Author |
Message |
DSPaudio
Joined: Nov 11, 2008 Posts: 12 Location: California/Japan
|
Posted: Mon Jan 25, 2010 4:45 am Post subject:
G2 16 -> 24bit memory upgrade possibilities? |
 |
|
Any ideas if the 16bit memory in the G2 could be upgraded to 24bit with a working result? I'm not familiar with hardware memory architecture but it seems like truncation is going on and if that truncation was occurring within the memory itself then we could just install proper 24bit memory no?
This and a DIY or aftermarket ADAT I/O board would really make my day.
Cheers |
|
Back to top
|
|
 |
Antimon
Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 371
G2 patch files: 100
|
|
Back to top
|
|
 |
dorremifasol

Joined: Sep 28, 2006 Posts: 823 Location: Barcelona, Spain
Audio files: 7
G2 patch files: 49
|
Posted: Tue Jan 26, 2010 11:48 pm Post subject:
|
 |
|
Even if you could make that modification, probably all the G2 code is designed to work with 16 bit memory, so in the end you will not achieve any difference at all.
But you can sue Clavia for advertising a device with 24 bit memory (in all the places where it's mentioned), and selling it with 16 bit memory instead.  _________________ Cheers,
Albert |
|
Back to top
|
|
 |
drapdap

Joined: Oct 11, 2004 Posts: 204 Location: London
Audio files: 1
G2 patch files: 1
|
Posted: Wed Jan 27, 2010 4:00 am Post subject:
|
 |
|
dorremifasol wrote: |
But you can sue Clavia for advertising a device with 24 bit memory (in all the places where it's mentioned), and selling it with 16 bit memory instead.  |
Could do, not a friendly solution though.
Wouldn't it be nice if they would just log in here as a user, and chat about these things, explain us kindly of their decision why using 16bit in the first place, how it could work modding it with new 64 bit memory, etc. sadly that's not the swedish style. their red is dead. |
|
Back to top
|
|
 |
ark

Joined: Mar 06, 2008 Posts: 679 Location: New Jersey
Audio files: 13
G2 patch files: 31
|
Posted: Wed Jan 27, 2010 8:18 am Post subject:
|
 |
|
Aren't we getting just a little ahead of ourselves here? What exactly does it mean to say that "it seems like truncation is going on?" Can someone post a G2 patch that shows this truncation happening?
If not, I don't understand what the problem is that the original poster is claiming. |
|
Back to top
|
|
 |
Antimon
Joined: Jan 18, 2005 Posts: 4145 Location: Sweden
Audio files: 371
G2 patch files: 100
|
Posted: Wed Jan 27, 2010 8:48 am Post subject:
|
 |
|
I think it's been established that the delay modules on the G2 work with 16-bit audio, while the rest is 24 bit. So if you pipe an oscillator into a delay, the output from the delay will be 16 bit resolution.
Is there such a thing as 24 bit memory?
/Stefan _________________ Antimon's Window
@soundcloud @Flattr home - you can't explain music |
|
Back to top
|
|
 |
ark

Joined: Mar 06, 2008 Posts: 679 Location: New Jersey
Audio files: 13
G2 patch files: 31
|
Posted: Wed Jan 27, 2010 9:19 am Post subject:
|
 |
|
Antimon wrote: | I think it's been established that the delay modules on the G2 work with 16-bit audio, while the rest is 24 bit. So if you pipe an oscillator into a delay, the output from the delay will be 16 bit resolution.
|
If that claim is indeed true, then the chance that a hardware change will fix it without a corresponding software change is zero. On the other hand, it is entirely possible that a software change might be able to fix it without a corresponding hardware change. I would think the main issue would be whether it would cause existing patches to fail to run because of increased resource requirements.
Of course, it seems very unlikely that Nord is working on such a software change. |
|
Back to top
|
|
 |
jksuperstar

Joined: Aug 20, 2004 Posts: 2503 Location: Denver
Audio files: 1
G2 patch files: 18
|
Posted: Wed Jan 27, 2010 9:47 am Post subject:
|
 |
|
The external bus of the 56k series is indeed 24 bits wide. And it is possible that truncation is happening in hardware, not software. But it would be a series pain in the ass to wire up 24 data + address bits for 4-8 DSP processors. |
|
Back to top
|
|
 |
ark

Joined: Mar 06, 2008 Posts: 679 Location: New Jersey
Audio files: 13
G2 patch files: 31
|
Posted: Wed Jan 27, 2010 10:01 am Post subject:
|
 |
|
jksuperstar wrote: | The external bus of the 56k series is indeed 24 bits wide. And it is possible that truncation is happening in hardware, not software. |
Seems unlikely--because in that case, I would think that it would affect all transfers to and from memory, not just some of them, and the software would have to take that into account for anything to work. |
|
Back to top
|
|
 |
DSPaudio
Joined: Nov 11, 2008 Posts: 12 Location: California/Japan
|
Posted: Wed Jan 27, 2010 10:37 am Post subject:
|
 |
|
ark wrote: |
If that claim is indeed true, then the chance that a hardware change will fix it without a corresponding software change is zero. On the other hand, it is entirely possible that a software change might be able to fix it without a corresponding hardware change. I would think the main issue would be whether it would cause existing patches to fail to run because of increased resource requirements.
|
There have been threads discussing the G2's 16bit memory and confirming it with patches. Someone even went so far as to implement a true 24bit delay line using multiplexing. I use the G2's delay memory in feedback configurations and to me it sounds like truncation, but I haven't investigated past listening.
Ark, I think you've overlooked that DSPs have paging memory. The G1 only had the local DSP page memory to work meaning super short delay times.
JKSuperStar, thanks for the insight!
If 24bit memory is possible, then the new expansion boards by Tony could be outfitted at the factory, leaving only 4 DSP's to swap them out on I love SMD.
And he'd have an upgraded, upgrade so to speak.
Do DSPs (like the Motorola 56xxx) use a memory controller or some sort of raw pipeline handled by the software?
You might think I should know this (with my name and all) but all our coding is native.
cheers, |
|
Back to top
|
|
 |
varice

Joined: Dec 29, 2004 Posts: 961 Location: Northeastern shore of Toledo Bend
Audio files: 29
G2 patch files: 54
|
Posted: Wed Jan 27, 2010 4:53 pm Post subject:
|
 |
|
For memory external to the DSP the G2 system uses a SRAM SMD IC (one per DSP) organized as 262,144 words x 16 bits (I've seen an Alliance AS7C34098-12TC part used). That memory chip (and the main and expansion board layout) only has 16 I/O data lines connected to the DSP data buss. That number of words divided by the G2 96KHz sample rate gives you about 2.73 seconds of data storage, which matches the 2.7 second max time of the delay modules.
So,the main and expansion board layout for the this 16 bit SMD IC make a 24 bit DIY "drop in" replacement not possible. The DSP code also is almost certainly written to do bit reduction for the memory writes and to fill the unused 8 bits of the 24 bit DSP register on memory reads, I would think.
The DSP code could have been written to use 24 bit data with that 16 bit chip, but it would at least double the read/write time (two R/W cycles instead of one per sample, maybe a show stopper?) and require a more complicated memory pointer scheme. And of course, that also would reduce your sample storage time by 1/3 down to about a max of 1.8 seconds. Many digital delay devices have options to reduce bit resolution and/or sample rate if longer delay times are desired. Too bad the G2 delays do not have these kind of options and Clavia failed to mention that the delays were only 16 bit.
Availability and price may have driven the decision to use 16 bit versus 24 bit memory. There is a larger supply and selection of 16 bit parts. I have not compared prices, but I would think that the 24 bit chips are a bit more expensive.  _________________ varice |
|
Back to top
|
|
 |
dorremifasol

Joined: Sep 28, 2006 Posts: 823 Location: Barcelona, Spain
Audio files: 7
G2 patch files: 49
|
Posted: Wed Jan 27, 2010 5:12 pm Post subject:
|
 |
|
The use of 16 bit memory is understandable for a number of reasons.
The false advertising is not. _________________ Cheers,
Albert |
|
Back to top
|
|
 |
3phase

Joined: Jul 27, 2004 Posts: 1189 Location: Berlin
Audio files: 13
G2 patch files: 141
|
Posted: Tue Feb 09, 2010 4:04 pm Post subject:
|
 |
|
dorremifasol wrote: | The use of 16 bit memory is understandable for a number of reasons.
The false advertising is not. |
what false advertizing?
too bad that the G2 dont has an additional page memory delay..even when the delaytimes are small the interslot busses of the G" would allow to go much ahead of that what s possible in the NM1... and would made NM1>G2 patchconversion much easier |
|
Back to top
|
|
 |
varice

Joined: Dec 29, 2004 Posts: 961 Location: Northeastern shore of Toledo Bend
Audio files: 29
G2 patch files: 54
|
Posted: Tue Feb 09, 2010 4:36 pm Post subject:
|
 |
|
3phase wrote: | dorremifasol wrote: | The use of 16 bit memory is understandable for a number of reasons.
The false advertising is not. |
what false advertizing?. |
Clavia did advertize that the G2 system has 96KHz 24 bit audio processing. In fact, the G2 1.4 Manual incorrectly mentions that the DSPs have access to “256 kWord of 24-bit RAM” (see page 138 of the acrobat file). Has Clavia ever publically acknowledged that the delays use 16 bit RAM? _________________ varice |
|
Back to top
|
|
 |
|