My Roland SC55 resets when it receives MIDI notes

I purchased my Roland SC55 in France circa 1992.

It worked very well until a few weeks ago, when I noticed the unit always resets when it receives MIDI notes.
I first suspected the power supply, but it was working just fine.

So, it's time to open the unit..
Hopefully, I could grab the Service Manual quite easily on the web, and I have an oscilloscope.

What is going on?

Ok, here is what I can see:
- the unit boots up just fine
- when it receives a few MIDI notes at low velocity, it still works fine
- when it receives MIDI notes with high velocity, it reboots
- when it receives other MIDI data (such as pitch bend), no problem.
- when I try to play a demo song, it reboots

From what I observe, the unit resets not when is has to process MIDI data, but when it has to generate a waveform with high volume. Also, after a certain amount of time, obviously when the unit gets warm, it starts resetting in loop.

My guess is something wrong with the power line (bad regulator, bad cap), or a bad join somewhere else in the unit.

The schematics

From the schematics, I could draw the power circuitry (which is quite complex):

The power circuitry provides 5 different voltages:

- +5V (Vcc) and a reset signal for the digital circuitry

- +5V for the analog circuitry; the voltage is controlled by the main CPU

- +8V for the analog circuitry (amplifiers)

- +5V (Vdd) for the LCD screen

- +3V for a muting circuitry (to avoid pops on power up)

IC1 is a L78M05R from Sony, a voltage regulator with a reset line. It provides a stabilized +5V, and also provides the reset signal during the boot up by maintaining the low level on the reset pin for a certain amount of time; the time is set by a capacitor (C7).

However, the datasheet indicates a high current drain can also triggers a reset (protection against shorts). 

What I did?

I could see on my oscilloscope a clear down on the reset pin when too much MIDI notes where coming in. My first guess was to replace this IC1 : bad guess.

I could also notice ~ 300mV fluctuation on IC1 output (input is clean). As the output is filtered by a few capacitors (2 are electrolytic, the others are ceramic), my second guess was a bad cap and I replaced all the electrolytics: bad guess again.

And now?

My first option would be to measure the output current drained from IC1; not easy, my DVM is not fast enough to capture the drop.

My other option is to find a bad join. I tried unsuccessfully to tip the board with a tool, but I didn't trigger any reset yet.

Stay tuned...

Update: problem found

I suspected the issue was linked with the DAC (Digital to Analog Converter). But the fact the problem occurred obviously when the unit gets warmer tends me to suspect a bad solder joint or a bad cap.
So I decided to redo the solders of the analog circuitry. While doing so, I found another clue: the solder does not glue to on leg of two capacitors (C18 and C19). These caps are responsible to bring the voltage reference to the DAC.
So here is what is going on: after 25 years, these two caps started to leak slightly, dissolving the track around the thru hole. When the unit gets warm, the solder becomes open and the DAC looses its voltage reference, inducing a oscillation on the power line (IC6, propagating to IC5 and to IC1). When this oscillation becomes important (above 440 mV), IC1 detects a voltage drop and triggers a reset.

Update 2: problem NOT found!

Replacing the 2 caps and repairing the PCB brought me late at night; I didn't have time to make tests before a  couple of days. At first, it seems Ok, but after a couple of minutes, deception stroke again: the unit had rebooted after receiving a loud chord.
However, I took the opportunity to make thorough testing: at first, the unit reboot after 24-25 notes (no matter notes are played sequentially or within chords, no matter the playing speed). And then, as time goes on, the number of notes I could play before a reboot is triggered decreases gradually until the unit reboots without receiving any MIDI. Unplugging-plugging the power cord has no effect: heat is definitely a factor.

As my scope has 4 channels, I could make some measurement:
- channel 1 (yellow) on +9V - power input
- channel 2 (cyan) on the output IC5 (+8V)
- channel 3 (purple) on the output IC6 (+5V for analog circuitry)
- channel 4 (blue) on the output IC1 (+5V for digital circuitry)

Update 3: problem finally found!

After making the above mentioned measurements, I realized something were wrong : I could see oscillations (~ 120 Hz) on probes 1, 2 and 3. Probe 4 was clean. But wait ! Why oscillation on +9V? And why 120 Hz (first harmonic of domestic 60 Hz)? The input should be at least a clean DC. I know some Roland products use AC input, but what about the SC55?
I gave a quick glance at the first section in the service manual: the recommended adapter should be ACJ-220 and I was using a PSA-220. The difference : ACJ-220 sources 1000 mA, PSA-220 sources 200 mA,
Because I purchased some of my instruments in France, I installed a couple of 220 V wall sockets at home, and I still have a bunch of 220 volts wall-warts. Most of the time, I use a generic 115V adapter, but on my workbench, as I have a 220V socket, I grabbed a 220V Roland adapter, but not the one recommended for the SC55. This was my mistake !
Once I change the adapter for the right one, everything was clean and worked perfectly.

A lesson painfully learned!

So, what did I learned? There was an issue, because the unit worked with a the same generic 115V adapter for years before it failed. I think the DAC capacitors were really wrong and had to be replaced. But because I did not take the good adapter, I was chasing an issue which was not real.
Well, we learn from our mistakes.
Anyway, I'm glad to see my Roland SC55 back in my studio !