Send more polyphony

[paynterr]

I wonder how the next versions of GS will fare on the polyphony front.
It seems that with Halion coming along quite quickly behind Tascam will not be able to afford to continue selling the product at different prices based around a polyphony fix of 96 or 160?
I\'d be interested to know what sort of \'factory\' polyphony values Tascam can get out of GS in development and whether we\'ll have access to taht... it seems that technology strides along in leaps and bounds... if I want a top of the range box with top of the range storage, I should be able to get more that 160 voices.
I\'m interesting in other peoples thoughts on this?

[thesoundsmith]

I\'m curious how many voices you think you would need (upon re-reading this, the tone sounds confrontational, but there\'s no sarcasm implied, just curiosity.)

Let\'s say we want a hundred-member symphony, each instrument independently in stereo, with overlapping decays-that\'s occasionally 400 voices, is it not? (I\'m new to Giga, and may be missing something.)

With each instrument on its own MIDI channel, you\'d need 7 MIDI ins INT(100/16)+1. If we made it 96, you\'d need 6 MIDI ins.

Then let\'s include a piano, harp and organ-assume 16-note stereo polyphony for each, plus decays- 32 * 2 * 3=192 voices, plus the other 93 * 2 * 2=372, total 564!

All those samples streaming off the disk makes for a LOT of throughput-I don\'t even want to THINK about calculating the bandwidth requirements.

The sofware could probably keep up, but I don\'t think the disks are going to be happy about it, so there\'s a long way to go for the Laptop Philharmonic to make it to real-time...

That\'s why God made multi-track recorders! http://www.northernsounds.com/ubb/NonCGI/images/icons/grin.gif

But seriously, what do you feel would be a reasonable expansion? What would let you feel you could do everything you want the great majority of the time? My current vote is 256-nice round binary number, and shouldn\'t put a major strain on system resources... Of course, I don\'t do symphonic, so my max instrument pool is a twenty-piece big band plus pop strings, well within the current limits.

I\'d settle happily for the current 160 voices with GUARANTEED no popping!!! http://www.northernsounds.com/ubb/NonCGI/images/icons/rolleyes.gif

Dasher

[This message has been edited by thesoundsmith (edited 11-29-2001).]

[Joris Vincken]

Whatever the math, in a few years we will have 1024+ voices and not even think about it anymore. Two years ago with Gigasampler 160 voices would be the ultimate dream. Now you don\'t even need a top system anymore to get that http://www.northernsounds.com/ubb/NonCGI/images/icons/smile.gif

[Chadwick]

What Soundsmith says is pretty right.

Even if you have 32 layers of velocity, if you play one note, and the sample in that velocity layer is mono, then you use only one voice. No matter how complex the instrument is, it doesn\'t use any voices up until you start hitting keys. Once you do, it\'s down to stereo, layers, crossfades, release triggers.

As SS points out with one very important polyphony chewerupperer is releases.

You can use up your whole 160 voices of polyphony with one single stereo sample of a ride hit, by playing 16ths, if the ride has a nice long release:

A stereo ride cuts poly to 80;
1 bar will use 32 voices if masking is off (16ths x stereo), and the release is long (first note still in release as the 16th note is hit);
By half way through bar three, you\'ve hit your poly limit(again, if the release is nice and long)!

Of course, that\'s totally impractical and stupid, and we all avoid it by using masking, shorter release times capturing to disk, and not playing like Buddy Rich abusing illegal substances.

More polyphony is always going to be great, because it means we can layer and crossfade without feeling the need to calculate whether it\'s worth the polyphony cost.

I think Joris is right. These things don\'t go up by small percentages, they seem to go up by nearly quantum leaps (sorry Nick).


[This message has been edited by Chadwick (edited 11-29-2001).]

[pantonality]

I think the math works out a bit differently. I may be mistaken, but I seem to recall that velocity layers count as one polyphonic voice each? Could somebody who actually knows offer some insight into this? What I do know is that the situation is more complex than than 160 voice polyphony means you can easily emulate a full orchestra (though you can get pretty close).

Steve http://www.mp3.com/stevechandler (\"http://www.mp3.com/stevechandler\")
aka Ettienne http://www.mp3.com/ettienne (\"http://www.mp3.com/ettienne\")

[thesoundsmith]

Thanks for that, Chadwick. I actually hadn\'t thought about multiple releases on the same sample ala cymbal rolls.

But you mentioned another possibility-crossfades-do these eat up max two adjacent layers simultaneously, or can they get more complex? (can\'t wait for David Govett\'s tutorial, so I can start diggin in!)

But I also think Joris\' point is valid, although it doesn;t seem to me that simple CPU speed is much help here. The bottleneck is more likely going to be the hard disk. Now if the molecular storage IBM is working on can be brought to market reasonably, THEN we\'ll really see some polyphony! http://www.northernsounds.com/ubb/NonCGI/images/icons/grin.gif

Dasher

[paynterr]

No - my comments were not supposed to be confontational... just curious really.
Technology does move quite quickly and I was just musing as to whether we have another great leap ahead of us just around the corner reference polyphony.
I suspect there is... I foresee - in a couple of years - us being able to have more polyphony on a single machine than we perhaps need, which will be nice...
CPU isn\'t much of a factor so it should be possible to have a single PC loaded with memory and CPU running all our applications. It is just the storage bandwith that is \'behind\'at the moment, although I use that term very loosely since we have seen quite a remarkable movement in the music arena in the last few years.
I\'m just speculating (read: dreaming http://www.northernsounds.com/ubb/NonCGI/images/icons/wink.gif about where we\'ll be going next...

[thesoundsmith]

Paynter,

I was referring to MY comment...

\"I\'m curious how many voices you think you would need\"

I wrote it \'curious-ly\' but when I read it back, it sounded confrontational, which was not my intent...

And I agree, the issue is not CPU speed, but storage and bandwidth.

I just read that IBM has created a single-molecule NOT gate (inverter). This is the basis for a new generation of REALLY fast hardware, once they can get the mass duplication/manufacturing details nailed down-could be up to 1000 times faster than current hardware (and much smaller-can you say terabyte CPU speeds in a wristwatch computer???) So the old \'notched-molecule\' storage talked about in the late seventies could actually become a reality, and THEN we\'ve got our \'Symphony-on-a-stick\' machines in out shirt pockets.

If we can just live long enough... http://www.northernsounds.com/ubb/NonCGI/images/icons/grin.gif

Dasher

[seclusion]

That\'s why when everyone looks at me stupid cause I have a midsport 8x8 installed on the giga system, I say you\'ll see.....
It\'ll come and most likely they\'d double the poly.. So I\'m ready if they open 4 more ports for 8 midi ports... Seems like the way they do things... Giga 96 only used 2 ports.. Though I go through the poly just on the giga piano at times.. If I really try to..
Double the poly, with the kick assed puter systems and that\'ll shut Halion fans up in a hurry..
Brian

[thesoundsmith]

And in the meantime, I\'m just using GS 96 and multi-track...
But I\'m going to upgrade while the price break is on! http://www.northernsounds.com/ubb/NonCGI/images/icons/grin.gif

Dasher

[Kenn159]

Quote

But I also think Joris\' point is valid, although it doesn;t seem to me that simple CPU speed is much help here. The bottleneck is more likely going to be the hard disk.

Quote

It is just the storage bandwith that is \'behind\'at the moment.

Reply

Actually I don\'t think the bottle neck or weakest link in a state of the art ,current system is the CPU or hard disk.
I think its the PCI buss .
Giga studio has to comunicate with A PCI sound card ,which is limited to the bandwith of the PCI buss, which is 33 megahertz.

[thesoundsmith]

Good point Kenn, but the output is still just digital audio @ 44.1 KHz; even with 64 stereo channels (4 ports X 16 MIDI channels), that only needs 64 X 2 X 44,100 bytes/sec = 5.6 MHz bandwidth, well under the 33 meg the buss can theoretically handle.

My concern is that we\'re asking the computer to read streams of data from a large number of locations (one per mono voice) from one 7200 RPM hard disk, locations which are constantly changing as note events begin, end, and change from layer to layer.


Dasher

[David Govett]

There is no such thing as too much polyphony. Hanz Zimmer is probably pushing the limit of his system (at least 20 GigaStudios, do the math) Another example is the harp glissandos using the GigaHarp. I have dedicated a small, dirt cheap computer just to do the harp because it will eat up your whole system if you play it like a real harpist using the Glissando variations.
Anyway, The most polyphony you can get under a single note is 8 (stereo x 4 layers) These layers can also crossfade of course. If you add a release trigger, then that can add another two notes briefly. Velocity switches and dimension switches don\'t use any extra layering type polyphony.
Have fun
David Govett