HP 3458A repair's (now there are two of them)

[TheSteve]

Picked up an HP 3458A earlier this week off the 'bay. It looked pretty good in the pics and had the well known "RAM TEST 1 LOW" error - no other testing was done(supposedly). The price was right so I couldn't resist(I blame TiN, he is our 3458A hero and source of constant inspiration).

I received the 3458A today, jumped right in and powered it up, when the NVRAM error was bypassed it passed the self test, yay!

So I opened it up and removed the A5 (digital) board. I desoldered the two DS1235YW's and the DS1220Y NVRAM's. I tried solder wick first but didn't have too much luck with it so I switched to an ancient Ungar HOT VAC. It worked great and removed the solder with no damage to the PCB. The main memory NVRAM's were totally dead however the DS1220Y with the CAL info was still working(another YAY). I soldered in some good quality sockets and replaced the NVRAM's courtesy of Maxim(ordered them a few days ago).
It now powers with no errors at all(my third YAY moment). This is also another win for the TL866 programmer - it has paid for itself several times over already.

So dare I say at this point there isn't too much "repair" to do. I doubt I am lucky enough that the NVRAM's are the only fault, this is a pretty early 3458A from the early 90's(serial 2823A 05xxx) and is covered by an HP service note that suggests replacing the A3 board because of the dreaded "ERRSTR 114 SYSTEM ERROR — multislope rundown convergence" error that has no cheap fix.

I won't go into huge detail into the 3458A's workings etc. This has been covered in amazing detail by TiN(and others) in various threads.
My unit does have REV 6 firmware, I will upgrade it to 9.2 at some point. It also has option 001(expanded memory) which seems to have been added later based on date codes.
I need to go through the various service notes to see what other upgrades/changes need to be made.

For now it is assembled again, received a quick cleaning to remove stickers and is warming up so I can run an ACAL.

My biggest challenge will be determining if it is drifty or good. I have no good reference on hand. Anyone in the Vancouver, Canada area have something suitable that wants to meet up? Or anyone in Canada that has something I can borrow, buy/rent etc?

If I can verify it is working properly after some months of constant operation to stabilize the reference it will go to Keysight Canada for calibration.

Here are some pics of the unit, they are not of TiN's quality and were taken with my cellphone. I will of course add to this thread as I make changes/updates and test the unit.

[TheSteve]

A few more pics:

[Vgkid]

Nice, so far a simple fix...
 

[TheSteve]

Nice, so far a simple fix...
 

Indeed, ACAL passes fine and the voltage at least matches my calibrated 34461A.

[TiN]

Glad to be of help to get rid of those useless color paper bills you had  .

I'd also replace electrolyte and X/Y-caps on A6 and get rid of that Shaffner-paffner. 
You can bodge up simple LM399 circuit first and monitor in parallel with your other meter.

Time to get GPIB and Raspberry Pi handy and prepare more papers for DC reference standard and 10K resistance standard. I think you can get both under price of one official calibration+shipping. If someone step up and measure your standards, you can get nice calibrations yourself.

[Kleinstein]

To check if there is a  drift caused by U180 like in TiN's unit, there is no absolute need for a super stable reference. A simple 9 V block would be good enough - such drift should be well visible in jumps, when internal recalibration is done. So the ref has to be stable only for the short time of the readjustment, not over many hours.

[vtp]

Congratulations on your new toy.

My biggest challenge will be determining if it is drifty or good. I have no good reference on hand.

You do not need an external reference to check U180 drift. Just keep terminals for example shorted (measuring zero - or anything for that matter) for about one week or so and do daily ACAL DCV before taking a CAL 72 reading. If U180 is drifting you will notice it in few days. U180 is compared against LTZ1000 in ACAL DCV - iow the meter assumes the reference is stable and calibrates the rest against it.

If you are concerned about the LTZ reference drift then there is need for an external stable reference to compare against. However, I would first assume the LTZ is ok and check for U180 drift.

[Dr. Frank]

To check if there is a  drift caused by U180 like in TiN's unit, there is no absolute need for a super stable reference. A simple 9 V block would be good enough - such drift should be well visible in jumps, when internal recalibration is done. So the ref has to be stable only for the short time of the readjustment, not over many hours.

Kleinstein, that's utter nonsense.

For checking the short -term A3 drift, caused by U180, you only need a temperature stable environment, but no external reference at all.
The CAL? 72 parameter is re-calculated by ACAL DCV, and monitoring this parameter therefore checks the internal 10V gain stability, which very probably also depends on the interior temperature (~0.5ppm/K). Therefore, CAL? 175, the ACAL DCV temperature, should be monitored as well, to differentiate possible temperature shifts.


To check the stability of the internal LTZ1000 reference, you need a very stable external reference, like a 732A/B, or a known good external LTZ1000.

If this 3458A is an elder instrument, it's very probably not affected by the long-term LTZ1000 drift, as described in AN-18, as this affected certain later S/N only.
As TiN found out, older units may be as well affected by the U180 defect.

The Steve, as you are now also joining the volt-nuts / 3458A club, I recommend anyhow to build your own volt and resistance references, at least for verification of the 3458A, but also as true standards, which the 3458A itself isn't.

Frank

[BFX]

Lucky man so far 
I'm waiting for my chance
Congratulation 

[TheSteve]

Thanks for the replies guys.

I will be performing a daily ACAL each evening and then recording the ACAL 72 value and the temperature, I will edit this post with the values.

Day 0 - Dec 17, 20:00 - 989.771605E-03 Temp 32.1C (4 hours after being powered on after who knows how long)
Day 1 - Dec 18, 20:00 - 989.771293E-03 Temp 32.1C
Day 2 - Dec 19, 20:00 - 989.771416E-03 Temp 32.8C
Day 3 - Dec 20, 20:00 - 989.771292E-03 Temp 32.5C
Day 4 - Dec 21, 20:00 - 989.771245E-03 Temp 31.8C
Day 5 - Dec 22, 20:00 - 989.771315E-03 Temp 32.0C
Day 6 - Dec 23, 20:00 - 989.771227E-03 Temp 32.0C
Day 7 - Dec 24, 20:00 - 989.771320E-03 Temp 32.2C - 7 day PPM deviation was only 0.04 PPM/day - nothing to complain about.

Day 9 - Dec 26, 22:00 - 989.771215E-03 Temp 31.9C

[TiN]

I'm joining too with my second box, which ended up way more expensive than TheSteve's. ...
 

Day 1 : 18.DEC.2016 20:48, 35.5°C, 983.313029E-03, Reference, 0.000 ppm
Day 2 : 19.DEC.2016 08:32, 35.5°C, 983.313098E-03 , +0.0702 ppm
Day 2 : 19.DEC.2016 18:31, 35.7°C, 983.313055E-03 , +0.0264 ppm
Day 3 : 20.DEC.2016 07:48, 35.3°C, 983.313080E-03,  +0.0519 ppm

[TheSteve]

Very nice!
I currently have my Prologix GPIB adapter connected to the 3458A but am working on firing up a spare raspberry pi and using a 82357B with it for data logging.

[tszaboo]

The Steve, as you are now also joining the volt-nuts / 3458A club, I recommend anyhow to build your own volt and resistance references, at least for verification of the 3458A, but also as true standards, which the 3458A itself isn't.

Frank

Would not be easier to start with the 10Kohm reference first, and check the stability of the meter with that? A bulk foil resistor is some 15 EUR on Digikey. 1 year DCV 10V is 8ppm, 10Kohm is 10ppm. 1 day is 2ppm.

[Dr. Frank]

Would not be easier to start with the 10Kohm reference first, and check the stability of the meter with that? A bulk foil resistor is some 15 EUR on Digikey. 1 year DCV 10V is 8ppm, 10Kohm is 10ppm. 1 day is 2ppm.

Well, definitely, no.
Resistance is not 'easier' to check, price of component is not everything, and especially what you suggest, is not sufficient at all.

At first, TheSteve wants to check, whether the 3458As internal LTZ1000A module is long-term stable or not, as described in AN-18. That can't be done with a 10k reference resistor, because U and R are independently realized in the 3458A. Additionally, there is no  known instability for the internal 40k reference resistor.
Otherwise, TheSteve should check, if his instruments already contains the VHP101, which is probably much more stable than in the specification, which relates to the older type.

It might be as good as < 1ppm/year, if one assumes the Vishay specification for  their oil filled type BMF resistors.
The T.C., anyhow is about < 1ppm/K, and therefore, compared to the voltage reference, much worse, as the interior of the 3458A might easily change by ten degree Celcius or more.

That's why I suggest, that the 3458A can't be used as a 'standard' instrument, especially for the mediocre resistance mode.

So you would need a very stable and/or well characterized external resistor reference, which I assume cost much more than these 15$.

Some volt-nuts have an SR104, or Fluke 742-10k, or DIY Vishay 10k VHP 202Z, VHP 101, with included thermometer, so the T.C. can be calculated back to normal R25.
See also 'T.C. measurements on precision resistors'.

As far as I can tell, from comparison of the 3458A against a group of 5 external VHP202Z, the 40k resistor inside my own 3458A, a VHP101 type, seems to very stable over the years, < 2ppm/2years, and its T.C. might be about 0.4ppm/K (variations are too high for precise determination of T.C.). But this ain't no 'easy' measurement, either.



You also see, that absolute Ohm measurements with a 3458A are not better than about +/-2ppm.
If you monitor its internal temperature, maybe a bit better.
Relative Ohm measurements (transfer) can be done to about 0.2ppm, though.

Frank

[HighVoltage]

Congratulations on your new toy and welcome to the 3458A volt nut club
I am looking forward to your stability findings.

What a great Christmas Present.

[TiN]

I think most of the reason right now for SN18 test is to determine if ADC drift is OK. Using SN18 procedure is good enough for that.

[TheSteve]

Checked on the meter this morning - heart sank when I saw "ERR" on the VFD. Thankfully was only a syntax error as my Win box decided to reboot itself after it installed patches and sent a character to the 3458A via GPIB in the process.

I now have the 3458A connected via a 82357B and raspberry pi - much nicer combo I can query from anywhere.

A reference 10K resistor would be very nice, they all cost more then I spent on the 3458A though(the 3458A was less then my 34461A). I am sure one will find a way into my lab eventually, same with some other voltage references.

[TiN]

Give us IP/login, we will query your 3458 too  .

...via a 82357B and raspberry pi

Did you have any issues getting it to work? What kernel verison you use? From what we see in other threads, new kernels cause 82357B's not work with linux-gpib, either real dongles or fakes, same.

I noticed that ACAL gives often calibration errors message if high-ish AC voltage left at terminals (e.g. 100VAC).
Perhaps some leaking signals cause meter anger (front terminals are normally disconnected from measurement circuitry during ACAL procedure).

[TheSteve]

Give us IP/login, we will query your 3458 too  .

...via a 82357B and raspberry pi

Did you have any issues getting it to work? What kernel verison you use? From what we see in other threads, new kernels cause 82357B's not work with linux-gpib, either real dongles or fakes, same.

I noticed that ACAL gives often calibration errors message if high-ish AC voltage left at terminals (e.g. 100VAC).
Perhaps some leaking signals cause meter anger (front terminals are normally disconnected from measurement circuitry during ACAL procedure).

Hah - yes I might just give you a login, would be easy to do.

No problems getting the 82357B to work. I am using a raspberry pi model B (original one). I downloaded the newest Raspbian Jessie Lite but it seems to have remote ssh disabled by default which is a pain - I don't want to bother with keyboard/monitor/mouse. I already had the 2016-05-27 build downloaded so I used that instead, it lets me login via ssh headless. I have 5 raspberry pi's running 24/7 at home, I have never plugged a keyboard or mouse into any of them. I then mostly followed your howto on xdevs.

root@gpib:~# uname -a
Linux gpib 4.4.0-1-rpi #1 Debian 4.4.6-1+rpi14 (2016-05-05) armv6l GNU/Linux

[TheSteve]

One more day to go to have a full 7 days of data - so far it looks pretty darn good to me, especially if we toss out day 1 which was right after power on.
It has been measuring voltage or a resistor 24/7 since being powered on and I run the full ACAL each night, so far not a single error - fingers crossed that continues.

And just to add a little more fun to things I have a second 3458A on the way. So now I really need some decent standards in the lab.

[TheSteve]

Unit number 2 arrived today(another ebay purchase). It is pretty good shape overall. Serial # 2823A004xx - so a very early unit. I see parts dating as early as 1986 in it. The part # on the serial tag indicates it is a "3458AR" - I also see a sticker which mentions HP financing so I assume it was a rental unit at some point. It was clearly not "refurbished" at all at any point, so I can't see the "R" meaning anything else. Option 1 was installed from factory.
It had a NVRAM error, the same as the first unit. I replaced all 3 NVRAM's and was lucky again that the CAL NVRAM was still working (date code 8827). This unit had firmware version 2 in it, it is now updated to version 9.
I had to break a CAL seal to open it which is always a nice thing - oddly enough while having a good look around I found the 5 volt regulator on the A6 supply board was missing its screw. I wonder how long it ran without proper heatsinking.
It needs a new fan(runs but is loud) and I need to apply fixes from 2 services notes, 3458A-01C and 3458A-06A - basically change two capacitors on the A3 board and add a resistor to the A2 board. Parts from Mouser should arrive tomorrow.

The self test and ACAL all pass - so fingers crossed she is good!

Being the CAL NVRAM was still good I was able to see the last ACAL 72 value(no idea when it was done of course) - it was .988869338, after running an ACAL today the value changed to .988869682 - which so far seems pretty reasonable.

[TheSteve]

Here is a shot of the two. They were as close as 200 nanovolts apart early when both were fully warmed up. I then reassembled the new unit and stacked them, it is still a few degrees cooler. Of course neither has a current calibration.
btw, cal count on the first unit is 150, cal count on the second unit is 10. The cal string has no readable data on either unit.

[Theboel]

You are so lucky. I am not dare to dream have one 3458A and You already have two 

[Towger]

The Volt Nut disease seems to take hold fast once it is caught.

I must get my HP3456A out of storage and check it still works.  I picked mine up for 50 euro, they are an affordable gateway into the world of a Volt Nut[emoji6]

[TiN]

Only meter better than 3458A is.... Two of 3458As.

But I need to note, whatever you doing, you are not doing in tin's way no replaced transformers, no replaced A3's? 28 year old box just works after NVRAM pill? Why everyone but me that lucky .

Let's see more guts photos! Afaik your second box is oldest we ever know across EEVblog crazy-nuts.