'd hazard a guess that a lot of those guys at Los Alamos were on legal amphetamines, probably not getting much sleep and drinking in the evenings together after long amphetamine fuelled days in the lab developing weapons to kill millions.

I find it fascinating to think of Slotin's experience, because he immediately knew he killed himself but wasn't dead yet, and wondering if he killed others in the room.

Actually this used to be my profile pic:

[/img]

[tisaalho]

I'd like to see MIDI any connection on the rear with a 3p KK header (on any module with MIDI, actually). Then everyone can wire that to whatever connector they like in whatever polarity is needed (which certainly wouldn't be a 3.5mm TRS jack in my case), on the front or the rear of the modular... No one does that though except Graham Hinton How does your module connect to the expander?

Cheers

The module and expander are connected with a 26 pin 0.05" pitch flat ribbon cable (included with the expander). Sorry to say it won't be possible to inject the MIDI from there. The THRU appears on the jumper pins (0.05" pitch) for polarity selection, but I suppose it's the IN you're more interested in?

[JES]

Thanks for patiently answering so many questions. Let me re-ask one you may have missed. Any planned suppor for MPE? Or, how would the module handle MPE?

Of course, sorry, I meant to answer but that one slipped my mind!

So the initial version of firmware will not have MPE support. At the moment, we don't have explicit plans to support that in a firmware update either, but on the other hand we don't want to exclude that.

What would you like to control with MPE? Per note pitch and volume are obvious targets, on the other hand Spread and Core Stability might give some interesting opportunities for expression

Hmm... maybe implementing MPE would be the perfect excuse for me to get myself a Linnstrument/Seaboard etc

Thanks tisaalho.

The only really important thing for the Demon Core at first is that the MIDI through passes MPE data correctly.

So here's what I think I'd need with a Linnstrument to make MPE work with an oscillator module. If you pulled this off, I think it would be the first MPE oscillator in Eurorack.

1. MPE is "note per channel" so 4 note polyphony takes 4 channels. So the Demon Core needs to be able to be addressed 4 channels at a time. This allows, within the MIDI standard, for each note to have its own velocity and aftertouch. This is super useful with things like piano multisamples. It's less critical with analog synths because that's not how people expect to hear them (so much of music is about expectation). But people could learn to change their expectations....

2. Aftertouch and Y axis could be assigned to Spread and/or Stability. Again per note/channel.

Here's a good explanation of MPE (Roli has one too): http://www.rogerlinndesign.com/mpe.html

Also happy to talk Linnstrument vs Seaboard if you are seriously considering one or the other (I own a Linnstrument and a Seaboard block).