Repair worklog : Old rusty HP 3458A

[Dr. Frank]

I understand the meaning and difference behind resolution and accuracy, just interesting how different products aproached this. Low ranges on DCV also provide less digits on screen.

I did never doubt that!

Your zero noise project comes in handy for this aspect.

Maybe we'll finally see some people delivering figures for the KEITHLEY DMM7510 (Shahriar) and also the 8508A (quarks)??

Frank

[Kleinstein]

With a fast comparator, the decoupling caps should be smaller. More like 10-100 nF on the board should be better. If really needed one might add the larger ones (e.g. 10 µf) with enough ESR (e.g. 10-50 Ohms) to damp resonances, e.g. by using electrolytics for these or with resistors and MLCCs.

To large a decoupling cap lowers the resonace too much and might gives ringing in the frequency range of interrest.

[TiN]

Problem video

[Kleinstein]

At high speed (low NPLC) the ADC is used in a different mode (kind of. dual slope). The same is true for AC measurements.

The message sounds like the ADC is sometimes not getting a stable state in the slow, multi slope mode.
There could be several possible defects / parts that make to ADC loop instable:
One could be a slightly too small integrating capacitor - could test with 10 % more.
Also a to large value for the larger of the integrating resistors could cause such a problem - in this case the error should depend on the voltage measured. In this case the second slower slope might take too long.
If it's a kind of analog error, more measurements of the ADC at work might help. However this might need a deep memory scope, to resolve a full conversion cycle at something like 1 PLC.

As there was some kind of mistery fix for this problem, it may be a problem that is rather difficult to locate. It took HP some time to find a fix for some kind of design flaw. So it might help of one could get the change in circuit between the old and fixed version. Worst case this is something like an updated FPGA / costum chip.

My guess is that there is something very close to the edge of just working or just not working. This might also be some kind of race condition in a digital signal.

[ManateeMafia]

3-4 years ago, one of the broken 3458A's I bought had a bad A3 board. Given the recent repairs and valuable info from TiN I thought I would give it another shot. All of the voltage checks were performed years ago, but I went ahead and did them again. This was a donor board as I had acquired a replacement from Agilent.

The board had a 202: Slave Test Convergence error. Based on TiN's recommendation, I went through the ZR_LO and ZR_HI inputs from the A9 reference board. Then I went through the +12ref, -12ref, and +5ref to confirm they were working.

I checked the Reset lines and clock to the 8051 (U220). That looked good.

TiN had mentioned the TARM command during his repair that I didn't previously try. Using the command, I watched the outputs of pin 6 (OUT) briefly change on U110,U111,U112, and U140. So far so good.
When I got to U142, I could see the negative transition on the input, but the output never changed from +4.2V. After testing it a couple of times, I decided to replace U142 and fired up the 3458A.

The meter came up and was reading DC. I ran the self-test and ACAL DC successfully.

The EL2018CN was sourced from UTSource .

Thank you TiN and all the others that have added to this thread. . Next step, repair defective A2 board with EL2039 from the same seller.

[TiN]

Great to read, that at least yours 3458A got progress Leave it running for a day or two, sampling 1Meg resistor, and perhaps it could become a lemon like mine   

202 Error I had was due missing jumper on header near reference on A1 board. So issues we have still are different.

UTSource replied today that they will ship EL2018's soon (still did not do it so far). I'll get LM6321 buffers by friday (maybe not an issue, but one on AC A2 board run's 50ish °C.)

P.S. got DS1245YP's for scope, hope will be back in business, so I can do proper measurements and debug.

Can I add your experience into article? Maybe with few photos, you can practice LED lights trick

[ManateeMafia]

Feel free to add it. I have more info on my A2 board, the el2039 'fixed' it, sort of. I have a missing C104 that was stolen for another board. I will find a sub and try to clean the board. The meter is reading 75V on highest range. Scal may or may not fix it. I will give it a shot once I can get a new cap. As long as I can reduce the error, I think the board will be a good enough backup. It was reading 4550 VAC before I changed the el2039.

I will see if I can capture some waveforms in trigger mode to simulate the pulses throught the chain of parts.

[TiN]

Yey, my scope is back in business, after replacement of pesky NVRAM chip.

Here are some initial shots

Setup is rather crude, I'll have to make few adapters to have proper probing.
U142 EL2018 comparator output to digital domain of ADC.

NPLC 1:





It does seem to be working, at least signal is there.



Above two blocks of pulses, first fast ramps and then slower ramps. This is when I get measurements before error happens. Sometimes there are only first half of pulses.
Square pulse train waveform is comparator's output.

A2 Board, U504 buffer (left input pin 2, right output pin 3), meter in mV range on ACV.



Looking OK.

Integrator capacitor waveform (test point TP112, left is NPLC 0.001 failing, right is NPLC 0.0001). You can see on faster setting it just takes 2 big ramps, while on slower setting it takes 12 smaller ramps.



Not sure if those little dips should be there...

[TiN]

Also on HP 3458A CLIP's schematics ADC comparator is drawn incorrectly! LM6361 U140 output going to noninverting pin 2, not 3! 

Pinout map of U180 ASIC:

[ManateeMafia]

I have added some scope screenshots of my recently repaired A3 for comparison to TiN's.

The NPLC .0001 doesn't look right, but that is probably an issue with my setup.

[TiN]

Thank you for measurements.

I got more parts, just to find out that Avago HFBR-2521Z is NOT pin compatible with old HP's HFBR-2501 
Got deceived by newer A3 03458-66513 board photos with HFBR-2521s 

[TiN]

More followup vid's. Layout boards for adapters.

[TiN]

Received EL2018's from UTSources. Even tho they were listed as "New" they are not, all pins tinned and little bent.
Chips also assorted, with different datecodes, from 1991 to 2000. Hope ones I pick would work and fix pesky error.
Better be, If not I'll use hammer for non-authorized express repair of evil thing... 

I wonder how's OldSchoolTechCorner doing? He's been suspiciously quiet for long time 

[crispy_tofu]

I wonder how's OldSchoolTechCorner doing? He's been suspiciously quiet for long time 

Same, waiting very impatiently for a repair! 

[Mickle T.]

"114 SYSTEM ERROR - balanced rundown convergence"
http://bbs.38hot.net/forum.php?mod=viewthread&tid=134921

[crispy_tofu]

I may be interpreting this wrong (lol Google Translate), but the forum post is saying to replace the AD637 (U601).

Never mind, unrelated.

[Macbeth]

They are for a fault in the AC section which TiN isn't having trouble with.

[TiN]

Unless fault on AC board cause somehow problems on DC. Time to try those EL2018's. 

[Dr. Frank]

Unless fault on AC board cause somehow problems on DC. Time to try those EL2018's. 

What's interesting about the failure in the AC path, it is not related to the A/D section, and especially not to the comparators.
The signal from the AC simply is not responding, and that causes this error message.

If you think this idea a bit further, maybe the DC path before the A/D input is broken or unstable somewhere ..
So, before changing the EL2018, which seem to be operable, as AC is digitized correctly, I would first give this DC path a try.

Frank

[TiN]

Well, seems something might be true in that. If I disconnect DC_AD from A1 board, I can run longer till error come, even like 5 minutes using NPLC 0.5.

Balanced error almost gone after replacement of all three EL2018 on A3. I'll more into A1 and A3. Looking on schematic, A3 also have DC function blocks, it's not AC only board there.

I was also scoping around and noted big spikes on DC_AD signal with repetive digitazing rate, like 30V peak peak, from negative to positive rail. I wonder if that normal. Tried autozero off and on, same result.

[ManateeMafia]

I found this in Chapter 4 of the service manual pg 72 (03458-90011)

"The AC circuitry is also used for track/hold measurements.  Although these
measurements are DC volts measurements, the AC circuitry is used due to
its higher bandwidth.  Track/hold requires a higher bandwidth than the DC
circuitry provides."

[Dr. Frank]

I was also scoping around and noted big spikes on DC_AD signal with repetive digitazing rate, like 30V peak peak, from negative to positive rail. I wonder if that normal. Tried autozero off and on, same result.

Strange thing. Sure that this doesn't come from outside? (Probably no)

Just an idea, about what has been changed essentially: Did you connect the shields of the transformer in the correct manner?
No ground loop from Dig to Ana PSU?

Btw.: Thank you very much for the schematics of the transformer!

Frank

[TiN]

Yea, tried either way. What puzzles me, that meter was working ok for a day before it failed...

[Kleinstein]

There should not be such large spikes. If at all something in the low mV range (so hardly visible on the scope) may be.

Teh delayed problem may be due to a hidden defect that inititially (e.g. cold) did no big harm but than did some damage somewhere else. Its well possible that such excessive spikes can damage some other part - up to the point of causing a latchup in some chips. So we likely look for more than one defect part. Those parts that where allready fixed (possibly by someone else before) are prime candidates to look for - not as the prime source of damage, but as a damaged part.

Finding the source of these spikes would be likely the first step.

[TiN]

Also I think it's worth to leave link to this patent, explaining exactly 3458A's track&hold circuitry. Question if it's being used for DC measurements is remain to be seen.

Brute force repair is not going to work this time, so need actually sit and learn how DC path in instrument working. To compare - i already spent more time on this Error 114 issue than whole rest repair project total (over 40 hours). Good example of how initial "easy" repair can go bad in no time.

Where is our second sunken 3458A owner. Reveal youself