Author |
Message |
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Fri Apr 23, 2004 1:35 am Post subject:
polyphony rise with DSP expansions? |
 |
|
I'm wondering what kind of polyphony gain to expect with the expansions. Because, as you probably know, patch polyphony is determined by several factors:
1) computational DSP load
2) module memory load
3) module output zero page usage
Point No.3 is what makes me suspicious. I quote Rob:
Quote: | Every module output eats away something like
0.8% of memory. This is not the RAM memory but internal dsp 'zero page' scratchpad memory needed to store the output value of a module. This zero page memory is the fastest memory in the dsp and must be used for outputs, so other modules will have these output values available as fast as possible. [...] The maximum number of outputs in a G2 patch is slightly over one hundred, I think it is now 104 while on the NM it was 111 or something. This starts to count when doing patches with modules that have more outputs, like the eight output switches, muxes and counters. |
I quote the Clavia website:
Quote: | The G2 Voice Expansion board makes it possible to expand the polyphony in your G2 patches to twice the original. The G2 Voice Expansion board features four extra DSPs and RAM circuits |
Ah, yes. But is this zeropage also expanded?
In order for an real polyphony increase to happen, not only should DSP power & memory (Point No. 1&2) be expanded, but also the zeropage array. Otherwise, it will be possible to use more DSP/memory-hungry modules, but not simply more modules per se. The amount of possible module outputs will remain constant.
Of course -if the zeropage resides on each DSP and is not shared with others , the amount of possible module output should also increase. Shouldn't it? I'm just not sure...
The question doesn't bug me that much, as I'm more into the DSP-hungry stuff (shape oscillators, anyone? ) Nevertheless, it interests me.
What can we expect from the expansion (or the G2X, for that matter)? |
|
Back to top
|
|
 |
mosc
Site Admin

Joined: Jan 31, 2003 Posts: 18240 Location: Durham, NC
Audio files: 224
G2 patch files: 60
|
Posted: Fri Apr 23, 2004 6:07 am Post subject:
|
 |
|
I haven't received my G2 expansion board yet, but based on my experience with G1 and on what I've heard, you can expect to double the polyphony.
The DSP boards also contain the memory used for samples (delays, reverb, etc). This doubles. I'm not sure what zero page memory is, unless it is memory associated with the non-dsp general purpose auxilliary processor. This is used for I/O scaning, patch managment, global synth settings etc. This won't change but it isn't related to polyphony. |
|
Back to top
|
|
 |
blue hell
Site Admin

Joined: Apr 03, 2004 Posts: 24422 Location: The Netherlands, Enschede
Audio files: 297
G2 patch files: 320
|
Posted: Fri Apr 23, 2004 12:53 pm Post subject:
Re: polyphony rise with DSP expansions? |
 |
|
shoshin wrote: | I'm wondering what kind of polyphony gain to expect with the expansions. Because, as you probably know, patch polyphony is determined by several factors:
1) computational DSP load
2) module memory load
3) module output zero page usage
Point No.3 is what makes me suspicious. |
Zero page memory is a part of the DSP internal memory, so it's doubled as well (as any other DSP or delay memory based thing is) when using an expansion board.
the only thing that does not double is the amount of patch storage memory (your sound banks).
Jan. |
|
Back to top
|
|
 |
Tim Kleinert
Joined: Mar 12, 2004 Posts: 1148 Location: Zürich, Switzerland
Audio files: 7
G2 patch files: 236
|
Posted: Sat Apr 24, 2004 5:34 am Post subject:
|
 |
|
Excellent news! |
|
Back to top
|
|
 |
Anig Browl
Joined: May 05, 2004 Posts: 20 Location: San Francisco
|
Posted: Wed May 05, 2004 9:54 pm Post subject:
|
 |
|
Zero page means that there is no paging cycle to read those memory values. Motorola DSPs are multi-threaded so when you are reading in memory from an external buffer (eg the G2 delay line) you can program the chip to do something else until the internal memory is full and has to be processed. So you can have ~104x4 modules in an unexpanded G2 using interslot communication.
Jan, I saw you commenting elsewhere on the possibility of very very large noodles using itnerslot communication. But it appears to me as if the expansion will simply provide two parallel G2s, rather than allowing the 8 DSPs to be manipulated seperately. Thus you could have two identical big 4-slot noodles going on at once, but not a single one utilizing all 8 DSPs (routing the signal in a loop on the rear inputs wouldn't work as the input routing would also be duplicated in the patch.
I kind of wish Clavia would let the extra DSPs be used freely, although interslot communication might become too complex - can't recall if the DSP has sufficient plumbing to handle 8 incoming channels. I really do not care all that much about polyphony, although I have succumbed to the temptation to play chords more often When I had a micro modular I hardly ever minded having limited polyphony (even though I didn't have much other gear), though I longed for more timbrality so I could play more than one patch at once. The expansion board seems very competitively priced, but if it is simply a polyphony expander then I think I'd rather get an Evolver.
Then again I wish they'd put on a sensible amount of RAM too instead of a mere 256 kwords, it's not like it would have added more than about $20 to the cost. Ah well you can't have everything. |
|
Back to top
|
|
 |
blue hell
Site Admin

Joined: Apr 03, 2004 Posts: 24422 Location: The Netherlands, Enschede
Audio files: 297
G2 patch files: 320
|
Posted: Thu May 06, 2004 2:34 am Post subject:
|
 |
|
Hi Eddy, good to see you pop up !
Anig Browl wrote: |
Jan, I saw you commenting elsewhere on the possibility of very very large noodles using inter slot communication. But it appears to me as if the expansion will simply provide two parallel G2s, rather than allowing the 8 DSPs to be manipulated seperately. Thus you could have two identical big 4-slot noodles going on at once, but not a single one utilizing all 8 DSPs (routing the signal in a loop on the rear inputs wouldn't work as the input routing would also be duplicated in the patch. |
Well, I think it should be possible, as 200% patches are possible when using the Poly (sorry, Effects) area (remember we discussed this on the NM list for a new OS ? Well they implemented it ! (and there is no micro anymore for which it wouldn't work)). But I still can't test it .. as of yet there are no racks and no expansion boards available over here...
Of course there will be a limit of four audio busses, but using the MIDI for inter slot communications gives a lot of possibilities as well.
Jan.
ehm, where I said Poy I meant Common of course ... Last edited by blue hell on Sat May 08, 2004 3:00 am; edited 1 time in total |
|
Back to top
|
|
 |
Rob

Joined: Mar 29, 2004 Posts: 580 Location: The Hague/Netherlands/EC
G2 patch files: 109
|
Posted: Fri May 07, 2004 5:20 pm Post subject:
|
 |
|
Anig Browl wrote: |
Jan, I saw you commenting elsewhere on the possibility of very very large noodles using itnerslot communication. But it appears to me as if the expansion will simply provide two parallel G2s, rather than allowing the 8 DSPs to be manipulated seperately. Thus you could have two identical big 4-slot noodles going on at once, but not a single one utilizing all 8 DSPs (routing the signal in a loop on the rear inputs wouldn't work as the input routing would also be duplicated in the patch.
I kind of wish Clavia would let the extra DSPs be used freely, although interslot communication might become too complex - can't recall if the DSP has sufficient plumbing to handle 8 incoming channels. I really do not care all that much about polyphony, although I have succumbed to the temptation to play chords more often When I had a micro modular I hardly ever minded having limited polyphony (even though I didn't have much other gear), though I longed for more timbrality so I could play more than one patch at once. The expansion board seems very competitively priced, but if it is simply a polyphony expander then I think I'd rather get an Evolver.
|
When the expansion is fitted all 8 DSP's are on the same 'buss', so it is not like two parallel G2's, in fact it is one 8 DSP G2 with the expansion. Still a voice or an FX area can use only one DSP. But one slot can use the 8 DSP's, seven for voices and the eighth for the FX area. There is now a nifty Status module that gives the voice number in a format that it can be used on the new multiplexer switches and the new CtrlSeq which allows each voice to use its own set of parameters to do different things, even routing signals through different modules.
See it like this: slots and dsps have no particular one to one relation, for one slots it might be that one voice is in dsp1 and another voice in dsp6, while te FX area for that slot is in dsp7 (with expansion fitted). |
|
Back to top
|
|
 |
|