Trying to troubleshoot Roland Jupiter-4 negative regulator cct

[MattPadden]

Hi,
Looking for some advice/understanding about a Roland Jupiter-4 power supply that I'm investigating. Schematic is attached for reference. On receipt, the UA723-derived -15V regulated rail was noisy and jittery, wandering randomly over several mV. I've fully rebuilt the board but the -15V rail is still noisy and still wanders over several mV, in both load and no-load conditions. By contrast, the +5V and +15V rails (both generated by 78xx 3-pin regulators) are smooth and stable.

For the rebuild, I uprated the bridge rectifiers to Vishay GBPC6 types (the originals were underpowered and ran very hot) and fitted Panasonic FR series reservoir caps throughout. New resistors, new trimmer, new series pass and small signal transistors, new UA723, new 7805/7815 regulators. The +5V and +15V rails are stable and smooth, but the -15V is exactly the same noisy jittery thing. Attached images are video stills of the -15V regulated rail (top) with either the regulated +15V rail (flat line) or the unregulated -ve supply (rippled line). All traces are at 2.5mV/div on the scope.

I'd like to think a 723-based regulator would be much quieter and more stable than this, but am I wrong? Is this expected performance? I'm hoping not, given the -15V rail stability directly affects VCO pitch in this instrument. The circuit as built is effectively the example negative regulator from the 723 datasheet (Fig 18) with the only differences being the external current limiting arrangement and the implementation of output voltage adjustment.

The only original component in the power supply is the mains transformer (providing secondaries of 40V CT for the +/-15V supplies and 12V for the +5V). But would a transformer issue manifest as this sort of behaviour?

[Audiorepair]

Have you checked for the conductive glue problem?

You might find the main caps have been glued to the PCB.
This glue might have turned corrosive and conductive, and found its way onto nearby components and started to eat them, or nearby tracks.
It will probably have turned dark brownish by now, you need to get ALL this off the PCB.

I have a Juno 106 in my workshop now that has a very similar circuit, adjustable -15v, fixed +15v regulator,  the glue had leaked onto a resistor and eaten through its lead.

[MattPadden]

Thanks audiorepair. The board was stripped clean (a previous 'repairer' had used high-temp lead-free solder to improve things) and all glue & flux removed before I repopulated it.

Foolishly I tried adding a Vref decoupling cap between pin 4 (inverting input) and 0V to see if that was the issue, but that killed the 723; I think I exceeded the error amp's maximum differential voltage at startup.

I need to breadboard a duplicate cct.

[Audiorepair]

I have 2 bench power supplies that I can rig up to provide the + and - 18 volts or so required to power up the circuit.

If you could possibly arrange that it would rule out, or in, the transformer to be the problem.


You could actually do this with 4 x 9v batteries if  you don't have bench supplies.

Now there's a thought.

This kind of thing:
https://triembed.org/blog/wp-content/uploads/2014/09/DSC_3015.jpg

[MattPadden]

Supplies aren't an issue; mainly I'm intrigued as to how such a comparatively straightforward cct should produce so much noise! I will triple check the PCB for contaminants but I'm pretty sure it's clean...

[Audiorepair]

It won't produce that noise.

You haven't found the problem yet.


Check all the grounds are good, and hard wire them anyway to test.

[MattPadden]

The only original component is the mains transformer (secondaries at 40V CT for the +/-15V supplies, 12V for the +5V supply). Would a short in the windings cause such noise?

[Audiorepair]

Yes, but you could rule that out or in using batteries or bench supplies.

[MattPadden]

Surely. In the current setup, the 0V line and the unregulated negative supply are the only jitter-free nodes. Everything else is exhibiting jitter.

[Audiorepair]

There is no more to be said here until you try either bench supplies or batteries to rule out the transformer.

[MattPadden]

Indeed. Report to follow, likely Friday.

[MattPadden]

There is no more to be said here until you try either bench supplies or batteries to rule out the transformer.

OK:
Detached the transformer and ran the board from a pair of DC bench supplies set for +/-23.7V (same as the unreg DC as measured at the transformer attachment points). No change.
Measured the DC resistance of the ground path between the 0V centre-tap attachment point and pins 11 & 12 (Vcc and V+) of the 723 at 0.028ohm.
Moved the 0V attachment point directly to 723 pins 11 & 12 to remove the ground path effects. No change.
Swapped the 470p disc ceramic compensation cap for 3 x 150p polystyrene in parallel (didn't have direct value to hand). No change.
Placed a 4.7uF (and later 100uF) electrolytic between Vref and V- to decouple the Vref voltage. No change.
PCB is clear of flux and has no visible damage.

I'm going to breadboard a duplicate cct next.

[Audiorepair]

Could there be some contamination you can't see under the output sockets?

Sometimes you can just lever off the plastic parts and slide them upwards off the pins and have a look, so you don't have to  desolder them.

[MattPadden]

Could there be some contamination you can't see under the output sockets?

Sometimes you can just lever off the plastic parts and slide them upwards off the pins and have a look, so you don't have to  desolder them.

I've just breadboarded a duplicate cct and am seeing exactly the same couple-of-mV DC wandering and 0.5mV-or-so AC noise on it, powered from DC bench supplies. The wander and noise persist regardless of whether the current limiting transistor arrangement is in cct or not, or whether the trimmer adjustment is in cct or not; in other words, I reduce the cct to a straight copy of the Fig18 negative regulator from the datasheet. Presence or absence of the compensation cap makes no visible difference.
Am I being all Don Quixote here and tilting at something that is in fact expected operation?

[Audiorepair]

The PCB is a component, and they do occasionally go wrong.

Marshall amplifiers infamously had a bad batch of PCB's that went conductive, but then they have 500v DC on it.
I recently had a Korg Poly 6 that I also believe had a conductive PCB - cutting a trace at both ends and running a wire cured a problem, but then it doesn't take much to put VCO's out of tune.

Can't quite see how a low voltage power supply can suffer from a bad PCB, unless there is still some of that conductive glue around somewhere you can't see.


Edit:  Oops, I posted while you posted.

[MattPadden]

But the breadboard cct is also exhibiting the same behaviour. So the PCB is not an influence here.

[Audiorepair]

Yes, I posted before seeing the post about the breadboard, the posts are less than a minute apart.

[MattPadden]

Yes, I posted before seeing the post about the breadboard, the posts are less than a minute apart.

It's tempting to set up another negative regulator cct built around an LT3015, just so I can eliminate the environment from the investigation. I've tried switching off every switcher-powered item apart from the DMM (an Agilent 34401a), but there's no discernible change in the wander & noise.

[Audiorepair]

Yes, I posted before seeing the post about the breadboard, the posts are less than a minute apart.

It's tempting to set up another negative regulator cct built around an LT3015, just so I can eliminate the environment from the investigation. I've tried switching off every switcher-powered item apart from the DMM (an Agilent 34401a), but there's no discernible change in the wander & noise.

The question why only the negative rail is adjustable is interesting.
The Juno PSU I repaired recently was very similar, and any change to the PSU requires a total recalibration of the whole keyboard.

The alignment procedure in the Juno says the -15v rail accuracy is critical, the +15v was don't care so long as its in the same ballpark.

[MattPadden]

Yes, the JP-4 is similar. There was a factory revision at some point which added separate direct cabling of the -15V rail to the key assigner board to mitigate tuning problems. So it’s a double shame that it’s the noisiest rail!