Falcosoft wrote on 2024-05-22, 10:56:

1. Yes, it is normal. FSMP by default uses the real 0-based Program Change numbering the same way as the Midi protocol while Sound Canvas uses 1-based numbering (that is real Program Change value + 1) . Although you can change it in FSMP's Main menu -> Compatibility setings -> Program Numbering.
2. Yes, it is because of a logic for Instrument preloading at the loading phase that is useful mainly for soundfont synths. But as you said it does not cause any side effects. If you press stop and then play again then you will get the empty/deafult instrument.

You never stop learning 😅. Perfect thanks.

EDIT
Maybe I understood what the problem is with System Shock music with SC-55 mk2. For example in THM4.XMI, "Program Change 17" is specified which the SC-55 interprets as "Program Change 18" which is not a valid instrument.
The SC-55 mk1 still chooses the "017 Power" bank on channel 10, while the SC-55 mk2 returns the error "No DRUM SET" and so remains the "001 Standard" bank.
If you use "Program Change 16" instead, both SC-55 models work correctly.

(forgive me if I make mistakes regarding the technical terminology).

Is there any downside to routing game midi through FSMP and enabling it's CTF patching before outputting to Nuked SC55 with a mk2 rom? On the surface it seems like the ultimate setup but are there any scenarios where this fails?

markanini wrote on 2024-05-23, 20:46:

Is there any downside to routing game midi through FSMP and enabling it's CTF patching before outputting to Nuked SC55 with a mk2 rom? On the surface it seems like the ultimate setup but are there any scenarios where this fails?

I tried it, it works but you have to start two software instead of one and you have to create another port with LoopMIDI. Not a big deal but we have both mk1 and mk2 emulation so I don't know if it's useful.
Are there any DOS games that use the SC-55 mk2's additional instruments?

I noticed a case where CTF in FSMP behaves differently than in Nuked SC-55 (rom v1.21). I have this Duke Nukem 3D track ("GrabbagOriginalVer_GS1.mid"), the main theme, I don't remember where I downloaded it. The README file indicates it is the original midi version composed on the SC-88 and uses instruments that don't exist on the SC-55, this drives the mk2 crazy, many missing instruments.
The SC-55 mk1 doesn't seem to have any problems playing this track, it assigns replacement instruments as expected (even if they are not the ones intended by whoever composed this track). In channel 2 it is assigned "031+Feedback Gt." (On SCVA with SC-88 map it is "031 Feedback Gt. 2"). With CTF by Falcosoft MIDI Player, in channel 2 "031 DistortionGt" is assigned instead.
I know this is an edge case but perhaps it denotes a subtle difference in behavior in some cases? I don't know.

Anyway, the System Shock case above seems like a small mistake by the composer that hasn't been fixed probably because, I assume, in 1993-1994 they were using a module that had some sort of CTF that hid the mistake (unless someone has a technical explanation that justifies what happens). If so, who knows how many other errors there are in other games that the SC-55 mk1 or the Sound Blaster cards fixed on their own...
Some say that CTF was used deliberately by the composers. Is there any evidence of this? Or maybe CTF was hiding mistakes?

Kappa971 wrote on 2024-05-23, 21:50:

... I noticed a case where CTF in FSMP behaves differently than in Nuked SC-55 (rom v1.21). I have this Duke Nukem 3D track ("Gr […]
Show full quote

...
I noticed a case where CTF in FSMP behaves differently than in Nuked SC-55 (rom v1.21). I have this Duke Nukem 3D track ("GrabbagOriginalVer_GS1.mid"), the main theme, I don't remember where I downloaded it. The README file indicates it is the original midi version composed on the SC-88 and uses instruments that don't exist on the SC-55, this drives the mk2 crazy, many missing instruments.
The SC-55 mk1 doesn't seem to have any problems playing this track, it assigns replacement instruments as expected (even if they are not the ones intended by whoever composed this track). In channel 2 it is assigned "031+Feedback Gt." (On SCVA with SC-88 map it is "031 Feedback Gt. 2"). With CTF by Falcosoft MIDI Player, in channel 2 "031 DistortionGt" is assigned instead.
I know this is an edge case but perhaps it denotes a subtle difference in behavior in some cases? I don't know.
...

It's called 'Sub Capital Fallback' and it's very mkI specific. It works this way: In case of instruments 1-120 and Bank MSB values 1-63 the fallback is not always to a capital tone but to the last valid variation tone separated by 8 banks.
That is invalid instruments at banks 1-7 falls back to bank o (real capital) but if there is a valid instrument at bank 8 then invalid banks 9-15 fall back to bank 8. If there is a valid instrument at bank 16 then invalid banks 17-23 fall back to 16 and so on. But if there is no valid instrument at bank 16 then invalid banks 17-23 do not fall back to valid instrument at bank 8 but to bank 0 (real capital).
FSMP's CTF feature does not follow these needlessly complex rules but implements a more generic approach that is similar to how SF2 synths (e.g. Creative SB HW synths, Bassmidi) implement CTF.
This works equally well for mkII and later Sound Canvas devices from Roland and also for different XG synths from Yamaha (e.g. S-YXG50 has no CTF support in XG mode for Bank MSB variation tones but it works with FSMP's CTF implementation).
That is in case of any invalid instruments at any banks FSMP's CTF implementation always uses the capital tone (at bank 0).

Falcosoft wrote on 2024-05-23, 23:40:

It's called 'Sub Capital Fallback' and it's very mkI specific. It works this way: In case of instruments 1-120 and Bank MSB val […]
Show full quote

It's called 'Sub Capital Fallback' and it's very mkI specific. It works this way: In case of instruments 1-120 and Bank MSB values 1-63 the fallback is not always to a capital tone but to the last valid variation tone separated by 8 banks.
That is invalid instruments at banks 1-7 falls back to bank o (real capital) but if there is a valid instrument at bank 8 then invalid banks 9-15 fall back to bank 8. If there is a valid instrument at bank 16 then invalid banks 17-23 fall back to 16 and so on. But if there is no valid instrument at bank 16 then invalid banks 17-23 do not fall back to valid instrument at bank 8 but to bank 0 (real capital).
FSMP's CTF feature does not follow these needlessly complex rules but implements a more generic approach that is similar to how SF2 synths (e.g. Creative SB HW synths, Bassmidi) implement CTF.
This works equally well for mkII and later Sound Canvas devices from Roland and also for different XG synths from Yamaha (e.g. S-YXG50 has no CTF support in XG mode for Bank LSB variation tones but it works with FSMP's CTF implementation).
That is in case of any invalid instruments at any banks FSMP's CTF implementation always uses the capital tone (at bank 0).

This explains everything, thanks.
So to the user's question above, no it won't fail but the behavior will be slightly different.

Regarding the test version of FSMP with EMIDI support, perhaps I noticed something wrong:
1. in ALFREDH.MID, the “GS Crystal” (94 Metal Pad) track on channel 2 is listed as disabled while the “Crystal” (“93 Bowed Glass”) track is enabled, but the “94 Metal Pad” instrument is assigned to the emulator. Should the "GS Crystal" or "Crystal" track be enabled? The same thing happens with "GS Effects" (disabled) and "Effects" (enabled). Also in "GS Effects" there seems to be a similar problem as in System Shock, an invalid Program Change (Program Change 1).
2. GRABBAG.MID is listed as "EMIDI" but I don't think it is, I don't see any duplicate tracks.
3. in STALAG.MID, the "FMViola" track is enabled.
4. in STALKER.MID the "GStrings" track is disabled (should it?).

Kappa971 wrote on 2024-05-24, 11:15:

Regarding the test version of FSMP with EMIDI support, perhaps I noticed something wrong: 1. in ALFREDH.MID, the “GS Crystal” (9 […]
Show full quote

Regarding the test version of FSMP with EMIDI support, perhaps I noticed something wrong:
1. in ALFREDH.MID, the “GS Crystal” (94 Metal Pad) track on channel 2 is listed as disabled while the “Crystal” (“93 Bowed Glass”) track is enabled, but the “94 Metal Pad” instrument is assigned to the emulator. Should the "GS Crystal" or "Crystal" track be enabled? The same thing happens with "GS Effects" (disabled) and "Effects" (enabled). Also in "GS Effects" there seems to be a similar problem as in System Shock, an invalid Program Change (Program Change 1).
2. GRABBAG.MID is listed as "EMIDI" but I don't think it is, I don't see any duplicate tracks.
3. in STALAG.MID, the "FMViola" track is enabled.
4. in STALKER.MID the "GStrings" track is disabled (should it?).

Hi,
Here is the Apogee EMIDI specification that I used for the implementation:

According to this the only mandatory controller that you can use as a 'marker' for EMIDI files is CC#110 (Track Designation) with valid values 0-9 and 127. Grabbag.mid uses this controller albeit only for assigning all tracks to all possible instruments(value 127).
If you have 'Use Tracks for Sound Canvas' selected as EMIDI option in FSMP then you have to look for tracks that are enabled for device/instrument 1. This means that CC#110 has either explicit value 1 or implicit value 127 (all). You should also look for CC#111 (Track Exclusion) controller messages. This can only use explicit values. That is in case of 'Use Tracks for Sound Canvas' a track has to be disabled only if CC#111 has a value of 1.
Notice that track names do NOT matter!
E.g. in case of your STALAG.MID example while it is true that the track name is FMViola it is explicitly disabled only for FM Synths! (instruments 4, 5, 6, 7) 😀. Of course this can be a bug in the Midi file, but who knows...
The situation is similar in case of STALKER.MID. Despite (because of ?) the name GStrings this track is only enabled for GUS (instrument 9).
in ALFREDH.MID channel 2 GS Crystal is assigned explicitly only to GM, AWE32, and GUS devices (0, 2, 9) and channel 3 Crystal is assigned to Sound Canvas (1). So if 'Use Tracks for Sound Canvas' is selected then only track/channel 3 is used (it does not matter that Metal Pad is assigned to channel 2 , it does not receive any notes from track 2).

Falcosoft wrote on 2024-05-24, 11:54:

Hi, Here is the Apogee EMIDI specification that I used for the implementation: Emidi11.txt […]
Show full quote

Hi,
Here is the Apogee EMIDI specification that I used for the implementation:
Emidi11.txt

According to this the only mandatory controller that you can use as a 'marker' for EMIDI files is CC#110 (Track Designation) with valid values 0-9 and 127. Grabbag.mid uses this controller albeit only for assigning all tracks to all possible instruments(value 127).
If you have 'Use Tracks for Sound Canvas' selected as EMIDI option in FSMP then you have to look for tracks that are enabled for device/instrument 1. This means that CC#110 has either explicit value 1 or implicit value 127 (all). You should also look for CC#111 (Track Exclusion) controller messages. This can only use explicit values. That is in case of 'Use Tracks for Sound Canvas' a track has to be disabled only if CC#111 has a value of 1.
Notice that track names do NOT matter!
E.g. in case of your STALAG.MID example while it is true that the track name is FMViola it is explicitly disabled only for FM Synths! (instruments 4, 5, 6, 7) 😀. Of course this can be a bug in the Midi file, but who knows...
The situation is similar in case of STALKER.MID. Despite (because of ?) the name GStrings this track is only enabled for GUS (instrument 9).
in ALFREDH.MID channel 2 GS Crystal is assigned explicitly only to GM, AWE32, and GUS devices (0, 2, 9) and channel 3 Crystal is assigned to Sound Canvas (1). So if 'Use Tracks for Sound Canvas' is selected then only track/channel 3 is used (it does not matter that Metal Pad is assigned to channel 2 , it does not receive any notes from track 2).

I thought it might be a bug in FSMP but from this explanation everything seems to work fine, so thanks for your patience 😅.

EDIT
In Rednukem there is the console command "mus_emidicard" (force a specific EMIDI instrument set [-1- 10]).
mus_emidicard "-1" (default) seems to be equivalent to mus_emidicard "0" (General MIDI) so if it follows this:

1        
2        0 -     General MIDI
3        1 -     Roland Sound Canvas (GM only)
4        2 -     Sound Blaster AWE32
5        3 -     Wave Blaster and Compatibles (SCD-10, etc.)
6        4 -     Sound Blaster and Compatibles (OPL-2 and OPL-3)
7        5 -     Media Vision Pro Audio series
8        6 -     Logitech Sound Man 16
9        7 -     Adlib and Compatibles
10        8 -     Ensoniq Soundscape
11        9 -     Gravis Ultrasound, Ultrasound Max, Ultrasound ACE
12        127 -   All (see Controller Definitions)

mus_emidicard "1" should be Sound Canvas.

I do not know if anyone else has noticed this but with the mkI 1.21 ROM set the emulator consumes less CPU resources (about 60% of mkII ROM set).
So overall the mkI ROM set is the better one in most aspects. The only disadvantage I have noticed so far is that it does not support the GM Master Volume SysEx message only Roland specific one (mkII does support it).
So volume settings does not work in FSMP 6.3 and earlier (even if 'Use SysEx for Volume/Balance' option is selected). But here is an updated test version that supports mk1 master volume settings.
It works when GS reset option is selected (there is no point in selecting GM reset option in case of mkI anyway since it does not fully support it. GM reset just does the same thing as GS reset).
https://falcosoft.hu/midiplayer_64_test.zip

I can't tell if I'm having a GZDoom (v4.12.1) issue, weirdly I'm getting incorrect default instruments using a 1.21 rom that resolves with a mk2 rom.

EDIT: It's fixed when routing through FSMP.

EDIT2: One more question on a different topic. If I play this midi using 1.21 rom the snare sounds off, but if I route through FSMP with it's CTF patching enabled, the snare sounds right. How can this FSMP CTF patching an have effect on a 1.21 rom in the first place?

markanini wrote on 2024-05-25, 22:09:

EDIT2: One more question on a different topic. If I play this midi using 1.21 rom the snare sounds off, but if I route through FSMP with it's CTF patching enabled, the snare sounds right. How can this FSMP CTF patching an have effect on a 1.21 rom in the first place?

According to the top post here, CTF should be fully supported up to v1.21 on the mk1. Version 2.00 is the first one that drops CTF due to the patent fiasco between Roland and Yamaha.
Update SC-55 ROM

Other than that, define "right" 😀 Maybe it's just a question of expectations on your part? Or you mean hardware mk1 v1.21 compared to NukedSC55 with the same ROM?

markanini wrote on 2024-05-25, 22:09:

I can't tell if I'm having a GZDoom (v4.12.1) issue, weirdly I'm getting incorrect default instruments using a 1.21 rom that resolves with a mk2 rom.

EDIT: It's fixed when routing through FSMP.

I don't use GZDoom so I don't know.
Anyway it seems strange, are you sure you have configured Nuked-SC5 correctly? In this comment https://github.com/nukeykt/Nuked-SC55/issues/ … ment-2121071281 I see "mk2" on the title bar with the mk1 rom set.

markanini wrote on 2024-05-25, 22:09:

EDIT2: One more question on a different topic. If I play this midi using 1.21 rom the snare sounds off, but if I route through FSMP with it's CTF patching enabled, the snare sounds right. How can this FSMP CTF patching an have effect on a 1.21 rom in the first place?

This is not the original System Shock track, it has been slightly modified and for some strange reason uses channels 10-16. At about 1 min, there is "103 Echo Drops" on channel 14 and the cut note is clearly audible.

Kappa971 wrote on 2024-05-26, 10:53:

...Anyway it seems strange, are you sure you have configured Nuked-SC5 correctly? In this comment https://github.com/nukeykt/Nuked-SC55/issues/ … ment-2121071281 I see "mk2" on the title bar with the mk1 rom set.

I think the explanation is that Markanini used the version based on Jcmoyer's fork that had exactly this bug:
https://github.com/jcmoyer/Nuked-SC55/issues/3

markanini wrote on 2024-05-25, 22:09:

.. How can this FSMP CTF patching an have effect on a 1.21 rom in the first place?

In general the the answer is here:
Re: [RELEASE] Nuked-SC55, low-level Roland SC-55 series emulator announcement
In case of '0 - inntro.mid' it should not have any differences regarding drum CTF. The Midi file incorrectly sends Program Change 17 to channel 10 (instead of the right Program Change 16). This results in invalid drum set error on mkII (and thus default drum set) but on mkI 1.21 it falls back to Power set correctly. The same is happening if FSMP's CTF is enabled both on mkII and mkI. You should check on the emulator what drum set is active on channel 10. It should be Power set in both cases.

Ps: I have just checked and I cannot hear/see any differences on channel 10 drum set between CTF OFF and CTF ON states when '0 - inntro.mid' is played back with FSMP and the emulator uses mk1 1.21 ROM set.

Kappa971 wrote on 2024-05-26, 10:53:

markanini wrote on 2024-05-25, 22:09:

EDIT2: One more question on a different topic. If I play this midi using 1.21 rom the snare sounds off, but if I route through FSMP with it's CTF patching enabled, the snare sounds right. How can this FSMP CTF patching an have effect on a 1.21 rom in the first place?

This is not the original System Shock track, it has been slightly modified and for some strange reason uses channels 10-16. At about 1 min, there is "103 Echo Drops" on channel 14 and the cut note is clearly audible.

Okay, I take it that it's only by chance that FSMP CTF did something desirable in this instance. @Ricewind42 I was using this capture of a real SC55 as a reference.

As for GZDoom the original music works as intended. The music in this user made map set had some weird discrepancies with mk1 rom. Using Jcmoyers fork and routing through FSMP yields correct midi playback. This is unexpected because it should be equivalent to routing straight into either nukeykt v0.31, or the fork shared here. Even between them there are some discrepancies in instrument mapping which I cant understand. In either case playing the extracted midis in FSMP yields correct playback. So If I could guess the issue is on GZDooms end, or some quirks in the midis or a combination.

markanini wrote on 2024-05-26, 11:54:

Okay, I take it that it's only by chance that FSMP CTF did something desirable in this instance. @Ricewind42 I was using this capture of a real SC55 as a reference.

As for GZDoom the original music works as intended. The music in this user made map set had some weird discrepancies with mk1 rom. Using Jcmoyers fork and routing through FSMP yields correct midi playback. This is unexpected because it should be equivalent to routing straight into either nukeykt v0.31, or the fork shared here. Even between them there are some discrepancies in instrument mapping which I cant understand. In either case playing the extracted midis in FSMP yields correct playback. So If I could guess the issue is on GZDooms end, or some quirks in the midis or a combination.

These are the original tracks from System Shock (attached).
There is a lot of misinformation on YouTube, even regarding Duke Nukem 3D (here on Vogons I discovered how EMIDI works, on YouTube they reproduce the DN3D tracks on SC-55 incorrectly https://www.youtube.com/watch?v=e7yBJVJTGMc&l … jJcUyHp&index=8).

@Falcosoft has already explained the difference of CTF between the SC-55 mk1 and FSMP above.
I don't know on what basis it is possible to establish who reproduces these tracks correctly, we don't even know which sound module the guy used to compose them 😆. Nuked SC-55 mk1 I imagine behaves as the original module would behave.
I also attach the latest development build of Nuked SC-55.

Kappa971 wrote on 2024-05-26, 14:44:

I don't know on what basis it is possible to establish who reproduces these tracks correctly, we don't even know which sound module the guy used to compose them 😆. Nuked SC-55 mk1 I imagine behaves as the original module would behave.

Fair point. That youtube clip you linked to would raise a red flag to me because it sounds different in character from sc-55 captures I'm used to. Overall it's shows that it's not wise to trust a single piece of data, thanks for sharing that.

I make some inferences on occasion based om my background in studying music composition. For instance it's strange to use a soft clap sound for a snare in a busy, drum heavy track. Or for instance, it's strange to use the default instrument mappings on a complex song with time signature changes and use of chromaticism and key modulation.
But after that I seek more means of confirmation either playing the music outside the game, of listening to captures of real hardware. In the case the Doom fan made map pack the composer has a youtube channel showcasing his compositions that the game midis are based on. So quite a bit of detective work to make an informed guess on what's correct.

Anyway I'm not here to argue what's correct, more raise what I believe is a potential issue others might have questions about down the line. I do believe Nuked SC55 is working as intended. Let's imagine GZDoom has optimized it's midi playback for the default built in FluidSynth device, maybe each device expects bank changes to be transferred in a slightly different way. A user noticing instrument discrepancies will google Nuked SC55 first because that's how the issues was noticed. So maybe this is an opportunity to more accurately attribute the behavior. First someone more knowledgeable than me will have to determine the cause. At best I'm offering my observations here for posterity.

Thanks for the dev build.

EDIT:
I've found another user reporting the same issue with GZDoom. As far as my own testing, the dev build from the post before has been without issues so far. It's possible that the UART delay commit has had a positive impact in this scenario.

Work on a UI is progressing nicely
https://x.com/nukeykt/status/1789262985615073 … FlD8n1-lyA&s=19

I'm a newbie here when it comes to this stuff. I'm interested in using this in dosbox for games, and was wondering if there was a dos build or something that i can easily use in place of say a soundfont or mt-32 rom?

MacDad wrote on 2024-06-11, 19:05:

I'm a newbie here when it comes to this stuff. I'm interested in using this in dosbox for games, and was wondering if there was a dos build or something that i can easily use in place of say a soundfont or mt-32 rom?

First of all this program doesn't work on DOS (maybe someone would be able to but I don't see the reason). If you want to use it with DOSBox on Windows, you should:
1. Download LoopMIDI and create a virtual cable for Nuked-SC55;
2. Download Nuked-SC55 (you can download it from a comment above) REMOVED and use it with the ROM you backed up from your hardware.
3. Start Nuked-SC55 (with the -mk1 parameter from batch file or shortcut, if you are using the rom set recommended above);
4. In DOSBox run the command "mixer /listmidi", note down the number of the port previously created with LoopMIDI and insert the number in the DOSBox configuration file in the "midiconfig" option (for example "midiconfig = 2").

If for example you want to use a Windows source port to play Doom, you can install Coolsoft MIDI Mapper and select the LoopMIDI port from there (and in this case you could set in DOSBox "midiconfig = 0" to use Coolsoft MIDI Mapper also on DOSBox but this is not important).

I would also suggest using mt32 when the game was made for it.

Someone with more knowledge can explain this but, it will just not sound right.

Kappa971 wrote on 2024-06-11, 23:16:

First of all this program doesn't work on DOS (maybe someone would be able to but I don't see the reason). If you want to use it […]
Show full quote

MacDad wrote on 2024-06-11, 19:05:

I'm a newbie here when it comes to this stuff. I'm interested in using this in dosbox for games, and was wondering if there was a dos build or something that i can easily use in place of say a soundfont or mt-32 rom?

First of all this program doesn't work on DOS (maybe someone would be able to but I don't see the reason). If you want to use it with DOSBox on Windows, you should:
1. Download LoopMIDI and create a virtual cable for Nuked-SC55;
2. Download Nuked-SC55 (you can download it from a comment above) REMOVED and used it with the ROM you backed up from your hardware.
3. Start Nuked-SC55 (with the -mk1 parameter from batch file or shortcut, if you are using the rom set recommended above);
4. In DOSBox run the command "mixer /listmidi", note down the number of the port previously created with LoopMIDI and insert the number in the DOSBox configuration file in the "midiconfig" option (for example "midiconfig = 2").

If for example you want to use a Windows source port to play Doom, you can install Coolsoft MIDI Mapper and select the LoopMIDI port from there (and in this case you could set in DOSBox "midiconfig = 0" to use Coolsoft MIDI Mapper also on DOSBox but this is not important).

/Removed ROM info and offer.

1st and last warning. Next will be a ban for 30 days.