digistorm wrote on 2022-03-23, 06:19:

The GUSpnp is quite a complex card, I wonder if it will ever be possible to reimplement it at a reasonable cost. If flooding the market with inexpensive GUSses were the goal, it would be better to recreate the functionality of the original GUS. What it needs to do is well known (because of all the emulators that implement it) and the Interwave functionality does not offer that much. Trackers that can use it also support a variety of other sound cards really well, and the in-game support and MIDI functions are done way better with other sound cards…

On the tracker code, there is one bit to set to put the card in interwave mode.
Then, the notes period and panning values are different, only this.

Support/emulate fully the interwave is probably not really needed.

In theory the Interwave is more accurate and has some extra features, but I mean, in practice… I own an GUSpnp and 99% of the time it is doing GUS classic stuff. It is nice if it could be done, but I suppose re-implementing a GUS ace would be cheaper then the full Interwave card because behind the scenes the Interwave chip is way more complex then the GF1 chip, when looking at the data sheets. All that functionality just wasn't used by tracker style software (like the envelope generators and LFO's), they did it all in software.

Also, I have this subjective feeling / memory that the GUSpnp applies a stronger brickwall filter when emulating the GUS classic, and I remember liking the GUS classic more when playing demo's that use a lot of channels (like the demo Dope). It just isn't fully 100% backwards compatible.

Shishkebarbarian wrote on 2022-03-21, 16:59:

just saw this on LGR Blerbs. if this project ever goes anywhere, or anyone makes more of these, i would really like to buy one.

Indubitably! I love this!

Berzerkula wrote on 2022-03-24, 03:00:

Shishkebarbarian wrote on 2022-03-21, 16:59:

just saw this on LGR Blerbs. if this project ever goes anywhere, or anyone makes more of these, i would really like to buy one.

Indubitably! I love this!

If your main goal is to play mod trackers on an older machine, why not going for an AWE32 for less than 80 USD?
I've seen some AWE32 patches for some of the games with GUS support.
I've been recently seeing people spending 400-700 USD on Gravis card which doesn't make sense.
There are only 10-20 game titles supporting GUS. They must be really hardcore fans or extremely serious collectors for spending that much money.

It is unfortunate, because it locks a lot of people out who want to experience the DOS demo scene legacy on original hardware. On the topic of MOD players; I use Open Cubic Player both with a AWE equipped machine and a GUSpnp equipped machine, and I find the AWE sounding a bit off in hardware mode, in contrast to the GUS, compared with software rendering. Maybe the interpolation method works differently on the AWE (it does not do linear interpolation like the GUS) or there is some processing that cannot be disabled (with Cubic Player). It is still neat to play XMs and ITs on a system that is way too slow to do it in software 😉

digistorm wrote on 2022-03-26, 06:24:

It is unfortunate, because it locks a lot of people out who want to experience the DOS demo scene legacy on original hardware. On the topic of MOD players; I use Open Cubic Player both with a AWE equipped machine and a GUSpnp equipped machine, and I find the AWE sounding a bit off in hardware mode, in contrast to the GUS, compared with software rendering. Maybe the interpolation method works differently on the AWE (it does not do linear interpolation like the GUS) or there is some processing that cannot be disabled (with Cubic Player). It is still neat to play XMs and ITs on a system that is way too slow to do it in software 😉

I listened to some recording of two cards, I say GUS sounds low pass filtered and AWE32 seems to be not filtered and sharper in good or maybe bad sense.
In many of the tracks, I find AWE32 sound more pleasing . Maybe being on or off is matter of taste or what we are used to.

I am wondering if anyone compared GUS, AWE32 with some objective measurements (SNR, etc). Is there anything in GUS technical specs that makes its sound superior to AWE32?
Is the original GUS with 1 MB ram as capable of AWE32 in hardware wavetable, at least for XM, S3M, MOD files?

I seem to remember that the AWE has a different interpolation algorithm, and it also has a filter that is always in the audio path, but set to a "flat" curve by default. But the AWE has a few problems, though: It can only process 16 bit samples (so 8 bit samples have to be "enlarged" to 16 bits) and it doesn't support ping pong loops (so these also have to be expanded to emulate this). This means that the GUS will need less sample memory to play the same module as an AWE. Finally it's pitch range is a bit limited for use with trackers, so some notes cut off or are played a (few) octaves lower (as in Cubic Player). But I find both enjoyable to listen to, especially with a bit of reverb on certain modules.

Makes me remember that I never updates my own module code to play over AWE hardware back in the days, when I finally had money to buy one.
I still reused my software mixing code from my previous SB16.

(And I can't for my life remember which AWE-enabled module player I was using)

Can anyone please clarify if 1MB Classic GUS can play any module especially XM, S3M with 16 or 32 tracks with large samples? If there is a limit how much is it?

terryfi wrote on 2022-03-27, 18:14:

Classic GUS can play any module especially {..} with 16 or 32 tracks

The main limit from what I've heard is that old GUS isn't as fast as a chip and the more voice playing at the same time, the lower the max mixing frequency:
CD-quality 44.1KHz sampling rate can only be achieved when mixing up to 14 voice.
At 32 voices as you're asking, the maximum sampling rate that can be achieved is 19.2Khz .

Later cards (Interwave chips in later GUS and in this project's ARGUS ; and in my experience creative's AWE cards) don't suffer from that.

terryfi wrote on 2022-03-27, 18:14:

if 1MB {...} can play any module especially XM, S3M {...} with large samples? If there is a limit how much is it?

Old classic ISA cards (GUS, AWE, etc.) cannot directly read main memory on their own (outside of whatever is sent down the DMA channel) unlike later PCI cards (SB Live!, Audigy, YMF7xx, etc).
So whatever RAM is available on the card is your maximum limit of samples that you can load and play.

1MB card could at maximum play 1MB worth of samples from your module. Using anything larger would require some trickery (e.g.: swapping banks of samples in and out as needed - such as for exemple using two different sound font for two different parts of a song. I don't know how often this happens with real-world module players), or mixing it in software on the CPU (how the extra 32 voice of a AWE64 are done).
(Also see WinAMP's AWE output plugin as another example of trickery).

https://siliconpr0n.org/archive/doku.php?id=m … ster:amd:94201b
DIE-Scan of the InterWave seems to be available since december 2021

And nopes, still no interest in going semi-commercial (which the last 2 pages seem to be all about), but currently negotiating with someone who might do a run for the american continents - they're currently deliberating my proposal.
If someone wants to do that for Europe/Asia/Australia hit me up. They'll have exclusive rights to the layout/gerbers for 6 months and get a supply of InterWave chips for a decent price, after that I'll release the gerbers to the public.

If that fails, I'll sell off large parts of my remaining InterWave stock for profit and go open source (gerbers [yes, that would mean either going with the current state of the PCB or rerouting the card yourself] + BOM + editable schematic + documentation).

Therefore at the worst case you'll have a virtually open source GUS PnP in 6 months +/- 1-2 weeks.

digistorm wrote on 2022-03-23, 15:38:

In theory the Interwave is more accurate and has some extra features, but I mean, in practice… I own an GUSpnp and 99% of the time it is doing GUS classic stuff. It is nice if it could be done, but I suppose re-implementing a GUS ace would be cheaper then the full Interwave card because behind the scenes the Interwave chip is way more complex then the GF1 chip, when looking at the data sheets. All that functionality just wasn't used by tracker style software (like the envelope generators and LFO's), they did it all in software.

Also, I have this subjective feeling / memory that the GUSpnp applies a stronger brickwall filter when emulating the GUS classic, and I remember liking the GUS classic more when playing demo's that use a lot of channels (like the demo Dope). It just isn't fully 100% backwards compatible.

If you can find a supplier for the GF1 + companion chip, sure, the GUS Ace layout is also quite streamlined.
Sample rate drop can be done by updating the GUS PnP drivers _I think_ the SDK at least has some commented out code that implies that this feature was planned at some point.

And, when this is all sorted out and dealt with, I could also throw in gerbers of an ACE like version of the PnP, for the slightly lower end price spectrum 😉

Since it's your layout/project you can do whatever you want with it 😉
(Wouldn't mind a screenshot of the layout via PM tho ... I think we've talked about this before?)

shock__ wrote on 2022-03-28, 11:19:

If that fails, I'll sell off large parts of my remaining InterWave stock for profit and go open source (gerbers [yes, that would mean either going with the current state of the PCB or rerouting the card yourself] + BOM + editable schematic + documentation).

Why do people do this?

If you're going to open source just open source. What benefit to you is there of forcing people to reroute the PCB from schematics if they wish to make a modification?

maxtherabbit wrote on 2022-03-28, 13:17:

shock__ wrote on 2022-03-28, 11:19:

If that fails, I'll sell off large parts of my remaining InterWave stock for profit and go open source (gerbers [yes, that would mean either going with the current state of the PCB or rerouting the card yourself] + BOM + editable schematic + documentation).

Why do people do this?

If you're going to open source just open source. What benefit to you is there of forcing people to reroute the PCB from schematics if they wish to make a modification?

Because we (I) can.

Why do people try to force useless discussions on me? Think this is of any aid? If anything it just puts me into the "fuck it, call it dead and make a maximum profit"-mindset even more.
So gerbers for production with an added hassle of modifying the layout for scumbag moves (i.e. removing credits - yes, I've seen that happen) are closed source?

shock__ wrote on 2022-03-28, 13:50:

Because we (I) can.

Why do people try to force useless discussions on me? Think this is of any aid? If anything it just puts me into the "fuck it, call it dead and make a maximum profit"-mindset even more.
So gerbers for production with an added hassle of modifying the layout for scumbag moves (i.e. removing credits - yes, I've seen that happen) are closed source?

I really don't care what you do, I'm not interested in this project personally. I asked the question out of genuine curiosity because I've seen this before and I do not understand the mindset behind it.

Continue to threaten whatever you like, I remain simply unbothered

maxtherabbit wrote on 2022-03-28, 13:53:

shock__ wrote on 2022-03-28, 13:50:

Because we (I) can.

Why do people try to force useless discussions on me? Think this is of any aid? If anything it just puts me into the "fuck it, call it dead and make a maximum profit"-mindset even more.
So gerbers for production with an added hassle of modifying the layout for scumbag moves (i.e. removing credits - yes, I've seen that happen) are closed source?

I really don't care what you do, I'm not interested in this project personally. I asked the question out of genuine curiosity because I've seen this before and I do not understand the mindset behind it.

Continue to threaten whatever you like, I remain simply unbothered

You've had your 2 answers then (authority over personal effort & potential for misuse) ... whether you can make sense of it or not, is up to you.

Yeah because omitting the silkscreen layer of the gerber stack is totally impossible (to remove credit)

maxtherabbit wrote on 2022-03-28, 14:03:

Yeah because omitting the silkscreen layer of the gerber stack is totally impossible (to remove credit)

Never said it was impossible (hence increased effort), nor would it make sense in this case (credits in copper layer - which of course can still be edited).