In what context is 'MPU-401' meant here? Just that the waveblaster interface talks MPU-401, or the actual MIDI ports, so you can connect it to other devices for MT-32, Sound Canvas etc? Either physical hardware or using another PC for emulation through MUNT or such?

Because in the latter case, I don't think MPU-401 would actually be a hard requirement. A lot of MT-32-targeted software needs an intelligent MPU-401 interface, where I don't know of any non-Roland soundcard (not just MIDI interface) that supplies this.
Which would imply that you would run SoftMPU on it anyway. And SoftMPU does not require MPU-401. It also supports the SB DSP MIDI interface. And it could also be made to support more interfaces (I want to add IBM Music Feature Card support to it at some point).
So in that case the requirement would simply be 'a MIDI port supported by SoftMPU'.

MPU-401 port, but wavetable header is a big bonus.

Can you actually emulate a fully functional MPU-401 on a Creative SBPro2 with SoftMPU?

James-F wrote:

Can you actually emulate a fully functional MPU-401 on a Creative SBPro2 with SoftMPU?

I believe so, yes. SoftMPU virtualizes the MPU-401 ports anyway, so it doesn't matter whether you have a real MPU-401 or an SB DSP-based MIDI port on the other side.
SoftMPU captures all intelligent commands, performs the MPU-401 emulation, and then outputs 'dumb' MIDI commands to a MIDI port. Whether that port is MPU-401 or SB DSP or something else, doesn't matter. It just writes single bytes to it.

Since my expertise in ISA audio cards is seriously lacking I'll just take the easy route and choose Yamaha Audician 32 as the all-arounder. Maybe ESS, Terratec or Turtle Beach might have better choices but Yamaha Audician covers most important specs fluently enough: SB compatible FM-support (OPL3), wavetable header, MPU-401 support, driver support for both DOS and Windows. Add a General MIDI daughtercard to it and you got almost all covered (all but LA-synth).

I've got the Audician 32 in two of my retro DOS PCs and I couldn't be happier. If a GM daughterboard counts as part of the "one sound card", then I'll take the DreamBlaster X2 on my Audician 32 😉

Outside of the interrupt-driven playback of Legend Entertainment titles, most DOS games requiring an intelligent-mode MPU-401 yet handle their own MIDI playback timing, and only leverage the "intelligent" features of the MPU-401 in a very rudimentary sense, looking something like the following:

MPU Reset
o 0x331 0xFF - MPU Reset
i 0x330 0xFE - Command Acknowledged

SysEx Transmission
o 0x331 0xDF - Want To Send System Message
i 0x330 0xFE - Command Acknowledged
o 0x330 0xF0 - Start of Exclusive
...
o 0x330 0xF7 - End of Exclusive

MIDI Message #1
o 0x331 0xD0 - Want To Send Data (on Track 1)
i 0x330 0xFE - Command Acknowledged
o 0x330 1st Byte of MIDI Message
o 0x330 2nd Byte of MIDI Message
o 0x330 3rd Byte of MIDI Message (if applicable)

MIDI Message #2
o 0x331 0xD0 - Want To Send Data (on Track 1)
i 0x330 0xFE - Command Acknowledged
o 0x330 1st Byte of MIDI Message
o 0x330 2nd Byte of MIDI Message
o 0x330 3rd Byte of MIDI Message (if applicable)

So on, and so forth.

The majority of UART-only MPU-401 implementations only recognize/acknowledge the "0xFF: Reset" and "0x3F: UART mode" commands. Even with the limited example given above, you can see why this would be an issue, as (poorly-coded) software expecting acknowledgement of unsupported MPU commands will simply wait forever, resulting in hanging behavior (Hi, Sierra On-Line!).

Some hardware vendors took a slightly different approach, and designed their UART-only MPU-401 implementations to acknowledge all MPU commands, irrespective of underlying support, while also allowing/sending subsequent data bytes in "non-UART" mode. Given the example above again, you can see why this would successfully allow software expecting an "intelligent-mode" MPU-401 to work with such interfaces.

Soundcards that I've found to offer this "faux" intelligent-mode MPU-401 support include:

• Ensoniq's entire Soundscape line
• Mediatrix' Audiotrix Pro
• Yamaha's SW20-PC (lacks external MPU-401 MIDI OUT - "SB-MIDI" only)
• Yamaha's SW60XG (lacks external MIDI OUT)

Crystal/Cirrus Logic chipsets are an interesting case in that they will similarly acknowledge all MPU commands in "non-UART" mode, but subsequently ignore any data writes, so while the hanging behavior is avoided, there's no MIDI output to be had.

SoftMPU arguably makes this a moot point, but given "hardware-based" MPU-401 criterion, and for several other reasons besides, an expanded Audiotrix Pro is about as close to a one-card solution as I've ever found.

aleksej wrote:

I hope i can use DB by rules?

Everything that isn't prohibited is thus allowed.

My list of games is this - I'll put a best [imho] sounding music device in brackets:

Heroes of Might and Magic II (SC-55 or CD music)
Ugh! (OPL only)
The Incredible Machine
X-COM: UFO Defense (OPL)
Civilization (MT-32)
Golden Axe (OPL)
Dungeon Keeper (CD music, but AWE32 is required for advanced ambient sound effects)
Magic Carpet II (AWE32)
Tyrian (SC-55)
Baryon (I prefer its PCM music over MIDI)
King's Bounty & King's Bounty II (PC speaker only)
Warcraft II (Miditemp DoX-1)
Space Quest V (MT-32)
Dangerous Dave in the Haunted Mansion (PC speaker only)
Vanguard Ace (tracker-ish music)
Time Commando (SC-55)
Dune II (MT-32)

As it can be seen, I am not much into Intelligent MPU-401 mode games so I'll live withour it. I'd like to have an AWE32 due to it's superior music in Magic Carpet II. It's weak in other General MIDI supporting titles though, so there's a need for something else to handle that. This leaves out the actual AWE32 cards due to MPU bugs and loud noise.

A MT-32 is needed for Space Quest V, Dune II and Civilization. If I were tight on budget, I'd leave the MT-32 out and live on SC-55 music for SQ5 and Dune II, and OPL for Civilization.

So, an AWE64 Gold with, say, a Roland SC-55 or Miditemp DoX-in-the-box, and a MT-32 would do it for me.

Strong sides of this solution:

- SB Pro and SB16 compatibility
- AWE32 music
- no MPU bugs
- very decent GM and MT-32 sound via MPU-401
- low noise

Weak sides:

- CQM emulation of OPL music
- no mini-jack audio output (not an issue for me though since I use an external mixer anyway)
- no Intelligent MPU-401 mode (not an issue for me - I'm not in that type of games)
- the overall cost is quite high, up to $500 if we count an AWE64 + DoxBox + MT-32. If we take a SC-55 instead of DoxBox, it'll be cheaper.

Did I forget something?

P.S. Note that this is all theory. I never had such a setup.

IBM Music Feature Card master race!

GUS Extreme for me.

F2bnp wrote:

AWE32 with OPL3, wavetable header and upgradeable RAM. 😊

Thats my opinion too. If only ensoniq has a wavetable, with its UART that would be my choice.

There are no AWE32's with OPL's that are also free of clicks and pops in single-cycle DMA (and later stuff improves on their SNR, too), and even the best SB16 chipsets still get some faint ringing that only got fixed with the AWE64. The SB32 CT3930 might be a somewhat interesting cut down option, though.

Getting into further minor details, YMF289 and Yamaha YMF7xx OPL cores have a different sample rate from the original YMF262 and CT1747 OPL3 implementation, which may be causing imperfections in their output.

Anything that isn't SB16 compatible is obviously losing a little bit with late DOS titles. AWE support is nice in a couple titles; similarly GUS, Pro AudioSpectrum, CMS, etc. have their titles.

Any card is a bit of a compromise in some way, unfortunately. And then there's stuff like Star Control II that'll wreck your day when attempting to put together a build that does everything you want.

firage wrote:

And then there's stuff like Star Control II that'll wreck your day when attempting to put together a build that does everything you want.

What sort of day-wrecking trouble are you experiencing?

Star Control II has a quirk in that it doesn't like to see a second IRQ specified in the ULTRASND environment variable. I created a batch file for the game that truncates/restores that variable, and includes manual soundcard specification (/s:gravis), and have never found the GUS support to be problematic as a result, even in a system having half-a-dozen accessible/initialized soundcards.

As I recall, SC2 looks for other sound cards at their default addresses despite the manual parameter, and would just hang with the GUS and Roland SCC-1 installed at the same time.

firage wrote:

Getting into further minor details, YMF289 and Yamaha YMF7xx OPL cores have a different sample rate from the original YMF262 and CT1747 OPL3 implementation, which may be causing imperfections in their output.

James-F looked into that and found no differences in his captures. Do you have some test patterns you'd like to try?

firage wrote:

As I recall, SC2 looks for other sound cards at their default addresses despite the manual parameter, and would just hang with the GUS and Roland SCC-1 installed at the same time.

That's exactly the behavior that truncating the ULTRASND variable prior to gameplay will prevent.

E.g.,
SET ULTRASND=240,7,7,3,3
to
SET ULTRASND=240,7,7,3

Thanks for the tip. I'll check it out when I have the system running again.

gdjacobs wrote:

firage wrote:

Getting into further minor details, YMF289 and Yamaha YMF7xx OPL cores have a different sample rate from the original YMF262 and CT1747 OPL3 implementation, which may be causing imperfections in their output.

James-F looked into that and found no differences in his captures. Do you have some test patterns you'd like to try?

Where was that finding? Last I know, he measured a 1 Hz frequency deviation at 440, and both Ace and I personally ABX'd differences between CT2290 and YMF719 output in an X-Wing clip (SETMUSE test, CD-ROM version). On my part, I do think what I heard could have other sources between the chip and the recording input, but who knows. I call the issue an imperfection, but not really an actual concern in use.

firage wrote:

Thanks for the tip. I'll check it out when I have the system running again.

gdjacobs wrote:

firage wrote:

Getting into further minor details, YMF289 and Yamaha YMF7xx OPL cores have a different sample rate from the original YMF262 and CT1747 OPL3 implementation, which may be causing imperfections in their output.

James-F looked into that and found no differences in his captures. Do you have some test patterns you'd like to try?

Where was that finding? Last I know, he measured a 1 Hz frequency deviation at 440, and both Ace and I personally ABX'd differences between CT2290 and YMF719 output in an X-Wing clip (SETMUSE test, CD-ROM version). On my part, I do think what I heard could have other sources between the chip and the recording input, but who knows. I call the issue an imperfection, but not really an actual concern in use.

Bah a 1Hz difference... Let's just say all music that is not tuned to A=440Hz ia an issue and thus imperfect.

I doubt tuning the OPL to 439 Hz is an artistic choice. 😁 Anyway, so the output being out of tune is just the easily verifiable part, the other part is aliasing artifacts. As insignificant as it is, there does seem to be *something* there that people can hear, and some people even claim to be bothered by the issue.

Pavarotti canceled a concert once for not being tuned to 432Hz but 440Hz.

clueless1 wrote:

I've got the Audician 32 in two of my retro DOS PCs and I couldn't be happier. If a GM daughterboard counts as part of the "one sound card", then I'll take the DreamBlaster X2 on my Audician 32 😉

Does the X2 actually fit on the Audician 32? I know that's a short sound card.

It should. The X2 is quite compact.