Easiest way to test the channel 10 mapping is to use some MIDI software to play the drums - the upper key ranges will not sound when you play them, but once the SysEx I posted earlier in the thread is sent to the device, this maps the sounds into their correct places and you'll be able to hear them. At some point I'll make a MIDI file that has the required notes to test.

RichB93 wrote on 2025-07-29, 11:28:

Easiest way to test the channel 10 mapping is to use some MIDI software to play the drums - the upper key ranges will not sound when you play them, but once the SysEx I posted earlier in the thread is sent to the device, this maps the sounds into their correct places and you'll be able to hear them. At some point I'll make a MIDI file that has the required notes to test.

I downloaded the application "Virtual MIDI Piano Keyboard" to try the sound effects, set it on channel 10. No extra sound effects until i load the SysEx you kindly provided. I tried to upgrade the CM-32 control rom to the 1.02 version and tried to power on the device without the RA-50 control rom, no difference and still the SysEx was needed.

I don't know if at this point it's worth trying changing the resistors' configuration that you tried before since it already made no difference. I wonder if other chips may have their own bit of programming that could be different between each products.

The last thing that looks not tried yet to the RA-50 it to wire keys and display of an MT-32, the same thing possible on a CM-32. I couldn't find the exact location of where to solder the wires and the display, and i wonder if some of the keys and display of the RA-50 could be repurposed to do the mod since the MT-32 parts are nowhere to be found

I wanted to do another test, i've red from this post: Re: Roland RA-50, RA-90 and RA-95 doubts about purchase or not that there is an MT-32 mode for the RA-50 activated by this sequence:

- Push and hold "WRITE" and power on. This will put the device in factory default (E-20 mode).
- Push: "MIDI", "7", "+", "MUSIC STYLE". Now the display should show MT-32

I have tried with either the mod enables and disabled, and it seems that only the first sequence worked, typing the second i have no change on the display. The goal was to see if that mode really change something in the behavior of a modded RA-50 and what changes in the unmodded one. Does anybody knows if this MT-32 mode exists in all revisions of the RA-50 and if is the sequence the same for all? Thanks!

I asked ChatGPT to investigate the curious missing samples, and after disassembling the control ROM it figured it out:

The CM-32L firmware reads bit 5 of the switch/configuration port at address 0x021A. That corresponds to SC0 bit 5, ultimately IC27 SI5, pin 6.

At ROM address 0x518A, the initialization routine does this:
518A ld r78,#8580 ; full CM-32L rhythm table
518E ldb r70,021A ; read SC0/configuration inputs
5193 jbs r70,5,519A ; if SI5 is high, use CM-32L table
5196 ld r78,#5FAE ; otherwise use reduced table
519A ld r76,#E090 ; rhythm setup RAM destination
519E ld r72,#0055 ; copy 85 entries

There are therefore two complete factory rhythm maps inside the same ROM:
0x8580: CM-32L map, including the 33 extra effects
0x5FAE: reduced map, with notes E5 upward set to rhythm timbre 0x7F—effectively unassigned

The tables are identical through the ordinary percussion range. The reduced table replaces all the additional effect assignments with 7F 64 07 01.

Connection to the RA-50 hardware

The RA-50 service schematic shows:
IC27 pin 8 / SI7 pulled high by R70
IC27 pin 7 / SI6 pulled high by R69
IC27 pin 6 / SI5 tied low
That low SI5 is precisely what makes the shared CM-32L ROM select the table at 0x5FAE.

The required modification is not merely disconnecting SI5 from ground: it must be positively pulled high.

The logical fix is:

Disconnect IC27 pin 6/SI5 from its ground strap—apparently the relevant R68 option.

Pull SI5 to +5 V through approximately 22 kΩ, matching R69/R70.

If altering the configuration input is undesirable, the ROM can be patched so both paths use the full table. At offset 0x5196, change:

A1 AE 5F 78

to:

A1 80 85 78

That changes the reduced-table pointer from 0x5FAE to 0x8580. SI5 can then be either state and the effects will always initialize correctly.

Apologies for the formatting. I also must stress that I have not yet tested the above!