VOGONS


88Emu is out (Roland SC-88/Pro/50 Emulator)

Topic actions

Reply 60 of 121, by daveiver

User metadata
Rank Newbie
Rank
Newbie
Karmeck wrote on 2026-10-06, 05:56:

it works but 88emu does not recognize it, probably a md5 mismatch. But hanks for letting me know.

I found that 88emu 2.2.25 will still take the rom as long as it's named "sc8820_internal.bin". The warning it displays is purely informational, it will still load it and work correctly in my testing. This dump is so new it happened after the 2.2.25 release, the new MD5 will need to be added to 88emu to get rid of the warning (and allow any filename) for the next release.

Karmeck wrote on 2026-10-06, 05:56:

But really, about the missing roms aspect, are they really missing, like really, or just not easy to access? because, if you have the md5 value, that file is out there.

I think not easy to access since the hashes are known as you pointed out. Hopefully more surfaces soon, but we are in a pretty good spot with true SC-8820, JV-1080 and XV-3080 being shared now.

Reply 61 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

still need the g-800, though. That one has actually working sc55 backwards compatibility (they didn't have to cut samples), as well as some unique samples added to the sc88 map (like StikyRhodes).

The other two (a waveblaster sc55 daughterboard and a sc88 pro daughterboard) are quite a bit less essential.

Reply 62 of 121, by Admiral_MIDI

User metadata
Rank Newbie
Rank
Newbie

I have downloaded all the available roms but the JV-1080 and XV-3080 are not showing up, everything else is working. What am I missing???

I have an MU50

Reply 63 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

did you extract the "not a 1080 or 3080" zip file?

Reply 64 of 121, by yes

User metadata
Rank Newbie
Rank
Newbie
Admiral_MIDI wrote on 2026-10-08, 00:38:

I have downloaded all the available roms but the JV-1080 and XV-3080 are not showing up, everything else is working. What am I missing???

They're known by their code names as NU-10B (JV-1080) and MIIG5 (XV-3080) .

Reply 65 of 121, by Karmeck

User metadata
Rank Member
Rank
Member
daveiver wrote on 2026-10-06, 20:32:
I found that 88emu 2.2.25 will still take the rom as long as it's named "sc8820_internal.bin". The warning it displays is purel […]
Show full quote
Karmeck wrote on 2026-10-06, 05:56:

it works but 88emu does not recognize it, probably a md5 mismatch. But hanks for letting me know.

I found that 88emu 2.2.25 will still take the rom as long as it's named "sc8820_internal.bin". The warning it displays is purely informational, it will still load it and work correctly in my testing. This dump is so new it happened after the 2.2.25 release, the new MD5 will need to be added to 88emu to get rid of the warning (and allow any filename) for the next release.

Karmeck wrote on 2026-10-06, 05:56:

But really, about the missing roms aspect, are they really missing, like really, or just not easy to access? because, if you have the md5 value, that file is out there.

I think not easy to access since the hashes are known as you pointed out. Hopefully more surfaces soon, but we are in a pretty good spot with true SC-8820, JV-1080 and XV-3080 being shared now.

Seems the 8820 rom situation is even more advanced. There seems to be be misconceptions about the ROMs correct size.

https://github.com/dsp56300/gearmulator/pull/315

Reply 66 of 121, by daveiver

User metadata
Rank Newbie
Rank
Newbie
Karmeck wrote on 2026-10-08, 09:54:

Seems the 8820 rom situation is even more advanced. There seems to be be misconceptions about the ROMs correct size.

https://github.com/dsp56300/gearmulator/pull/315

Yes, this was uncovered after looking at the real cpu rom dump that was just made available the other day. My PR was merged yesterday to add the new MD5 (no more warning with real rom) and this new PR will fix the remaining issue with CPU and rom size assumptions. The next release will have these fixes, but if you don't want to wait or can't build yourself, you can still load the real rom in the current release (2.2.25) by naming the file correctly and it will work just fine (although CPU will be SH7017 instead of SH7016 without this fix, but that shouldn't affect accurracy).

Reply 67 of 121, by daveiver

User metadata
Rank Newbie
Rank
Newbie
zaphod77 wrote on 2026-10-07, 00:13:

still need the g-800, though. That one has actually working sc55 backwards compatibility (they didn't have to cut samples), as well as some unique samples added to the sc88 map (like StikyRhodes).

The other two (a waveblaster sc55 daughterboard and a sc88 pro daughterboard) are quite a bit less essential.

XPGS / G-800 and VE-GS Pro have surfaced, only things missing from the pack now are:

CM-300 / SCC-1 internal rom
SCB-55 / RLP-3237 internal rom
RLP-3237 program and wave roms

Reply 68 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

they aren't in the same spot as the 8820 rom, though..

Reply 69 of 121, by Ishisashi

User metadata
Rank Newbie
Rank
Newbie
Karmeck wrote on 2026-10-08, 09:54:
daveiver wrote on 2026-10-06, 20:32:
I found that 88emu 2.2.25 will still take the rom as long as it's named "sc8820_internal.bin". The warning it displays is purel […]
Show full quote
Karmeck wrote on 2026-10-06, 05:56:

it works but 88emu does not recognize it, probably a md5 mismatch. But hanks for letting me know.

I found that 88emu 2.2.25 will still take the rom as long as it's named "sc8820_internal.bin". The warning it displays is purely informational, it will still load it and work correctly in my testing. This dump is so new it happened after the 2.2.25 release, the new MD5 will need to be added to 88emu to get rid of the warning (and allow any filename) for the next release.

Karmeck wrote on 2026-10-06, 05:56:

But really, about the missing roms aspect, are they really missing, like really, or just not easy to access? because, if you have the md5 value, that file is out there.

I think not easy to access since the hashes are known as you pointed out. Hopefully more surfaces soon, but we are in a pretty good spot with true SC-8820, JV-1080 and XV-3080 being shared now.

Seems the 8820 rom situation is even more advanced. There seems to be be misconceptions about the ROMs correct size.

https://github.com/dsp56300/gearmulator/pull/315

Yes, I am the one who opened the PR. It has been confirmed that dumping the CPU ROM with a size of 128 KiB results in mirroring.
I also checked the chip markings recently and confirmed that the CPU is indeed an SH7016.
This actually stems from an error in Roland’s Service Notes. In the Service Notes for the SC-8850, and for the SC-8820, the Block Diagram correctly identifies the CPU as SH7016, but the Parts List and the Circuit Diagram incorrectly list it as HD64F7017F28. This does not match the markings on the actual chip.
The reason why the CPU model of the SC-8850 is correctly identified in the code of 88emu may be explained by the following page:
EDIROL SD-80: The Adventure Continues
In the table ‘More detailed comparison of various Roland PCM-based sound modules’, the entry for the SC-8850 has been corrected based on the actual motherboard (HD6437016E09F, 13th footnote), but the entry for the SC-8820 has not.
The correct code for the SC-8820 is HD6437016E11F.

Reply 70 of 121, by daveiver

User metadata
Rank Newbie
Rank
Newbie
zaphod77 wrote on 2026-10-09, 23:00:

they aren't in the same spot as the 8820 rom, though..

Please try searching on the most common site to find such files.

Reply 71 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

okay, i've located them.

now if i could just throw the g800 into sc55 mode and play test midis, like from warcraft 2 on it to see if they sound right. this will probably need editing the midi files to force bank lsb 1.

Reply 72 of 121, by Admiral_MIDI

User metadata
Rank Newbie
Rank
Newbie
yes wrote on 2026-10-08, 09:46:
Admiral_MIDI wrote on 2026-10-08, 00:38:

I have downloaded all the available roms but the JV-1080 and XV-3080 are not showing up, everything else is working. What am I missing???

They're known by their code names as NU-10B (JV-1080) and MIIG5 (XV-3080) .

Thank you. Got it sorted.

I have an MU50

Reply 73 of 121, by Karmeck

User metadata
Rank Member
Rank
Member
zaphod77 wrote on 2026-10-10, 23:23:

okay, i've located them.

now if i could just throw the g800 into sc55 mode and play test midis, like from warcraft 2 on it to see if they sound right. this will probably need editing the midi files to force bank lsb 1.

Can't you use a SysEx message for that?

Reply 74 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

I don't see an option to send sysex directly in 88emuplayer.

Reply 75 of 121, by Ishisashi

User metadata
Rank Newbie
Rank
Newbie
Karmeck wrote on 2026-10-11, 06:58:
zaphod77 wrote on 2026-10-10, 23:23:

okay, i've located them.

now if i could just throw the g800 into sc55 mode and play test midis, like from warcraft 2 on it to see if they sound right. this will probably need editing the midi files to force bank lsb 1.

Can't you use a SysEx message for that?

If a GS Reset is sent after using SysEx, the former SysEx will be ineffective.

Reply 76 of 121, by shandavid

User metadata
Rank Member
Rank
Member

Can't find the ROM file, and the demo audio doesn't seem to sound as good as the original hardware...

for comparison:
https://www.youtube.com/watch?v=z1P3YYHf4RY
https://www.youtube.com/watch?v=LFDmDiaVk8o&t=254s
https://www.youtube.com/watch?v=wO3KEwceJTs&t=2310s

I’ll just stick with the SC-8850, it’s cheap to pick up second hand anyway.

(A lot of details in software can affect sound quality, even different compilers (Intel compilers sound better than Microsoft ones) and different compiler settings can produce different sounds, but most people don't believe it, and I’ve stopped bothering to bring it up.....)

Reply 77 of 121, by DX7_EP

User metadata
Rank Newbie
Rank
Newbie

Saw that 88Emu got a new stable release that adds MT-32/CM-32L/CM-64 support. The last one interested me most as a CM-64 owner (plus this is the first CM-64 emulation option AFAIK), so I had to compare both.

Using various X68000 games, 88Emu held up quite well, feeling more accurate than Munt (maybe it's the analogue emulation, which I enabled) and working well with addon cards and the two games that support them. The experimental analogue emulation worked quite well too, even including the right channel bias. Lastly, the new part editor feature has been a fun way to see how devs used the LA synths. Though when comparing hardware and emulator recordings, it seems signal is sometimes inverted in the latter.
On that note, I'll put FLAC recordings up on a GDrive folder for the time being. It includes other emulations like S-MU2000 as well, and may be updated whenever I feel inclined to do so.

Too many ideas and not enough time.

Synth modules: CM-64, FB-01, SC-55ST, SC-8850, SD-20

Reply 78 of 121, by zaphod77

User metadata
Rank Member
Rank
Member

So turns out the claim that compiler affected sound of 88emuplayer was *correct*.

It's fixed now...

[Fix] SC-88 family, SC-8820, SC-8850, NU-10B, MIIG5: A few calculations in
the XP sound chip relied on behaviour that C++ leaves undefined, which
an optimising compiler is free to break. The sound does not change.

Reply 79 of 121, by Ishisashi

User metadata
Rank Newbie
Rank
Newbie
zaphod77 wrote on 2026-10-18, 17:30:
So turns out the claim that compiler affected sound of 88emuplayer was *correct*. […]
Show full quote

So turns out the claim that compiler affected sound of 88emuplayer was *correct*.

It's fixed now...

[Fix] SC-88 family, SC-8820, SC-8850, NU-10B, MIIG5: A few calculations in
the XP sound chip relied on behaviour that C++ leaves undefined, which
an optimising compiler is free to break. The sound does not change.

In C and C++, undefined behaviour, unspecified behaviour, and implementation-defined behaviour are three categories of behaviour for which the standard does not fully specify the exact outcome.
Undefined behaviour (UB): The standard imposes no requirements on the behaviour. Programs should always avoid invoking undefined behaviour.
Unspecified behaviour: The standard permits several possible valid outcomes but does not specify which one is chosen. The implementation is not required to document its choice.
Implementation-defined behaviour: The standard leaves the definition of the behaviour to the implementation and requires the implementation to document its choice.

Code that invokes UB should never be written in the first place, and if it is, it must be fixed. Code that invokes unspecified behaviour is generally not used in contexts where it can affect the outcome. In my view, code that invokes implementation-defined behaviour is the kind of code that people genuinely do write, and whose results can actually differ between compilers.
There are many cases of implementation-defined behaviour involving signed integer arithmetic. It was not until C23 and C++20 mandated two’s complement representation for signed integers that many of these behaviours became specified by the standards. Unfortunately, however, gearmulator requires C++17.
Even with two’s complement mandated, the standards still seem conservative in how they define the behaviour of signed integer arithmetic. The C standard is even more conservative than the C++ standard.
That said, modern microcomputer architectures all use two’s complement anyway.