PhoeniX98 wrote on 2023-01-05, 03:58:

Hello everyone. I wanted to test some midis that i found on the web on my SC-55 mkII, and i found that some files sounds kinda s […]
Show full quote

Hello everyone. I wanted to test some midis that i found on the web on my SC-55 mkII, and i found that some files sounds kinda strange. Is not the Capital Tone Fallback problem, my problem is more related to drum parts.

In this file http://www.vgmusic.com/music/console/nintendo … KC2_Mine-KM.mid for example, there's three drum parts, but the third drum part sounds as a piano, i didn't recorded it on the unit but i recorded it on SC-VA, where also is the problem. This behavior is strange and only occurs in Sound Canvas, because in the MS Wavetable sounds normal, in soundfonts and MU modules too.

There's a way to fix this without rewriting the file? Another Roland modules has this behavior too? For example, the SD-20, 80, 90?

Is not the only midi file with that problem, also there's these ones. All are GS-compatibles.

https://www.vgmusic.com/music/console/nintend … ghway-KM-GS.mid
https://www.vgmusic.com/music/console/sony/ps … /MM8TitleGS.mid

I attached audio files too.

Hi,
I have investigated the case of DKC2_Mine-KM.mid.
The problem is that in track 1 named "GS Reset" there are 2 additional SysEx messages that are supposed to set channel 11 and channel 12 to be drum channels:
F041104212401A15020FF7 -> set channel 11 to be a drum channel
F041104212401B15020FF7 -> set channel 12 to be a drum channel

But as you can see even at first glance one of them MUST be incorrect since the last but one byte is the Roland Checksum and it cannot be the same for SySex messages where the payload differs in one byte (0x1A vs. 0x1B).
As it turns out the checksum is correct in the 1st SysEx message but not in the 2nd. Thus only channel 11 is switched to be a drum channel on Roland gears (and channel 12 remains a melodic channel so you hear a piano).

The checksum byte is mostly ignored by clones/non-Roland GS compatible devices. But real Roland gears consider wrong checksum as an indication of an error in data transfer so ignore such SysEx messages.
I have not tested the other files but the problem must be something similar.
Unfortunately the only way to fix such problems for real Roland gears is to fix the wrong checksum byte.

Here is a fixed version of the Midi where the 2nd SysEx is corrected to have a proper checksum byte. Namely:
F041104212401B15020EF7

Far better. I didn't known about the checksum bytes. I forgot to mention that in the screen of the modules (even in Yamaha), appears a "Checksum Error" message, i used to ignore it, until i played it in a Roland gear, and i believed that behavior is related to that message, and from what i read, i realize that it's true.

I don't know if Roland programmed it intentionally or they don't noticed that detail. Thanks for fixing it, in any case, there are very few midi files with that checksum errors.

PhoeniX98 wrote on 2023-01-05, 16:35:

...
I don't know if Roland programmed it intentionally or they don't noticed that detail. Thanks for fixing it, in any case, there are very few midi files with that checksum errors.

1. Roland intentionally programmed its devices to ignore SySex messages with wrong checksum.
2. I do not think that Roland has anything to do with these Midi files. These badly programmed Midi files must have been created by someone else.