Repair worklog : Old rusty HP 3458A

[TiN]

Captured bunch of diagrams on A1 board.
http://doc.xdevs.com/doc/HP_Agilent_Keysight/3458A/scope_data/

U107 2.5V REF output is somewhat -16.11V. Based on schematic, I don't see how that is correct.
Perhaps I can get reference measurement to check on this one..

[ManateeMafia]

I can check that in a couple hours.

[Dr. Frank]

Captured bunch of diagrams on A1 board.
http://doc.xdevs.com/doc/HP_Agilent_Keysight/3458A/scope_data/

U107 2.5V REF output is somewhat -16.11V. Based on schematic, I don't see how that is correct.
Perhaps I can get reference measurement to check on this one..

Without knowing that chip, I assume that this circuit part is referenced to -18V, not to GND:

That seem to create a CC for the input amplifier, i.e. constant 2.5V over R103 => 1.4mA

You should measure constant 2.5V between -18V supply (pin3 of U107) and output, pin2.

Or your -18V should read about -18.6V

Frank

[TiN]

Doh, sure you right. Should have noticed minus sign on voltage rail there. No problem there.

[TiN]

Just a heads up. No progress was archieved so far, was back to A3 A/D board replacing more parts without any improvement:

So far replaced U142,U181,U170(to LT1013), U230, U303, U304, U110, U112, U150, U400, Q403, CR405, CR406, CR111, added OSCON capacitors around C122,C120,C181 for sake of trying, no any sign of improvement, it still does fail with Error 144 - multislope rundown convergence or Error 114 - balanced rundown convergence.

There are not much components left to try, and it could be suspected unobtanium suspect ASIC U180, which handles all charge mangment around integration and switching.

I'm pretty sure it's not A2 or A1 board, as I had excluded them from analog DC path and still was getting same issue (used my LTZ1000 reference module externally, disconnected AC_AD and shorted DC_AD to ground on A3 PCB).

It's possible we have to go plan B in this case... 

[Dr. Frank]

Just a heads up. No progress was archieved so far, was back to A3 A/D board replacing more parts without any improvement:

So far replaced U142,U181,U170(to LT1013), U230, U303, U304, U110, U112, U150, U400, Q403, CR405, CR406, CR111, added OSCON capacitors around C122,C120,C181 for sake of trying, no any sign of improvement, it still does fail with Error 144 - multislope rundown convergence or Error 114 - balanced rundown convergence.

There are not much components left to try, and it could be suspected unobtanium suspect ASIC U180, which handles all charge mangment around integration and switching.

I'm pretty sure it's not A2 or A1 board, as I had excluded them from analog DC path and still was getting same issue (used my LTZ1000 reference module externally, disconnected AC_AD and shorted DC_AD to ground on A3 PCB).

It's possible we have to go plan B in this case... 

Ok, before you replace further components, especially U180, maybe it's possible to test a bit further.

I was under the impression, that the instrument maybe stuck itself into this failure mode, by improper ACAL (?) calibration constants.

Maybe you bring the instrument back to some initial state first.


I assume, that you stored the content of the CAL nvRAM (2kx8).
And you may also have the old DALLAS parts, which you may erase to FF in all cells on your ext. programmer, and then insert into the instrument.
(Store the new nvRAM in a safe place)

Also, please remove these 4x 32kx8 additional RAM, U123-U126, as you used nvRAMs also, which may  confuse the instrument on booting.
Maybe, you might also erase the other both 32kx8 nvRAMs to delete all possible stored states.

If you then start the instrument, you will have a virgin state, causing many CAL and ACAL errors.
The FW will initialize all calibration constant to a default, and an initial ACAL will be necessary.
On 2nd boot up, there should be no error any more, provided, the old nvRAMs still work.
(Otherwise you'd have to use the new nvRAMs for that)

Maybe, this may cure the problem.



In case there's no success, you might first measure precisely the +/-12V and +5V ADC references, TP160, 165, 151, for correct ratio (1.69x) relative to LTZ value, and for stability.
(If there's a noteworthy deviation, that may also cause your error).

I'm not sure, if the Time Interpolator is used for normal DCV operation (to my opinion, only in Digitizing Mode, like ACV Sub- and Random Sampled)
But who knows.. Inside there's another EL2018, U405, maybe you check this IC also.

These convergence errors have not yet been understood well in the past (for other instruments).. but this might have to do with the matching of the different slope resistors , as described in hph4/1989, page 12. It's necessary, that the absolute values of these internal resistors fit each other, not extremely precisely, 0.1% only.
They cannot be accessed externally, that's the problem.. But maybe you try carefully to measure the 50k between pin 6 and pin 11/13 of U180, and maybe the minor slope resistors, pins 18,19,21,23.

And you also did not leave any fingerprints there, I assume, which may cause parasitic currents at the A/D.

Give all that a try before giving up ..

Good luck

Frank

[TiN]

I was under the impression, that the instrument maybe stuck itself into this failure mode, by improper ACAL (?) calibration constants.

Does not look like that. It did run SELFTEST fine about 5 times I did, then passed complete ACAL on all functions, then I did CAL against 10V from K2400 SMU and against 10K resistor from ESI DB52.
Turn meter off, turn meter on, measured 10V from K2400, values matches, all okie, run SELFTEST couple more times, turn off unit.

Next day - power it back on, measured 10V, all ok, connected 1Meg VPG resistor and leave it sampling NPLC 10, all settings default. Was fine for at least 30 minutes, then I left room, come back in 5 hours - measurements halted, ERR message 114. Since then it's all the same, throwing these two errors on anything >100us conversion time on all ranges/modes in DCV,DCI,Ohms, no matter what other settings I fiddle with. Sometimes can do 3-5 samples at NPLC 10, sometimes cannot do even 1 count on NPLC 0.01. Usually locks up less the a second.

I assume, that you stored the content of the CAL nvRAM (2kx8).
And you may also have the old DALLAS parts, which you may erase to FF in all cells on your ext. programmer, and then insert into the instrument.
(Store the new nvRAM in a safe place)

Tried all that, no effect. New NVRAMs, old NVRAMs, erased, or in original state, all same. Just tried once again, same. I'd assume digital outguard domain working OK at this point.

Also, please remove these 4x 32kx8 additional RAM, U123-U126, as you used nvRAMs also, which may  confuse the instrument on booting.
Maybe, you might also erase the other both 32kx8 nvRAMs to delete all possible stored states.

No effect, just did this, and tried this week ago. Regarding CAL ROM, I tried that (just poped in old NVRAM chip, which still have original data), no difference.

If you then start the instrument, you will have a virgin state, causing many CAL and ACAL errors. The FW will initialize all calibration constant to a default, and an initial ACAL will be necessary.
On 2nd boot up, there should be no error any more, provided, the old nvRAMs still work. (Otherwise you'd have to use the new nvRAMs for that)

Old SRAMs dead, but as all above, I don't think there is a problem. Yea, I even tried to run without CAL NVRAM, believe me. On this matter meter tells you to change jumper JP132 and power cycle .

Maybe, this may cure the problem.

Nope, no magic pill on A5 board .

Even removed all ROM chips and checked them to make sure all data verified 100% match to firmware binaries. (I had unit running open cover those two days when it covered, was thinking maybe bit or two got corrupted from sunlight UV. Nope all were 100% match.)

In case there's no success, you might first measure precisely the +/-12V and +5V ADC references, TP160, 165, 151, for correct ratio (1.69x) relative to LTZ value, and for stability.
(If there's a noteworthy deviation, that may also cause your error).

Now we have something.

ZR_HI to ZR_LO (entrance of zener voltage from A1, at A3 connector): +7.07411, stable up to 5th digit.
ZR_LO to TP160 (+12VREF) : +11.93936 V, stable to 5th digit. Ratio to LTZ is 1.687754
ZR_LO to TP165 (-12VREF) : -11.92274 V, stable to 5th digit. Ratio to LTZ is -1.685404
ZR_LO to TP151 (+5VREF)  : +4.960405 V, stable to 5th digit. Ratio to LTZ is 0.701205

Also probed with scope, did not see any spurious large-scale noise.

I'm not sure, if the Time Interpolator is used for normal DCV operation (to my opinion, only in Digitizing Mode, like ACV Sub- and Random Sampled)
But who knows.. Inside there's another EL2018, U405, maybe you check this IC also.

Replaced U405 as well, forgot to mention, sorry. I have now both U142 and U405 with collet sockets, tried bunch EL2018, none show any difference.

These convergence errors have not yet been understood well in the past (for other instruments).. but this might have to do with the matching of the different slope resistors , as described in hph4/1989, page 12. It's necessary, that the absolute values of these internal resistors fit each other, not extremely precisely, 0.1% only.
They cannot be accessed externally, that's the problem.. But maybe you try carefully to measure the 50k between pin 6 and pin 11/13 of U180, and maybe the minor slope resistors, pins 18,19,21,23.

50K network read as 50.2095 kOhm. Added pair of 22M RN7C resistors between pin 6 and pin 28 - got 49.982K, still same behaviour.

I tried to measure minor resistors, got wrong values, but that's likely to opamps in loop. Only way is to desolder U180 and measure separately.

And you also did not leave any fingerprints there, I assume, which may cause parasitic currents at the A/D.

Yep, leakage are bad in low-current charge balanced circuit, but cleaned multiple times, IPA, distilled medical grade water, IPA again, no any visible difference.

Give all that a try before giving up ..

Nobody is giving up so easily, just more radical, less enjoyable approach would be taken

I also reflowed pins on U180, no effect (just heated up joints to solder point, without removing chip)

I also tried voltage margining on all supply rails (+5, +18, -18) and voltage references (-12, +12, +5) one at a time, injecting little current with SMU into feedback nets, did not see difference, so it's not like something is bit off tolerance, talking about voltage domain. Problem likely lies in current domain or in timing/sequencing.

[Macbeth]

..erm... did you try?? Give suspects like U180 a blast then switch on and see how long before errors. 

( ...goes and hides from TiN's ire, frustration and rage! ) 

[TiN]

I cooled it (U180) to 0°C with 40x40mm TEC while running, nope

[Dr. Frank]

Another odd idea, analogously to the AC converter error in the bbs38hot thread:

Have you ever re-aligned the jumper P100, located near the LTZ assembly,which initially was missing? (see picture)
What happens, if you set this to OV, 7V, back to normal?

Is the signal out of U100 stable?

(I know, that you already fed DC signal directly into DC_AD, I assume)

Frank

[Macbeth]

I cooled it (U180) to 0°C with 40x40mm TEC while running, nope

Oooh.. Get you! Even your thermal diagnostic tricks are an another level! 

I tell you what, I wish OldSchool' didn't have to flog his old Riglent gear to try and raise funds for his 3458A restoration. He could be joining in right now.

Hmm... Seeing as he's busy with all that legal shenanigans, maybe he would be a gentleman and lend you his ADC board for comparison purposes? 

[TiN]

Macbeth
So we end up having not one, but two meters broken? Nah
Feel bad for OldSchool, that's nasty bork on his ebay acc. 

Dr.Frank

Same stuff, meter just reads near zero if P100 set to GND or 7V, but conversions are still locking up with E114.
That's why I'm not taking A1 into account.

I might have another piece of gear to come (not 3458A, but HP-made too), so feel like kinda failed on this project.
I spent less than 30 hours on all repairs total, and wasted already over 60 hours on this A/D convergence stuff  [emoji14]alm:. Eek.. 

[MadTux]

Have you considered cleaning the boards?

That faults seems so typical for electronics with water damage. Everything seems nice and dry and the instrument is powered on. But there is still some salts or water in some hidden spot. This water is then electrolysed into hydrogen and oxygen, that causes corrosion damage. Salt residues might be hygroscopic and draw water from atmosphere and also cause increased leakage. Then after a while the instrument fails without clear evidence of a failed component (shorted tantalum capacitor, dried electrolytic or obvious failed components somewhere like a bad transistor or diode that shifts all voltages related to it or shorts something out) because of leakage or corrosion damage somewhere.

I would guess that the previous owner(s) quickly flushed the boards with (deionised) water to prevent most of the damage that the frame received from long time water exposure but didn't dry it well afterwards. Then when the faults shown on ebay became evident, they trashed it.

 So before declaring the project as failed, I would consider cleaning all critical boards with normal water/soap, blow dry with compressed air from compressor (something like that: http://www.honestraders.com/images/air-compressor/2.jpg compressed air cans are crap), flush with deionised water, dry again and then optionally flush again with isopropyl alcohol. I cleaned lots of boards that way, even dishwashered a dirty 3456A analog board and they all work fine so far.

[Macbeth]

Have you considered cleaning the boards?

I do believe TiN has a plan B. I also believe (considering the lengths he has gone so far and his amazing efforts to reverse engineer the 2001 PCB) that this will involve not just that but also desoldering every damn component on the board and testing out of circuit! Then scrubbing up the PCB. In fact - why not make a whole new one? Only double sided...  Keep some in stock to sell on... 

[Mickle T.]

 

[Theboel]

@ Mickle T, You are mad 

[crispy_tofu]

T...th...that's not soapy water, is it? 

[TheSteve]

Several 3458A's were just listed on ebay - several with errors. Anyone else want to give a repair effort a go? The one for 2199 with error 209 might be a decent option.

[OldSchoolTechCorner]

I cooled it (U180) to 0°C with 40x40mm TEC while running, nope

Oooh.. Get you! Even your thermal diagnostic tricks are an another level! 

I tell you what, I wish OldSchool' didn't have to flog his old Riglent gear to try and raise funds for his 3458A restoration. He could be joining in right now.

Hmm... Seeing as he's busy with all that legal shenanigans, maybe he would be a gentleman and lend you his ADC board for comparison purposes? 

I ended up upgrading to new much faster mixed signal scope, plus a better signal generator and reason sold the Riglent gear, which I would have never expect what they did, at least they made it right, just hope they won't do it again.     

Could lend you the ADC board if you need it for the time being, since I have to finish up other projects I have on the bench currently before starting the 3458A restoration.

[TiN]

No need man, glad you got upgrade and sorted out selling adventures. I have something coming already, to fix my 3548.

I also got another piece of HP gear , which worth own restoration thread.

[OldSchoolTechCorner]

No need man, glad you got upgrade and sorted out selling adventures. I have something coming already, to fix my 3548.

I also got another piece of HP gear , which worth own restoration thread.

Wonder what the next piece of HP gear will be?

[helius]

Thanks TiN for the repair log, very interesting and helpful.

Oh-ho-ho. Glad I removed them. And glad HP 3458A using only 2-layer PCBs, so I'll just clean it to shiny copper, add thermal-resistant mylar cover around pins , just in case.

How do you decide when the PCB needs to be scrubbed down to copper, and how wide to go?
I have some 4-layer boards with NiCad leak damage, in some cases there is visible darkening around the PTH that can be seen through the board. If there are no traces in the darkened area, would it work to simply mill it out?

[krivx]

Thanks TiN for the repair log, very interesting and helpful.

Oh-ho-ho. Glad I removed them. And glad HP 3458A using only 2-layer PCBs, so I'll just clean it to shiny copper, add thermal-resistant mylar cover around pins , just in case.

How do you decide when the PCB needs to be scrubbed down to copper, and how wide to go?
I have some 4-layer boards with NiCad leak damage, in some cases there is visible darkening around the PTH that can be seen through the board. If there are no traces in the darkened area, would it work to simply mill it out?

I would mask out the area and go to town with fiberglass brush. See if you can clean it once the soldermask is removed.

[TiN]

helius

Well, think about end result you want to get. Task is to try remove all electrolyte from PCB surface, so it would not cause more damage than it already did.
In this case it was simple as PCB is only dual layer, and pin hole is really big and sturdy, not like those tiny 20 mil vias on 8-layer PCBs.

Removing solder mask to good unaffected zone is just a measure to guarantee that electrolyte soaked in between copper and mask effectively removed and cleaned off.

If you have multilayer via oxidized, I usually clean everything on top and bottom surfaces, get via pad surface to shiny tin/copper and then clean it's drill hole with thin strand of wire and solder in jumpwire inside affected via, so I would not have to worry about bad connection (in case of internal via's barrel got eaten away). It takes some time to do, but it's the way to save yourself from debugging random rare intermittent issues.

[helius]

then clean it's drill hole with thin strand of wire and solder in jumpwire inside affected via, so I would not have to worry about bad connection (in case of internal via's barrel got eaten away).

Thanks. If I understand you correctly, you mean use a patch wire inside the hole to connect the component lead to its net, because the plating could be bad. But if the PTH has an internal layer connection that has corroded, you need to find another point to connect the wire to?