Need help with Jupiter 6 CPU reset circuit

[sqweebking]

I'm chasing an intermittent no-boot on my Jupiter 6 but I'm not 100% sure the issue I think I have is what is causing my problem. If that makes sense. Could someone more knowledgeable than myself tell me if I'm on the right track or if I should be looking somewhere else?

My Jupiter 6 won't boot sometimes. In checking the CPU reset circuit I found that the duration of the on time is only about 35-40ms. The service notes say that this time should be about 80ms, is this true? Their picture shows the ramp up taking ~20ms and the ON time ~80ms. My whole cycle is done in ~50-55ms.

Yellow = Reset @ CPU Blue = Reset @ PSU



I've replaced C33 and C43 and while the reset duration is a more consistent 40ms ON time. Overall it still seems short, so what else could cause this? OR am I looking in the wrong place?

[Rasputin]

I'm definitely not "someone more knowledgeable" but my thinking is that if you are trying to fix or alter an RC delay and you've replaced the caps then the resistors are the next place to look. I know resistors are generally assumed to be good, but you know where assumptions get you in situations like this...

Barring that, you could probably substitute a larger capacitor for the 0.1uF and really notch up the delay.

Seems like the op amp and the transistors for the reset timing should be working okay since the reset is working to some degree, it's just the timing is off, and that timing should be set by the combination of capacitors and the resistor network. Right?

Not something I know much about, but maybe my musing will inspire you.

[sqweebking]

Yea the resistors were the next thing I was looking at, some of them can be measured in circuit but at least one or two need to be pulled to verified. I figured before I started pulling more components I would try and check with the more experienced people here. I'd mainly like to confirm that what the service notes are showing is indeed true, my waveform does *look* like theirs but the duration's are throwing me off. They don't do a good job of indicating that 20ms ramp up is shorter than the 80ms on time

Anyway, I'll probably pull those resistors tonight and check/replace. If still nothing changes after that I'll begin to assume the service notes are just poorly drawn and written. (I did order some replacement M51201's but like Rasputin said, it and the transistors appear to be working)

[madtheory]

Ya that's a funny representation of time in the service manual. I think that intermittent would imply a temperature related problem such as a dry joint, or a failing joint in the IC, or the trannies. Try heating up the IC with a hair dryer, while scoping the start up. See if you can re-create the no boot. You could try freeze spray as well.

[sqweebking]

I've scoped the start up numerous times, catching both a normal boot and non-boot. The only difference I could see was about 5ms difference on the non-boot but even still that wasn't super consistent. When the CPU is in non-boot fault mode all of the data pins are held high @ 5v (at least all the ones I checked, most of them) so it seems either the 40ms on time I'm seeing is either normal and I'm chasing the wrong thing or its just on the edge of "long enough to work" and "not long enough to work" once its warmed up.

Freeze spray isn't a bad idea, it does seem that once the board has been on for a while the no-boot issue is more persistent. Usually on a cold start, it starts right up, after 30min to an hour of being on it won't boot 4/5 times.

[madtheory]

OK well I think if the difference was only 5ms and it's not consistent then that is not the source of the problem. If we assume the diagram in the service manual is drawn with intent, then they're saying approximate really is approximate, so 5ms is well within the error margin for normal functioning. Freeze spray other relevant parts of the circuit perhaps?

[Rasputin]

When the CPU is in non-boot fault mode all of the data pins are held high @ 5v (at least all the ones I checked, most of them) so it seems either the 40ms on time I'm seeing is either normal and I'm chasing the wrong thing or its just on the edge of "long enough to work" and "not long enough to work" once its warmed up.

You never specifically said, but I'm assuming that the CPU operation is stable once booted, right? If so, then you've got good clock, power, ground, etc. That leaves the RESET line for intermittent boot (which is the most obvious candidate, anyway).

Given that some CPUs are quite sensitive to duration of RESET signal, I'd be inclined to say you're barking up the right tree.

I believe that the CPU RESET has to be held active (high) long enough for the CPU to register clock and VCC to stabilize. I'd think the RESET signal is dropping low before one or both of those events occurs. To test this, you could probably manually jump the RESET line to high and then release it.

[sqweebking]

You never specifically said, but I'm assuming that the CPU operation is stable once booted, right? If so, then you've got good clock, power, ground, etc. That leaves the RESET line for intermittent boot (which is the most obvious candidate, anyway).

Yep, the synth is rock solid once booted. I did have some major audio issues before but they've since been resolved by a combination of cleaning all of the connectors at the module boards and replacing the 3 EPROM's.

Going to pull the CPU board again and check those resistors and reflow the CPU socket itself.

[Rasputin]

As clarification to my above point, the service manual says the pulse should last at least 24 clocks [two machine cycles], but that doesn't *really* tell you anything because how long does it take the crystal to stabilize? The 24 clocks is a minimum for the CPU but that doesn't speak to the crystal or the CPU voltage which has to reach the correct state before those cycles.

Additional thought: If the capacitors have residual charge then they'll reach their charged state faster than when they have no residual charge. Therefore, on a cold boot, they will take longer to reach the state in which the RESET line goes low. If they go low too soon then the pulse will be too short.

That could explain the cold/warm boot issue, especially if the RESET pulse is right on the threshold. The thing is... the RESET pulse length isn't constant in behavior, and the needed RESET length isn't a constant either. If the crystal and voltage take too long then it doesn't matter if the RESET pulse is the "correct" length, and even if the crystal and voltage are "correct" then the RESET pulse has to be long enough. The pulse length has to always exceed the maximum time of crystal/voltage ramp up for the CPU to boot 100% of the time. Both conditions have to form within a specific window. It has to be put within a "sweet spot" because RESET timing is a moving target not a fixed value because the behavior of the circuit is variable (albeit within a certain range) every boot event.

[sqweebking]

You gave me an idea, I measured the clock signal and reset signal at boot time to see if it was cycling at least 24 times before reset goes low.


Yellow = Reset @ CPU
Blue = 12mhz clock pin 19


650.0us per cycle x 24 = 15600us = 15.6ms

38.4ms = 38400us = 59 cycles

[Rasputin]

The only thing left is the +5V line to the CPU then, it seems. Maybe there's some issue with the regulator taking time to fully kick in or it has some oscillation issue or something.

What if you feed the CPU +5VDC from a source other than the stock power supply?