Yamaha GS1-inspired synth, revisited.

[k brown]

Hello, friends - long time no see.

Having an extended gap between recording gigs, I was playing around with SkematicGS, which I like a lot, but I have to admit I've tired of the Skematic-style GUI - it was a fun thought-experiment, but I always winced a bit when I went to use it; for one thing, I'm not crazy about high-key GUIs.

So I had the idea to redo the GUI to reflect the appearance of the GS1's wood piano-style case for the area of the performance/global controls, and do up the operator programming area in a CS-80 style.

I like looking at / working with this much better.

All that's really left to do is fine-tune some of the presets from SkematicGS, that got a bit out of kilter in the transition.

I added the additive carriers feature from SkematicSC, so now it hints a bit at the structure of the Synclavier (4 mod/carrier pairs with additive carriers). Clicking on the 'ADD' buttons below the carriers will display the additive osc harmonics controls.


One bit of fine tuning I'd like to do, that I need some help with is the clockwise two positions of the Detune control (Random Detune), which is described thusly in the manual:
"With the GS1's random detuning, overall depth is determined by the Depth setting ("1" or "2"), but an additional "random" factor is introduced on each note. Notes played one at a time will always be detuned by the same amount, but notes played in groups will each be be detuned by a different amount that depends on the order in which the notes are played."

I'm not even sure I know exactly what that means, but does anyone who might get it, know if such is even possible with FS? The way I have it now, the detuning is random, but all the notes in a chord are detuned exactly the same; has a very static sound.

gs1a crop.jpg (128.52 KiB) Viewed 13178 times

Screenshot copy.jpg (143.91 KiB) Viewed 13171 times

FirstFM-a v1.01.fsm

(5.87 MiB) Downloaded 645 times

[billv]

long time no see

Welcome back...Mr Brown...thanks for the synth...always interesting.

[tulamide]

One bit of fine tuning I'd like to do, that I need some help with is the clockwise two positions of the Detune control (Random Detune), which is described thusly in the manual:
"With the GS1's random detuning, overall depth is determined by the Depth setting ("1" or "2"), but an additional "random" factor is introduced on each note. Notes played one at a time will always be detuned by the same amount, but notes played in groups will each be be detuned by a different amount that depends on the order in which the notes are played."

I'm not even sure I know exactly what that means, but does anyone who might get it, know if such is even possible with FS? The way I have it now, the detuning is random, but all the notes in a chord are detuned exactly the same; has a very static sound.

From the description, I, from a musician's perspective, would take it as follows. How much the detuning spreads is determined by the depth setting, and it seems to be 2 fixed values, that we don't know. When I play a note, it will hit the depth/spread detuning value exactly. However, when I play some chord it notices that it is not a single note and adds some random amount (positive or negative) to each single note, on top of the fixed detuning. The order of the notes in the chord is used to determine the random value.

Without knowing the exact values for detuning and random max amount, it is hard to guess. But, should you find out those values, it is pretty straight forward. The first note coming in is given the fixed detune value. If a note comes in, while the first is still sustained, you add a random amount of detuning to the fixed detune value and apply that. And so on. This is simple, because midi is a serial protocol. Even if it sounds like all notes play at the very same time, technically they are sent one after the other. So all there is left for you is to determine if there still is a note sustained, to add further randomized detuning. As soon as it is the last note from the chord (last sustained stops and no new note came in before), you reset to just the fixed detune value, until the game described above repeats.

[k brown]

That sounds like a fuller, clearer description of what I had only vaguely formed in my head ; - )

Now, just need to figure out how!

Playing around with the synth a bit, it seems to me like the wave read osc prim used by the additive oscillator doesn't do phase modulation ('FM') nearly as cleanly as Martin's sine; and feedback even less well. So I just finished re-doing it with the additive module from Spogg's 'Mystery' (with added oscs to bring the total to 22; close to the Synclavier's 24), and I think it sounds much better. Only drawback is the loss of the ability to save/load additive waveforms and copy them to other oscillators. The use of regular sliders is much less 'twitchy' than the toolbox additive osc, though.

I also expanded the feedback options. There's no info on the net that I could find, about implimentation of feedback in the GS1's programming computer. The algo diagram I found doesn't show any feedback, just the cross-mod between modulators. I find it hard to believe that the programmer would not have allowed for feedback (afterall, the thing had EGs with something like two dozen break points!). So I added two controls in the carriers to allow feed back from the carrier to it's modulator, and from the carrier to the modulator in the neighboring pair. Added to the cross-mod which amounts to a 3-op stack, this provides much of the feeback options of the DX7.

[Spogg]

Great to see you here again Kevin!

I may have the wrong end of the stick, or indeed no stick at all, but I think you could make use of the Random Poly modules in the attached schematic. I struggled with this until Adam shared his secret with me and I now make extensive use of the module to produce variations from note to note of whatever parameter I want.

When a poly channel opens for a new note (and you get a whole new channel for each and every note in a chord) the ASM operates just once and gives out a random value between the set limits for each channel opened. That random value is retained until the channel is finally closed by the last envelope generator to complete.
So if you play a four note chord, each pitch (or whatever) will have a new and fixed random offset. If you play the exact same chord again (even if the notes appear in a different order) you’ll get a new set of random offsets, one for each note. So it would be a matter of scaling the random values. For example, setting 1 could be a maximum of +/-10 cents and setting 2 could be a maximum of +/- 20 cents.

Maybe this will help…

[k brown]

Hey, that's great - certainly sounds like what tula was describing.

I'll give it a shot, thanks!

[k brown]

I think that did the trick - sounds good to me. Thanks, Spogg (and Adam).

Here's v1.02, with your (expanded) 'Mystery' additive osc module, and addtn'l feedback controls.

FirstFM-a v1.02.fsm

(8.19 MiB) Downloaded 616 times

Screenshot copy 2.jpg (147.89 KiB) Viewed 13069 times

[tulamide]

Great to see, that there is a stream channel solution! I already wrapped my head around how to make use of midi to poly for that to no avail!

Out of curiosity: If I understood correctly, wouldn't it apply randomness to EACH note, rather than only to notes in a chord?

[Spogg]

... Out of curiosity: If I understood correctly, wouldn't it apply randomness to EACH note, rather than only to notes in a chord?

Yes that’s right. That’s why I wasn’t too sure if it was exactly what was needed, because the GS1 quoted paragraph didn’t seem too clear to me.

It says that single notes are always detuned by the same amount, which leads me to ask “why?” A single note has nothing to beat against. Then it says that detuning is determined by the order in which notes played in a “group” (chords) will be detuned according to the order in which they were depressed. This leads me to think that the detuning is actually done by a lookup table which has pre-made “random” entries. The assignment system would index the first “random” value for the first note played. Subsequent notes added to make a chord would index subsequent values in the lookup table and the values indexed would be determined by the note depression order. That would mean that any staggered chord played in an identical order would always have the same “random” detuning per note.

If that was the case, there would be 2 “random” tables, chosen by that knob and each having a different set of “random” entries.

We need to think about the technology of the time. All sound generation was done in hardware, controlled by a microprocessor with severe limitations, by modern standards. To generate truly random numbers digitally in real time may well have been beyond what was possible back then. Pre-made tables would be a solution because indexing a value would be very fast. Plus I think it would give a similar impression to the random poly module; play a chord and get slow beating which would be slightly different every time. That could be really nice for strings for example.

It’s interesting to consider that an analogue synth would have little trouble doing this. Just sample and hold a random noise source at note-on and apply it to fine detuning. If my analysis is right, I must say that Yamaha’s digital solution is really clever.

If Kevin was looking for an accurate simulation we would need those detuning tables and a fairly complex solution to index the appropriate values based on note order, but I suspect it wouldn’t offer any advantage in terms of musical result.

[k brown]

I guess all I can say is that what you provided produces a result that seems to me what that quote from the manual describes. Anyway, I like it and it sounds eminently musical.

Thanks again.