Another K2010 repair thread

[Cyclotron]

U148 control over U149 appears to be the problem when in failed state.

When working the average signal voltages towards the gates on U149 are averaging 2.5vdc. In a failed state, they are sitting against the 5V DC supply rail of U148, but the signal is still in there.
The issue is that it doesn't swing the required amount to switch the gates of U149, and so control is lost.

Now I need to sort out why these voltages are sitting up against the rail of U148 instead of averaging in the middle of 0-5vdc.

[Cyclotron]

Ok, the signals on the ADQ bus are coming out of U165, but they have only 450 mV swings.  Per my prior captures, they should be 0-5vdc swings. Soooo... hmm. Starting to worry about my PLCC chip. Time to take a serious look at its rails again to be sure it's getting what it's supposed to.

[Kleinstein]

Judging from the signals for pins 20 and 21 and the integrator signal the reference switching and the integrator are working correct. The ADQ0/1 signals are high to activate the ref. signals. I don't seen an issue with U148 or U149. The references are clearly switched - it would be a very odd defect to change the polarity of both signals.

The problem are the signals coming from U165.
Not necessary every step should trigger the comparator, but the voltage should not move away from zero for more than 2 steps in a row - already 2 steps should rarely happen, if at all.

The gate signals should not be stuck, but still have the short pulses. At least the intergrator ouput as the short steps down. So the switches work, just not the right control signals.

[Cyclotron]

Judging from the signals for pins 20 and 21 and the integrator signal the reference switching and the integrator are working correct. The ADQ0/1 signals are high to activate the ref. signals. I don't seen an issue with U148 or U149. The references are clearly switched - it would be a very odd defect to change the polarity of both signals.

The problem are the signals coming from U165.
Not necessary every step should trigger the comparator, but the voltage should not move away from zero for more than 2 steps in a row - already 2 steps should rarely happen, if at all.

The gate signals should not be stuck, but still have the short pulses. At least the intergrator ouput as the short steps down. So the switches work, just not the right control signals.

I just realized I have a bad scope probe for the last few tests so I'm going back to do those measurements again.
Sigh....

[Cyclotron]

ICB+ and ICB-

Not sure why the ringing is on the ICB+

I have assumed that the sawtooth on U137 should normally cross zero on each tooth segment, not just climb 7 steps and then repeat.

I'm trying to sort out what should trigger it's polarity swap.

I'm going to let the K2010 rest and see if it returns to working as I've missed some observations that I need to compare with. At this point, I no longer suspect the PLCC as it seems to be getting signals and responding but the AD/integrator stuff appears to be what's going off the ranch.

Thanks!

[Cyclotron]

ok, in this failed state, pin 2 on U138 has a constant -0.503 VDC, and I'm guessing that's what is keeping the integrator from sawing its way through zero crossings.
With that constant bias, it starts to move towards the positive rail on U137.

I lifted R287, and the errant voltage remained.  Lifting L103 eliminated the -0.503 VDC voltage and the signal, of course. The voltage remained at pin 11 of R271.

At this point, I suspect the R271 or PLCC is the cause, and I'm leaning towards the TF245 given its history.
U148 has been replaced, and so have U149, U153, Q175-178, and U138. I had all of these lying around, so I put them together to make some eliminations.

When putting L103 back, I just swapped it with L104 to have another variable to kinda check off.

The U148 area is either getting the voltage from a bad R271, which I'm not sure about, or the signaling to it from U149-U154-U165 is the cause.
Maybe after it sits tonight, it will come back, and I can measure the R271.  I measured it in the circuit today, but I have nothing to compare it to.

[KungFuJosh]

You should definitely measure R271 in and out of circuit. Voltage & Ω in circuit, and Ω out of circuit. Values for reference in the other thread.

[Kleinstein]

Unsoldering is quite some stress for R271. So I would definitely avoid it. It should be good enough to check the +-12.8 V reference voltages.
There is no indication that R271 is causing the issue.

The integrator waveforms shown look very much like the reference, the reference switching and the integrator are OK. The intergrator signal perfectly follows what is expected from the control signals send to the switches.  The problem is the signal that is shown for pins 20 and 21 of U165 (the ADC ASCIC).
This could either be an issue with U165 itself, maybe it's supply (there are multiple supply pins) or the komparator signal not getting in fully (supposedly is there). I would consider resoldering U165, just in case there is a cold solder joint (e.g. with the supply) or tin wisker near the chip. If it is a loose bond wire, the chip may however also fail fully from the thermal stress.

I don't think it is the optocoupler or control signal, as the pattern show the normal period, just as if the chip does not realize the state of the comparator.
This pattern should be produced by the ADC chip and not controlled from the µC side. The only point is that we don't know how much of the function is fixed and how much can be programmed from the µC.  From the die photo shown on "the signal path" channel it looks like U165 is a gate array (kind of a mask programmed logic array) and thus more or less fixed logic and not an FPGA or µC that would be programmable or a CPLD that might have lost some content.

[Cyclotron]

ok, today it is back to working. I have also gone ahead and re-soldered the PLCC and the vias near it.

The errant -0.504 VDC is gone, and that's expected. The sawtooth from U137 is what I expect, though I think I may have misunderstood how the pattern works.  I'm going to capture that sequence more precisely while it's working.

[Cyclotron]

I had the parts in stock so I just replaced everything in the analog AD integrator. 
U148, U149, U152, U153, U137, U138, U142, U145, CR101, Q175-178.

U139 p 6 is green in these caps. It is working right along and you can see U142 p 6 following along as it should.
Yellow is watching ADQ0 on U154. 

It works along for a while, then you see U139 p6 start to go high, and it goes right off to the rail.
If I leave it on, it will soak enough to stay there until it is left off for a while.  I can cycle it near this threshold and keep repeating the failure start.
I'm trying to find a source of the push to high.

[Kleinstein]

It looks a little like the run-up part is working - looks like some 20 ms. The non working part seems to be the run-down part. The integrator is going up rather slow in the failing phase. This could well be the slow slope part of the run-down.

The analog part of the ADC was working before and now also seems to work OK. It is odd that the failure from the digital part changed - now working for nearly all the conversion, just not detecting the very end.

I would suspect U165 and maybe it's supply. It is still odd to have digital logic to be flaky and somethimes work and sometimes partially fail.
It might be a dependence of the supply, like work with 5 V and fail with 5.05 V.

[KungFuJosh]

Tek said:

U165, PROGRAMMED IC that appears to be the 2000-802A02 which sells for $36.

If you don't have a direct contact yet, you can email buy@tektronix.com. I recommend getting a spare TF-245 also while you're at it.

Thanks,
Josh

[Cyclotron]

Tek said:

U165, PROGRAMMED IC that appears to be the 2000-802A02 which sells for $36.

If you don't have a direct contact yet, you can email buy@tektronix.com. I recommend getting a spare TF-245 also while you're at it.

Thanks,
Josh

I just bought another bit of equipment that has the U165 and the TF245 in it so I can use those.
Shoudl be sitting in the lab in a few days so I can start my work again to solve the K2010 issues.

Thanks for the info!