Joe
I created a test file and had a look in Mix Master, quite instructive as you'd suggested. I am only looking at the voices used in the midi file song parts, not the R1/R2/R3 voice parts.
I recorded part 1 using 'Soft Mini'. Then I changed the cutoff / filter (didn't save as a user voice) and recorded part 2. Then I edited some DSP parameters (again not saving as a user voice) and recorded part 3. Finally, I reslected the 'Soft Mini' voice, changed the velocity offset and recorded part 4.
The result was that the MIDI file (attached in case anyone is interested) contained
+ SYSEX messages to assign the DSP to each of the 4 parts
+ SYSEX messages to set up the custom parameters on DSP3
+ SYSEX messages to set up the velocity offset on part 4
+ CC messages to set up filter/cutoff on parts 2 and 3.
There were also SYSEX messages to set up some things that I had not touched - velocity offset on parts 1-3, EQ on each part, and one parameter on all 4 DSPs. Without running another test I can't be sure if they are always there or just there for the voice I picked on.
So, finally, to answer the OPs question ... you can open the file in MixMaster and you will see the instructions to select a starting panel voice, and the various voice set and DSP settings that were applied. It might take a little time to understand what you are seeing, but it is all there. This will let you call up the panel voice on R1, then apply the exact same settings to get the voice you want.
(NB Toril - you lost the DSP from the song part because you used too many -, either on too many song parts with DSP or by calling up a panel voice after setting up the song parts. PSR-S only have 4 DSP so are vulnerable to this. Tyros has 9 and Genos 28 so it's less likely).
[attachment deleted by admin]