Forgive the cross-posting if you\'ve already seen this on the Tascam GS forum, but I was advised to post here.
Whilst trying to resolve a recent batch of GS technical issues, I noticed something weird happening in Gigastudio. Although the piece being worked on only uses about 6 fairly simple GS parts, the peak voice count seemed to be getting to the max voices set very quickly. So I kept turning the polyphony setting up, and the \'Peak\' count kept climbing higher.
Eventually, I narrowed it down to a couple of parts: I have two Garritan string parts playing very quick stacatto 16ths in one section, just single notes. However, if I solo these two parts (or either of them individually), instead of the \'voices\' counter in GS showing 1-2 voices only, it keeps climbing up as high as the polyphony limit I have set - ie. every note played adds a voice to the voice count, despite the fact that another note is finishing. Clearly not right.
I checked all the note lengths and they are all much less than a 16th. So I figured MIDI note off data may not be getting through to Gigastudio, or that it is not responding to it perhaps. Alternatively, I figured it could it be that the Garritan parts just have unecessarily long sustain settings, despite them being short bowed sounds (variations of 1st Violin Sul Tasto). Other parts not using Garritan sounds don\'t seem to cause this problem.
So, I ran some tests with the Garritan strings patches in question. They have a typical release time of approximately 0.58 seconds, meaning that at a tempo of 130 BPM, there should never be more than 5 notes and their associated release curves playing simultaneously. With a stereo patch (as this is) that means 10 voices being used at any one time. I confirmed that a single hit was using two voices, so no excessive voice usage there.
However, a single part playing repeated 16ths at this tempo is using between 24 and 28 voices in Gigastudio, according to the voice indicator, instead of 10. If I set the relase time to 0, the voice count is correct - never showing more than two voices (stereo) being used simultaneously.
With my limited technical knowledge, this seems to suggest that Gigastudio is not releasing voices
properly, for whatever reason. Perhaps somebody with more familiarity of the software could shed some light. Or is might there be some specific programming in the GOS bansk which could be causing this phenomenon?
Fo testing I\'m using a single part and whether I play different notes or the same not repeatedly, the results are the same. One note uses 2 voices, repeated single note, very short 16ths at 130BPM uses 24-28 voices.
By my calculations, and assuming my assumptions are correct, with a release time of about .5 seconds, it should be using around 10 voices.
I should add that my understanding of the programming side of GS is limited, so somebody please feel free to jump in, tell me I\'m a **** and explain what\'s causing the anomoly.
Incidentally, my assertion (if I\'m even making one) is not that Garritan Strings are at issue here - although I guess there could be some programming quirks causing the phenomenon - but that Gigastudio is not releaseing voices properly or quickly enough, which is severely limiting polyphony.
This is most likely a result of the “length control” feature of the standard short bows. You are getting overlaps that reflect the actual length of the raw samples, so the overlap remains relatively consistent except when you make modifications for significantly smaller release settings. To tell you the truth I hadn’t noticed that before. Thanks for pointing it out. It should be an easy thing to improve. I’ll try to add that to the current update before it goes out. By the way, the update’s short bow variability features will help with just this kind of rapid passage and, due to different programming for these new instruments, doesn’t have the polyphony problem you describe. Contact Gary by email and he can pass you on to me for further information.
I\'ve dealt with this Gigastudio voice-stealing bug with some of my clients. Two things I\'d suggest would be:
1) Set the max # of voices to 150-152 with 6-8 transition voices. I\'ve had some systems crash when the voices would get maxxed out.
2) Many of the GOS short bows use controller 83 to adjust the decay and release times. You can make adjustments within your sequence and see if this improves things.
Thanks Tom, are there any workarounds to improve the problem? I also use GSt96. You\'ve mentioned controller 83 and some other things. It could be really helpful to post a few more specific suggestions for improving the polyphony issue.
I wanted to return to this subject for a moment. I had hoped to make these modifications for inclusion in the present update but just ran out of time (next time it comes up for a vote I\'m going to favor having more than 24 hours in a day). Anyway, the line had to be drawn on this update package to get it out the door, so the short bow polyphony changes will be addressed in a future update. In the meantime, the new VAR instruments in the present update will be a good choice for problem passages, and they come with some valuable new features.
“For testing I\'m using a single part and whether I play different notes or the same not repeatedly, the results are the same. One note uses 2 voices, repeated single note, very short 16ths at 130BPM uses 24-28 voices.
By my calculations, and assuming my assumptions are correct, with a release time of about .5 seconds, it should be using around 10 voices.”
(TH) Set up this test again. Make the 16th notes loop indefinitely. While they are playing, go to the MIDI Control Surface page in GigaStudio and lower the “length control” GPC-8 (cc#83) fader. Watch the “Voices” polyphony meter in GS. It will read fewer and fewer notes as the fader is lowered. Depending upon the particular instrument, it will probably drop from 20-30 at the top to 6-10 with the fader all the way down. This is because you are applying a smaller value for the decay envelope as you lower the fader. A smaller decay value translates into shorter notes and reduced note overlaps – hence, the lower polyphony demands. This also gives a clue to the correct use of this feature. Faster passages require shorter note values and using this control to reduce the note length can keep polyphony demands to a minimum in faster passages. The default setting is 127 and represents the longest note value for a particular articulation – and the greatest amount of overlap.
The change I plan to make in a future update concerns the release envelopes (not the decay envelopes). I will set up a test very similar to Jules\', lower the GPC-8 fader to zero and adjust the release envelope until the polyphony meter reads “4” (the overlap of two stereo samples). Once that adjustment has been made, the maximum polyphony use should be around 10-12 with the GPC-8 fader at maximum. It should have minimal impact on the actual default sound of the instruments but give much better polyphony performance. Each of the short bows will be adjusted to give this kind of performance. By the way, if you would like to do this yourself in the Instrument Editor, you will probably find that the final release envelope values will lie in the 140 to 180ms range. If you need more detailed instructions, let me know.
As long as we’re on the subject of “length control,” I think I’ll talk a little about its application. “Length control” is primarily intended as a tool for use during editing to make detailed subtle adjustments to note lengths within an existing passage. This can most easily be accomplished by setting your sequencer for “drawing” cc#83 data into the track. This is especially convenient during editing because it allows very quick changes for note lengths. For example, all notes in a track can be shortened by a chosen amount with a single change in cc#83 data placed at the beginning of the track – no need to edit individual MIDI note lengths, whether it’s 10 notes or 1000. Even more useful is the ability to tailor a phrase, groups of notes within a phrase, or even individual notes, by quickly drawing controller data into the track. You could adjust a group of notes to gradually change from short to longer note values by drawing an ascending diagonal line over the span of notes. You could try this with the 16th note test above, using the 1st Violin Marcato, only change the tempo to 60bpm or less to make the effect clear. In general, the shorter the bow stroke to begin with (sautille, sul tasto), the smaller the contrast, because there is less room for change in the shortest bows. And keep in mind that a note can never be made longer than the raw sample length.
For those who prefer the standard note length control (by MIDI note length), the new short bow VAR1 and VAR2 instruments, added in the soon-to-be-released update, will be the short bows of choice. They tend to feel more intuitive as you play them. They will also give you a number of variability tools to combat the dreaded “repeated note syndrome.” They make, by their nature, polyphony demands closer to the proposed after-adjustment numbers shown above for the “length control” instruments.
We are trying to supply choices that satisfy a variety of possible working method preferences. Different situations may also call for different methods to achieve the most effective results. We try to put these choices in the hands of the user. I hope this information helps.