Early 3458a repair

[martinr33]

I am working on a very early 3458a - paper label, serial number just above 00100. As it stands, it won't calibrate - but is only 2uV off of my standards. I have to figure that this unit, being so early, got pretty special attention. It is about 30 years old, judging by the date codes.

The NVRAMs are original, and were still holding. Gone now, socketed with new parts (and old data).


It has two problems, so far.
 - Open circuit x100 gain resistor. HP won't sell me a new one, but they will swap my DC board for a (hefty) fee. I have hacked in a temporary replacement while I wait for a custom Vishay divider.
 - Somethng bad in the trigger DAC circuit on the AC board. Most likely the DAC, but it could be the op amp or CMOS shift register. I am waiting on parts.

I'll write this up in more detail. I like to troubleshoot by probing, rather than replacing parts. But in this case, I have socketed all the suspect parts on the AC board, so that it only gets heavy handling once. These are pretty trick sockets; you can (barely) see them under U006 and U501. They are turned pin sockets that fit inside the existing PCB hole.
 
Here's the Google album, high-res.
https://photos.app.goo.gl/BGxNEMijzuKswMmi9

I'll have to create some smaller images for this post.



A few points of note:

The AC board has component designators onthe rear side.
It needs another wash. I'll do that when I have it working.

[TheSteve]

Does your unit by chance have option H01?
One of mine is early, but still a few hundred after yours. Does your LTZ reference have the standard part number or an internal number?

[TiN]

We were wondering in volt-nut chatbox who will snatch that meter. Now we know (and glad it's one of "ours"). 

Open circuit x100 gain resistor.

Which one is it? I still have blown up A1 PCB, which I could in theory take the resistor from . 

[martinr33]

@TheSteve:
The unit does have option H01. The serial number is 2823A0000113. I'll pull the cap off of the reference when I replace the voltage divider resistor, BUt I would expect it to be an HP part number. Many of the parts on  the boards are marked with house numbers.


@TiN
Thanks for the offer, I should have asked here first! I have a couple of Z-foil precision dividers on order from Texas Components TX144Z (these folks are really good, BTW - they will do one offs, and deliver in a couple of weeks. They build Vishay parts). The spec on these is +/- 1.5ppm per degree C, but I can't imagine that the Vishay ceramic divider that failed would be any better. Matching on these resistor packs is only critical for each divider pair, and not the pack of two. One pair is divide by 10, and the other pair divide by 100 (well, not quite 100). For now, it has a selected 100K (99.6K measured) resistor doing duty

[TiN]

I'm surprised you tried to elude EEVBlog. 
And this thread is crying without photos. 

[TheSteve]

@TheSteve:
The unit does have option H01. The serial number is 2823A0000113. I'll pull the cap off of the reference when I replace the voltage divider resistor, BUt I would expect it to be an HP part number. Many of the parts on  the boards are marked with house numbers.

Mine had an internal LT part # on it which I had never seen before, not an HP part number on it.

[martinr33]

The ADC appears to work fine, with no obvious drift. I haven't been able to fully test it yet, but the meter is withing 2ppm of where it should be without being able to autocal.

The socketed 8751 is original. HP must have been finishing up the code, but this code has never changed since. The new Atmel part is the same code.

Top right you can see where I have added the missing tantalum caps. HP's recommended repair is to replace the two 0.1uF decouplers with 2.2uf parts. I decided to do what HP did, and add the tantalum electrolytics. My parts are a little older than the meter...

[martinr33]

@ TheSteve: here's the reference. T106382, built in 1987. So about a year older than yours. Otherwise, the board looks pretty much identical.

[ClavinovaGuy]

For the early units, there have been multiple service notes on updating components. If this is the first repair in 30 years, it might be worth it to contact Keysight Colorado for the factory refurb on the unit. Yes, it is expensive, but the calibration it receives will include a burn-in on the components and a factory and primary lab cal.

[martinr33]

For this unit, I really wanted to keep it close to original - I am curious as to how the first units performed, HP would have been all over these - they were (and are) an exceptional design. As it stands - without cal - this one is only a couple of ppm off of my references.

The updates are reasonably manageable. Caps on the ADC board; new code; socketed and updated NVM (all done).  Replace suspect electrolytics - in process. 110K on the AC board - next time I pull it. Upgraded 40k resistor standard - arrived today. Powerline filter - probably not an issue on 110V. Mostly, the  design is unchanged from 30 years ago. Plus, I can get the resistance and voltage calibration close enough that a Keysight cal would only be better for a short period.

[TiN]

Upgraded 40k resistor standard - arrived today.

I'm curious what "upgraded" means and how you will validate that.
I didn't dare to replace 40K in my units, but I did test 40K long ago on my dead A1 board, which resulted in +1.4 ppm/K TCR.
K2002's give me better TCR for 2k/20k/200K ohm ranges, if i'm going after tempco.

[coromonadalix]

I ve seen few notes on the shop who calibrate my company meters,  they charge minimum 1200$ usd to switch the old adc board
to an refurbished one and lots of $$ to do an complete calibration and certification outch.

But it is not impossible to do the mods and find an proper resistor has you said / wrote.

Unless you want a clean / total working calibration, and the 3458 be up to 100% working has it should. It's a very good beast to tame.

have you seen this thread on XDEV'S : https://xdevs.com/fix/hp3458a/

very informative

[martinr33]

I have a 40K VHP202Z from Texas Components in hand.

http://www.vishaypg.com/docs/63120/hzseries.pdf

0.2 ppm typical TCR, and 2ppm over 5 years shelf life stability. As I am not loading it in any meaningful way (it is rated at 0.2W), I would expect to see something like the 2ppm / 5 year number.


And I agree on the K2002: it does seem to hold readings stable across a wider temperature range than the HP, even if it is noisier.  There again, the HP was designed with the idea that autocal would trump thermals.

[martinr33]

Wow - December already. I finally had time to shape those Texas Components metal foil resistors into a replacement for the resistor array - RP100 - that controls the main DC amp gain. It did not go well - I randomly assumed that they had the same pinout as the HP, so I ended up doing some crafty bending on the pins. I would have saved a lot of time by getting it right in the first place.


I glued the two parts together to make a replacement RP100. Ths is part of the main divider in the DC amplifier stage I could have replaced just the bad half of the divider, but I prefer to replace the whole thing. The Texas metal foil matched resistors should be better than the original.   



Now I can button up the DC board and get back to the DAC error on the AC board. I'll check to see what the DCV error is with the new array in place. It won't autocal until I fix the AC board.

[TiN]

Those dividers look interesting, didn't see them before.

And what happen to high-voltage Coto 7005 relay? 

[martinr33]

The divider is actually two resistor networks glued together, so maybe not as interesting as it looks. I am figuring it won't affect my 10V range, which I presume to be unity gain. However, as I cannot yet complete autocal, the other ranges should be off.

At one point (due to incorrect installation by me) the error reported was in the flatness DAC, and pretty much instantly upon attempting autocal.
With the divider working correctly, the error returns to the Level DAC circuit.

It now almost completes the DC cal cycle, dropping out at the very end with the DAC error. As the DAC appears to be inactive (0 output), I need to look at the control signals - the chips appear to be good.  It will complete autocal for resistance.

On the high voltage relay - it came with those marks. There's no evidence of ham handed work inside the unit - or any work at all - so I wonder if  the unit was made that way. With SN 127, it has to have been in the first few months oof production.

[TheSteve]

My early unit has some funny looking spots on the relay too - so I wouldn't worry about it.