John Bowen Synth Design

Just wanted to mention it's wonderful to be an owner of this synth, not only do we have a fantastic instrument that keeps getting better. We can be part of a community where you can read a dev (Jim) & John discussing low level internal workings of this instrument. Keep up the good work guys, these discussions are super interesting imho.

[quote=stefanovic post_id=24218 time=1661695849 user_id=1122]ok, and for those pairs, only the osc tuning would be slightly different, right?
[/quote]
You almost got it. It’s hard to be clear because I don’t have the right words for it. I guess you can call it voice tuning. When a note is started its voice tuning can be set by internal code and can be different for each voice.
But [i]oscillator [/i]tune must be the same for the two voices, as with all other parameters.

[quote]Still, it would be possible to set different voices to have slightly different offsets to some parameters like filter, EG times, etc.[/quote]

That would be nice, indeed, if it's not too complicated to do!

[quote] But pairs of voices would have to have the same offsets because of the way the Solaris is designed with each DSP producing two voices.[/quote]

ok, and for those pairs, only the osc tuning would be slightly different, right?
Well, it might be useful anyway for 4-voice polyphony, and even more so with more voices...

[quote=stefanovic post_id=24216 time=1661526616 user_id=1122]Yes, he is using the seqs in KeyStep mode because without having access to the firmware code, that's the most practical (and maybe the only) means to implement VCM on synths with such seqs. [/quote]
What I am saying is that I don't think he is achieving what he calls VCM using sequencers in these other synths. It can be done internally such as what the Prophet-5 rev 4 might be doing, I don't know the internals of that. But I don't see how it can be done by a user with a sequencer. Try this example on the Solaris:
1. Start with a simple INIT preset.
2. Set the oscillator to saw wave.
3. Set EG6 Release time to 20 seconds. Optionally set EG curve to linear to hear it longer.
4. Set the filter cutoff to 126. Set the filter mod source to SeqA, Amount to 120.0, Destination to cutoff.
5. Now set up the sequencer. Set Mode to key step. Set pattern length to 2. Set step 1 value to 0. Set step 2 to value -127.

Turn on the sequencer. Play the first note. The filter is open and the note lasts a long time. Play the note again to get to the second step, The filter closes for both voices not just the second voice. So this shows modulation affects all voices. This may seem obvious, sorry, but I'm just trying to be clear. With a typical synth there's no way to simulate what he calls VCM using a sequencer that I'm aware of. Can any synth do this test and leave the first voice with an open filter? Anyway.....

[quote=stefanovic post_id=24216 time=1661526616 user_id=1122]So if we completely leave seq aside (and be completely off-topic wrt to ratcheting, sorry), would it be possible for you to extend RndTune to other parameters than "Tune" only? Maybe not in the next OS release, but in a following one? And with values that keep the same for each voice?[/quote]
On the Solaris, the individual tuning of a note/voice---which is separate from the oscillator tune!----can be done but only by internal code. This is how Unison mode and RndTune can be easily achieved. Other parameters are not like that. Still, it would be possible to set different voices to have slightly different offsets to some parameters like filter, EG times, etc. But pairs of voices would have to have the same offsets because of the way the Solaris is designed with each DSP producing two voices. This has nothing to do with the sequencer of course. Maybe it should be a different thread.

Thanks Jim for chiming in.

[quote]Do you have reason to think that these parameters are being set per-voice? On a synth such as the Rev 2? I would rather think this changes the parameter for all voices. That’s how it works in the Solaris. I mean, what if it did work per-voice but you wanted to modulate the parameter for all voices how would you tell it to do that?[/quote]

A guess: it seems to me he only mentions analog synths, which may have voices with unique (non-shared) parameters per voicecard. With a KeyStep mode, the firmware could set the parameters for the next voice to be triggered, without affecting the other voices. With all-digital synths, one could share the parameters between voices, which would make it impossible to change a voice without affecting the others.

That would also mean that VCM in KeyStep mode is impossible on the Solaris? It seems to me that I could definitely tell the difference when I tried it on my unit, but maybe I deceived myself :) ...

[quote]Actually, to implement his original premise, if you do it internally, you don’t need to use the sequencer. You just need to set up each voice’s parameters slightly differently at the start and leave them there. The only reason he’s using the sequencer is for key step mode so he can make the parameter change at each note. If all you need are random changes per note, and if this is all implemented internally then you don’t need the sequencer. We know when notes are played and can make the parameter change without the sequencer. The only reason you’d need a sequencer is for specific, user-controlled changes per-step.[/quote]

Yes, he is using the seqs in KeyStep mode because without having access to the firmware code, that's the most practical (and maybe the only) means to implement VCM on synths with such seqs. Since you have access to the code, I agree it could be implemented in a more straightforward way. (Still, implementing it with user-defined seqs allows for experimenting, fine-tuning and assessing what works and what doesn't).

So if we completely leave seq aside (and be completely off-topic wrt to ratcheting, sorry), would it be possible for you to extend RndTune to other parameters than "Tune" only? Maybe not in the next OS release, but in a following one? And with values that keep the same for each voice?

I read that guy’s web page. I think his original point makes some sense. With an analog polysynth, once it warms up, each voice settles to be more or less stable although the voices may settle into a different random place. Whenever voice #1 plays it always plays the same as it did before but it may not play the same as voice #2 (slight randomness aside). So if you set up the voices of a digital synth in such a way it should sound more like an analog poly.

But apart from that, either he is missing something or else I am. He is using the sequencer to set different values at each step for parameters such as oscillator tuning, filter cutoff, EG attack, EG decay, and more. Do you have reason to think that these parameters are being set [i]per-voice[/i]? [i]On a synth such as the Rev 2[/i]? I would rather think this changes the parameter for all voices. That’s how it works in the Solaris. I mean, what if it did work per-voice but you wanted to modulate the parameter for all voices how would you tell it to do that?

Actually, to implement his original premise, if you do it internally, you don’t need to use the sequencer. You just need to set up each voice’s parameters slightly differently at the start and leave them there. The only reason he’s using the sequencer is for key step mode so he can make the parameter change at each note. If all you need are random changes per note, and if this is all implemented internally then you don’t need the sequencer. We know when notes are played and can make the parameter change without the sequencer. The only reason you’d need a sequencer is for specific, user-controlled changes per-step.

Regarding voice assignment, the sequencer knows nothing about voices. It just generates notes (and is a modulation source). Voice assignment happens further downstream.

The Solaris does have a change regarding voices that was not in the original. A user once pointed out that some older polysynths allowed you to set it so that playing the same note multiple times would use the same voice. So we added this to the Solaris. If you set the PlayMode to Poly and the Legato to Reassign then repeating the same note will allocate the same voice to that note. If you set a long release time (like 6 seconds) it will cut off that release because it’s re-using the same voice. It’s hard to tell, but the way you can tell is to set the RndTune parameter to a higher value. Then as you repeat the same note with Legato Off you will hear the detuned phasing as it uses different voices. Set Legato to Reassign and you won’t hear that because it reuses the same voice. The RndTune parameter works because the tuning of a note can be set differently for each note….for each voice (it's not Osc tune). It’s a unique parameter in that way. Other parameters currently are not unique per voice but are the same for all voices. The Unison function also works in a similar way: each note consists of multiple voices each slightly detuned.

Jim

[quote="John Bowen" post_id=24213 time=1661373119 user_id=2]
For example - let’s say you want keystep to give you a slight pitch variation using one of the mod lanes, so you set it for a very small pitch mod amount, then a long decay, so that multiple pressings of the same key will give you a slightly varying overlay of detuned notes. With keystep mode, each key played will restart the current voice, and you do not get the kind of “overlay” you’d expect.
[/quote]

It seems to me there two phenomena :
- the first one is your example, pressing multiple times the same key, hoping that a slight variation will prevent some artifacts. I think this was the problem that @alexg tried to address, back in the days: [url]https://forums.johnbowen.com/viewtopic.php?t=875[/url]

and...
[quote="John Bowen" post_id=24213 time=1661373119 user_id=2]
I can ask Jim about how the voice assignment works with the sequencer - maybe there’s an option? (Currently, changing the Legato settings doesn’t do the trick.)
[/quote]
...that would be very nice!

- a second one is with polyphony? Say, playing a chord? With the actual KeyStep mode, we will hear a stack of slightly detuned voices (not at the same note, of course), won't we?

Anyway, my suggestion is that there could be some internal seqs in KeyStep mode (or a specialized version of it for use case one), unavailable to the user, which would serve as the implementation of VCM. Of course this could eventually lead to a dedicated page where some parameters would be available to control VCM, but this is much more work to do.

[quote=stefanovic post_id=24212 time=1661367519 user_id=1122]
John, Jim, since ratcheting is about the seq, is there any chance that you implement what's now known under the name "Voice Component Modelling", i.e. using multiple internal seqs in keystep mode to slightly modulate parameters per voice as discussed here?:

[url]https://forums.johnbowen.com/viewtopic.php?t=17164[/url]

This would free us from programming the seqs for the sake of poly analog feel, and would prevent the waste of user-defined seqs.
[/quote]

The 4 modulation sequences that are there now (the current gated sequencer rows A B C D) will still be in the synth - just as ‘modulation lanes’.
I understand what you are talking about - using the mod lanes to create variances with each keystep - but there is still the issue that any key played is part of the current step, and voices are reassigned.

For example - let’s say you want keystep to give you a slight pitch variation using one of the mod lanes, so you set it for a very small pitch mod amount, then a long decay, so that multiple pressings of the same key will give you a slightly varying overlay of detuned notes. With keystep mode, each key played will restart the current voice, and you do not get the kind of “overlay” you’d expect.

I can ask Jim about how the voice assignment works with the sequencer - maybe there’s an option? (Currently, changing the Legato settings doesn’t do the trick.)

John, Jim, since ratcheting is about the seq, is there any chance that you implement what's now known under the name "Voice Component Modelling", i.e. using multiple internal seqs in keystep mode to slightly modulate parameters per voice as discussed here?:

[url]https://forums.johnbowen.com/viewtopic.php?t=17164[/url]

This would free us from programming the seqs for the sake of poly analog feel, and would prevent the waste of user-defined seqs.

I would understand that this would require another OS iteration to be properly implemented though.
And I hope there is still room space for that...

Ratcheting (2-3-4-6-8 times) is very important along with a probability parameter (chance that a programmed ratchet happens). 8 times ratcheting allows close-to-the-audio-domain sounds...

Glad you mention things already done - the control lanes are now separate, and there’s a new, single Note lane. There’s also modulation provided in a couple of ways, just as you suggested! There’s still some work to be done on the new functions, so that’s why I was wondering what else to add, such as ratcheting.

I think it could add some interesting possibilities, esp. if it were configurable per step.

Another interesting idea would be if the 'control' lanes were separate from the note lanes - meaning one could have an "n"-length step control track that sets the transpose, step division, ratchet, etc. for a given note step, and a "y" length note track that is controlled by the control track. This could lead to interesting evolving patterns with offsets between the n and y pattern lengths.

Dear Solaris Owners,

Jim is working on some improvements to the Sequencer in the Solaris. There’s already some fun things in the works, and now I am wondering if it’s worthwhile to add a ratcheting function to each step, so I thought I’d post it here to get your responses.

What do you think?

John B.