Author Topic: Are the 3458A's low current ranges calibrated? The spec's say typical.  (Read 4793 times)

0 Members and 1 Guest are viewing this topic.

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Hi,

 I'm new here and have a question on the 3458A low current ranges. The specifications list the 10µA, 1µA, and 100nA ranges and give accuracies but have a little 6 next to the range number and that note says "Typical". Also the certification doesn't include those ranges. My question is can I use those ranges and if so what would be the accuracy for each be? I've seen those ranges on scopes of accreditation so I would think that question has come up. Thanks for any information.
« Last Edit: November 13, 2018, 09:31:02 pm by 2874 »
 

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #1 on: November 13, 2018, 09:22:26 pm »
I've attached a Keysight cal cert showing the current ranges
 
The following users thanked this post: Vgkid

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #2 on: November 13, 2018, 11:19:24 pm »
The question is, what "calibration" means..

For the 3458A, this is a two step process, you apply 10V and 10KOhm, and that's enough to adjust all other modes and ranges to the 24h specification.
There is no reason, why these low µA ranges should not be as well adjusted, and meet their according specification, not only as typical.

The term 'Calibration' has the background, that you verify all these modes and ranges against another standard, having  equal or better lower uncertainty, as documented in the calibration report, you've mentioned.

On first page, you'll find a Fluke 5720A or similar, which also is an AUTOCAL instrument.

But these calibrators do not have appropriately accurate source currents of 10µA down to 100nA, so it's not possible to verify and calibrate the 3458A in these ranges.. The specification originates from 1988, so there'*s even lesser chance, that an appropriate current standard existed. (Don't remeber, if the 5700A was already available.

So you have to live with typical uncertainty only.

The only chance to improve the specification, would require a more precise way of measuring these ranges, e.g. by a CCC, and by validating a certain amount of 3458As beforehand

Frank
 

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #3 on: November 14, 2018, 01:47:09 pm »
Thanks for the reply and your insight on these instruments. I'm about to start the process of accreditation for a calibration lab and need to define the scope. I've seen other scopes of accreditation that list a 3458A being used at these low current ranges and was wondering how they justified that to their assessor if the manufacturer doesn't give a specification or verify those ranges. My understanding is they needed to have an uncertainty from the manufacturer or calibration vendor to determine their expanded uncertainty for their scope of accreditation. Thanks again for any information and help on this.
 

Online splin

  • Frequent Contributor
  • **
  • Posts: 999
  • Country: gb
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #4 on: November 15, 2018, 12:01:14 am »
But these calibrators do not have appropriately accurate source currents of 10µA down to 100nA, so it's not possible to verify and calibrate the 3458A in these ranges.. The specification originates from 1988, so there'*s even lesser chance, that an appropriate current standard existed. (Don't remeber, if the 5700A was already available.

So you have to live with typical uncertainty only.

The only chance to improve the specification, would require a more precise way of measuring these ranges, e.g. by a CCC, and by validating a certain amount of 3458As beforehand

Frank

Hmm, I don't buy the lack of suitable calibrators explanation. The use of a multifunction calibrator is a convenience not a requirement - HP could very easily have created their own reference current sources for factory calibration. Presumably they didn't bother because they believed so few customers were sufficiently interested in those current ranges being calibrated, to justify the extra production cost.

Personally it seems a bizarre decision to not even provide wider 'guaranteed by design' limits (due to auto cal), having gone to the expense of providing those ranges. Stating 'typical' means nothing at all without any specified statistical data/confidence levels - the customers are left with the task of calibrating / characterising those ranges themselves by whatever means.

A 10uA reference should be almost trivial with nominal a 10V source and 100k ohm resistor given the 3458A’s 24hr 2.2ppm 1V and 1.8ppm 100k ohm specs, compared to the 17ppm spec for 10uA. The lowest ranges might be a bit more challenging, but then the accuracy specifications get much wider.

PS. Why the odd choices of burden resistors for the ranges below 1mA – eg. 730 ohms for the 100uA range rather than 1k?

 
The following users thanked this post: e61_phil

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #5 on: November 15, 2018, 06:40:47 am »
"Odd" burden resistor values are due to use of all higher range resistors in chain to provide total divider ratio, not just single resistor.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #6 on: November 15, 2018, 08:27:07 am »
The shunts used for the low current ranges are also used in the internal ACAL procedure for the higher (e.g. 100 K, 1 M) ohms ranges. So they had to include those low current ranges anyway.
So given that there are specs for the higher Ohms ranges, not guarantying the low current ranges is odd. They kind of have to work to make the high ohms work. There might be still leakage current from the input protection that are not included in ACAL use. 

Even when considering the chain of resistors, the values are a little odd values. This could be to minimize errors due to self heating. A fixed 100 mV burden voltage is not the best choice and the ACAL case might use a higher current than normal use.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #7 on: November 15, 2018, 10:55:35 am »
I think what helps is a look at a real calibration report of dc current on 3458A. So attached is a snipped of our report.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #8 on: November 15, 2018, 05:10:50 pm »


Hmm, I don't buy the lack of suitable calibrators explanation. The use of a multifunction calibrator is a convenience not a requirement - HP could very easily have created their own reference current sources for factory calibration. Presumably they didn't bother because they believed so few customers were sufficiently interested in those current ranges being calibrated, to justify the extra production cost.

Personally it seems a bizarre decision to not even provide wider 'guaranteed by design' limits (due to auto cal), having gone to the expense of providing those ranges. Stating 'typical' means nothing at all without any specified statistical data/confidence levels - the customers are left with the task of calibrating / characterising those ranges themselves by whatever means.

A 10uA reference should be almost trivial with nominal a 10V source and 100k ohm resistor given the 3458A’s 24hr 2.2ppm 1V and 1.8ppm 100k ohm specs, compared to the 17ppm spec for 10uA. The lowest ranges might be a bit more challenging, but then the accuracy specifications get much wider.


I append a document, how HP tried to verify the Artifact Adjustment by themselves.. DCI on page 4.. similar to what you propose.
Maybe that you could really demonstrate the accuracy, but it's probably not sufficient for a traceable verification / calibration.

To require a 5700A for verification of the other ranges / modes, allows to reproduce this process in any other calibration facility, not only in a very specialized manner in a special hp location.
Mention also, from the hp document, and from more elaborate 3458A calibration certificates, that a 'good' traceable calibration really requires: 'As Found' - Adjustment - 'As left', so the use of a calibrator is a necessity, not only for convenience.

HP may also not have trusted the traceability of these ranges, and Keysight also does not trust the reliability of these ranges, either.
For the 34465A /470A, they also specify 'typical' accuracy only, even for their 100µA range!


A further argument is the observation, that no similar precision instrument on the market has specified traceable DCI ranges of lower than 100µA, like the Fluke 57xxA and Wavetek 4808 calibrators, Datron 1281, Fluke 8508A, KEI 2002, PREMA 6048, and the 3458A.
They mostly have a lowest range of 100 / 200µA, or the uncertainty is undefined.

Your idea of 'simply' using a precise 100k Ohm resistor and 1V measurement also might not do the job, due to additional errors, and again - not really being traceable.

Instead, you might send the 3458A to an NMI for traceable calibration of these ranges, according to the typical specification.
The capabilities can be found here: https://kcdb.bipm.org/appendixC/search.asp?reset=1&met=EM

That's a complete overview for all countries, just chose UK (NPL) and U.S. (NIST), and open their calibration tables.
They do not even offer DCI calibration at all.

Germany (PTB) and Suisse (METAX) offer that, even low current measurements.
PTB uses a shunt / DCV technique, as proposed.



As of today, they can achieve 3, 20 and 60 ppm for 100µA, 1µA and 100nA .. so that's really very tight, or not that trivial, to verify the claimed typical accuracy of the 3458A .. so probably no chance at all for HP themselves, any other lower grade cal lab, or even the end user to verify / calibrate these ranges precisely.
And don't forget, that the design of the 3458A is dating back 30 years.. as the discussion and verification oft Artifact calibration

I also refer to the discussion about the 5700A II calibrator. Its Artifact Calibration was quickly accepted in the U.S. , but not in the EU.
That's described in the FLUKE document from 1998.

It's referring to a 1996 verification study by PTB, SP and NMi, which gives a good idea, how Artifact Calibration works, what the problem is, and how it is verified..
Just search for: "Artifact Calibration: An Evaluation of the Fluke 5700A Series II Calibrator", it's hosted on the FLUKE site (91 pages).

So, noise and leakage currents might prevent these low current ranges to be easily adjusted by Artifact Calibration.

Frank

« Last Edit: November 15, 2018, 05:14:39 pm by Dr. Frank »
 
The following users thanked this post: e61_phil

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #9 on: November 15, 2018, 06:35:23 pm »
Thanks for the articles and explanations, Frank!

But I don't agree in "not sufficient for a traceable calibration". My understanding of traceability is a unbroken chain of uncertainties. And that would be possible with the mentioned resistor method. Surely you have to account for many things (as always if you build your equation for uncertainty calculations). But in the end you can end up with a traceable calibration. It might be the case, that the uncertainty is really high, but it is traceable. There is no "good enough" involved in the uncertainty calculation.

Additional point: In many documents HP says one can use a 3458A opt 2 within 90day spec to verify another 3458A against 1year specs. Therefore, the argument with the missing reference standard (for example Fluke 5700A) doesn't count imho.
« Last Edit: November 15, 2018, 06:41:26 pm by e61_phil »
 

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #10 on: November 15, 2018, 07:10:31 pm »
Thanks for the reply and your insight on these instruments. I'm about to start the process of accreditation for a calibration lab and need to define the scope. I've seen other scopes of accreditation that list a 3458A being used at these low current ranges and was wondering how they justified that to their assessor if the manufacturer doesn't give a specification or verify those ranges. My understanding is they needed to have an uncertainty from the manufacturer or calibration vendor to determine their expanded uncertainty for their scope of accreditation. Thanks again for any information and help on this.

There are specs for 10µA (100 ppm), 1µA (820ppm) and 100nA (0,8%) if you use the calibrated 100µA range ;)
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #11 on: November 15, 2018, 07:28:30 pm »
Thanks for the articles and explanations, Frank!

But I don't agree in "not sufficient for a traceable calibration". My understanding of traceability is a unbroken chain of uncertainties. And that would be possible with the mentioned resistor method. Surely you have to account for many things (as always if you build your equation for uncertainty calculations). But in the end you can end up with a traceable calibration. It might be the case, that the uncertainty is really high, but it is traceable. There is no "good enough" involved in the uncertainty calculation.

Hello Philipp,

For sure it's possible to get traceability here, as PTB claims and specifies. Though process description, calculus and proof is not available, especially not for artifact calibration. So that is a theoretical claim only.

Instead, do you have a better idea, why hp did NOT succeed in specifying a traceable uncertainty? That would be more helpful, maybe also to make up a method to verify that on our own.


Another miracle is, why hp did not specify the Transfer Accuracy for the Ohm ranges, but for FLUKEs 3458A/HFL, it has been done (about 0.2ppm, if I remember correctly, and what can be measured by e.g. measuring StD). And there are several other oddities in the 3458A specs.

Maybe that's the real culprit, that hp/AGI/KS did not use proper metrological methods to characterize their instruments, although the 3458A is advertised as being metrology grade. (including the 95°C LTZ1000 reference  :-- )

Fluke may be much better in that aspect, as they are more inside metrology, than hp.

Frank
« Last Edit: November 15, 2018, 07:49:53 pm by Dr. Frank »
 

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #12 on: November 15, 2018, 07:59:12 pm »
I think (and that is only speculation) there are many things driven by the marketing department, like nowadays. Perhaps there is something in the uncertainty expression which gives really bad specifications under bad circumstances and they can't gurantee that will not happen on a few instruments. Or it is like you said and there is/was no "cheap" way to calibrate these ranges and marketing says "we don't want an instrument that is expensive to calibrate". I don't know how companies have worked in that time, but nowadays many product decisions/specs come from marketing and not from R&D.

Please, don't get me wrong, but I think many things on that instrument are done to reduce costs and to be cheaper than the 1281 for example. The whole ACAL stuff is a nice ADC stunt, but in the end it is only to reduce the costs of the parts imho. Proper parts (like in the 8508A) are expensive. Therefore, they decide to rely only on two good references and derive everything else from there. I also always wondered what benefit the artifact calibration has for a DMM? You need proper sources for the calibration anyhow and it will destroy the history of the whole instrument if a single range is out of spec and you need to adjust the meter.

By the way: It should be easy to test how stable the lower current ranges are. One can run severeal ACALs and monitor the calibration values of that ranges.
« Last Edit: November 15, 2018, 08:04:04 pm by e61_phil »
 

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #13 on: November 15, 2018, 08:21:54 pm »
I think I'm going to go through my procedures and make sure the 3458A is not used on those ranges. I know of two that I've automated, Keithley 2410 and 2400.
 

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #14 on: November 15, 2018, 08:29:19 pm »
I think I'm going to go through my procedures and make sure the 3458A is not used on those ranges. I know of two that I've automated, Keithley 2410 and 2400.

The 10µA spec of the Keithley SMUs could be verified in the 100µA range with a TUR of ~5, I think. But for 1µA it is only ~1:1.
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #15 on: November 15, 2018, 08:51:39 pm »
I also would welcome a combination of both the 3458As ACAL and the 8508A more stable components, its MUCH better Ohm mode (TruOhm), and its direct comparison feature.

In one of the documents about the 5700A, there's a remark, that AutoCal will bring the instrument back to its 24h specification, apart from the 10V / 10k standards drift.

This is a true advantage, and no stunt, as this can be observed in the specifications.

The uncertainty of the different ranges of a given mode of the 8508A  diverge over time, but the 3458A do not.
Therefore, these better components of the 8508A effectively make it more stable on short to mid term basis, and over temperature, but not necessarily on a long term basis.

If the 3458A only would have a more stable LTZ1000 and VHP101 / 40k reference, it would be nearly on par with the 8508A on the 1yr. specifications, w/o having these stable components.

Frank
 

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #16 on: November 15, 2018, 09:08:12 pm »
ACAL is only a benefit if the range itself is less stable than the "main" reference + transfer from the main reference to the specific range.

I don't miss the ACAL on the 8508A. If one looks on the 1year specification of 8508A the ranges don't diverge. There is 1.5ppm from 200mV to 20V. 0ppm from 2V to 20V and 1.5ppm from 200/1kV to 20V. The 3458A specs are more or less the same (only looking on difference between the ranges). Therefore, it seems to me that ACAL isn't needed for DCV. And for other ranges the improvement might be even less.

ACAL on the 5700A isn't the same as ACAL on the 3458A. If Fluke says ACAL will bring the 5700A back to the 24h specs, they mean using a 732x and the two 742A to adjust the calibrator. For that, the 10V and 10k internal drift doesn't matter. That is also true for the 3458A (10V and 10k external, bring it back to 24h spec). But you have to verify that and you will destroy/disturb your cal history.


I think ACAL would make the 8508A worse than it is. If one looks into the transfer spec of the best range (10V) the say 0.05 + 0.05ppm. That means a transfer from 10V to 1V has an uncertainty of (0.1ppm + 0.55ppm = 0.65ppm). You need this transfer twice from 10V to 1kV, but the 24h distance between 20V and 1kV on the 8508A is only 0.5ppm.
I often heard from metrology people that they don't use ACAL on the Fluke 57x0A and only calibrate the calibrator instead. Because the calibrator is more stable than the transfer.


An ACAL based CAL check without changing the calibration on the 8508A would be nice. Like it is done in the Fluke 57x0A.
« Last Edit: November 15, 2018, 09:21:13 pm by e61_phil »
 

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #17 on: November 15, 2018, 09:56:21 pm »
I guess I can get a standard resistor and calibrate those ranges.
 

Offline 2874Topic starter

  • Newbie
  • Posts: 7
  • Country: us
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #18 on: November 16, 2018, 01:34:19 pm »
I called Keysight to see if they have a optional calibration of those ranges, they don't offer that. I spoke with their technical assistance and was told it's been like that for 30 years.
 

Offline alm

  • Super Contributor
  • ***
  • Posts: 2840
  • Country: 00
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #19 on: November 16, 2018, 02:54:13 pm »
I guess it makes sense that if the designers were unable to specify the tolerance of the low current ranges for whatever reason, the service department aren't going to do that either. I imagine you could get a certified measurement (at time X, this instrument indicated 1.23 uA when connected to an 1.22 uA +/- X ppm current source) at a third party cal lab.

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #20 on: November 16, 2018, 04:37:00 pm »
But even if he gets a certfied measurement of the ranges, without specs it is not usefull for further calibrations.
 

Offline alm

  • Super Contributor
  • ***
  • Posts: 2840
  • Country: 00
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #21 on: November 16, 2018, 06:17:04 pm »
I agree, but it is the best you'll be able to get from a cal lab without manufacturer support. A history of these may help towards characterizing the current ranges. Whether that's worth the effort over acquiring a good quality shunt resistor is a different matter.

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #22 on: November 17, 2018, 12:09:22 am »
If Fluke says ACAL will bring the 5700A back to the 24h specs, they mean using a 732x and the two 742A to adjust the calibrator. .... But you have to verify that and you will destroy/disturb your cal history.

Not always, you still can run ACAL on 57xx and preserve history. You just perform ACAL to obtain calibration report, store report externally on PC/paper, but do not store it into calibrator's NVM, so on power cycle/reset original coefficients still used. Calibrator have three copies of calibration in it. Only issue is that AC/AC transfers are not traceable (flatness calibration sensor is used as reference, but not tested against external reference), so for ACV/ACI you still need run "traditional" calibration versus external reference (5790A or 792A + A40 shunts).
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 
The following users thanked this post: e61_phil

Offline e61_phil

  • Frequent Contributor
  • **
  • Posts: 962
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #23 on: November 17, 2018, 07:06:44 am »
If Fluke says ACAL will bring the 5700A back to the 24h specs, they mean using a 732x and the two 742A to adjust the calibrator. .... But you have to verify that and you will destroy/disturb your cal history.

Not always, you still can run ACAL on 57xx and preserve history. You just perform ACAL to obtain calibration report, store report externally on PC/paper, but do not store it into calibrator's NVM, so on power cycle/reset original coefficients still used. Calibrator have three copies of calibration in it. Only issue is that AC/AC transfers are not traceable (flatness calibration sensor is used as reference, but not tested against external reference), so for ACV/ACI you still need run "traditional" calibration versus external reference (5790A or 792A + A40 shunts).

Interesting!
Nevertheless, there is no ACAL instrument which brings itself back to 24h specification without known external standards.
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: Are the 3458A's low current ranges calibrated? The spec's say typical.
« Reply #24 on: November 17, 2018, 08:50:23 am »
For the low current ranges on the 3458 the amplifiers input current and the leakage current from the current input protection can be significant. These are not checked with the ACAL procedure (not absolutely sure with the amplifier). So a stringent uncertainty estimate for the low current ranges would  give a rather high worst case value.  It does not make much sense to give something like a 1% limit if the typical accuracy level is much better. The 1 µA range even uses the internal 40 K reference resistor as the main shunt - so for the internal procedure it must be stable and is even the best one (though used with a higher burden and thus not 1 µA but more like 10 µA). I think it's just difficult to get top notch performance in the µA range with a protection device good for some 10 A. The low current ranges also use a reduced burden voltage - this makes sense to reduce the effect of the protection.

For the ACAL procedure it depends on which ranges / functions to apply it. The ACAL72 step can be rather accurate, as it use some 70% of the ADC range.  AFAIK some other meters also use this step. The lower / higher voltage ranges are more difficult, as they only use 10% FS.

The part with amps and ohms is even more tricky, as the voltages involved are small (some 10 mV in case of 10% of current range) or there may be extra self heating. The 100 mV range is in some aspects limited by the amplifier and not so much the ADC. So 10 mV in the 100 mV range are worse than 100 mV in the 1 V range.  So with the amps ranges the ACAL way is primarily a way to reduce costs, and better shunts could give similar to better performance. It might still be a good thing to compare the ranges to each other - as an extended self test and maybe to simplify calibration / adjustment if needed.

If used in calibration process, it might be a good idea to also log the constants determined by the ACAL procedure. Just to make sure that worst case one could apply corrections and as an additional check. Any larger change in these parameters should be a warning for a critical instrument.
 
The following users thanked this post: e61_phil, alm


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf