High Poly Octave-Phase-Coherent Drawbar Organ

[Tim Kleinert]

This patch (a performance actually) contains some of my newest designs and delivers a full classic 8-drawbar organ model, with true octave phase coherence, at 26 voices, plus FX.

It's based on a new technology that allows me to multiplex 64 audiorate sawtooth oscillators through 2 interslot busses.

(This isn't an april fools joke, btw.)

It's a performance, since one DSP (the FX area of slot 1) is in charge of calculating the octave-phase-coherent oscillator drivers as well as the drawbars and piping them into the interslot busses.

Slot 2 is the poly area to be controlled from the keyboard.

There's a factory preset called "YetAnotherOrgan" which has the best Chorus/Lesie emulation I've ever heard from the G2. And since I royally suck at patching such things, I allowed myself to cannibalise that and put it into the FX section. I couldn't credit the author since no name is mentioned in the original patch. However, I explicitly credited the patch itself. Thank you very much, whoever you are.

Of course, there are compromises to be made. In this case, the patch doesn't feature drawbar folding or key click. And the organ percussion is not derived from the rest of the engine and therefore might not always be in phase. I don't find this terribly noticeable however.

On the plus side, I personally find the effect of perfect phase-coherence to be quite noticeable especially when more overdrive is employed (which makes sense if you think about it).

Since the relevant parameters are spread over two slots, I assigned them to the global pages to have everything in reach.

cheers
t

EDIT: Some minor optimisations in the voice area, and upper C (C7) now has its proper octave footage. New version uploaded.

---

HighPolyOrgan_TK.prf2
Description:	High Poly Octave-Phase-Coherent Drawbar Organ (26 voices); IMPROVED VERSION	Download
Filename:	HighPolyOrgan_TK.prf2
Filesize:	16.5 KB
Downloaded:	4529 Time(s)

---

[drapdap]

[Tim Kleinert]

What I mean is that all identical notes in different octaves (eg. C2, C3, C4 etc.) are in perfect phase with each other. This occurs in tonewheel and transistor organs as well as string machines and is an important sonic characteristic of those instruments, especially noticeable when playing chords with multiple doubled octaves. Also, multiple keystrokes on the same key never cause phase cancellations (as happens when using a conventional synth architecture with free-running oscillators and a standard voice allocation scheme). This was considered impossible to patch on the G2, or any synth with free running oscillators for that matter, because the oscillators in one voice would "need to know" what those in other voices are doing phase-wise.

I managed to patch it in a way that is maximally economical from the polyphony standpoint. I threw in dynamic waveform generation tech in there as well, so that the individual voices only need to play that back at the desired pitch rather than having their own 8 drawbar oscillators eating up all the DSP. This opens up the voice count even more.

The result is a full 8-drawbar organ with not only 26 voices, but also full octave-phase-coherence. Two things that -even by themselves- would be considered impossible to patch.

Nothing this red box can't do.

[ZeeByeZon]

Hi Tim
I had a look on this patch, but I do not understand how it works exactly. In particular, when it comes to switching and mixing the sawtooth oscillators (seems like it's not multiplexing here...). Could you please explain it?
Thanks a lot

[dorremifasol]

Very nice, thanks Tim!

[Tim Kleinert]

[ZeeByeZon]

It seems far more complex than just multiplexing.
I really would like to understand this one

To my understanding, the 12 rampLFOs are tuned as 12 half tones for one octave (with quite low frequencies). Their control inputs are all connected to the same signal, which is an audio rate step increment looping from 0 to 11. I would say that the first switch is always on, and the others accumulate until the value 11 where they are all on, resulting in the addition of the 12 signals, and cycle back to 0...
The control signal cycling from 0 to 11 is derived from a master clock cycling from 1 to 64. This leads to 5 phases (for 5 octaves) plus 4 extra steps (from 61 to 64). A gain is applied to the previous "ramp LFO mix": 4 for the first octave, 8 for the second, and so on (but 64 for both octave 4 and 5?!). Then you transform this with the second sequencer module with 100% fade, I do not understand what it is.

On the busses, the 4 signals are:
1 - Audio rate 1 to 64 numeric cycle
2 - The misterious stuff described above
3 - Another cycle from 1 to 64, but step size is 8 times smaller (0.125)
4 - The drawbar mix

Note that the drawbar frequency are strange too: the 5 1/3' should be 562.56Hz, the 2 2/3' should be 1125Hz and the 1 1/3' should be 2250Hz...

Then to gather all this you make complex combinations with delays...
Am I wrong? What do I miss?

[Tim Kleinert]

The two control sequencers deal with transposing the base LFOs to the corresponding octaves. The first one contains the frequency multiplicator (I started from 4 for convenience reasons) and the second one is my weird audiorate sawtooth modulo waveshaper that actually performs the frequency multiplications.

I use the delay line as waveform memory, so i need a sample-accurately lock-stepped counter to access the data properly and pipe this to the voices together with the actual audio data.

Since I generate all 64 sawtooth readout pointers simultaneously, the update rate is a lowly 1500 Hz. But in the voice area, I developed a modulo interpolation circuit that is able to accurately reconstruct, from the coarse 1500 Hz signal, a perfect 96kHz audiorate signal, which is then used to play back the waveform memory in the delay line at the desired pitch.

Sorry if this sounds like total gobbledigook , I have difficulty putting this stuff in words.

[Tim Kleinert]

Update: Some minor optimisations in the voice area, and upper C (C7) now has its proper octave footage. New version uploaded.