Supersaw!

[Urs]

Next up:

Example1 (little detune)
Example2 (quite detuned)

I've changed some settings in the soft (XMF was in Zebra 2.3 mode, arrrgh), and I added some resonance to the JP to get rid of some of its dullness (the soft is still much brighter now). Note that the soft is slightly more detuned than the JP.

Also note that the JP exhibts way more aliasing in the high notes, best heard in the little detuned version.

Oh yes, I think the order is JP-soft-JP-soft for each note.

[TiUser]

I'm just curious. I think waveform comparisons turn often out to be very coarse.

I've written my own scope & analyser software.

The very distinct shape of the JP8000's Supersaw can easily be identified in a scope. It not only sounds different, it also looks different from all other Supersaw implementations I came across.

You're right - it's not anything exact though, and ears are more important than eyes

Cancellation tests are close to impossible to perform. The stacked oscillators in the JP8000 can not be reset in phase, and they can not be completely tuned to the "master sawtooth oscillator", nor can they be completely faded out. There's always something going on, so the only chance to do a cancellation test is by recording a long, held note of the JP and then manually set phase and tune for each slave oscillator. To do that one has to observe the movement of phase wraps over seceral minutes of audio. Not even I would go that far.

Urs

That's an interesting insight. So you basically compare in the digital domain.

I thought you'd probably go the opposite way comparing the analogue output of the Roland to the analogue output of the emulation.

But anyways, both methods will involve A/D or D/A conversion in one or the other way... and I admit as well I don't share too much passion in the question how accurate an emulation technically is - I agree to your statement that ears count most.

I think you are picky enough committed to quality... Zebra shows this off very well.

[pac0r]

Here's what I hear..

In example one..

There is more low end... more low mids.. and sounds like theres a scoop in the upper mids 2-3 kHz or so... Its airier but its not like an eq boost in the highs..
More like white noise in the highs... modulating somehow :p

In example two...

I hear fairly static oscillators and it sounds edgy (2-4 kHz) in comparison to example 1....

Example 1 seems to have some sort of compression/saturation as well. When I compared the waveforms, example 2 had quite a few peaks, where as example 1 was very even looking and all the peaks were obviously controlled.

Try pitching the 2 waves down an octave or two then listen, the difference is quite obvious there

Two octaves down is really obvious! Example 1 has some crazy sounding resonance but its not harsh..
Also the envelopes seem different, Example 1 has a tighter decay/sustain making it more punchy... Maybe the white noise has a delayed envelope a few milliseconds after the envelope on the oscilators? Im just speculating though..
But anything is possible!! Think about it??? Open your miiiiiindddddddddd to meee...

I really hope example 1 is your algorithm, I prefer that one by far.

Hope that in-depth analysis helps.

[pac0r]

Woops ... Didn't realise..

I was referring to the example with the crazy trance melody (ferry corsten)

named: supersaw1,supersaw2

[hakey]

Here's my take:

Example 1
Example 2

And here's mine:

Super?

Anywhere near the ballpark/carpark/Yellowstone National Park?


(FWIW, I preferred Urs' softsynth version.)

[bmrzycki]

Here's my take:

Example 1
Example 2

Both sound great. For what it's worth I prefer example 2, the bass notes tend to sound rounder and fuller to me with more presence. NIce work though, they're almost indistinguishable to my ears.

[brok landers]

there wasn't any supersaw before the one of roland. everything before was unison, something entirely different, technically and soundwise. true unison stacks complete voices, not only osc's. therefore, if unison is 2x, you end up with 2x osc count you used without unison, 2x filter, 2 times lfo's, 2x times env's, etc... it sounds completeley different if you stack osc's, then feed them into a filter (if the filter has any feedback saturation), than if you stack voices, where all the individual voices have their own seperate signal path.

before VAs arrived unison was limited to 3x at most...so the core of the sound was made with oscillators anyway (pwm+saw+saw)

?? sorry, not true. real unison was made by using all availlable voices at once. so when the synth had 6 voices, all 6 voices were played with one note, just as simple as that.

the kind of sound the roland supersaw oscillator could produce was nothing new to my ears ..it was just more saws together than what i heard previously ...and an an easier way to achieve that sound (easier than doing it on the 12osc Cheetah for example).
the roland supersaw was a new oscillator...new technology...but a new sound ? very arguable.

it was a new sound. not _extremeley_ new or unique, but that wasn't the point of roland at all. the point was, to create an osc that is capable of doing very rich sounds using one osc only. and it was a processing power issue: real unison eats a lot processing power (which was a big issue back then on digital hardware synths, as is today on vsti's), as all voices are stacked. the supersaw was one osc that used the processing power of only two independant saw osc's, by that it was _very_ efficient, due to the fact that not the whole voices signal flow had to be multiplied/calculated. another benefit was, that you could have this "unison fake" fully polyphonic.
i could go more into deltail to explain, why a supersaw (in fact any superwaveform) is different from unison, you can do a lot with a supersaw that you simply couldn't do with unison (given there's more possibilities in audiorate modulation whatsoever in the synth) and vice versa, due to the pure natural differences of the making of both.

and about unison vs supersaw oscillator
i know the technical differences between stacking voices containing saws and stacking saws only at oscillator level.....but how can u say it sounds completely different ? ....

then do the following:
send a supersaw into a filter that is capable of feedback saturation. by this the whole conglomorate of all saws is saturated by only one filter saturation.
now do that with a real unison, where _each saw_ has it's own filter, thus it's _own_ saturation seperateley. completeley different sound...
or use osc1 as a supersaw as fm modulator for osc2. not even possible with real unison, as if you'd fm osc2 with osc1, you'd have only one single wave modulating osc2, after that happened the unison voice stacking is simply multiplying that result. sounds entirely different - because it _is_ completeley differentley done.

http://www.infekted.org/virus/showthread.php?t=26647
are u telling me , during the years 1998-2003 when trance was popular global phenomenon , only sounds made with the jp8000 were referred to as supersaw? i don't recall any other synth of the time which had a supersaw like oscillator. improperly or not u will find even sounds made with then Virus and the WaldorfQ were being addressed as supersaws, yet all were made thanks to the 8x unison and 32 polyphony (maybe u wanna go to the vengeance website and listen to virus C and waldorfQ demos.. no supersaw sounds to be heard right ?)

i never said that. and yes, during that time period the only synths that were able to mimic a supersaw were the virus and the q, (when it comes to polyphonic melodies), then later on the supernova... however, most of the trance songs were done with the jp and the virus during that period. ferry corsten started the typical supersaw sound back in 1996, he was the one that used the jp8000s supersaw pronounced as main melody on every song he did...

i'd be curious if u could recognize in a song whether a supersaw sound is made through stacking voices containing saws (unison)...or stacking just saws at oscillator level (supersaw oscillator style).i seriously doubt it.

i don't think i could, and i don't give a rats ass either. i personally think i found a lot of synths that actually do a better supersaw than the jp/virus, of which one is zebra and before all others it's helix... and i don't even give a rat's ass on using the supersaw for the typical trance melo's, i can't hear that stuff anymore, it's over for me... i'm completeley saturated with that sound, i had to create that stuff way too long, in productions and sounddesign wise.

that's actually why most analog synths that featured unison were monophonic then

how can a synth having unison (therefore capable of stacking voices) be monophonic. *confused*

in all analog synths that is a natural behaviour, it's due to the nature of what unison is. again, if you turn on unison on any old analog synth, it stacks the availlable voices to one note. again, if a synth was equipped with 6 voices polyphonic, when switching on unison, all 6 voices were used stacked on one note, therefore it _had_ to be monophonic. there'were rare synths that had an enhanced unison feature in the way that, if you hit only one note, all voices are used for one note, if you hit 2 notes simultanesly, the voices were split equally on the two notes, etc... using the 6 voiced synth as example again, this would mean: one note = 6 voices, two notes = 3 voices each note, and so on...

edit:ah ok u mean only available in monomode

nope. real unison on analog synths was _always_ monophonic, due to the very nature of it. with implementation-variations i described above, but those were very rare...
it seem you have to read up the very basic structure of a synth as well as what unison was on analog synths, to understand this better...

[Urs]

Example 1 seems to have some sort of compression/saturation as well. When I compared the waveforms, example 2 had quite a few peaks, where as example 1 was very even looking and all the peaks were obviously controlled.

Yes, I sent it through the wrong lowpass filter, with too much saturation. That explains the funky stuff as well as the more tamed envelope.

The new examples with different detune settings use a lowpass filter with less distortion, and a different highpass filter that boosts the fundamental a bit better to match the low end of the hardware.

[pac0r]

Wow the new examples sound much closer. In fact the differences are not significant enough for me to care which is which.

Nice job.

[bmrzycki]

Wow the new examples sound much closer. In fact the differences are not significant enough for me to care which is which.

Nice job.

I was just about to say that.

[pac0r]

I was just about to say that.

Ye, Urs Heckkenmaen is pretty talented. Much credit is due.

[mcnoone]

I can't get Urs examples to download.
It's not even giving me 1kb, then stops after a while.
bummer

[Urs]

Wow the new examples sound much closer. In fact the differences are not significant enough for me to care which is which.

Nice job.

Thanks!

[Urs]

I can't get Urs examples to download.
It's not even giving me 1kb, then stops after a while.
bummer

Try the mp3s I just uploaded:

http://www.u-he.com/music/Suse1.mp3 (little detune)
http://www.u-he.com/music/Suse2.mp3 (quite some)

[mcnoone]

I can't get Urs examples to download.
It's not even giving me 1kb, then stops after a while.
bummer

Try the mp3s I just uploaded:
http://www.u-he.com/music/Suse1.mp3 (little detune)
http://www.u-he.com/music/Suse2.mp3 (quite some)

Thanks Urs, but it's slow here tonight, I'll try in the morning.
Probably just traffic delays, or something.