Recipe for the best supersaw.

[Architeuthis]

definately tracking down the other supersaws and stabbing them helps to make one the obvious choice.

?

[xalama qo]

Phasing is the reciproce of beating....

No Urs-dude, you've misunderstood. The OP said RECIPE....

[Architeuthis]

alright doods

https://soundcloud.com/elanhickler/a-st ... aveshaping

more supersaw using the same technique. proving this statement to be untrue:

but if you listen to trance music, it is mostly the filter cutoff which is being modulated, or the frequency is modulated to increase spread, but it's not done the whole time, rather like a "riser". I don't see much use for the sound in your sound demo, as it is modulated quite extremely. (Fair enough to say that you probably only wanted to showcase the effect, and therefor made it that extreme though i guess)

[neotec]

:No offence:

I'm still trying to figure out, what you want to tell us. Is it: 'The original SuperSaw was made like this' or 'That's the way you all should do SuperSaws'?

Well, your version definitely sounds interesting, though I'm not a big friend of such strong pitch modulations. Sometimes the sound in your first demo sounds like some weird chorus effect ... maybe that's because it's very similar to a chorus (I said similar, not identical).

I for myself like the harsh sound of the JP8000, and I also love the original SuperSaw. I can't stop smiling when I think about that Roland's VA OSCs in the JP8000 were just brute-force digital oscillators with a pitch tracking high-pass or shelving filter.

Still it worked pretty well and it shaped the sound of electronic music for decades.


About phasing: The original SuperSaw definitely has phasing artifacts, but they're part of its unique sound. (Citation needed)


About beating: When you look at the SuperSaw's various examinations, which can be found on the web, it gets pretty clear that Roland used maximum pitch spreads expressed in prime numbers.

I chose maximum detune values in cents and these here should come pretty close to the JP8000: -191, -109, -37, 0, 31, 107, 181. Beating is reduced to an absolute minimum when using prime detune values. Phase randomization (on note-on) is also definitely needed or you won't get a strong attack on every note-on.


Mix and detune: I won't go into those settings, as knob/slider-to-value mappings always are a matter of taste. Just some thoughts: Adam Szabo came up with some pretty complicated equations for mix and detune and IMHO he's definitely wrong. We're talking 90's DSPs!

Things were dirty back then so the mapping must be simple. Logarithm tables, squared values, ... The 'drop in volume' on the fundamental oscillator compared to the side oscillator on higher mix settings might also be caused by the tracking filter. My guess is that the main oscillator stays constant in gain and the side oscillators are simply between >0.0 and 1.0 in gain. This would cause the main osc to be at 0.7 and the side oscs at 1.0 on full mix settings which comes close to Adam's observations (as far as the ratio is concerned). Maybe there's also some 'normalization' in the end (dividing by sum-of-osc-volumes ... well, more like multiplying by some value and right-shifting, but you get the point I hope).


Sample rate: As one can see from the JP8000 schematics, the word clock for the DAC is 88.2kHz (67.7376MHz / (16 * 3 * 16)). So the JP8000 renders its audio at 44.1kHz, stereo (if they don't use any digital down-sampling inside one of the 4 DSPs (which I strongly doubt)).


So: my SuperSaw is just 7 non-bandlimited, digital oscillators, rendered at 44.1kHz, detuned, mixed and post-processed by a 24dB SVF/0D HPF tuned to/tracking the fundamental frequency. The whole thing uses less than 1% of CPU resources on my crappy MacBook Air and pretty much sounds like the original SuperSaw.


Maybe it would also be a good idea to explain what a 'good sounding SuperSaw' is or a 'crappy' one. Is the original 'good-sounding'? If it is, then your chorused sawtooth is the wrong way to go.


That's my two cents concerning the SuperSaw. As stated above: no offence. ... but I had to reply to this

[xalama qo]

Hi guys, i can recommend a listen to Archetype Instruments Arc5 synths 'Super7' oscillator. Apologies if OT, but i find it to be a very smooth/warm flavour of supersaw. I have no idea how it was created though! So now i'm outta here.

[Architeuthis]

I'm still trying to figure out, what you want to tell us. Is it: 'The original SuperSaw was made like this' or 'That's the way you all should do SuperSaws'?

The point of this thread is to let developers know good practices of implementing stacked voices, that's all. So any information beyond mine is good. I'm just telling you what I know to be good. My information has nothing to do with any research of old algorithms. I use only my ears, and I'm offering the most braindead solution to a good supersaw. No calculations needed, no special algorithms, no sample rate specifications.

Well, your version definitely sounds interesting, though I'm not a big friend of such strong pitch modulations. Sometimes the sound in your first demo sounds like some weird chorus effect ... maybe that's because it's very similar to a chorus (I said similar, not identical).

I can do strong and light pitch modulation for heavy and fast for harsh sounds or slow and light for classic sound. The demo has both, unless you simply disagree that I never hit the "classic sound" mark.

Adam Szabo came up with some pretty complicated equations for mix and detune and IMHO he's definitely wrong. We're talking 90's DSPs!

If it sounds good, it's right. If it sounds bad, it's wrong. You're not wrong just for doing things differently to get similar or better results.

Maybe it would also be a good idea to explain what a 'good sounding SuperSaw' is or a 'crappy' one.

It's good if it pleases your ears. It's bad if you take any old vsti synth (or even analog synths), use basic frequency spreading, and call that a supersaw. Doesn't sound good. If that pleases your ears then you probably need to have your ears checked. Obviously some good supersaw synths only have one knob "frequency spread" but those synths do more under the hood to get a good sound.

As stated above: no offence. ... but I had to reply to this

Glad you did, and in a respectful manner!

[whyterabbyt]

I'm still trying to figure out, what you want to tell us. Is it: 'The original SuperSaw was made like this' or 'That's the way you all should do SuperSaws'?

The point of this thread is to let developers know good practices of implementing stacked voices, that's all.

given how many times you've fallen flat on your face with similar proclamations in the past, you'd probably be better off not claiming that you have a 'best' solution which people with years more experience than you should line up to implement solely on your say-so.

i kinda recommend you start rephrasing it as "here's some supersaws as I prefer them" and leave the "I know better than anyone else" shite out of it.

[raphx]

Has anybody compared stacking a bunch of oscillators to the PadSynth technique? http://zynaddsubfx.sourceforge.net/doc/ ... Dsynth.htm

I'm interested in ways of making a thicker sound without consuming lots of CPU, that seems promising.

[Architeuthis]

i kinda recommend you start rephrasing it as "here's some supersaws as I prefer them" and leave the "I know better than anyone else" shite out of it.

Sorry, that's just how I do things. Take it or leave it.

[mystran]

Has anybody compared stacking a bunch of oscillators to the PadSynth technique? http://zynaddsubfx.sourceforge.net/doc/ ... Dsynth.htm

I'm interested in ways of making a thicker sound without consuming lots of CPU, that seems promising.

If you're after the "7 saws" type super-saws then that won't really work well. If you're after smoother "enough saws that they morph toget" type stacking then I suppose the PADsynth algo would work. How much is "enough saws" depends on the detune too (more detune, the more saws you need to blend it into one sound).

[bjporter]

I for myself like the harsh sound of the JP8000, and I also love the original SuperSaw. I can't stop smiling when I think about that Roland's VA OSCs in the JP8000 were just brute-force digital oscillators with a pitch tracking high-pass or shelving filter.

Interesting...

Hey where did you find that info? I've always wanted to talk to the engineers who built it.


Does anyone have primary source documented evidence or hardware analysis of how the orignal JP8000/8080 did the supersaw?

[mystran]

Has anybody compared stacking a bunch of oscillators to the PadSynth technique? http://zynaddsubfx.sourceforge.net/doc/ ... Dsynth.htm

I'm interested in ways of making a thicker sound without consuming lots of CPU, that seems promising.

If you're after the "7 saws" type super-saws then that won't really work well. If you're after smoother "enough saws that they morph toget" type stacking then I suppose the PADsynth algo would work. How much is "enough saws" depends on the detune too (more detune, the more saws you need to blend it into one sound).

I'd like to further add that the authors ideas about the profiles of spread harmonics is pure gold though. The "all saws the same amplitude" is not necessarily that great, unless you're going for raw-sound rave-leads and varying the profile (ie varying amplitudes depending on how much particular voices are detuned, you can apply similar ideas to panning as well) can make the same thing sound quite different, and allow you to use tons of detune (and stereo spread) yet still have a very coherent sound.

In fact, the JP8k (which IMHO sounds pretty horrid by today's standards, but that's probably just me) has a limited form of this since it (IIRC) let's you can vary the amplitude of the unison with respect to the main saw. You can go further though, and vary them all individually.

[Architeuthis]

How much is "enough saws" depends on the detune too (more detune, the more saws you need to blend it into one sound).

Very good observation, didn't think of that.

You can go further though, and vary them all individually.

This is exactly the kind of thing I want to encourage. Treat each saw as if it's an individual violinist. Give it different properties, different legato speeds, different pitch, different volume (well, different volume could be carefully implemented so that the loudest saw is most in tune and the quietest is least in tune, that would give some interesting panoramic effects)

[mystran]

Oh and this thread kinda inspired me to make some pointless "music" again after a while of just messing with code without attempting to compose anything.

Not exactly very interesting, nor terribly relevant to the thread (sorry), but thought I'd post as a "thanks for the inspiration" anyway.

https://soundcloud.com/mystran/pointles ... aw-wankery

Oh and the lead is 32 BLEP-saws (per note) with equal spacing on log-freq (pitch) scale, with a bit of volume (and pan) profiling (basically the proto-synth goes from flat profile to a sort of gaussian affair). The magic ingredient is obviously "Phazor" (and EQ and delay and reverb, but those are basic).

[bjporter]

Hey guys, here are 3 examples of Supersaws created from synths that don't have one. Used various techniues (layering/stacking, panning, detune, delay phasing, etc). DID NOT use chorus, flanger, doublers, etc

Example 1: Aalto Synthesizer (skip to 30 seconds) - 7 saws, detune spread by 12cents, stereo delay phased about 1ms difference per layer
https://soundcloud.com/bjporter/bjporte ... venture-on

Example 2: Feldspar Synthesizer (skip to 2 minute mark for chords, and see 1:30 for pluck) - 2x 7 saw layers (bussed and panned out to mimic stereo supersaw), detune spread by 12cents, stereo delay phased about 1ms difference per layer
https://soundcloud.com/bjporter/bjporte ... -down-to-a

Example 3: Podolski Synthesizer (skip to 2 minute mark) - 7 saw layers (bussed and panned out to mimic stereo supersaw), detune spread by 12cents, stereo delay phased about 1ms difference per layer - and gated using automation (this one didn't come out as well)
https://soundcloud.com/bjporter/bjporter-nature-was-the


The first 2 are a bit better. The third is a bit rough.