This page is available in English only

CetoneSawteeth

Even more supersaws

 (6 Votes)
1.3 (Updated 5 years ago)
376.9kB
December 28, 2016
Reaktor 6
Subtractive

DESCRIPTION

First: this is not a JP-8000 emulation, neither did I ever lay hands on a real one, nor did I ever try a plugin that emulates the original Roland.

This is my take on a supersaw-only synthesizer that roughly copies the control layout from the JP-8000. You could call it a re-interpretation. This instrument MUST run at 44.1kHz sampling rate, just as the JP-8000 did.

As for the presets: I'm pretty bad in coming up with presets and I also don't normally use them, so there are only pretty few contained. If you come up with cool sounds, tell me and I integrate them.

And: feedback is highly appreciated.

Features:
- 2 JP-8000-style supersaw oscillators
- 12/24dB ZDF multimode filters
- analogue envelopes
- parametric bass and treble controls
- stereo chorus
- stereo delay

Update v1.1: Fixed a bug in mix computation, clamped oscillator pitch.
Update v1.2: Reduced filter CPU load, limited maximum resonance for 24dB.
Update v1.3: Unused voices are automagically turned off now, to reduce CPU load again.

A few notes on the supersaw oscillator emulation:
We're talking 1996 hardware here. In my ten years of working with audio DSP I did a lot of sound chip emulations for 80s and 90s hardware, so you get a feeling for what is possible and what not. The JP-8000 had 4 DSP chips, each clocked at ~67MHz, this leaves 768 clock cycles per sample per voice. You can't do much with that. The DSPs must have featured pretty fast integer multiplication units (yes, no floating point). So, they probably had 24 bit phase accumulators and derived the naive saws directly from this (as the C64 SID chip did). As naive saws generate a lot of aliasing, Roland just put a low shelving filter after those, to filter out the aliasing components below the fundamental frequency. Without the filtering you would need 13 or 15 additions, 1 multiplication and 1 right shift to generate and mix all 7 sawtooth waves.

So, if you just look at the stuff from a 1996 hardware perspective, things get a lot clearer.

COMMENTS  (6)

Please login to add a comment
René Jeschke
5 years ago
No, a hard clip to limit self oscillation is everything I put in there. I must admit that I like my filters clean. Also, my filters are already quite heavy on the CPU and adding iterations would really be too much. :-D I still mostly prefer my filters because I can configure them to have any frequency response, which makes them suitable for a high number of applications.^^
Colin Brown
5 years ago
Did you incorporate any nonlinearity? (hard clip?) I've had my best results using newtons method with tanh saturation, but some of the other methods are still eluding me. The factory filters in Reaktor 6 are superb, so I'll probably be working with them for most things.
René Jeschke
5 years ago
The filters are Linkwitz-Riley 2/4 pole crossover. Each path was 'solved for out' to remove the delay. This is basically the idea of ZDF, compute the output before it would normally be available (because of the unit delay). I'll have a look at your supersaw oscillator, and yes, I cheated a bit with mine, as I use floats and an extra multiplication. :-D
Colin Brown
5 years ago
Sorry, didn't take long enough to look at the structure. Is that a newton Raphson or something else? FWIW, I built a supersaw osc that might be closer to the real thing - at least in terms of internal process - than most of the freely available versions out there. It's built on a super efficient process that would have been practical in a 90s system. It's in a block in my blocks polysynth demo. https://www.native-instruments.com/en/reaktor-community/reaktor-user-library/entry/show/10343/
René Jeschke
5 years ago
Point taken concerning the 'genuine'. :-D Regarding the filters, they ARE ZDF, they are just not built using the Reaktor ZDF toolkit. I was involved, quite some time ago, in the discussion about ZDF at kvr-audio with Vadim and others, Vadim also mentions me in his ZDF paper (I'm Neotec at KVR), by the way.
Colin Brown
5 years ago
"2 genuine JP-8000 supersaw oscillators" O RLY? :-D... seriously, there are whole threads dedicated to trying to understand/replicate the original supersaw oscillator, and it's an elusive beast - definitely not _just_ 7 detuned naïve saws mixed together. Also, it might be probably be worth using one of the newer ZDF filters for better quality (even if the original Roland didn't have these).
now