zaphod77 wrote on 2026-08-05, 13:29:

Here's what i would actually like to see among the output soundfonts. […]
Show full quote

Here's what i would actually like to see among the output soundfonts.

1) sc55mkII with CTF
2) sc88 with ctf
3) sc88pro with XG mode and ctf
4) 8820 with xg mode and ctf.

number 1 is actually the one i'm most interested in. something that with a fully compliant synth (fluid/bassmidi/spessasynth)should playback the three midis i've linked correctly.

CTF is not a property of the SF2 soundfont itself. But all known SF2 based synths (including Sound Blaster's HW SF2 synths) support a simplified form of the SC-55 CTF by default. That is in case of missing instruments they fall back to Bank 0 (Capital Tones) in case of melodic channels and Program 0 (GM/GS Standard Kit) in case of a Drum channel.

It can certainly be put into a soundfont directly, if the synth doesn't already do it, simply by adding instrument definitions that implement the fallback directly.

And yes, i remember that real creative cards did have a fallback. I'm not sure if it was direct to bank zero, or if it would pick the highest bank that was lower than the one with the missing instrument.

zaphod77 wrote on 2026-08-05, 16:18:

It can certainly be put into a soundfont directly, if the synth doesn't already do it, simply by adding instrument definitions that implement the fallback directly.

And yes, i remember that real creative cards did have a fallback. I'm not sure if it was direct to bank zero, or if it would pick the highest bank that was lower than the one with the missing instrument.

Most likely it's because the original AWE hardware had fixed ROM presets at Bank 0 all the SB hardware I tested (SB LIve!, Audigy, Audigy 2 ZS) always fall back to Bank 0 in case of missing melodic presets and always fall back to Program 0 in case of missing Drum presets. There are no in-between fallback steps even if there are in-between instruments. BassMidi behaves the same way. That's why I told that nothing should be done in the soundfont itself (besides providing the default GM/Capital tones at Bank 0) since most SF2 engines work this way (and only this way) anyway.

Okay, the TVF seems properly hooked up now. right in the night sounds awesome.

rtft is now sounding great as well.

i'm very impressed.

so pretty much everyone buy Roland was able to have capital tone fallback anyway. that's... stupid.

Only issue i've spotted now is that seems some samples are slightly out of tune at some pitches in sc55 mode only. opened an issue with a test midi for it. you can tell right at the start what's wrong.

Now we just need a clap plugin, a yst2, and a vst3. and some insertion effects figured out, though i haven't found something that uses those yet. but note that variation in XG counts as a single insertion effect that can be set many ways, with send levels for individual channels.

So i still think this needs to be figured out at some point. But for playing typical retro gs midis (which rarely, if ever use the sc88Pro insertion effects), it works great.

Okay, the previous issue i've found is resolved, and my mother 1 pollyanna midi plays just like it does in the actual Sound Canvas VA

Okay, this is insane. i now have a midi that both this and sound canvas va screw up exactly the same way in sc55 mode (!) I was about to open an issue, but decided to test in the actual SCVA and the same error happens there, ONLY in sc55 mode!

Now i'm curious if the real SC8820 will do the same with this midi file in sc55 mode. Anyone have one to test? The midi file is

https://gifx.co/music/MIDI%2FCorona%2FBaby%20Baby.mid

There's supposed to be a downward sweep one beat after the "baby baby" synth. Works correct on real sc55, nukes sc55, and on va on everything but sc55 mode. Those exact results are replicated in this web app. I'm not sure if the innacuracy is in scva, or in the sc8820's sc55 mode.

RTFT sounds significantly different on nuked vs sound canvas VA as well, but the new project sounds like sound canvas VA, so i can't report this as a bug.

By the samples are usable for their intended purpose loophole, i believe this project is likely to be legal to use if you own SCVA.

I can confirm the sound canvas va behavior is incorrect by comparing vs nuked-sc55 on a mkII rom.

zaphod77 wrote on Yesterday, 20:48:

Okay, the previous issue i've found is resolved, and my mother 1 pollyanna midi plays just like it does in the actual Sound Canv […]
Show full quote

Okay, the previous issue i've found is resolved, and my mother 1 pollyanna midi plays just like it does in the actual Sound Canvas VA

Okay, this is insane. i now have a midi that both this and sound canvas va screw up exactly the same way in sc55 mode (!) I was about to open an issue, but decided to test in the actual SCVA and the same error happens there, ONLY in sc55 mode!

Now i'm curious if the real SC8820 will do the same with this midi file in sc55 mode. Anyone have one to test? The midi file is

https://gifx.co/music/MIDI%2FCorona%2FBaby%20Baby.mid

There's supposed to be a downward sweep one beat after the "baby baby" synth. Works correct on real sc55, nukes sc55, and on va on everything but sc55 mode. Those exact results are replicated in this web app. I'm not sure if the innacuracy is in scva, or in the sc8820's sc55 mode.

RTFT sounds significantly different on nuked vs sound canvas VA as well, but the new project sounds like sound canvas VA, so i can't report this as a bug.

By the samples are usable for their intended purpose loophole, i believe this project is likely to be legal to use if you own SCVA.

I can confirm the sound canvas va behavior is incorrect by comparing vs nuked-sc55 on a mkII rom.

This bug is not exactly about the mode but the specific instrument. That is Program 33 Fingered Bass in the SC-55 Map. If you select SC-8820, SC-88Pro or SC-88 mode and you also set Bank LSB to 1 (SC-55 Map) on channel 2 it will sound bad in all modes on SC-VA. It's most likely about a badly programmed ADSR envelope that reacts a strange way to Portamento (CC#65) and Portamento time (CC#5) controller messages on channel 2.
Here is a version that sounds bad in all modes on SC-VA 😀

And here is a version that also sounds right in SC-55 mode on SC-VA (uses SC-88 Map on channel 2):

Going through these as I am also working heavily on tests right now, and have some uncommitted changes.

Re: Right In The Night, the three filter sweeps at periodic points aren't the reverse cymbal channel, they're the "Rain" instrument in a different channel. I didn't even know that was an issue that I caught somehow. Probably the fixes to the bank MSB 64/65 overrides, but I can try to trace the exact fix.

Also trying to go for correct PRNG draining order so that files which depend on random behavior produce consistent power-on state behavior compared to SCVA itself.

Oh yes, there was also other behavior relative to the filter: Due to the way that the filter coefficients are ramped on reassign, they also re-excite the filter resonance. Also, the resonance control was not being implemented properly, due to some disassembly misinterpretation.

For testing any specific map and channel issues, could you also verify with the TUI for solo or mute, and use the short seeking to repeat portions?

It just needs VCPKG to locate dependencies, indicated with VCPKG_ROOT set to the folder with your clone of vcpkg.

1cmake --preset player
2cmake --build --preset player

and find the player in ./build/player/apps/tui

Pass either environment TS_SCCORE_DLL=<path to DLL> or use the --dll switch to locate the ROM data.

Another interesting finding: Just like the original SCVA module, the tone bank vintage selector is just a default, not a forced selection. Bank LSB and the 40 4x 01 SysEx address are a guidance, the last writer always wins and overrides whichever vintage it wants.