Roland Jupiter 8 VCO1 drift

[rich_de]

I thought of one other thing in the meantime: If the drop is (mostly) a constant frequency (between low and high octaves, say 3 Hz), it is likely related to the integrator reset circuit, which corresponds with my assumptions. If it is a constant ratio (say 50 cent), it's probably something outside of my guesses (but you'll still want to know what effect the switching has to make make further guesses).

And the "opposite check" with grounding pin 14 of the IC17 4052 in square mode should be safe because it has 10k on one side and 100k after the output, even with the IC built in, so it can be done before getting down & dirty; I was under the impression that you might have sockets for the CMOS ICs anyway from servicing.

For visual inspection you might have a very close look at pins 4, 6, and 7 of IC 3A and 4A (power and output lines) for hair cracks. IC4A works against IC3A and that current might cause trouble over 40 years of service.

ps Forgive me if I lack a bit of respect for the vintage celebrity. In my "shop" project, after I've soldered 16 JP8-signal compatible voices myself, I outsourced the next 16 to two assembly houses and one of them delivered exactly zero good boards (solder splatter and a handful of wrong components per board), which I had to debug - with lots of swearing and resorting to ever cruder methods over time

[DesolationBlvd]

I thought of one other thing in the meantime: If the drop is (mostly) a constant frequency (between low and high octaves, say 3 Hz), it is likely related to the integrator reset circuit, which corresponds with my assumptions. If it is a constant ratio (say 50 cent), it's probably something outside of my guesses (but you'll still want to know what effect the switching has to make make further guesses).

Just had the chance to try switching the 4052s. No effect. Upon closer listening, it sounds like the pitch drop is in cents and not hertz.

[rich_de]

Just had the chance to try switching the 4052s. No effect. Upon closer listening, it sounds like the pitch drop is in cents and not hertz.

Alright, that eliminates a good part of my guessing. Can you clarify that we're still talking about one particularly bad voice here, and the problem did not "spread" to thee others, there was somewhat conflicting information:

The autotune always leaves one voice's VCO 1 noticeably flat (only objectionably so in the top octave), and swapping the card causes the flat voice to move with it.

and

On my JP-8, the saws on the odd numbered voices are slightly flat compared to the pulses, and this is both before and after auto-tune. The auto-tune seems to work fine for the pulse

If four voices are affected in the same way, we'd probably have to look at the module controller, but I assume for now that one voice is particularly bad. The offset being in cents rather than Hz leaves us an error source either in front of the expo, or with the reset threshold, which would hint at some odd ground issue. To rule out anything the JP does that we overlooked, can you do the test with the manual switch installed in the socket instead of the 4052, and a precise frequency counter hooked up (best to the square). Set the patch to square and tune, it should be fine now. Does the frequency then jump when you flick the switch? If so, we can continue guessing about the steep pulse rising edge interfering somehow from down the signal chain. If not, we have to figure out what else the JP does when it sets the 4052 to saw.

[Thunderbird]

Hi, There was a fix for a ground loop on the roland boards it comprises of cutting one track and putting a reistor in on each voice board.
I have attaached the relevant roland details. Worth looking into that maybe?
Regards.

[DesolationBlvd]

Update on my issue (saws being flat but pulses OK), I have since taken the JP-8 back to Bell Tone Synth Works (and had Encore MIDI installed while it was out). They found that the selection of the saw itself loaded the whole oscillator down. They buffered an offending voice between the saw output and waveform selection, which fixed it. I am picking it up next Friday.

Before I took the JP-8 back to them, I performed the "Hum and Sound Leakage" ECO myself. Quite nerve-wracking, taking a razor to the circuit boards of an instrument worth more than my car. No effect on tuning, but it helped reduce noise.

[DesolationBlvd]

Epilogue: More of the oscillators started sagging again. I made and installed more buffer modules, but one of them I made was faulty, and I replaced the 3046 trying to diagnose it. Reverting the change and it seems to be good. Maybe the 3046 was on its way out. Knock on wood.

I also fixed the sync on mine having a weird hiccup, it's an early first-500 unit. C17 A and B were 15p versus the 33p the service manual says. Correcting the value made the sync much smoother.

[rich_de]

Epilogue: More of the oscillators started sagging again. I made and installed more buffer modules, ...
I also fixed the sync on mine having a weird hiccup, it's an early first-500 unit. C17 A and B were 15p versus the 33p the service manual says. Correcting the value made the sync much smoother.

The thing with the required buffer modules is really weird. There must be cause why it worked before and now only with them installed. I'd now suspect the integrator opamp itself,

Also, with your MIDI installed, would it be possible to set the JP to "mono" (hoping that it will be confined to one voice) and run the VCOTuner program discussed here in https://www.modwiggler.com/forum/viewto ... ?p=3771874? I'd be curious to see the deviations over the frequency range, and also the "uncertainty" of the oscillators at the different frequencies. This might give an idea what's at the root of your issues, and would also be interesting to me to see how different the JP-like oscillators in my poly project are from the vintage version.

Also, if you somehow can feed external CV from MIDI, either setting a low fixed frequency and feeding external CV through 100k to the IC1 side of R46, or doing a bit of analog computing with the input side of R46 and feeding the required difference, tuning could be made much, much easier, faster, and more precise with that VCOTuner app.

With the sync, I'm a bit baffled that it works at all. I'd guess there are around 28V swing at TP1, and dividing that through R49/R48 would just barely be enough to turn on TR9A alone, not accounting for other losses and the C17 discharge through R48 during the edge. Having a meg or so parallel to R49 might also help.

[DesolationBlvd]

The sequence of events:

Before I sent it back out, all odd saws were sagging.
Then they installed one buffer module and apparently it fixed all the voices.
Then the three voices that didn't have the buffers sagged again.
Then I made two buffers, and the buffered voices were fixed while the remaining one sagged.
I realized the two buffers were floating around in the case, so I ordered standoffs and made one more buffer.
This last buffer I made was faulty, and I replaced the 3046 along with socketing IC1, IC3, and IC4 before thinking to revert to the resistor that was there before.
That last voice sagged before I attempted to buffer it, but stopped sagging after I replaced the 3046 and socketed the other ICs.

[rich_de]

Then the three voices that didn't have the buffers sagged again.

So three voices started sagging at once? That's some high probability it's not voice related. If that's indeed the case, the 13V regulation on the module controller board would be a primary suspect. If you have a reasonybly precise voltmeter, you could give it a check.