AC current calibration - a couple of questions

[binary01]

Hello all,
I'm hoping a few calibration experts might be able to offer some advice on calibrations of AC current.  I have performed basic calibrations in DCV and DCI for some years, but I am trying to expand my capability in to ACV and ACI.  I have found good literature on ACV and corresponding uncertainty terms, however I have found it harder to find good information on ACI or examples of robust uncertainty budgets.

Some questions I have are:
- Does anyone have any guidance on how to estimate uncertainty associated with leakage and AC impedances such as lead inductance?  My NMI provide guidance on how components such as lead capacitance load/impact the ACV case and complete uncertainty budgets (Keysight also provide some info, i.e. https://www.keysight.com/au/en/assets/7018-01084/application-notes/5988-5513.pdf), but I can't find an equivalent calculation for the ACI case.  I am also unsure how to estimate any potential current leakage for the ACI case.  I have considered measuring the voltage across a shunt placed in series at the calibrator "high" side and then repeating with the shunt placed on the "low" side to quantify any differences - is this a reasonable approach or is there a better method?

- Perhaps linked to insufficient knowledge of the above questions, but I have tried measuring the ACI output from a calibrator using a single DMM, and then also using two DMMs in series.  When I use the DMMs in series, the second DMM in the circuit always reads higher by some small offset compared to the first.  It is as if the first DMM injects some current in to the circuit, but I'm not sure what causes this.  My very basic understanding is that the current should passed through a current shunt inside each of the DMMs, but I'm not sure why current would be injected in this process.  Can someone explain what causes this, how to avoid it, or if it is an unavoidable problem when two DMMs are used to measure ACI in series?  I haven't observed this problem in DCI measurements using two DMMs in series (or parallel DCV/ACV).

Thanks for any advice in advance.
Regards, Binary

[jonpaul]

Barry:

A few questions please so we may better respond:

a/ What frequencies, a 50/60 Hz test will be very different than a 1 kHz or 1 MHz test!

b/ What current magnitudes eg 1 mA , 1 A or 100 A?

c/ Resolution and accuracy of you meter or calibrator: 4 1/2 5 1/2 6 1/2 digits?

Shielding, guarding, magnetic fields and cable type affect will be different for different applications.

Kind Regards,

Jon

[Kleinstein]

A mains powered meter will usually inject some AC current from capacitice coupling. This current would go the the meter on the lower side, as the calibrator as current source should have a higher impedence.
With some meters one can reduce the injected common mode signal by using a guard  terminal, but not all meters offer this and even than, no guard is perfect.

[binary01]

Thanks Jon - a few answers below.
I am testing over the range 10 Hz to 10 kHz, 0.1 mA to 20 A
I am using a Fluke 5522A and experimenting with a variety of DMMs 6.5 to 8.5 digit.  I am only nominally targeting uncertainties around 0.1% for AC current, at which level lots of the small unkowns, such as the injected current to the second DMM are not really significant, but I would still endeavour to understand or reduce them.
For basic connection to the calibrator in AC, I normally use Pomona 1167-24 or Pomona 2BC leads, but when using two DMMs in series these leads don't really work, so I have been using Pomona 1756 series in these instances.  I have also tried with my normal "DC" test leads, which are AB-Precision LEC-1, for low currents which seem to give the same numbers within my ability to determine.


Thanks Kleinstein.  In this instance the calibrator and both DMMs are mains powered from a common outlet.  The 8.5 digit DMMs (Fluke 8588A) have selectable internal or external guard, which I have tried a variety of approaches.  Could you recommend an approach for guarding, for example connecting 2 x 8588As in series from a 5522A?

Ultimately it may be simpler to compare the DMMs indirectly, by just characterising the 5522A output with one DMM only, disconnect it, and repeat the test with the second DMM within a short time, with some allowance for very short term stability of the calibrator.

[ap]

Besides the items you are using and the overal frequency/amps range you specified, I am still not clear what you are trying to do. Calibrate the calibrator, calibrate the DMMs, compare one calibrated to a non calibrated DMM?
Why would you connect two meters in series? Maybe some more thoughts of why you did what you did would help.

Generally speaking it is not a good idea to connect two precesion meters to a source at a time and expect a correct result. The setup may cause interferences, and you cannot exactly determine what (parasitic) effects have what impact. That even applies to DC. If you want to do standard or even slightly improved uncertainty measurements (better than your stated 0,1%) and use an appropriate meter/calibrator you can stick to the specs and your calibration certificates if you use a standard setup (good quality and short coaxial cables). Usually with good meters and in the lower frequency ranges that can give you better than 0.1% (see specs). Within the uncertainty and frequency range of the calibrator/meter used, you get the related results by applying the relevant spec-sheet/cal-certificate data. Always provided your environment does not have strong sources of dicturbance, e.g. magnetic fields. But you can check for that.

If you want to go beyond, the setup involves characterised AC shunts (and AC meter). Holt is e.g. known for that, and at high currents and lower frequencies e.g. a Ultrastab current transducer will give you low uncertainty. If you want to build shunts yourself, that gets complicated, but you can achive good uncertainties at high frequencies as well. See e.g. https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=31690 . Other similar docs cover alternative designs. Baseline is to develop a calculable standard with low drift rates and low reactance. Lots of work needed, so buying off-the-shelf units is maybe the better option for you, but costly, depending where your preferences are. You could also have a look at some older calibrator cal manuals for the Datron/Wavetek 4708/4808 calibrators, available online, where some setup details are descibed. Vague answers to a vague question.

[binary01]

Hi ap,
I should clarify that my first question relates to the uncertainty estimation for the typical case, using a calibrator connected to single DMM or DUT, and is not specifically related to the second part.

Using two DMMs simultaneously is not something that I am doing routinely or during calibration, but is something that I have experimented with purely for the purpose of comparing and cross-checking my two 8588As.  They are calibrated externally at different times during the year, and when one comes back from calibration I like to use the fresh calibration to check drift of the remaining unit.
My observation has been that parallel measurement of DC voltage has worked well for this, as has series measurement of DC current although for the bulk of the DC current scale I use DC shunts in series with a DMM.  My focus and lowest uncertainty remains in DC, and I have found this method useful enough.
For AC current in particular, I have found the stronger differences when two DMMs are used in series.  I will likely abandon this approach and instead measure with one DMM and then the next DMM seperately in quick succession to avoid any interaction but still allow comparison.  I wondered if the cause of interaction might be something that could be easily explained or reduced with the advice here, but the impression I am getting is that I should just avoid it when a high quality comparison is needed.  Although the 8588A is capable, I don't expect to ever need to work to better than 0.1% in AC current.
I'm sorry that my questions were vague but thanks for your information nonetheless.
Regards.

[ap]

When you compare two meters you should always (AC but also for DC!) use a single connection. E.g. if you have a recently calibrated meter, use this as the comparison reference, measure the current of the calibrator, and then connect the calibrator to the DUT meter, transferring the MU of the calibrated meter to the DUT meter. In any parallel connection, you never know what parasitic issues you have, even if in some cases it may work out well. It is just not good practice.

[RYcal]

Hey Binary01

We tried this with two long scale DMM's (8508A's) to determine lead effects and loading effects we essentially connected one meter and took a series of readings then connected the second same make model meter (same leads same length) and observed the change to give us some ideas of what the DMM and leads were doing to the Fluke Calibrator. It worked reasonably well but I still wasn't 100% convinced the results were accurate.

I to would like a decent (simple) publication/guide about how we go about connecting say a Fluke 8588A to Fluke 5522A to take advantage of the guard and earth posts to reduce errors i.e what should go where and what the instrument should be setup like, Which instrument should have the "guard" activated etc etc.

Can I ask where you get your 8588A's and 5522A's calibrated?....

Regards

Nathan

[TiN]

I'd agree on using REF DMM => characterize calibrator => use calibrator to verify DUT DMM approach. However 5522A even freshly calibrated does not have good enough overall performance to meet low-period 8588A targets, to be certain in results.

Something like 5720A or fixed standard setups much more suitable to the task. There is suitable paper from Fluke showing 5522A vs 8588A aspect better.

[binary01]

I to would like a decent (simple) publication/guide about how we go about connecting say a Fluke 8588A to Fluke 5522A to take advantage of the guard and earth posts to reduce errors i.e what should go where and what the instrument should be setup like, Which instrument should have the "guard" activated etc etc.

Can I ask where you get your 8588A's and 5522A's calibrated?....

Hi Nathan - yes a guide like you describe would be ideal.  So far I have been referencing documents like the 5522A service manual (and other calibrator service manuals) and the 8588A settings defined in each section of its calibration report to develop a strategy.  I have also refered to the Fluke Calibration Philosophy in Practice book, but the guidance in this book is naturally very general.  If anyone could recommend a good detailed guide on guarding, shielding and grounding, please let me know.

For recalibration, for the 8588A the uncertainty requirements are too challenging local labs, except perhaps from our NMI, so at this stage it goes back to its birth place, Fluke UK.  The 5522A is less than 1 year old and I am still undecided as to whether to also send it to Fluke UK for consistency, or if one of the higher level local labs would be suitable.  The 8588A is my best traceability to SI, so it is my most important external calibration, and I can use it to check my 5522A well at most points.  Who do you use for your external calibration?

I'd agree on using REF DMM => characterize calibrator => use calibrator to verify DUT DMM approach. However 5522A even freshly calibrated does not have good enough overall performance to meet low-period 8588A targets, to be certain in results.
Something like 5720A or fixed standard setups much more suitable to the task. There is suitable paper from Fluke showing 5522A vs 8588A aspect better.

Thanks TiN.  Yes, I agree that the 5522A is no match for the ultimate performance of the 8588A, especially its transfer/24 hrs specs (<1 ppm on DCV!).  I would love a 57xx series, but they don't come cheap, and in reality I am not working at a high enough level to justify it.  I can always dream...
In my application, I am working at a level that is broader than the 1-year 8588A specs on all ranges, and I am just checking some key points on each range between my two DMMs to get some confidence that the "older" meter is still within drift allowances, and doesn't have some undetected fault. At this broader level, the 5522A short-term stability doesn't seem to be a big limitation in the cycles of checking that I have performed so far, but its early days.  My feeling is that is stable to within a few ppm on DCV and DCI (except for higher currents) within a very short period.  Is this consistent with your experience?
For true drift calculation I am relying on the lower uncertainty external calibrations.

Thank you for the link to the 5522A vs 8588A document - I have read that document and found it helpful when deciding what reference equipment to buy, and determining a checking strategy for my 5522A.
Although it's certainly not a 57xx competitor, at 10 months old the 5522A long term stability is actually looking quite a bit better the Fluke specification on DCV and DCI, which I am pleased about.

[TiN]

Well, I feel like you are running this not for craze home science experiment but as a business, so maybe buying used 5720A is an option I've seen some go for $10k-ish levels which is a fraction of 8588A cost. Otherwise it seems that you just replacing external calibration cost of second 8588A with your own labour and equipment downtime cost.

And please, lets talk numbers and actual data? If you can reproduce data you observe and agreements between instruments, it would be educational for interested public, including myself. There is little information available or datasets available about 8588A and even old 5522A. Maybe I can obtain some similar datasets from this grade instruments and we could compare the results.

I am still very firm on idea that Keysight 3458A is unrivalled for DCV performance, no matter how many times Fluke marketing says "best multimeter in the world" in their press-releases. I did not have any experience with 55xx series MFC but use two 5720A's daily and can share or run anything with them later. Also I know quite a number of enthusiastic people who own 57xx series units and long scale DMMs to collect larger datasets.

[bdunham7]

For AC current in particular, I have found the stronger differences when two DMMs are used in series.

When you said the 'second' DMM in a series circuit, I had to wonder which one would be 'second' in an AC circuit.  Perhaps you could draw a diagram or schematic of how your connections are set up, including guards?

Can you tell us exactly how much difference you see between the two meters at given currents and frequencies?  Unlike parallel voltage measurements, I don't see any real theoretical issue with measuring current in series as long as the combined burden voltages are below the compliance voltage of the calibrator.  I would think that any effects that cause the readings to vary should be included in the rated uncertainties of the meter, as similar effects will occur in actual measurements.  What those effects are I don't know and perhaps some specific data might give us more to think about.  The first thing I can think of is to sort of divide up the possibilities into two types--those that actually cause the current in the shunt of each meter to be slightly different and those that cause the meters to read the exact same current slightly differently. 

Have you tried selectively changing the polarity of the leads at each meter and at the calibrator to see what effect it has on the difference between readings?

Also, just curious--if you are having the 8588A calibrated, how old are they and what type of actual performance (drift, etc) are you seeing?  When did those come out--it has only been a year or two, hasn't it?

[RYcal]

Binary01

As my instruments (5522A and 8588A) both arrived from my suppliers with traceable accredited calibrations I have yet to send them out for re-calibration, But I'm picking the 8588A will go back to the UK/USA and the 5522A may end up in a lab somewhere in Australia.

I have written a nice little MET/CAL procedure to compare both instruments and put the data into an excel template that calculates errors and % of spec for the 5522A, I plan to run this procedure often to track any drift and appease my auditors.

I am now curious as to what lab you are working in in Australia.

Cheers

Nathan

[Kleinstein]

It absolutely makes sense to use the 8588 to verify / check the 5522 (+ the used cables for AC) performance. It would be no way good enough to calibrate a 8 digit meter, but should be OK to get something like the short term stability of the 5522. Especially for AC it makes sense to compare meter to meter and thus a direct substitute of reference and DUT.  It is more like the exception for DC that one can calibrate a voltmeter to an accurate voltage source.

There is slight downside though - the 8588 is still relatively new and thus not much experience on how well it handles the stress on transport. Not sure how much it is made as a traveling standard. It would probably need a few cal / transport cycles to gain a little more confidence here. Ideally one would need a 2nd relatively stable (for the time the unit comes back from cal.) reference that does not move, so one could  see the changes caused by transport.

[binary01]

Wow, lots of feedback today.  Thanks all.

TiN - you are right that the use is for business, but my questions are largely out of personal curiosity, so I guess it is partly crazy experimenting!  If you come across a decent 5720 in the $10k-ish region again, let me know     I have had a pertpetual eBay search on 5720s (and 5700 EP3) for ages, but I haven't seen them cheap enough unless they have some errors.  I don't have the confidence to build one up like you have done - impressive stuff.

I'm away on holidays this week (bad timing for this thread!), but I'd be happy to put some numbers up and collect more data when I get back to the lab. I don't have any sources with better stability than the 5522A, so I might not be able to give significant data on the 8588A performance, but I could collect and present some data on the 5522A stability if that would be of interest.  It sounds as if RYcal is also setup well to collect this data with the same models.

Regarding 3458A vs 8588A - to be honest I was dreaming of 3458A for many years, and I understand it is still the DCV gold standard on EEVblog.  However, when it finally came time when I had a budget available, the 8588A (at least by its specification) seemed better suited to my application. For example, I mostly support the process industry where 4-20 mA is an important meaurement for low uncertainty, and the 10mA range on the 8588A is ideal for this, where the 3458A would jump up to 100 mA range.  If DCV was my primary focus, the 3458A would still be my pick based on the opinion in this forum and the high regard given by NMIA when I attended a course their a few years ago.

bdunhum, in the two DMM scenario I am considering the 1st DMM to be connected closest to the high terminal of the calibrator, and 2nd connected closest to the low terminal.  I have tried many combinations of guarding and grounding, such internal guard on both DMMs, external guard connected to calibrator guard, and other combinations in between.  The guarding does impact the comparison.  I note that Fluke calibrate them with internal guard on ACI, and this does offer the closest comparison between the DMMs.  From memory, the difference is perhaps .07uA at 0.1mA 50Hz, with the second meter always reading high.  When I get back to the lab next week I will look at the data again across a few currents/frequencies and provide more data.  It is most obvious at lower currents where the resolution makes it easy to observe.  I have swapped the DMMs between the 1st and 2nd position, and confirmed it is the position in the circuit rather than the DMM.  I am of the opinion that the cause is that the second DMM does actually have more current flowing across its shunt, rather than it reading in error, but I don't know why/how.  Anyway, I'll probably just pack-up this experiment as time wasted, and stick with the characterisation approach does provide a excellent agreement between the DMMs.

RYcal - if you establish any short or long term stability info on the 5522A (or 8588A), I would be very interested, as would the community I'm sure!

Kleinstein, thanks for you feedback.  Yes, it is early days for the 8588A.  My understanding is that the 8508A was a "good" DMM, so my hope is that the 8588A will at least be a "better" DMM.  Time will tell with regards to how well it transports...

[e61_phil]

Our 8588A will also go for calibration in the next weeks. I'm very curious about the stability. I'm afraid that it isn't really improved over the 8508A and some cost cuttings may have been introduced.

The 3458A is a great 10V digital VOLT meter. For everything outside fast, low noise or high linearity measurements in the 10V range I prefer the 8508A or 8588A. In the end it is the best if you have the ability to choose the right tool for the job.

Would be very interesting to collect some results from 8588A calibration reports after the first year.

I was also searching for a replacement of our broken 5440B. I ended up with a 5730A, because a 5720A was really hard to get and is out of service. If you need it in a professional enviroment the service is really important. For private use it is completely different thing.