Voice allocation and dispersion behavior

[Sans Nom]

Hi,

Very nice job with this synth but I feel that the way the voices are cycled using voice dispersion is a bit strange.

In poly mode, repeatedly playing one note will only trigger voice 1 (1..1..1). Playing 3 notes chords will only play the 3 first voices (123..123..123). As the detuning is not random but setup for each voice (as it should be), that behavior results in a static detuning while playing chords or single notes.
I don't know how a real JP8 is managing voices but it feels wrong. On my MophoX4/Tetra it's not working this way. Voices are cycled round robin. Playing a 3 notes chord will result in 123..456..781..234 etc... That way a detuning is less predictable.
What is interesting is that with the arpeggiator ON the JP8-V voice allocation becomes round robin. Repeatadly playing one note in this mode will cycle through all voices 1..2..3..4..etc, producing a nice detuning pattern.

Could you please clarify if that's the correct behavior?

[LBH]

Hi,

I think the real Jupiter-8 allways play the lowest availble voice. I think it's one of the synths that did'nt do round robin voice allocation. Have you seen otherwise?

But the Poly 2 in Arturias Jup-4 does work like exspected. It should stop all notes playing in release phase, no matter how many voices is played after. If you in example have 3 voices playing in release phase (no notes is pressed on the keyboard), then if just play a single note, then the 3 voices should stop playing. In Arturia version you have to play at least the same amount of notes to make this happens, and that's afaik not correct, as it would not stop the muddiness when using sounds with long release time, which should be the advantage of Poly 2 when using that mode.
I hope Arturia will correct this.

EDIT: BTW: Also the original Jupiter-8 did not steal notes, when a ninth note was played, while the 8 others also was pressed. Notes in release phase was stolen. I also think this was the case with Juno synths. I hope Arturia also will correct this. EDIT END.

[Yann_Arturia]

Hi, thanks for your feedbacks !

The real JP8 does not have a cycling voice allocation, it is kind of a reassign mode. If possible, it will use the same note for a given note. That's why if you play play three times the same note it will only use one voice, but if you play three different notes it will use 3 different voices. This voice allocation method is very logical with the particular JP8 envelopes. The envelope's level are never reseted to 0 when a note is retriggered, the envelope will begin its attack from its current output level.
For example if you have a long attack and a long release, if you play one note, release it and then play the same note before the release time has ended, you will notice that the envelope starts its attack from the current release level and not from 0 (leading to a shorter attack).
That's why the voice allocation is not cycling, to benefit from the property of the envelopes, which can be really useful especially for pad sounds.
(If you want to highlight this behavior, I suggest that you try modulating the pitch with envelope 1 set with a long attack and a long release, then play one note and repeat it).
You are right, the voice allocation works always like this except when arpeggiator is enabled.

As for the Poly 2 mode, that's how our JP8 unit is behaving and we decided to keep it the same way. And yes indeed the real unit does not steal voices while all 8 notes are pressed. But we saw this as a limitation and decided to allow voice stealing

[LBH]

Thanks for your reply Yann_Arturia.

About the poly-2 mode, then the original jupiter-8 manual say this:
"Poly-2 is very similar to Poly-1, again assigning only one synthesizer (Voice LBH) to each key played. The primary advantage of Poly-2 is that only the last note or notes played together recieve their natural Release length. Any notes played previously will have instantaneous Release regardless of the patch programs."
And;
"Poly-2 has the distrinct advantage of reducing muddiness in musical passages played with patches having long release lengths."

I believe my old Juno-106 Poly-2 mode worked like this (But it's long ago).
If the advantage is to reduce muddiness, then all notes in release phase need to be cut off, no matter how many notes is being played and how many notes was playing. Otherwise the muddiness can not be avoided, unless you play the excact same amount of voices all the time, and that's very limitating.
Pressed notes will not be cut off, and thus playing legato will also prevent a note to be cut off.
That's also a reason that the voice stealing of pressed notes is a problem.
Also if you wan't to play more notes than 8, then you in Arturias version can set a higher voice number. So it's not a limitation to not steal a pressed note.

The original behavior give the player the best control. So i still hope this will be corrected.

best

[Sans Nom]

Hi, thanks for your feedbacks !

The real JP8 does not have a cycling voice allocation, it is kind of a reassign mode. If possible, it will use the same note for a given note. That's why if you play play three times the same note it will only use one voice, but if you play three different notes it will use 3 different voices. This voice allocation method is very logical with the particular JP8 envelopes. The envelope's level are never reseted to 0 when a note is retriggered, the envelope will begin its attack from its current output level.
For example if you have a long attack and a long release, if you play one note, release it and then play the same note before the release time has ended, you will notice that the envelope starts its attack from the current release level and not from 0 (leading to a shorter attack).
That's why the voice allocation is not cycling, to benefit from the property of the envelopes, which can be really useful especially for pad sounds.
(If you want to highlight this behavior, I suggest that you try modulating the pitch with envelope 1 set with a long attack and a long release, then play one note and repeat it).
You are right, the voice allocation works always like this except when arpeggiator is enabled.

As for the Poly 2 mode, that's how our JP8 unit is behaving and we decided to keep it the same way. And yes indeed the real unit does not steal voices while all 8 notes are pressed. But we saw this as a limitation and decided to allow voice stealing

Thanks Yann for taking the time to explain how the original JP8 works.

On a slightly different topic, I feel that the max dispersions on certain parameters are too small : env, cutoff, reso and mod.

Cheers