NRPN incompatibility

moz

I found some bugs in AWE NRPN implementation:

Filter Cutoff value (16277): values 0-127 doesn't works, filter is always on lowest value. I found, that any efect has values greater than 8192, but all songs from AWE uses ranges 0-127

ENV2 to Filter Cutoff (16280): original AWE NRPN uses values 64-127 for negative and 0-63 for positive modulation.

I'm use Audigy2 with kX3535.

Eugene Gavrilov

thanks for your report -- I'll try to reproduce this bug soon

/E

Eugene Gavrilov

I've tried to fix filter cutoff issue -- re-check it as soon as 3537 is out

however, concerning 16280: (taken from SB Audigy NRPN Implementation):

NRPN LSB 24 (Envelope 1 to Filter Cutoff) -- 16280

Realtime : No
Range : [-127, 127]
Unit : 56.25 cents

Data values greater than 0 cause a positive phase (from 0 to maximum) filter modulation via Envelope 1 with magnitude of 6 octaves at envelope peak. On the other hand, values smaller than 0 cause a negative phase (from 0 to minimum) filter modulation via Envelope 1 with magnitude of 6 octaves at envelope peak.

(compare, for instance, with LSB 23):
NRPN LSB 23 (LFO1 to Filter Cutoff) -- 16279

Realtime : Yes
Range : [-64, 63]
Unit : 56.25 cents

Data values smaller than 64 cause a positive phase (from 0 to maximum) filter modulation via LFO1 with magnitude of 3 octaves at LFO peak. On the other hand, data values greater than or equal to 64 cause a negative phase (from 0 to minimum) filter modulation via LFO1 with magnitude of 3 octaves at LFO peak.

-=-=-=-

note that range should be -127..127, not -64..63
so, what is the desired behaviour and is there any additional NRPN information available?..

/E

moz

I try 3537 and cuttof (16277) have same effect as in older versions.

I'm using SB0244, GUID: 576b1080-7b89-11d8-9b5d-0000b4a8dde5

seskanda

It seems kX drivers do NOT playback NRPN's correctly. I played a midi called The Nervous Filter.mid, and with Creative drivers it sounds right, but kX does NOT. You can clearly hear the bass filter swelling with the CL drivers, whereas with kX, the filter is barely applied to the bass. I've recorded the differences here:

RapidShare: 1-Click Webhosting

RapidShare: 1-Click Webhosting

Google for the 'The Nervous Filter.mid' it should be the first hit you find. The kX version was recorded with 3534f, AFAIK, it sounds the same with 3537 & 3538, i didn't try later versions though. Any idea what's wrong?

klerg

I'm surprised nobody responded to this. I, for one, can confirm that "The Nervous Filter.mid" file does sound wrong with kX. This may be because NRPN's work differently with kX...i dunno just a wild guess (more of a hope). Rapidshare has "Happy hours" right now, so it'd be easy to download your examples. I know kX can use MIDI CC's for filter control, but that leaves modulation, LFO's, & pitch left. It's about time for a mod or admin to chime in here....

Russ

You should send an email to Eugene (link at kX web site).

seskanda

I was considering to PM him, but since it seems he doesn't frequent this forum often anymore, i'll e-mail him, yet it says on kX's contact page "we cannot provide personalized support via e-mail."

Russ

Right, you shouldn't contact E. about stuff that can easily be answered by other kX users, but there is some stuff that only he can answer...

Eugene Gavrilov

I got your message, but need additional input to proceed.

E.

Maddogg6

hmm, I was playing around with kX NRPN's - and, becuase its been sooo long since I used CL drivers its hard for me to do a comparison, and state what should be, or whats different.

But I had an Idea - it would be very cool to be able to say, edit a text file that allowed advanced users to make custom 'maps' - or even better, edit a 'master' map to tweak how NRPNS behave:
Something like:
16080, (NRPN Data Range), (kX Register Range)

The 16080 being the kX register
the NRPN Controller Data Range - that can include the NRPN MSB and or NRPN LSB
Ex. 0 - 127 (all Positive, or adding to set values) or
-63 - 63 (use kX existing range) or
-8191 - 8192 (use maximum 2 byte resolution range)

And the kX register Range is the ammount of possible counts - like say, if ENV1 > Pitch, we can use the full resolution possible of NRPN data to set how much the envelope affects pitch.

This doesnt even have to be in decimal, go ahead and use hex as that may make things a little simpler... ???

A 'default' mapping is included with kX by default, and users can edit it them selves and share their findings that best emulate old CL behaviour, or ...

Maybe this list is reset with a sysex reset command or something....
Maybe a sysex command to re-read a users list, or these mappings are encoded in sysex it self some how.

Just some ideas I had, probably not good ideas, but what the heck, I wont know until I share it.

But I digress - as long as they (NRPN's) work, for me, is good enough. I can usually adjust/work around (use another FX buss out for a single midi channel and use kX filter - not preferred, but possible)

*wonders how SFz VSTi compares with CL drivers on this subject*

seskanda

I replied to your e-mail over a week ago, but have NOT received a response. Basically, i said 16278 (Filter Q) 16280 (Env 1> Filter Cutoff) are the NRPN's NOT working, since the bass filter is being modulated with CL drivers. I was NOT sure if 16277 (Filter Cutoff) works or not, because compared to "Microsoft Wavetable SW Synth" I could hear a more filtered bass with kX. Also, i was wondering how kX could have support for 'AWE32 NRPNs', when the document that described this (Soundfont 1.0 spec) was NOT released to the public domain?

Max M.

>how kX could have support for 'AWE32 NRPNs', when the document that described this (Soundfont 1.0 spec) was NOT released to the public domain?

The 'AWE32 NRPNs' were specified not only in SF1.0 spec. It was widely spread separate document in 90's.
(Additionaly, you should not hardly head for the 'availability to the public domain' of certain documents...
The Emu10k1 and Emu10k2 specs also were never released to the public domain - How can kX exist at all? )

seskanda

>The 'AWE32 NRPNs' were specified not only in SF1.0 spec. It was widely spread separate document in 90's.

If this document is so 'widely' available, can you post a link to it?


>The Emu10k1 and Emu10k2 specs also were never released to the public domain - How can kX exist at all?

That's news to me. How, then, could kX model those DSPs?

Russ

Not released to the public domain does not mean not available...
There is such things as non-disclosure agreements, patents and other publicly available documents, experimentation and reverse engineering, etc, all of which are sources of info.

Max M.

>If this document is so 'widely' available, can you post a link to it?

Sorry, i don't have a link.. As i said it was 'widely' available ~15 years ago...
Many of links of that time were lost...

edit: well, a little searching helps:
AWE & SB Live! Tips
This is the one i talked about. If i'm not mistaken originally it was a plain txt file (like this: http://archive.cs.uu.nl/pub/MIDI/DOC/FAQS/AWE32-FAQ) shipped on certain AWE32 CDs and downloadable from Creative web site. It may not look too much like a "spec." and not too much as "official" - but it does have all the necessary info.

>That's news to me. How, then, could kX model those DSPs?

maaaagic