Restoration / Repair: Roland JX-1

[Raven Luni]

I've been working on restoring my old synth (Roland JX-1) and it is throwing up some interesting challenges. The CPU usually crashes and loses memory settings when powered on. I've replaced the battery and done a partial recap and it has made little difference.

On looking at the schematic it looks like the power circuit is rather poorly designed so I've been thinking about replacing it. At present it takes an unregulated 12V AC input, uses half wave rectification to get the digital 5V rail and a charge pump configuration for the +12V and -12V analog rails. Something else I remember from years ago is that it was more stable with the screws out. This suggests some kind of ground interference since the screws appear to connect the ground planes to the metal case.

Here is a portion of the schematic covering the power circuit (top half).


I've also included some pics of the PCB (flipped so the sides match up)


I thought of replacing it with a rail splitter powered from a more standard 9V DC input. The DC-DC converter modules I have are adjustable ones that put out very stable voltages at over an amp.
Let me know if this virtual ground circuit will do the job or if if it needs separation of grounds or the 5V rail needs to be tapped form a different point etc (I'm still learning when it comes to analog stuff)

[floobydust]

There's nothing wrong with Roland's power supply design - I'd just leave it alone and supply it with a high quality 50/60Hz 12VAC transformer.
Changing to SMPS (or single rail) will add lots of noise to the audio chain and not really work here because the +/-12V supplies come from a voltage doubler. Note C29, C30 are high stress parts and will have a short life due to the very high ripple currents.

I would look at the 5V rail to see it is stable, and the CPU RESET circuit and SRAM memory disable to see if it is working. As far as CPU problems, I'd look for bad solder joints as keyboards get banged around a lot. Give it the tap test, with a plastic pen (no metal) and see if anything makes it crash. Stare at the board with a magnifying glass. Roland's
 soldering can fracture. Otherwise, it might be bad memory - SRAM or ROM that is malfunctioning but this shows up when it's warmed up.

[Pete66]

The 5V regulator has a reset pin that is supposed to reset the CPU on power on and off.  This is probably what the problem is.
I would replace the 5V regulator. 

[shakalnokturn]

Something else I remember from years ago is that it was more stable with the screws out. This suggests some kind of ground interference since the screws appear to connect the ground planes to the metal case.

That would be very surprising.
OTOH it would point towards weak solders or bad contacts.

[elecdonia]

If the power supply voltages are within spec and free of AC mains ripple and noise, then try this next:

I’ve encountered more than one faulty/intermittent MCU clock oscillator crystal in items like yours. A typical symptom is intermittent crashing and/or failure to start up when power is applied. Oscillation at the crystal can be verified with an oscilloscope. Use a x10 scope probe to avoid loading down the oscillator circuit.

Most crystals are labeled with their oscillation frequency. Always use a replacement of exactly the same frequency. That’s the only spec that’s really critical.

-E