Fluke 8500A Fixer Upper - story so far

[manicdoc]

Hi,

Managed to pick up a 'not working' Fluke 8550A for $100 AUD, figured it was either fixable for extending my fixing skills or worth that for parts alone. On initial visual examination found it was switched to 120v, changed that to 240v, and then had another look around in case you know what had occurred and found a blown-to-bits tant. Changed all the tants out, did a quick rail resistive check, and then got it starting up but showing a fixed distinct value per range. hmm... Then confirmed the isolator was passing the measurement bit stream via my oscilloscope, this should mean the digital backend of the ADC is good as well (at least it was reading something).

By the look of things the analog side of the ADC converter is good, everything probed good, managed to power it up off a bench +/- 15v supply and got -7.0000v on the nose on the test point - so the 'black magic' voltage reference circuit is good. I then checked the PSU, looks like either the -15v or +15v linear regulator is toast, ran very hot (they share a heatsink..). On the common failure principle, I'm going to replace all the PSU caps and linear regulators and double-check the rectifiers. Plus put the -15v regulator on its own heatsink and put in the recommended 1uf tant for stability.

I'm also replacing all the tants with new ones the next voltage spec up as I find them - given this is a known component of magic smoke doom. The board components don't look at all aged from what I can see, so I'm not going mad doing a swap out - everything I've measured is in spec, and no strange shorts so far.

I'll dummy load the PSU once everything is replaced and confirm its operation, plus resistive check all the rails on the motherboard for shorts, with emphasis on the 15v rails... Fingers crossed this should lead to a working unit.

As for how it failed, I think a heavy mains spike, given the damage on the PS side and 'weak' filtering by modern standards. But maybe someone gave it '240v surprise' whilst it was switched for 120v - wouldn't be the first time... Also, nothing wrong with the front end that I've found so far. I've also put a beefy 275v MOV across the mains for good measure.

The parts for the PSU will turn up next Thursday, by then I should have confirmed whether all the subboards are good or not.

[Kleinstein]

Linear regulator do get hot when running normal. It is very rare that a linear regulator blows in a way that it gets hotter than normal. A hot regulator is more due to excessive current drawn or maybe rather high an input voltage. A larger filter capacitor would make things a little worse.

The old days instruments often had quite some voltage headroom to allow for mains lower than nominal, and maybe to also account for 220 V operation  in much of Europe.
A 275 V MOV accross 240 V mains would be way on the low side and more like a liability, waiting to give smoke signals. A transformer and linear regulator supply is quite resiliant to mains spikes and short time overvoltage. They often had a fuse that would even protected it, if accidently run with 400 V and likely also in the 240 V in the 120 V setting case.

[manicdoc]

I think either the +15v or -15v line wasn't working or overloaded, I measured the -7v test point when it was running off the PS, it was +1.1v...

I'm doing a divide and conquer on the rails, I've brought out leads to allow me to measure them all and swap out the subboards to find if one is playing up. I'll also check across common ground lines as well.

the MOV is MOV-14D431K: 430V, 275VAC, 350VDC
https://www.farnell.com/datasheets/2360419.pdf

[manicdoc]

Quick update, found a shorted tant between the -15v rail (Vcc Logic Supply) and the Analog Return on the AC/DC Converter Board - in effect shorting the -15v regulator across its GND and OUT lines... Right next to a part in the circuit driving some relay coils with back EMF protection diodes... Hmm..  one coil measures 0.01v different from the other two in diode mode. Now either the tant died of old age, or a protection diode or two died - I'm going to have to get these relays off the board as the diodes are right underneath the relays to test them.

I'll test the other boards tomorrow. 2 bad tants down...

[manicdoc]

Didn't find anything else wrong, the protection diodes were good. Decided to power it up and we are getting measurements with no error codes, not correct measurements but consistent, i.e. same value for a given resistor within a very rough ballpark (over by about ~50% I'd say) and it detects an open circuit (flashing HHHHHH). Looks to be auto-ranging okay. So there is a signal flow from the front with ADC to the display and control signaling back.

Time to try the diagnostic flowchart...

[bdunham7]

Start with DC voltages and see how they work.  Then manually select the 10V range and give it 0-20V both positive and negative in 1V increments.  If the PSU voltages are all good and the ADC indicates 0-20V without any serious errors, then you're well on your way to fixed. I have one of these if you need measurements to compare, although given the state of my bench and desk, it might take a little while.

[manicdoc]

thank you, I'll try that.

[manicdoc]

Okay, so the majority are within 1% agreement with my DMM, the odd ones (6v and 10v) were out by whole volts up. A bit shifter problem? I'll look at the digital side of the ADC and see if anything jumps out.

Correction: got the +/- 15v rails in agreement to 2dp and what I'm finding now is a zone 10v to 13.8v where it jumps to 13.8v, then carries on fine beyond that.

[bdunham7]

Okay, so the majority are within 1% agreement with my DMM, the odd ones (6v and 10v) were out by whole volts up. A bit shifter problem? I'll look at the digital side of the ADC and see if anything jumps out.

Correction: got the +/- 15v rails in agreement to 2dp and what I'm finding now is a zone 10v to 13.8v where it jumps to 13.8v, then carries on fine beyond that.

Yeah, that's not good.  If the PSU is good, then you have to get that ADC at working at least nominally correctly before you go any further.  It's possible that someone twiddled the calibration trimmers or it might have some damage somewhere.  Spend some time reading the manual, it is a unique system.

[Kleinstein]

AFAIK this DMM uses the recirculating remainder ADC. This is a kind of successive approximation ADC. If something goes wrong in the coarse step (e.g. 1 bit value wrong or failing swtich) the ADC gets stuck over quite some range.  A similar error likely reapeats on a smaller scale (e.g. 1/16 the range of constant readings) several times.

It may be a good idea to look at the procedure for the ADC / linearity  adjustment. In a frist step still without doing any changes, but taking readings only.

[manicdoc]

AFAIK this DMM uses the recirculating remainder ADC. This is a kind of successive approximation ADC. If something goes wrong in the coarse step (e.g. 1 bit value wrong or failing swtich) the ADC gets stuck over quite some range.  A similar error likely reapeats on a smaller scale (e.g. 1/16 the range of constant readings) several times.

It may be a good idea to look at the procedure for the ADC / linearity  adjustment. In a frist step still without doing any changes, but taking readings only.

Okay, I've already diode range checked all the logical board drivers, and checked the resistor sets, everything seems consistent there. I'll get a set of measurements and if that does not reveal anything put my 4-channel scope on the driver lines first and see if anything is sticking there then proceed to the analog side.

Certainly getting my monies worth in fixing experience...

[bdunham7]

I would think your problem will be found on the analog side, but you never know.  The troubleshooting guide is fairly brief, but to the point.  If nothing on this page pans out, you might just want to run the A/D calibration procedure with the instruments you have on hand and see if you can get it back to a functioning state.  A stable PSU and a 4.5-digit DMM will at least be good enough for a probably-go/no-go decision.

[manicdoc]

I would think your problem will be found on the analog side, but you never know.  The troubleshooting guide is fairly brief, but to the point.  If nothing on this page pans out, you might just want to run the A/D calibration procedure with the instruments you have on hand and see if you can get it back to a functioning state.  A stable PSU and a 4.5-digit DMM will at least be good enough for a probably-go/no-go decision.

Thanks for that, I've got a 4.5d DMM. I'll go through the various tests this evening and let you know; hopefully, the problem can be cornered. I think the PSU is good, readings certainly don't jitter once they settle, although will be replacing the caps with modern ones of equal or better spec.

[manicdoc]

Did the -/+7v check first, I swept through 0 to 20v and was able to capture the following..





as you just go over the 10v mark, it sticks for a good 3v. Yet, I did see right around 0v, that the first 'bar' went to 7v. I'll take out the optional boards and see if it clears up before I dig deeper.

[bdunham7]

You're looking at TP5 on the analog board of the ADC here?

[manicdoc]

Yep output from TP5. No change with the optional boards. My gut is saying something up with ADC sampling circuity 'A'. The a +/- 7v transition is per sampling round of 5 switched samples. So what I'm seeing is it hunting for the value by addition or subtraction as needed per round.

I'll see if I can get the X and Y values next and zoom in...

[bdunham7]

Try just running through the calibration procedure from the beginning.  If you get to a step that fails (can't be adjusted to the requirements) then maybe you'll have a better starting point for troubleshooting.  If I get time/space tomorrow I'll look at TP5 on mine.

[manicdoc]

Try just running through the calibration procedure from the beginning.  If you get to a step that fails (can't be adjusted to the requirements) then maybe you'll have a better starting point for troubleshooting.  If I get time/space tomorrow I'll look at TP5 on mine.

I'll get screenshots of the signals for its state now first and see if I can deduce from that first. Anything sticking should be clear.

[bdunham7]

I'll get screenshots of the signals for its state now first and see if I can deduce from that first. Anything sticking should be clear.

You'd have to look at the digital interface signal and figure out how to interpret it to make sure any 'sticking' wasn't actually commanded.  Because of the way the ladder works on this, if you were to have on section of it far enough out of adjustment (or broken) you get these issues.  I've not seen one badly enough off to not change over a 3 volt range, but I've seen them looking pretty ragged in the last 3-4 digits due to someone trying to calibrate a broken meter.

I popped the lid on mine and scoped TP5.  With a 10V input I get the first shot, expanding that to look at the slopes you see the part that I think is important.  If your slopes don't look like that there might be an issue in that area.

Edit:  These are single-shot captures for clarity.  With color grading, you see quite a bit of variance in the timing and duration of the signal.

[Kleinstein]

Try just running through the calibration procedure from the beginning.  If you get to a step that fails (can't be adjusted to the requirements) then maybe you'll have a better starting point for troubleshooting.  If I get time/space tomorrow I'll look at TP5 on mine.

I would first go through the ADC calibration procedure without doing any adjustment, just measure the testpoints / signals. With the described symptoms chances are that one of the earlier steps is way off and chances are high that the trimmer would not be sufficient to adjust anyway (unless the trimmer itself is the problem).

[manicdoc]

Try just running through the calibration procedure from the beginning.  If you get to a step that fails (can't be adjusted to the requirements) then maybe you'll have a better starting point for troubleshooting.  If I get time/space tomorrow I'll look at TP5 on mine.

I would first go through the ADC calibration procedure without doing any adjustment, just measure the testpoints / signals. With the described symptoms chances are that one of the earlier steps is way off and chances are high that the trimmer would not be sufficient to adjust anyway (unless the trimmer itself is the problem).

Once I'm off a Zoom call, I'm going to do this...

Okay, just ran through the steps on 4-8 and 4-9  A/D Ladder Adjustments:

Step          Target             Actual
3)             +10.1             +13.88
5)              -10.1             -13.77
8 )              +0.5              +3.86
11)             +5.1             +5.23
13)              +2.6             +3.86
15)              +1.4             +3.86
17)              +0.75            +3.86   


I have oscilloscope on TP5,1,2 & 3 in order and have the following:

[ Specified attachment is not available ] [ Specified attachment is not available ]

To my eye, it looks like some overflow/noise on the 10.91v (which read as 13...) compared to 8.48v (which read correct).

Actually, if you look at the TP3 (bottom) on the first one, it hits the 15v rail as the 16x multiplier kicks in, so after that everything will be the same until its gets beyond the 13.383.. value that happens to require the 'bad' bits to be set. So this only occurs also on the 0 to 3.83, and I don't see it occurring on the 5v, so 10v 'bit' is okay, and probably 5v bit as well.. I notice the rail hit on other cycles, so some bits are stuck off I'd say for sure. Checking the jfets will essentially require desoldering as they are all wired in parallel... or this could be a fault on the logic board around U22. Tomorrow I'll check if the pulses are getting off the logic board, that will steer me to one board or the other.

[manicdoc]

I'll get screenshots of the signals for its state now first and see if I can deduce from that first. Anything sticking should be clear.

You'd have to look at the digital interface signal and figure out how to interpret it to make sure any 'sticking' wasn't actually commanded.  Because of the way the ladder works on this, if you were to have on section of it far enough out of adjustment (or broken) you get these issues.  I've not seen one badly enough off to not change over a 3 volt range, but I've seen them looking pretty ragged in the last 3-4 digits due to someone trying to calibrate a broken meter.

I popped the lid on mine and scoped TP5.  With a 10V input I get the first shot, expanding that to look at the slopes you see the part that I think is important.  If your slopes don't look like that there might be an issue in that area.

Edit:  These are single-shot captures for clarity.  With color grading, you see quite a bit of variance in the timing and duration of the signal.

What I saw:



Looks comparable I'd say on TP5. That's for 4.0056v and is within 1% of actual.

[manicdoc]

Update: measured the control lines for everything but the 10v stage on the ADC, the 2 smallest look good, with proper mirroring going on, but the next 2 don't seem to be having the mirroring as it should, the negative is 'weak' or not there.  I'll do more analysis this evening. Could be bad jFETs, bad cable interconnect, or bad logic, but the fact two negative channels both have the same problem is leaning me towards the logic board and the circuit around U22... I'll test the interconnect cable as well..

[bdunham7]

Okay, just ran through the steps on 4-8 and 4-9  A/D Ladder Adjustments:

Step          Target             Actual
3)             +10.1             +13.88
5)              -10.1             -13.77
8 )              +0.5              +3.86
11)             +5.1             +5.23
13)              +2.6             +3.86
15)              +1.4             +3.86
17)              +0.75            +3.86 

OK, I don't think calibration will help that.  If you can't get the 10.1V step dialed in, then it's broken for sure.  It looks like you've dived in headlong and are on the right track.  You'll be the forum resident expert on these ADCs before long.  If I get a chance or you get stuck I'll try connecting the LA on my scope to the control circuits and see if I can make sense out of them.

[manicdoc]

Make sure your LA can deal with +/- 15v range without smoking... the logic drive is relative to the SD fet pins beyond the transistor arrays on the logic board