Advanced Unison

[tulamide]

Unison per se is pretty easy. Take a wave and play it several times parallel, slightly detuned against each other and distributed over the stereo field.

But, to achieve that, I would need to use x oscillators. I am thinking of unison for that wonderful SWM-Sine Oscillator that Martin made. But, say, 4 of them just for unison adds to the load (because in total there might then be 8 or 12, which increases CPU load at least 3 times).

Wouldn't it be better to send out 4 streams from the oscillator, while the oscillator calculates the detuning? Or doesn't it make a difference in terms of CPU-load?

[adamszabo]

As far as I understand Flowstone is using SSE which means 4 voices will take the same cpu amount as only 1 voice. So if you create a unison using 4 voices, (so then you get 1 mono key) it should cost no more cpu.

[martinvicanek]

If CPU load is a concern then try this (somewhat) optimized oscillator version.

[tulamide]

Somewhat? It is noticable improved! With 3 oscillators playing chords, so that at least 9 notes are playing:
Before 10%, now down to 8% - that's a 20% improvement! Thank you very much!

The question is still there, I know, I can be annoying.
Say we have a synth with three oscillators.
Playing notes triggers sounds created from those 3 oscillators in combination.
I would want to set one of these oscillators in 4-voice-unison mode (but not monophonic). Behind the scenes I would now use another 3 oscillators to get unison on that one oscillator, so in total there are now 6 oscillators playing.

But, for unison you wouldn't neccessarily need to generate waves from an oscillator. I could, in theory, just take the result from one calculation, and add a little detuning for 3 more. I was thinking that this additional detuning in the oscillator would need less cpu-cycles than generating 4 complete waves. Is that wrong? Is it even (theoretically) possible?

[adamszabo]

There is a "Midi to Multi Voice" component already which should do what you are after (i think), I have atached a schem how you can make unison with it (courtesy of Myco)

[tulamide]

There is a "Midi to Multi Voice" component already which should do what you are after (i think), I have atached a schem how you can make unison with it (courtesy of Myco)

Thanks a lot for this example. It is another approach to unison, that I didn't think of! Streams are poly-components, of course you can send several detuned frequencies per midi note. That's good to know.

But (please don't shoot me ), that's still not quite where I want to go. So I made an accoustic example, hearing may say more than a thousand words. I'll be as accurate as possible in my description:

You will hear one saw-oscillator. A bandpass filter is applied, set to mid-frequency. Also, midi-velocity drives the filter's envelope amount as well as the filter envelope's decay to certain degrees.
3 chords (of 3 notes each) are played with velocity 75, and a melody with velocity 125. The 4-bar-loop is repeated 4 times. On the first time you will hear the raw sound, the second time the same sound is sent through a unison module that is set to 16 voices and 50% of the possible detuning strength. For the last 2 loops it will switch back and forth per bar.

unison.mp3

I think this will make it clearer. The stereo field is used in total. With the current techniques I can't use the stereo field in FS. With x dedicated streamouts I could mix them in stereo.

[adamszabo]

With the unison example I posted its very easy. Just assign the voice tags (unison voices) to different outputs. For example, you can split your signal and create a L and R and have a unison of 4 voices, but have voice tag 0 and 1 only go through L, and voice tag 2 and 3 only go through R, then you will get the stereo effect.

[adamszabo]

Here is an example with a stereo spread knob. Its mono in the middle.

[tulamide]

Thank you, Adam, you saved my day! I totally forgot about channel IDs. That solves it and it is just as I imagined it (using one osc), perfect!

I hope I can return the favour soon (you know what I'm talking about )!

[Spogg]

Hi guys

I've been following this with great interest.

I have a problem with the schematic that I haven't yet been able to resolve. The MIDI input module works fine with the PC keyboard input selected but ignores my MIDI keyboard (Keystation 61 es). The keyboard is fine on other synths in FS. I've tried replacing all the main prims with those from my toolbox (FS 3.08.1) but the keyboard is always ignored.
I've never had this issue before.

Also, a more general question: This approach is very elegant but do we actually save CPU cycles over using individual oscillators (as I did in my Quilcom Unison synth)? The intuitive thought is that we still have to create the inividual waves to make the sound so if we do benefit in CPU terms how so?

Cheers

Spogg