An interesting fact about the GM standard is that effect processing is optional.
So beside the instrument samples In actual implementation the effect processing can be quite different. It can be processed by channel and/or globally. There can be different algorithms for Reverb etc. and also fixed parameter settings or tunable.
This gives a wide range how wavetables could sound and often it helps to play with the parameters.
e.g. in default settings the effect processing of an EWS64 (Dream SAM9407) is set very weak, while on the Maxi Sound 64 series using the same chipset the default parameters prefer stronger effects. However both can be tuned and there have been posted settings here on vogons that move the SAM9407 effects close to the strength known of a sound canvas. I don't know if anyone applied this on later Dream Chipsets though (like 9733).

boxpressed wrote:

Thank you for the excellent explanation. Very helpful. So, if a card is advertised as "General Midi" compatible (often with the logo), then its synth will have at least those 128 specified instruments. Did the Gravis Ultrasound ever advertise itself as GM-compatible, or did it try to "fake" GM with its patch set?

Since the GUS is not actually a MIDI synthesizer, GM doesn't really apply, and I don't think they advertised it as such.
However, with MegaEm, it is compatible with a Sound Canvas connected to an MPU-401 interface. Since the SC is GM-compatible, the GUS is also GM-compatible this way.
I'm not sure what you mean by "faking"... all GM-compatible wavetable synths use a GM-compatible patch-set. That's not faking, that's just how it works.

boxpressed wrote:

Also, I know that some cards have a real MPU-401 controller, and others emulate it in software with a TSR. I assume that the difference between the two has to do with compatibility, quality, or both?

It all makes sense if you look at the history.
The MPU-401 interface was just an ISA card to connect your MIDI devices to your PC. Devices such as the MT-32 and Sound Canvas were not specifically designed for use with PCs/games. They are 'modular synthesizers', basically synthesizers without the keys. They were originally designed for use in a studio. You could have a single synthesizer which you use as your keyboard, and you can connect it to various other synthesizers/modules to use their sounds. This is a cheaper and more compact way than using full synthesizers with keyboard.

Because the MT-32 was a rather cheap modular synthesizer, it became a popular option for gamers as well.
Roland also integrated an MPU-401 and MT-32 on a single card, the LAPC. Later cards did the same, and retained MPU-401 compatibility, because that was the standard way to connect MIDI devices to PCs, even if technically you didn't actually connect a MIDI device to your PC anymore, but it was all on a single card, without any physical MIDI cables (or even an actual MIDI-interface, but rather a software layer which understood MPU-401 commands and sent them directly to the native synthesizer).

Thanks for the knowledge, elianda and Scali. Interesting about how different processing on the same samples can make a difference. Would the MIDI controller IC determine the level of processing? For example, I have two cards each with the Crystal CS4112-CS IC patch set. But one has a Crystal CS9233-CQ controller, the other has a CS9203-CL.

About the GUS, I wasn't sure if it had a GM-compatible set or not, so if it didn't, I wondered if it tried to "fake" GM by matching the 128 instruments as best it could to its own set.

I added another card to my collection today. This is the Reveal Sound FX Wave. It uses the Opti 82C929 controller and the Crystal CS4248-KL codec. It has a Crystal CS4112-CS patch set and has a discrete YMF262-M. No TSR required.

@Scali - you mean the MT32 (, etc.) is a 'synthesizer module' rather than a 'modular synthesizer'. 😉 Those are two very very different things. Sorry for the pedantry.

boxpressed wrote:

About the GUS, I wasn't sure if it had a GM-compatible set or not, so if it didn't, I wondered if it tried to "fake" GM by matching the 128 instruments as best it could to its own set.

GUS uses RAM for instruments and samples are loaded dynamically. So there is no fixed 'own' set. The GUS software does this quite simple as there is a text file that defines which Instrument Sample file is loaded for which instrument number (for different RAM amount on the card).
MegaEM joins this e.g. to a 1 MB memory image, using the UltraMID functionality it is loaded dynamically per song (however it is still GM).
The flexibility beyond the default instrument sets means that you can either change the instrument samples or the mapping with the text file. As long as it fits within the GUS memory it will work.

Right -- how could I forget? I own a GUS Classic and ACE. Somewhere along the way I guess I confused the GUS with the AWE32, which uses RAM but also has a patch set in ROM. However, I didn't know how the GUS maps the instrument patches, so thank you again.

Here is my latest acquisition. It is based on the Opti 925, which uses Opti's OPL3 clone. Not good. It uses the QDSP QS702/QS703 patch set for GM, but the controller is an Opti/QDSP 82C941. The GM is quite different from the Admos QS700 that often goes with this patch set. The Opti 925 DOS/Win 3.1 drivers on vogonsdrivers work just great with this card.

The card says OPT32W3D for its FCC ID at the top. The back has a sticker that says S/N: OP 9250051515. If anyone knows who manufactured this card, I'd like to list it in the first post.

EDIT: I think this card may be the Super Sound Origins 32 by SIC.

I see a YAC-512 in thar card. Isn't that meant to be paired with a discrete OPL3 chip?

The FCC ID will give you a company name:
http://transition.fcc.gov/oet/ea/fccid/ (enter it in product code)

Not sure if they actually produced it or only distributed the cards.

Some website gives further info on the company name:
http://www.perfectleads.com/company/964179/Sic-Resource-Inc

"Ftb Suspended" sounds like something bad, so the company should be long gone.

Did a little sleuthing. It might be a Super Sound Origins 32W-3D by SIC Resource.

The codec in the photo in the link below is a Crystal (not AD like mine), however. And the text also mentions an OPL3.

https://web.archive.org/web/19970110163025/ht … e.com/sso32.htm

alexanrs wrote:

I see a YAC-512 in thar card. Isn't that meant to be paired with a discrete OPL3 chip?

AFAIK it can be used as the same D/A converter for similar OPL3 circuitry.

If you ever wondered how OEMs personalized those Opti 924/925/929/931 drivers. 😀

To be honest, I like "Supper Sound" much better than "Super Sound."

(Also the init files seem to be a little more recent than the reference Opti 925 drivers on vogonsdrivers.)

Maxi Sound Game Theater 64 has 4mb memory, and it can be expanded to 16mb by edo ram.
One cool thing about it is that it Works just fine with pure MS-DOS, Windows 9x (not at all with me), Windows 2000 and EVEN Windows XP 😳

Funny you mention the Maxi Sound Game Theater 64. One showed up on eBay a few days after your post, and I grabbed it.

Great card. As others have mentioned, the ESS1868 has a good OPL3-like sound. I like the sound of the Dream 9407 synth and the 4MB patch set quite well. I need to get a 16MB SIMM so that I can try out the larger patch sets.

A new addition to my onboard wavetable collection: the Turtle Beach TBS-2000.

* PNP
* Crystal CS4232-KQ
* OPL3
* ICS WaveFront wavetable (2MB, I think)
* Needs TSR for wavetable
* Wavetable has separate I/O (0340 default) from MPU-401 I/O (0330 default)

GM music is average. Still nice to have a different patch set in the collection.

I wonder if there's a way to hack flash ROM or SIMM capability onto this board. On the Tropez+, I believe the sound font can be changed within DOS which is a big advantage over the AWE32.

Like the Tropez Classic, the Turtle Beach TBS-2000 uses a 2MB stripped down version of the VoiceCrystal soundset which is 4MB. Only the Tropez Plus (and the RIO daughterboard) uses the full 4MB soundset.
Apart from that, the sample quality is above average. The GM synth performs really well despite the lack of an effects processor. A discrete OPL3 circuit is also a nice thing to have on a card. An advantage over the Tropez classic is the very good SB Pro compatibility of the classic CS4232 codec. The Tropez Classic has the OPTi929 sound chip which is also a good one but not as good as the Crystal.

I have a CS4232 based card as well, although not as nice as your TB. The CS4232 has very clean sound, much better than any of my SB16 cards.

My latest acquisition: the Turtle Beach Maui wavetable upgrade card.

This card is in interesting alternative to those looking to upgrade to GM music from FM synthesis but are concerned about the SB16's hanging note bug or stuttering issues with Duke3D.

The Maui's only function is to add wavetable MIDI to your digital audio (and FM synthesis) sound card. It has no capability for digital audio or FM synthesis. Because your SB16 will only be used for digital audio and not MIDI, you avoid the hanging note bug.

The ICS WaveFront 2MB ROM patch set is average, IMO. Better than the EMU8K and AdMOS sets in the cards I showed earlier in the thread, not as good as Roland, Yamaha, or Kurzweil.

Messing around with Awave, I'm having trouble exporting sound fonts to WFB due to bank limitations of the WaveFront chipset. Does anyone know anything about this?