Another 3458A on board

[PartialDischarge]

Last week I picked up an 3458A, it was a totally unexpected purchase. I was happy with my 3 6.5 digit dmms and positively decided some time ago not to go any further. But life has its surprises and when I saw it
at someones place I decided to make him an offer.

Well, the unit is old but like new, it seems that it was never used. The cal sticker on the side was still intact and I was the one to break it. Here's the unit and the inside:











NVRAMS are dated '98 and amazingly the are still working. Anyway they are already on sockets and backed up.

However the unit has a problem.... more on next post...

 

[PartialDischarge]

And the last two pics...






So, as I was saying when the unit boots up it does the RAM test, then it goes into "HARDWARE TESTING", then waits a few seconds, a beep goes on, and there it stays for ever...



The unit has 2 CPUs, one for the front panel and display and the other is the main control CPU. At first I though that one of them was frozen but it gets even more interesting to
test that both are working and responsive!

The front panel CPU has a test pin (17), that when pulled down displays test characters on-screen:



Hmm, there is activity in the main CPU too, and it seems that something is withholding the unit to continue its normal operation. Indeed, there is a watchdog timer
that generates and interrupt when it is not being taken care of properly, and this is the case after the unit starts beeping, an IRQ is generated and it is not attended:



Moreover and to my surprise, during all this time the crossguard controller of the A5 main CPU is still receiving and transmitting data (commanded by the MC68000 CPU) over the fiber optics in to the isolator board:



At this point one may be tempted to think that the CPU has gone into some kind of wild uncontrollable loop, but no  .
I decided to create a fault by removing the transmitter fiber optic that connects the outguard power supply A6 and the A/D converter and logic assembly. What the A5 CPU does puzzles me,
it turns off the beeper (on at all times up to now) and displays "ISOLATOR FAILURE" which is actually what it should do.

So to recap.

-NVRAMs passes the test, so the data buses and address lines are fine, at least address lines A1..A16 which are the ones needed to access those memories
-Both CPUs at A5 and A6 are working and responsive
-Timer seems ok
-Crossguard controller seems ok
-Front panel controller seems ok

I will continue the tests as time permits. Any idea or suggestion will be appreciated

[Kleinstein]

The meter has even 3 CPUs: there is one on the input referenced part (ADC board) too.

For a test one could interrupt the optical connection to the analog part - this might give an error message earlier in the HW test procedure and thus the option to do a few more tests with the other two CPUs.
The message slightly point towards a problem with the ground referenced CPU (68K) board - testing the supplies would be a first point. Maybe also check soldering on the sockets - AFAIK the 3458 does not come with sockets for the RAMs. So someone already added those and this could be a source of error (e.g. short, cold solder joint).

If looks very clean, but is still old. So more like old but not used very much.

[CalMachine]

Very interesting issues to be having...   I will definitely be following this thread  Looks like could be some very good troubleshooting info. 

Did you know there were issues with the unit when you purchased it?

[PartialDischarge]

The meter has even 3 CPUs: there is one on the input referenced part (ADC board) too.

For a test one could interrupt the optical connection to the analog part - this might give an error message earlier in the HW test procedure and thus the option to do a few more tests with the other two CPUs.
The message slightly point towards a problem with the ground referenced CPU (68K) board - testing the supplies would be a first point. Maybe also check soldering on the sockets - AFAIK the 3458 does not come with sockets for the RAMs. So someone already added those and this could be a source of error (e.g. short, cold solder joint).

If looks very clean, but is still old. So more like old but not used very much.

I was updating the thread as I was writing. I already interrupted the optic fiber link. The socket on the NVRAMS were added by me, the other empty ones I believe are optional.
It looks 99% like its an A5 board fault and yes, this unit has been unused in a shelf probably for 10+ years, corrected that. Also I added in the OP the fact that the sticker cal was intact so I know for a fact that no one before me
fiddled with the unit.

[PartialDischarge]

Thanks CalMachine, yes I did know about this issue, and for the price paid it was a better don't ask questions kind of situation.

Forgot to mention, I've seen some weird stuff like a few 4V logic signals instead of 5V, and some rare logic outputs... will have to investigate deeper

[Dr. Frank]

Congratulations and welcome in the 3458A club!

The unit is about 19 years old, and so is the CAL RAM, DS1220Y.

You urgently need to save the calibration constants, and replace all nvRAMs.

Frank

[rastro]

Is there any flow chart available for the power up or HW test sequence?

[TiN]

I'd start from refreshing NVRAMs too. I had similar issue in my Keithley 2001 when fake ebay  NVRAM died (MEM2 option). Meter could work for hour or a day, and then die with digital brains becoming completely unoperational and VFD controller just showing "No comms" as it does not get any data from main CPU. Turned out dodgy NVRAM was latching up some data lines, and since SRAM,firmware ROMs are on same databus -> meter go kaput. Replacement of NVRAM fixed the issue completely 

Also while I agree on needs of saving/transfering calibration data, if you have access to 10V/10Kohm standards preserving 19 year old calibration have no much means to your meter (if we talking about getting actual specified performance out of the box) 

[PartialDischarge]

I'm focusing my attention to the A3 board now after seeing that A5 and A6 boards are fine.

I did remove the cal NVRAM and turned on the meter, it shows the msg "switch jumper j132 and power cycle". After I put back the memory, power on again, it now displays "ACAL SCAL CALL ALL required, secure required", so it knows that I played with the memory and it must have stored that change in the other nvrams.

Will have to wait for the new nvrams to arrive from mouser.

[grumpydoc]

Two other 3458A listings on eBAY:

http://www.ebay.com/itm/Hewlett-Packard-3458A-Multimeter-Option-002-Parts-Repair-As-is-/272579790187?hash=item3f7705856b:g:T5EAAOSwXYtYvbh2
http://www.ebay.com/itm/Hewlett-Packard-Multimeter-3458A-Parts-Repair-As-is-Option-001-/351998675651?hash=item51f4c18ac3:g:U10AAOSw32lYvbiH

For sure one should ask the seller more technical questions if all boards are still inside and authentic to this unit (not swapped).
In the end there always remains a risk on the buyer's sholders ...

Even from what you can see, one has probably had the NVRAM batteries fail and lost it's cal data (the one with error 110) and the other, well, it could just be a missing jumper but who knows?

How much does a full re-cal on a 3458A cost?

[TheSteve]

Cheapest Keysight Cal starts at $600 USD.

[rastro]

Two other 3458A listings on eBAY:

http://www.ebay.com/itm/Hewlett-Packard-3458A-Multimeter-Option-002-Parts-Repair-As-is-/272579790187?hash=item3f7705856b:g:T5EAAOSwXYtYvbh2
http://www.ebay.com/itm/Hewlett-Packard-Multimeter-3458A-Parts-Repair-As-is-Option-001-/351998675651?hash=item51f4c18ac3:g:U10AAOSw32lYvbiH

For sure one should ask the seller more technical questions if all boards are still inside and authentic to this unit (not swapped).
In the end there always remains a risk on the buyer's sholders ...

Even from what you can see, one has probably had the NVRAM batteries fail and lost it's cal data (the one with error 110) and the other, well, it could just be a missing jumper but who knows?

How much does a full re-cal on a 3458A cost?

The unit with the 110 error also says that most the keyboard was inoperable.

[TiN]

Somebody is in for 1300$ AD board replacement for 002 box  .

[mimmus78]

Somebody is in for 1300$ AD board replacement for 002 box  .

[HighVoltage]

This is one of the best looking "older" 3458A I have seen, Congratulations.
It will be interesting to follow your repair

[TiN]

Who? I didn't buy it 6221 is enough for me 

[mimmus78]

Who? I didn't buy it 6221 is enough for me 

Why every-time a beaten up 3458a is unlisted from ebay we think at you ... 

[PartialDischarge]

A little update on my findings,

After a while of testing and probing more or less everything in the A5 board I started looking at A3. It seems that the main uP in A5 is trying to start the A3 board, and the thing is A3 is not dead and does indeed reply back via the fiber optic, but for whatever reason normal operation does not continue. Back to square 1 again 

A small detail made me focus my attention on A5 again: With the original NVRAMS in place, the equipment boots, checks the RAMs, goes into TESTING HARDWARE and stays there forever.
However with the new NVRAMS, the equipment shows all the ACAL SCAL errors, goes into TESTING HARDWARE and after a few seconds it reboots itself. This reboot is caused by the timer throwing an interrupt, interrupt that is ignored with the old NVRAMS. This led me to believe that there is some kind of weird interaction in the address/data bus lines. Maybe a register is not meeting hold/setup times or maybe a logic gate has a rare problem.

So I decided to replace 29 logic ICs in the board, all of them account for around 20€ so its mostly a matter of time,



While I was at it and with almost every chip removed I also inspected all the traces, the resistors and the capacitors. After replacing them all results are still the same 
Next I did extract the EEPROM, read its contents in a binary file and run a unix 'diff' against the version v8 posted at xdevs. They match and next week I intend to burn another EPROM with version v9 to see what happens.

So, it really annoys that I don't even know what's wrong with the equipment as there is not a clear fault. I already see me selling this meter 

[TheSteve]

Don't give up yet!
Is there anyone close to you that could test out your A5 board?

[TiN]

I can test your A5 if you want to ship it over. It sounds like there is power (caps on A6?) or digital problem.
Also there is A5 for sale on ebay now, for 450$ auction. About similar I paid for mine for second meter, 400$.

Only interface digital board talks to inguard guts is A3, so if 8051 or gate array logic there got sick, that could be a reason, but that's a long stretch. I have dead A3 which in theory could be donored for parts though  .

Before swapping gazillion of chips, I'd check resistance on power, sometimes those yellow MLCCs go bad.

[PartialDischarge]

To better understand the quandary that I find myself into with this repair, I'll post some scope screenshots of the Tx and Rx lines going out of the A6 board, a picture is worth a thousand words.

After power up, A6 board checks the RAM, GPIB... and then starts communicating with A3. Every 1.25s there is something sent and something received. After 5 tries, the 5th transmission is different and nothing is received, a timer goes out and the equipment resets itself:



And here is a closeup of frames #1 to #5, it seems that there is an common opening command (yellow Tx, blue Rx):












So you see that whatever the fault of this equipment it is something deep inside its inner workings. If anyone could probe these signals it would be great to compare, they are taken here

[TiN]

Since seeing dead 3458A feels like a crime for me, hereby I'm sending MasterTech 8051 MCU from my first convergence failured (bad U180, I don't even need to look that up, still giving shivers) from A3. Also included few more bonuses in bag  .

[PartialDischarge]

Since seeing dead 3458A feels like a crime for me, hereby I'm sending MasterTech 8051 MCU from my first convergence failured (bad U180, I don't even need to look that up, still giving shivers) from A3. Also included few more bonuses in bag  .

Thank you Tin  , as we talked, at this point all my suspicions lead to that 8051. Hope to,test it soon

[plesa]

Another 3458A with 002 available
http://www.ebay.com/itm/hp-3458a-multimeter-option-002/282445864081

[chris_11]

Helll,

hope that I can hijack this thread. otherwise I will open a new one.
Got an used HP3458A yesterday. Serial is 2832A04913. CAL all and ACAL all work. However the strange thing is, that the DCV reads about 0.15% too high. Which is way out of spec. With the assumption that somebody had played with the CAL I did put in a CAL value with some agreement with my Keysight 34461.
Measuring a 9V battery has even on NPL10 only the last digit move by one digit. You can literally see the battery discharge over the combined input resistances. With a ACAL All now the DCA range looks reasonable. However I have no experience if the reference or any other thing is still faulty. So any advice will help.
Before I take it to the cal lab I like to make sure that nothing is broken. For that I am willing to build some LTZ1000 boxes if someone can direct me, where to buy the precision wire wound resistors needed.

Thanks in advance
Christian

[plesa]

Helll,

hope that I can hijack this thread. otherwise I will open a new one.
Got an used HP3458A yesterday. Serial is 2832A04913. CAL all and ACAL all work. However the strange thing is, that the DCV reads about 0.15% too high. Which is way out of spec. With the assumption that somebody had played with the CAL I did put in a CAL value with some agreement with my Keysight 34461.
Measuring a 9V battery has even on NPL10 only the last digit move by one digit. You can literally see the battery discharge over the combined input resistances. With a ACAL All now the DCA range looks reasonable. However I have no experience if the reference or any other thing is still faulty. So any advice will help.
Before I take it to the cal lab I like to make sure that nothing is broken. For that I am willing to build some LTZ1000 boxes if someone can direct me, where to buy the precision wire wound resistors needed.

Thanks in advance
Christian

What is last calibration date? What is self test result? Try to log the cal values over time
https://xdevs.com/doc/HP_Agilent_Keysight/3458A/service/3458A-18A.pdf
If you have GPIB adapter try to log the cal values by Python script
https://www.eevblog.com/forum/metrology/raspberry-pi23-logging-platform-for-voltnuts

[chris_11]

Last cal was by Fluke in 2012. However the cal seal on the side screws are broken so anything could have been swapped. Self test does pass. Have it now on for the next days and will Autocal it at least daily to get drift data. The VFD seem to be quite (too) good for such an old instrument. There is an older Agilent cal sticker from 2009 on the rear.

Christian

[TiN]

I agree on plesa's sugesstion. Test per SN18 procedure for stability , and also run meter at NPLC100 24/7 (DCV 10V range) to make sure ADC works well. Some more detail and example test result can be find here.

[TheSteve]

I would also open it up and check the date on the nvrams. If they are old you'd certainly want to replace them before having it calibrated.

[TiN]

Good idea too, I didn't mention NVRAMs because that's obvious and has to be done regardless of the chip dates.
4913 S/N unit is around 1990-1991, so it's already ancient, however I'd expect Fluke replace NVRAMs if it was serviced by them in 2012.

[chris_11]

Hello,

today I opened up my Metuselah. Those unit was last calibrated by Agilent in April 2011 and  later Fluke in March 2012. Self test passes as well as ACAL all.
CAL?72 over two days was 0.25ppm max between three ACAL ALL.
Reference board is the original with LT1013 date code 8851 which is week 51 in 1988, so that matched with the vintage of the serial. The  topside analog board where the reference is plugged on reads mostly 1990 date codes, so that will be the original too. The processor board with the Dallas NV RAMS seems to be a bit younger It still Rev A with  09-9217 USA Made and has two Dallas NVRAMs date code 0640C   below the EPROMS but has still one Dallas NVRam date code 9814D.
So the 9814D seems to be the most scary one which is the closest to the rear side.

The calibration seals on the handle screw from Agilent and Fluke were broken so there could be some change after 2012.

How can I upload pictures I have some from the unit?

Christian

[chris_11]

Here some pictures:

[Dr. Frank]

Hello,

..

CAL?72 over two days was 0.25ppm max between three ACAL ALL.
..
The processor board with the Dallas NV RAMS seems to be a bit younger It still Rev A with  09-9217 USA Made and has two Dallas NVRAMs date code 0640C   below the EPROMS but has still one Dallas NVRam date code 9814D.
So the 9814D seems to be the most scary one which is the closest to the rear side.

The calibration seals on the handle screw from Agilent and Fluke were broken so there could be some change after 2012.

How can I upload pictures I have some from the unit?

Christian

Hello Christian,

the oldest nvRAM, date code from 1998, labeled DS1220, is the calibration RAM, and really is the crucial one.
You should at once read its content into a file on your PC, and later program all these calibration constants in a replacement chip.

The other two nvRAMs DS1230 are not critical, but should be replaced also, as they are already 11 years old.

Your A/D hybrid seems to be healthy.


Frank

[rastro]

Christian,

The instrument looks in excellent shape physically.  Great find!   

[chris_11]

The NVRAMs and Eproms are now new. The Eproms used are Atmel AT27C512R-45PU  (45ns) which seem to run fine.
What is a bit strange is that the readable string in the CAL RAM around 000005D0 reads like  2823A08118 which would be a later unit than my serial 04913. My assumption is, it is strange that Agilent in 2009 and Fluke in 2012 calibrated this unit with a 9814 CAL-RAM which the unit had when I bought it. That is an explanation too why the calibration was way off.

Christian