Topic: Limited polyphony: voice allocation error?

[voodoo]

Hi together,

In a german Genos facebook group someone posted a video showing that polyphony limit is reached easily when KinoStrings and KinoStringsLow are played layered on right 1 and right 2:

  https://youtu.be/sin4JYqtO5U

I found that this effects even occurs with playing only KinoStrings on right 1:

* Hold five notes in left hand, for example a C7 accord.
* Play a repeated 5 notes accord in the right hand
* After short time the sound is cut, leaving out the release phase of the notes.

How can this be?

* A regular voice can use up to 8 elements
* With a polyphony of 128 for preset voices, 16 notes should play at the same time.
* Perhaps the voice assignment algorithm if the Genos is not optimal? Least recently used notes should be cut off, leaving most recently used notes sounding.

Can anyone confirm this behaviour?

Uli

[Dromeus]

No, I can't confirm. When I try your example or the one demonstrated in the YT video, there's absolutely not cut of any sound.

I did a different test. I layered  KinoStrings, KinoStringsLow and KinoStringsNatural (octave shift +2). Recall that if you use the sustain pedal on such voices they will sound as long as you hold the pedal. Now I did the following:

1. Press continously (don't let go) a key (e.g. c)  in the bass area with the left hand.
2. Then press and hold continously the sustain pedal (use your foot for that purpose  ).
3. Now play a series of (diatonic) notes (c, d, e, f, ...) with the right hand. All will sound at the same time as the sustain pedal is held. Count the number of notes you are playing. Play exactly 20 notes. You end up hearing 21 notes simultanously.
4. Now let go the sustain pedal. The sound of the key you pressed in step 1. is still sounding!
5. Now do the same with a series of 21 notes. When you release the pedal, there is no sound. The very first note played in step 1 has obviously been cut when you pressed the 22th key.

Assumed that KinoString voices use two elements, a layer of three voices would use 6 parts of the polyphony. Then 21 keys would use 6 * 21 = 126. When you press one key more, 6 * 22 = 132, so with polyphony of 128 a note has to be cut which happens to be the very first one.

Makes sense to me, but it's kind of speculation as I have no information about the internal structure of the KinoStrings. What do you think?

[voodoo]

Micheal,

Thanks for your report. This is a very precise voice count.

Uli

[markstyles]

Genos has split the 256 voices into 2 categories, so for most of the instruments it is 128 voices, same as before.  A patch may have from 1 - 8 voices.. with most being very low. I believe probably in the data manual, it explains this a bit more. So a particular sound even though it sounds like one voice may have many voices in it. each contributing a little something, a lip smack. a release noice on guitar, etc.

If you pick the most complex voices, you indeed will run out of notes very quickly..  Yamaha is very good at 'voice stealing'.  That is as new notes are played, the machine scans thru and decides what voices it can end, without being audible. (just get lost in the over all sound).  Most of the time they do a very good job at this.

If you are working with a DAW and getting to the point, where you can hear these artifacts (and I've gotten there).. You can fix it by changing note lengths, attracts, the length of some notes etc. 

In short it's an artifact of the hardware, and design.. Otherwise they' have to up to 512 voices. then of course the machine would be ridiculously expensive.. no check that, they would never design it that way.

I have no idea of how the kino strings are designed or work.. Suffice it to say, this is the way the machine is going to work.