Unplug the synth.
Turn it face-down (propping it up so you don't damage the joystick), remove all the screws around the perimeter, and the five large screws in the centre section.
Carefully lift open the synth. It should look like this.
If, as with mine, there is a biro rattling around inside, now is a good time to remove it. :-)
We need to remove the middle PCB. This one.
Carefully unplug all of the connectors from the board, EXCEPT for CN12A and CN13A which are soldered to the board - these need to be unplugged from the other end.
Label the connectors so you know where they go afterwards. For the larger ones I wrote the connector number on the side with permanent marker, for the smaller ones I attached a label to the wire.
One of the nice things about working with old kit, is that the manuals actually give you tons of technical information. In this case, the Service Manual has a detailed schematic. The relevant bit is on page 18:
So, MIDI IN goes through an Opto-Isolator TLP552 (as per the MIDI standard, to prevent ground loops), then gets sent:
a. through a couple of inverters to MIDI THRU outpit, and
b. to pin 12 (NOT 23 - carefull with that diagram!) of the CPU.
Since we don't want to affect MIDI THRU, but we do want to affect the MIDI stream before it reaches the CPU, then be need to do our hack somewhere between pin 12 of the CPU, and the first solder connection.
Here is a close-up of the circuit board, after I cut the track to pin 12. (Note that 10 is helpfully labelled on the silk screen!)
The next step was to scrape off some solder resist, and solder up 4 pins worth of 0.1" header:
Black goes to ground (connection not clearly visible here, but it is the large copper area).
Red goes to +5V, which can be conveniently picked up on the side of R73 furthest from the CPU. ('A' on the schematic is +5V).
Yellow goes directly to pin 12 of the CPU. This will go to the Tx pin on the arduino.
Pink goes to the side of R73 nearest to the CPU.
(In retrospect, I should have used thinner wire, for less danger of lifting anything).
If a jumper is placed across the yellow and pink pins, then the synth should work as before. This is a good thing to test. Of course, DO NOT bridge the power pins by mistake, or you might not have a synth any more.
Here it is with a bit of heatshrink added to prevent any embarrassing short circuits.
At this point I decided it was a good time to replace the battery, before it leaked or died. This wipes the patches from memory, so if you want them then make sure you have saved them and know how to restore them using SYSEX or tape.
The original battery was soldered in place. Not awesome. I removed it with a combination of cutters, soldering iron and solder sucker.
You then need to install a CR2032 coin cell holder. I'm not aware of any that fit the footprint, so I ended up having to drill a new hole in the PCB. Luckily, there is plenty of board area free of tracks in this area, so it wasn't hard to find a suitable place.
There it is, just below R53:
And the other side, after soldering the cell holder. Note that the new hole happily lines up with the +ve track, so no need for a bodge wire.
SUPER IMPORTANT: Check at least five times that you have the polarity correct before you solder in the battery holder, otherwise very bad things will happen.
Insert a new 3V cell into the holder, and the interlude is finished.
Reinstall the PCB, and reconnect all the connectors.
The synth is now ready, time to prepare the Nano.It needs four connections to connect to our 4-way header in the synth. I used some 'Dupont' cables cut in half, and taped together to form the 4-way socket. The other ends connect to the +5V (red), GND (black), Tx (yellow) and Rx (pink) pins like this: (Notice the non-standard CH340G USB to Serial chip on this clone)