HP 3458A ; A/D board spare part availibilty and cost

[niner_007]

No, i have not run out of arguments. Just wrote down some information to show that there is no good reason why a 3458A ADC should drift at 11 ppm/year. Can't others contribute any information to solve this riddle?

Regards, Dieter

Get it fixed, get a service repair agreement, what's the problem?

[dietert1]

Apparently nobody understands how to fix it or they keep it secret to charge again and again. They just replace the ADC module but the new one may behave similar.

On the xdev.com site there are thermal images of a 2015 ADC board (IR000326 and IR000330) and it has a hot spot: The fast comparator on its patch board, which was measured to run at 75 °C (after opening the cover). With the covers closed it probably runs at 80 or 90°C and one could apply an IC cooler to bring this down and see whether it helps. Older ADC board revisions have a different comparator chip and it sits in the ADC board, but it also has a consumption of about 600 mW (20 mA x 30 V).

Regards, Dieter

[niner_007]

Do you or we (in the forum) have statistics on the failure rate? If we don't, it could well be anecdotal, sorry, but it's just that simple

[MiDi]

Agilent had quality problems with U180, look at SN18A.

[Kleinstein]

The Elantec comparators are a different critical point. They are fast and use quite some power for this, so they run hot. These comparators are also know to fail sometimes, though not sure why.  The daughter-board is a fix to replace 2 single comparators (with very poor availability) with a single dual version that for some reason had better availability. The dual version naturally runs even hotter.  The 3458 has a fan and thus may not run hotter with a closed lid - it may even run cooler for some parts.
Anyway the new SMD version changed the comparators to a different type.

[martinr33]

If you have a 732a or silimar super stable voltage source, a drifty ADC shows up very quickly. A ppm of drift is 10uV.

I don't know that sellers really know whether or not the ADC is bad. If you do a regular autocal, and are not monitoring stable sources over several days, you would never see it. Also, if you use the machine only occasionally and keep it turned off, you would not  see it.

Certainly, when you buy a unit on eBay, you might have been looking at a $1200 ADC repair. Now, you cannot get the ADC. Based  on the units from eBay I know about, 25% or so are bad.

[ap]

Unfortunatelly, the ADC drift issue seems not to be anecdotal. My personal experience, this affected more than 50% of the units I had in hands (used, older and newer ones), and it is all over the place, including A3 units bought from KS (which they then replaced). Have not done a formal statistic (should have), my personal summary. So beware of buying used 3458As without having the right to return drifty ones.

[Dr. Frank]

Unfortunatelly, the ADC drift issue seems not to be anecdotal. My personal experience, this affected more than 50% of the units I had in hands (used, older and newer ones), and it is all over the place, including A3 units bought from KS (which they then replaced). Have not done a formal statistic (should have), my personal summary. So beware of buying used 3458As without having the right to return drifty ones.

Hello Adrian,
I know and appreciate your experience already, as you also trade these used instruments. So what is your 'hard' criterion (boundary specification) of a 'drifty' U180? How many ppm/timescale is ok for you? What do you base your judgement on?

I think, w/o well-founded arguments about what is really relevant for the performance of the 3458A, namely its final specifications, such diffuse statements, also such as from Dieter and Kleinstein, that will only cause confusion and uncertainty on owners and possible buyers of this instrument. Also, please claim your use case, what a more or less stable U180 should be good or bad for.. every other discussion is only of academic nature.

In this context, I'd like to suggest to again read the hp journal 4/'89, p. 13 ff, where the design engineers nicely explained the design goals of the A/D circuit, especially the U180 drift and matching requirements. That may help to get some realistic expectations about this circuit.

P.S.: Your advice, to imply the right to return 'drifty' 3458As might only be worthwhile, if there would be a reliable, commonly accepted criterion. Maybe you can share your personal experience, in discussions with business partners, or with Keysight.
Frank

[dietert1]

Maybe that criterion can be derived by comparing the 3458A ADC to that of other long range DVMs. For example i can contribute an ACAL check measurement from our 3456A multimeters: Connect input to its own voltage reference. For me that is something easy to do. Make some measurements at different temperatures during the next days and repeat after three months to determine the drift.

It will take time to get the result, but in the end i think everybody would agree that a 3458A ADC should not drift a lot more than a 3456A ADC. Anyway there would be some numbers for Midi to compare to. That is the use case. Of course it would be interesting to have results from other 3458A owners, too.

Regards, Dieter

[Kleinstein]

The test HP suggested from the ACAL drift is 0.43 ppm per day, which is 3 ppm for a week. Because of temperature effects is may take more than 1 day to get a reliable drift rate. A week sounds more realistic for the test. Good meters are supposed to be much better, even the 0.03 ppm / day reported by Midi are more on the high side if persisting for a long time.

For practical use 0.4 ppm over a day may already be a problem and require quite frequent ACAL.

The ADC can tolerate quite some deviations for the slope ratios (e.g. 0.1 % for high speed, 3% for low speed). However the ACAL measurement checks a different resistor ratio. It is only an indirect test in this respect - so the limit for the total drift may have to be tighter (e.g. 100 ppm total drift) to be reasonable sure the other resistors have not changed 1000 ppm.

[dietert1]

Dr. Frank wrote before that the 0.43 ppm/day number is based on the assumption of random walk (0.43 ppm/day random walk gives about 8 ppm after a year which is the basic instrument accuracy spec). If the drift is continuous and persistent for more than a year as in Midis case, that assumption does not hold. In other words: The observation of a long term continuous ADC drift is far away from what Agilent engineers had seen and means that an ADC board maybe in risk.

Regards, Dieter

[Dr. Frank]

Dr. Frank wrote before that the 0.43 ppm/day number is based on the assumption of random walk (0.43 ppm/day random walk gives about 8 ppm after a year which is the basic instrument accuracy spec). If the drift is continuous and persistent for more than a year as in Midis case, that assumption does not hold. In other words: The observation of a long term continuous ADC drift is far away from what Agilent engineers had seen and means that an ADC board maybe in risk.

Regards, Dieter

Sorry again, Dieter, but this 0.43 ppm/day criterion is not based on 'Random Walk' phenomenom, and I have never said so. 

Again and again, there was no , and there is no specified criterion for the annual drift of the resistor array inside U180.
When this serial batch problem occurred in about 2006, the agilent engineers have used the timely DCV drift specification of the 3458A, which is mostly caused by the LTZ1000A drift, and boiled that down to its theoretical 24h drift. The typical behavior of any thermally induced drift is a square-root one, see article from hp journal 4/89, with an annual drift of 8ppm/yr.
So the drift function can be written down as: dU(t)/U_ref = 8ppm/yr * SQRT(t(hours)/8742h).
For t = 24h you get these ominous 0.42ppm.

First, this crude criterion has another flaw. If you have a 4ppm/yr. reference built in, this criterion is two times too high.

Second, this usual SQRT law will vanish after some years, and instead effects like popcorn noise, or random walk effects will get dominant.
So the drift of such references will probably never come to an end.

For comparison, I show the timely drift of one of my LTZ1000 references (at 50°C), from 'First Light' to the actual status.
The measurement is done relative to the group of the other 7 references, using the 3458A only for comparing them.
So this measurement is an approximation only for the absolute timely drift, but supported by several baseline comparisons to calibrated standards (e.g. on MM2019).
The comparison is done on a monthly basis, so you see a lot of small jumps and humps on the order of 0.2ppm.
These are caused by the intermittent measurements, maybe also by other drift phenomenons.
In the end, this is my worst drifting reference.. if hp would implement the LTZ1000 as intended, i.e. at 50..60°C oven temperature only, for metrology purposes in a metrology grade environment, an annual drift rate of typically -0.8ppm/year could easily be achieved.

To demonstrate that such stability measurements can also be performed much smoother, I also show a continuous 24h experiment, comparing my 3458A versus an ultrastable Datron 7000 reference at constant room temperature (after 2h). There's no ACAL during these 24h. A combined noise of about 280nV_eff, and small random jumps of +/- 0.1ppm can be observed. Both effects are caused mostly by the LTZ1000 / LTZ1000A references inside both instruments.

In the end, these are practical and realistic use cases for the 3458A, showing the general limitations of DCV stability measurements.
Another hint, my 3458A obviously does its job nicely, as no practical relevant drift of its U180 can be detected in this case.

Looking at these diagrams, this discussion, whether the theoretical 0.03ppm/day drift for the U180 of MiDi's 3458A, is' healthy', or not, to me seems to be really academic.

Frank

[dietert1]

You need to understand that the square root formula is a characteristic of random walk. I expected a physicist to know that ("Braunsche Bewegung"). When you read the technical note linked by Midi, it is completely clear that the authors are treating ADC drift as random walk and they could not imagine that an ADC will exhibit a continuous and persistent 11 ppm/year drift. The 8 ppm/year or 4 ppm/year instrument spec is exceeded at a continuous drift of 0.03 ppm/day as observed by Midi. I know this drift can be hidden by the ACAL procedure, as long as it works.

Seems like you want to divert the discussion away from Midis findings about the 3458A ADC to your favorite LTZ1000 hobby. But the question of ADC drift is different from the question of voltage reference drift. It is not academic, because the most sophisticated and linear ADC is useless once it does not work. With a voltage reference, we know how to fix that. If you want to contribute something useful, you can provide a log of ACAL 72 drift of your instrument, so we can see if it is nothing or 0.03 ppm/day.

[ap]

Hi Frank,

I stick to the 0.43 ppm spec, measured over the time specified (ok, when there is no drift after say 4 days, one can also stop there, provided stable ambient temperature). I check the meters continuously against a set of references (732s/4910), so no ACAL needed to determine the drift. Making an ACAL-based drift test every day does not supply meaningfull data imo (and as per spec).
When working with a 3458A, it is not about personal feelings or so, the instrument needs to meet its spec. The 0.43ppm is the only spec related to the A/D. That being said, a meter showing 0.4ppm/day (and I have not seen that) would be considered bad, altough well usable. You either have a drift of some ppms/day if bad (seen up to 20ppm/day), or in the 0.1ppm/day and below range if good (all averaged over the specified time).
For me the value of the 3458A is not to try to achieve sub ppm measurement uncertainties, even with a 2ppm/a low drift version (and most aged instruments today are in that range anway), if I need tight uncertainty I always use the meter in transfer mode. Thats most of the measurements when working with high accuracy stuff.

[Villain]

You need to understand that the square root formula is a characteristic of random walk. I expected a physicist to know that ("Braunsche Bewegung"). When you read the technical note linked by Midi, it is completely clear that the authors are treating ADC drift as random walk and they could not imagine that an ADC will exhibit a continuous and persistent 11 ppm/year drift. The 8 ppm/year or 4 ppm/year instrument spec is exceeded at a continuous drift of 0.03 ppm/day as observed by Midi. I know this drift can be hidden by the ACAL procedure, as long as it works.

Seems like you want to divert the discussion away from Midis findings about the 3458A ADC to your favorite LTZ1000 hobby. But the question of ADC drift is different from the question of voltage reference drift. It is not academic, because the most sophisticated and linear ADC is useless once it does not work. With a voltage reference, we know how to fix that. If you want to contribute something useful, you can provide a log of ACAL 72 drift of your instrument, so we can see if it is nothing or 0.03 ppm/day.

I really enjoy the discussion here and i am happy lurking but i gotta correct this real quick as we want to pay tribute to the great people before us correctly. Its Brownsche Bewegung not Braunsche.

[Dr. Frank]

You need to understand that the square root formula is a characteristic of random walk. I expected a physicist to know that ("Braunsche Bewegung"). When you read the technical note linked by Midi, it is completely clear that the authors are treating ADC drift as random walk and they could not imagine that an ADC will exhibit a continuous and persistent 11 ppm/year drift. The 8 ppm/year or 4 ppm/year instrument spec is exceeded at a continuous drift of 0.03 ppm/day as observed by Midi. I know this drift can be hidden by the ACAL procedure, as long as it works.

Seems like you want to divert the discussion away from Midis findings about the 3458A ADC to your favorite LTZ1000 hobby. But the question of ADC drift is different from the question of voltage reference drift. It is not academic, because the most sophisticated and linear ADC is useless once it does not work. With a voltage reference, we know how to fix that. If you want to contribute something useful, you can provide a log of ACAL 72 drift of your instrument, so we can see if it is nothing or 0.03 ppm/day.

Dear Dieter,
Random Walk obviously is understood as a drift in random directions, like up /down/left right, like the Brownian Motion, to designate it correctly.
The initial drift of a voltage reference shows no random, but a definite direction, i.e. for the LTZ1000 in most cases always in the negative direction.
So you probably have something different in mind.

I already contributed my CAL?72 stability data years ago, in another thread, and just again inside this 24h stability measurement, just in front of your nose.
You can easily estimate the max. 24h drift of U180 from that, it's well below these 0.43ppm/24h.

Please also verify your ideas about drift and ACAL of the 3458A against the recent statements of Adrian.

But anyhow, as you now finally are getting personal: 'I expected a physicist to know..' and '..want to contribute something useful..', it's now evident for me, and what I supposed the whole time reading your comments, that you just try to play bullshit bingo, here and in other threads. 

To follow the saying 'Don't feed the Troll', I will neither consider, nor answer any of your statements any more. 

best regards Frank

[Dr. Frank]

Hi Frank,

I stick to the 0.43 ppm spec, measured over the time specified (ok, when there is no drift after say 4 days, one can also stop there, provided stable ambient temperature). I check the meters continuously against a set of references (732s/4910), so no ACAL needed to determine the drift. Making an ACAL-based drift test every day does not supply meaningfull data imo (and as per spec).
When working with a 3458A, it is not about personal feelings or so, the instrument needs to meet its spec. The 0.43ppm is the only spec related to the A/D. That being said, a meter showing 0.4ppm/day (and I have not seen that) would be considered bad, altough well usable. You either have a drift of some ppms/day if bad (seen up to 20ppm/day), or in the 0.1ppm/day and below range if good (all averaged over the specified time).
For me the value of the 3458A is not to try to achieve sub ppm measurement uncertainties, even with a 2ppm/a low drift version (and most aged instruments today are in that range anway), if I need tight uncertainty I always use the meter in transfer mode. Thats most of the measurements when working with high accuracy stuff.

Adrian, thank you very much for these quantitative explanations.
Seem to agree with mine
Frank

[dietert1]

Amazing how engineers have difficulties with this. How can one say the limit is 0,43 ppm/day? If that was a continuous drift, after a year that would amount to about 157 ppm and technically speaking that ADC is junk. Such drift would be a clear indication that something is wrong.

...
Also don't go by SN18 criteria 0.43 ppm/day. It's not a good measure to test ADC stability. Some forum members will surely disagree on this, but I'd say anything with drift over 0.1 ppm/day is a bad/drifty ADC. 

If a potential buyer sees an ADC drift of 20 ppm on the first day or for some days, that case is simple. Maybe if he sees a total of 20 ppm drift after a week, after a month or after a year the conclusion should be the same. Apparently judging the state of a 3458A requires some patience.

Regards, Dieter

[Villain]

I feel like what you are saying dieter is that "a unit is drifting lets say 0.4 ppm/day which is roughly 150 ppm a year which is out of spec". That is the feeling i am getting here. But this is simply not the case because ACAL will take care of that and compensate for this drift (if something really is broken with the unit even ACAL will not be able to compensate and probably fail, then you send in your unit for repair, its quite simple). Specifications are only valid with last ACAL within 24 hrs and +-1 degree of current TEMP?. This will keep the unit well within spec. I am not saying that 0.43ppm/day is good, but the unit is fully working because it performs to its specifications and + does not violate the service note referred to here. That many people in the metrology section have tighter specifications or expectations in mind regarding to 3458A has nothing to do with the unit being broken if our expectations are not met.

I feel like i do not understand the point you are trying to make here (and i am not saying that is your fault, maybe i personally can just not grasp it right now)

I am not saying drift issue of 3458A is "no issue", i am just saying that a unit drifting 0.3ppm/day 0.4ppm/day is not necessarily broken.
If your point is that many units have U180 drift issues and you want to say "how can you buy such a product" then just don't buy it no one forces you to.

edit 150 per year obviously yes dieter, does not change my answer though

[dietert1]

No, 0.43 ppm/day of continuous drift translates into 157 ppm/year, not 15 ppm/year. Yes, ACAL may take care of that. With some luck, because an ADC that drifts that much, may have other problems, too. Bad resistor tracking of a thin film resistor array can not explain more than 100 ppm drift.

https://www.eevblog.com/forum/repair/agilent-3458a-repair-232680/msg3017320/#msg3017320

I found a CAL 72 log that user franklin posted in another thread in April and put his data into a diagram. There is no continuous drift but something like noise (probably temperature related). That log extends over one month and 30 days * 0.03 ppm/day = 0,9 ppm, that's definitely excluded.

Regards, Dieter

[TiN]

That many people in the metrology section have tighter specifications or expectations in mind regarding to 3458A has nothing to do with the unit being broken if our expectations are not met.

This! 

I use 3458's for days and days without ACAL, so 0.1ppm/day drift would drive me mad. That's a reason why I hunted down golden ADCs and polished instruments until they don't show any drift (just random noise/walk jumps) in days terms and don't have any tempco for DCV. Hence requirements are far tighter vs manufacturers spec. Obviously, I also monitor ADC constants on every ACAL between experiments to keep confidence that it's still all happy.



And A3 with large resistor ratio issues also have other problems which nobody talks about, other than ominous drift that's not very important to actual practical work

[Villain]

No, 0.43 ppm/day of continuous drift translates into 157 ppm/year, not 15 ppm/year. Yes, ACAL may take care of that. With some luck, because an ADC that drifts that much, may have other problems, too. Bad resistor tracking of a thin film resistor array can not explain more than 100 ppm drift.

https://www.eevblog.com/forum/repair/agilent-3458a-repair-232680/msg3017320/#msg3017320

I found a CAL 72 log that user franklin posted in another thread in April and put his data into a diagram. There is no continuous drift but something like noise (probably temperature related). That log extends over one month and 30 days * 0.03 ppm/day = 0,9 ppm, that's definitely excluded.

Regards, Dieter

Yes was typing to fast, forgot one zero there. Does not change my answer though.

That many people in the metrology section have tighter specifications or expectations in mind regarding to 3458A has nothing to do with the unit being broken if our expectations are not met.

This! 

I use 3458's for days and days without ACAL, so 0.1ppm/day drift would drive me mad. That's a reason why I hunted down golden ADCs and polished instruments until they don't show any drift (just random noise/walk jumps) in days terms and don't have any tempco for DCV. Hence requirements are far tighter vs manufacturers spec. Obviously, I also monitor ADC constants on every ACAL between experiments to keep confidence that it's still all happy.



And A3 with large resistor ratio issues also have other problems which nobody talks about, other than ominous drift that's not very important to actual practical work

Thats very impressive performance TiN!

[maat]

I feel like i do not understand the point you are trying to make here (and i am not saying that is your fault, maybe i personally can just not grasp it right now)

I believe what dietert1 is saying, is, that a 3458A showing 0.03 ppm/day linear drift is living on borrowed time. It can be brought back into spec using ACAL every day, but sooner or later that ADC will die and will have to be replaced. If you have a spare device and running on warranty - not a big deal if you use the device in a lab.

Since the 3458A is running 24/7 on assembly lines and test setups all around the word and there is not much off an outcry, I do believe the issue is not that big. I would rather assume that the devices on Ebay do show a tendency for failing ADCs. My ADC died as well and before doing so it also developed a horrid tempco of about 0.85 ppm/K (https://www.eevblog.com/forum/metrology/hp-3458a-ad-board-spare-part-availibilty-and-cost/msg3001606/#msg3001606).  (Note: I do not know if the tempco was there before. I saw the drift first, then dug into the matter.)  So there might be even more problems coming along with the drift, resulting in the device ending up on Ebay.

Back to dieters point: A lot of people on this forum bought their unit on Ebay and if their ADC is showing drift, they should be prepared to put down another 1.5k to have the board replaced by KS. So sum it up, I guess dieters point is not so much about the device not performing to spec, but rather about protecting one's investment.

[Villain]

I feel like i do not understand the point you are trying to make here (and i am not saying that is your fault, maybe i personally can just not grasp it right now)

I believe what dietert1 is saying, is, that a 3458A showing 0.03 ppm/day linear drift is living on borrowed time. It can be brought back into spec using ACAL every day, but sooner or later that ADC will die and will have to be replaced. If you have a spare device and running on warranty - not a big deal if you use the device in a lab.

Since the 3458A is running 24/7 on assembly lines and test setups all around the word and there is not much off an outcry, I do believe the issue is not that big. I would rather assume that the devices on Ebay do show a tendency for failing ADCs. My ADC died as well and before doing so it also developed a horrid tempco of about 0.85 ppm/K (https://www.eevblog.com/forum/metrology/hp-3458a-ad-board-spare-part-availibilty-and-cost/msg3001606/#msg3001606).  (Note: I do not know if the tempco was there before. I saw the drift first, then dug into the matter.)  So there might be even more problems coming along with the drift, resulting in the device ending up on Ebay.

Back to dieters point: A lot of people on this forum bought their unit on Ebay and if their ADC is showing drift, they should be prepared to put down another 1.5k to have the board replaced by KS. So sum it up, I guess dieters point is not so much about the device not performing to spec, but rather about protecting one's investment.

Well, i agree that for sure you have to be careful where/from whom to buy your 3458A. Or just buy a random one and put money on top for possibly new A3 board etc.

[martinr33]

But that was in the days when you could get a replacement A3 board.

On the subject of drift, the ADC board is not really part of the specification because of the 24-hour autocal requirement, which can mask a lot of drift.

I say that becuase all of the 3458A accuracy specifications require an autocal within 24 hours before the reading.
The ADC board would have to be truly horrible to drift by enough to be outside of the 2.6ppm post 24 hour specification.

It is only when watching precise references, without autocal,  over a long period that you start to see these errors. 

Also - whan a 3458a is sitting in a semiconductor test rack, I suspect it is using its other skill - super fast lower resolution measurements. The ADC errors won't show up there.