[3458A] Yet another 3458 repair thread.

[ap]

Frank,
Well, if a 3458A and 5700, artifact-adjusted to the same traceable standards do agree in all their compared ranges and measurements within the specified limits, there may be a good point in arguing this is good enough, I cannot judge however.  I could not explain why this is ok (or not). I read a doc of the US DOD (or a subco) who did an analysis some time ago and it seems it is ok for them, in a nutshell, that a self-adjusted 3458A is verified only every second time. But not always. Same may (or may not) apply for the 5700. Now based on that, where is the limit of what is acceptable in such a 'bootsrap' approach? Always? Every second time? Never? I guess from an ISO9000 perspective it would be hard to explain to the certifying body that two self-adjusting machines agreeing within to be calculated limits (and never verified by a an independent method) are good enough (unless they accept the Fluke paper as-is, but which also claims a 2 year external cal cycle, mostly for AC though) that each of them meets its spec; (another story is the specific 5720 history, built from parts. We should leave this aside here, but people servicing these units have warned me to buy mixed units combined of boards of unknown version configurations, which sometimes are arround. I can not judge this statement).

On the other hand, since the 752 and SR1010 are so simple and follow simple math in their self-cal / math models, this is easily traceable and accepted practice therefore.

[TiN]

We should leave this aside here, but people servicing these units have warned me to buy mixed units combined of boards of unknown version configurations, which sometimes are arround. I can not judge this statement

I can only comment on this, that mix-matching different boards/revisions is definitely not a good idea and I would not recommend doing so for any customer-related lab/business.
During my project build I have learned quite a bit why, however I do not feel comfortable to disclose that publicly for various reasons.
However I am confident in current performance data enough to theoretically send my self-built calibrator for calibration, if not the cost and risk of shipping damage.

Also following legal way, 752/1010 self-calibration would fall into same pitfall as ACAL, because there is no connection to SI w/o external verification.

[e61_phil]

I talked to the guy in our local cal lab here a few years ago and he said EVERYTHING is calibrated which is used for calibration. This includes their 752A, 792A and 720A.

Their Fluke 57xxA calibrators will also go to calibration every two years. The same thing said CalPlus as I ordered the 5730A.

Addionationally I think there are two big differences in the 57xx ACAL and 3458A ACAL:
1. The 57xxA will verify the DAC linearity during ACAL. Imho the 3458A doesn't verify its ADC linearity
2. The 57xxA has CalCheck and does a real verification with separate standards

If the 3458A ADC would have 3ppm INL due to a defect ACAL doesn't notice that and every range is corrupt. The 57XXA will detect that. In a first step it will detect the non-linearity and as a second step the CalCheck would report huge drift.

[TiN]

So then i'm good , as this is precisely where discussion did start - I have here perfect specimen and setup. Dodgy A3 ADC with corrupt linearity, that is confirmed by three good 3458As, even though none of those three did ever visit traceable calibration lab.

On second point - Calcheck procedure uses separate standard (internally) only for resistance. Calcheck does not verify 1.0 and 1.9 ohm, and using same LTFLU dual reference DAC as internal standard for DCV. Null meter ADC (very same one used in 8842A) compare all other functions/ranges (except AC/AC transfers) to the same LTFLU dual ref DAC. Only external ACAL will calibrate and report traceable shifts for 1.0/1.9 ohm and LTFLU.

I can do this experiment next week, run external ACAL of 3458A with bad ADC. Let it run few days to drift away 4-6 ppms. Then perform performance verification procedure using 5720A as source, and see how many of ranges/functions we fail and how bad. Then do 3458A ACAL external again and run same test without waiting, which in theory should still give broken data/out of specs functions. Then we can use good 3458A as final sanity check that 5720A itself did not drift away. My perfval procedure using python app takes about 15 hours.

[e61_phil]

Yes, the dual reference itself isn't verified, but all DCV ranges are checked against the transfer with the DAC. This wouldn't show a drift in reference but it shows drift in DCV ranges. To be honest, that is even with the 3458A possible, if one reads the CAL values before and after an ACAL.

Btw. If you have a traceable 10V source than you could do a traceable chain in my opinion if you transfer the 10V to all other DC ranges with a 3458A which is specified for such transfers.

I don't understand your experiment. Why do you want to let it drift?

[TiN]

Dual reference is verified with MFC's ACAL first thing, before anything else.

So we see how bad INL ADC is easy to find without any traceability. Maybe?

could do a traceable chain in my opinion if you transfer the 10V to all other DC ranges with a 3458A which is specified for such transfers.

Okay, then what difference does that make if I use ACAL DCV which give same end result, instead of measuring 10V on higher ranges, divided 10V on lower to calculate gain error?

As you mention need of verification on all ranges to get the valid calibration. Feels like we have arrived back to station of departure, so I'm puzzled now with this quote, sorry.

[Kleinstein]

Doing the transfer in the 10 V range externally is not that much different from internal ACAL. However one could first check the 10 V range is working well - which is a little more difficult for the internal ACAL connections - though less circuit and not very likely,  there might be some different source of a defect in the ACAL paths.

The main point of checking the 100 mV and 1 V ranges is to check that ACAL is actually working (e.g. INL of the ADC is OK) and also to check that the amplifier in the 100 mV and 1 V range works correct. This is more like a check for defects and less a classical calibration reading. To have the check to the same level as a normal calibration it would still need high accuracy.  However the expected failure modes are different - so one could argue for different tests:  As an additional test one should check if the ACAL constants are reproducible. On the other side slow drift and a temperature influence is less likely and could thus relax the test requirements here. 

The ACAL history would in theory also give additional testing of the instrument - so ideally defects could be detected (e.g. like a dirft in the CAL72 values that is related to the dreaded bad A3 boards). This could in theory allow for longer cal intervals in an ACAL instrument, especially if more than one range is rather stable even without the ACAL.  So ideally one should have both ACAL and at least a few more very stable dividers / shunts. With just 1 long time stable shunt in the 3458 it is more like cost reduction. With 2 stable resistors in the 5700 there could be additional consistency tests for added confidence.

At this level of resolution calibration can only do spot checks anyway - there is no way to check all possible values. So one has to assume the meter is kind of working in between the test points.

[e61_phil]

A last sentence to the different ACALs:

I think one cannot compare the two systems (57xx and 3458) from metrology point of view, because the intention was completely different. No matter what is technical behind. The 57xxA CalCheck was intended to be a CalCheck. The rest of the instrument is temperature stabilized (where needed) to operate over the wohle range (+/-5°C) without any further compensation due to transfers.
The 3458A instead uses ACAL to compensate not only for long term drift. It is also intended to compensate for temperature variations. Therefore, even if everything is fine the CAL constants will move with temperature. That isn't a flaw, thats the whole idea here.


My wishlist for a new 8508A wouldn't include an ACAL like the 3458A does. If they would be able to get an ADC which is as good as the 3458As one. I would like to see a fast 8508B with CalCheck. That means keep all the expensive resistors and so on and then do transfers like the 3458A with the intention to report drift and not to correct anything.

[TiN]

Calcheck is part of complete ACAL, as it rely on external trace to standards. Resistances in 57xx are not ovenized either, so any drift (thermal or temporal) is also corrected by ACAL. So tempco of the resistance networks at least will have direct effect on resistance, DC current and AC current functions.

If one to look at units as black boxes, there is no much difference between the 3458 or 5700, as even with range adjust function present in calibrator you cannot compensate/calibrate many aspects of the calibrator (leakages, flatness corrections, loading effects) that are currently corrected automagically by ACAL procedure. 

If you want to avoid corrections in 3458A, you are not required to do them at all. Perform ACAL ALL during the external traceable calibration, and never run it again. That will render operation just like 8508A or any other traditional meter, with constants preserved whole time. So I don't see a problem here. ACAL is not a requirement and there are specified TCR errors for each range listed with and without ACAL correction. If you maintain environment temperature stable and equal to calibration temperature, there is no need to apply ACAL or temperature correction, just like 8508A. 

A last sentence to the different ACALs:

I think it is valuable discussion and both points of application are equally important to talk about. 
Maybe mod can separate last dozen posts into new thread called "Fluke/HP Artifact Calibration and it's merit" or something like this.

[e61_phil]

Drift in the 57xxA will only be corrected if you run an artifact calibration. On daily/weekly/whatever use there is only CalCheck. That differs from the 3458A use. With the 3458A you run ACAL every day (at least).

It would be interesting how stable a 3458A is without ACAL. There is no clear statement in the datasheet, that says that ACAL is needed within the last x hours to fulfil the specs. Expecially the ranges which are not 10V are interesting.

[TiN]

With the 3458A you run ACAL every day (at least).

  Why?

3458B stable to <0.1ppm over a week, no ACAL. Beware, heavy SVG URL page.

p.s. okay, to follow legal word and manuf spec, yes, you right

[TiN]

Well, quick search only, sounds like Microprecision has a lab in Taiwan and can calibrate the 10V to 0.6ppm, 10k only to 2ppm. That saves you shipping to the US. And better than Keysight uncertainties on paper.

Well, as I suspected, MPC Taiwan lab is unable to calibrate my 5720A box, as they don't have the required set of standards. I knew there would no easy way on this. 

[TiN]

Replaced original dim worn out VFD to a 3rd party glass. Now it's even nice and quite bright.



Was more fiddling with pins to get them align with holes more time than actually soldering/desoldering.



Now just waiting for replacement part on 3rd unit, while 4th is under SN18 testing for few weeks. 

[michael_liu4444]

Hello TiN, I know you are an expert repairing 3458A. I am, as a starter,  doing it too now. I need a chip A3U180 (1NC1-0017). But I don't know where to buy it. Will you help to let me know where I can get the part ? I appreciate your effort and time spent helping me on this. Hope to hear from you soon.  Wish you a nice day ! Michael Liu

Sent from my SM-N950F using Tapatalk

[MiDi]

You have to buy complete new A3 board from Keysight.
Hopefully you have big pocket 

PS: the ADC of my unit seemed to have drift sickness, but it settled to acceptable rate in 1/2 year and is still settling

[michael_liu4444]

 I checked price for a complete A3 board. Oh, my... too expensive... No where to buy U180 itself  ?

Sent from my SM-N950F using Tapatalk

[TiN]

U180 is designed specially for 3458A by HP, it is not available anywhere. You need to get whole board, there is no other way.
What is the problem with your board?

[michael_liu4444]

Thanks TiN.  This chip U180 faulty. I confirmed it by swapping  good U180 from a good A3...

Sent from my SM-N950F using Tapatalk

[martinr33]

Is your A3 board dead, or just drifting?

[michael_liu4444]

my A3 board is dead: 3458A will fail some self test at power up ... it is con formed A3U180 caused it...

Sent from my SM-N950F using Tapatalk

[michael_liu4444]

Hi everyone, I have one more question: if I install a good old version A3U180 (1SJ8-0108) into a new version A3 board which used 1NC1-0017 before,  can 3458A still pass self-test ? What is the difference between 1SJ8-0108 and 1NC1-0017 ?
Hope to hear from you soon.
Thanks.


Sent from my SM-N950F using Tapatalk

[syau]

Hi everyone, I have one more question: if I install a good old version A3U180 (1SJ8-0108) into a new version A3 board which used 1NC1-0017 before,  can 3458A still pass self-test ? What is the difference between 1SJ8-0108 and 1NC1-0017 ?
Hope to hear from you soon.
Thanks.


Sent from my SM-N950F using Tapatalk

I just did the other way round for fixing the drift rate  The only issue is my bank account 
 

[TiN]

Gotta pay to play with 3458A. These are not cheap to keep alive, but there is no replacement on the market for best 10V meter.

Different U180 chip likely is just a batch, as I've seem different boards with different part numbers, but they are following same design and are interchangeable.

[michael_liu4444]

Thanks a lot to all of you.. I really encouraged by learning from you all.. wish you all a great weekend !

Sent from my SM-N950F using Tapatalk

[dennis123]

>>>>Unit 4 fixed, issue with A2's Flatness DAC convergence: 198 was dodgy EL2039 wideband opamp. Replaced one with NOS chip and she worked as expected. 

Hi TiN,

Can you share more info on this? How did you run the troubleshooting? Which U is this EL2039 in A2 board?

I have a similar Flatness DAC error with no 197 instead.