Gigasampler problem with sustained notes

[Robert]

I have been experiencing some problems whith the Gigasampler when I try to play sustained or long notes.

I was looking for a remedy to this problem on this forum and found a posting describing exactly the problem I\'m having.

In short it said, and I quote \"Playing the midi file, the long strings notes are cut after few seconds. Muting the piano or drums and the strings are playing ok.\".

Does anybody knows any solution for this problem?

---------------------------------------
This is the complete post....


Hi,

We are using BC Drums for BD,SNR & HH (4 note max).
Giga piano light version (6 note max).
Strings (3 note max)
Hammond organ (3 note max).

Playing the midi file, the long strings notes are cut after few seconds.
Muting the piano or drums and the strings are playing ok.

As I wrote, we checked : DMA setting, Read behined cache setting, Virtual memory
manualy set to 256MB and other staf we have as a defult for DAW setting.

On one of our coustomers computer we re-installed GS and it improved polyphony but still there is nothing like a \"check list\" to do just to insure 64 note polyphony.

The \"voodoo\" thing is that other customers,
which we installed the software the same way, does not have any problem with polyphony...

I read some of the posting here in the forum that users have 16 tracks of audio + 16
tracks of heavy midi arrangment playing thru
one card (SB Live...).

Hope I\'ll find a way to have this running for me.
Thanx to all, please keep posting any idea.
Dovi.

[Chadwick]

-----------------------------------
BC Drums for BD,SNR & HH (4 note max).
Giga piano light version (6 note max).
Strings (3 note max)
Hammond organ (3 note max).
------------------------------------

A few factors which affect polyphony:
If a sound is stereo, this equals two voices not one.
If a sound is stereo and layered, this equals two for the stereo, and add another two for each layer.
I\'m not talking about switched layers, I\'m talking about crossfaded layers, which gigasampler regards as plying at the same time, but with variable levels in the mix.
Velocity switched layers don\'t have the layers being played simultaneously, they\'e discrete.
Even a crossfaded level which you can\'t hear probably uses up whatever voices it needs.

Also, notes which overlap (even the first note\'s very end reverb tail with the front of the next hit) cannot use the same giga voice. Even though they appear consecutive and not overlapping in the sequencer, if the envelope release on the actual voice makes the sound continue past the beginning of the next note, you double up the voices required.
Hi Hats are a real killer here because they\'re often chugging away at speed using four or more voices, especially when they\'re stereo. Snares aren\'t as bad until you do a 32nds jungle fill!!!!

Strings often use up double the voices you expect because with the more paddy stuff you\'re often looking for longer releases and a touch of overlap.

It seems that even if you allowed four voices for each sound you\'re using, you should just reach the limit of 64 poly for Gigasampler.

Maybe you should try the same piece with a simple mono/no layers stripped down version of the patches. I\'m not saying this is a good way to work, just a way to isolate the reason for your dropouts.

It might also be worth checking the releases on the envelopes of the sounds you\'re using.
Even though you hear a drum sound finish, if its release time is huge the voice doesn\'t become available until that release time is over. So you if play a bar worth of 16s on the snare, even when you hear the last snare fall away, the first one could conceivably still be \'playing\' as far as the Giga is concerned.
Mind you, no self respecting programmer would put kits together with release times like that.

The other possiblity I\'m not too sure about is dynamic voice allocation.

Certain dumb synthesizers used to grab the next voice slot not currently subject to a note on. That meant that if a sound was in release, it became available to steal for a new voice - even though it was still sounding its release phase.
Smarter synths don\'t do this. They wait until the sample has played to its end.
I really doubt that Gigasampler has this old fashioned allocation. Surely more people would have been complaining about losing notes by now.

Good luck
Rick

[This message has been edited by Chadwick (edited 07-31-2000).]