Metrology: Methodologies, Uncertainties, and Calibration Certificates

[Kleinstein]

How much thermal EMF develops at a shunt depends on the symmetry. Ideally the temperature would be the same on both sides and thus no error from thermal EMF. It does not take that much to create some asymmetry, e.g. from different heat losses at the cables or a not ideal electrical contact for the current path that heats up. Good shunt should use low thermal EMF materials, but not all do.

[Overspeed]

How much thermal EMF develops at a shunt depends on the symmetry. Ideally the temperature would be the same on both sides and thus no error from thermal EMF. It does not take that much to create some asymmetry, e.g. from different heat losses at the cables or a not ideal electrical contact for the current path that heats up. Good shunt should use low thermal EMF materials, but not all do.

Hello

I agree but I will add this point

The power in the shunt is a critical parameter , make 50 Amp continuous with a 300 Amp Shunt is not the same as make 280 Amp with a 300 Amp shunt and this point is also linked to the fact to run comparison measurement with the same type of shunt .

A critical point on shunt measurement is the connection , surface shall be clean with a good shape contact shape and also a similar / same screw torque as contact pressure is a important parameter .

Industrial shunt are not always to not often '' high quality material '' as Manganin but more often in cheaper copper nickel alloys so be aware of wild tempco .

These problem can be partially solved by a robust procedure ( torque , contact checks ..) and a temperature / resistance value table .

Regards
OS

[Overspeed]

Hello

To avoid make massive quote , I have copied the paragraph

Regarding procedure : As your goal is to run calibration and generate valid results , a written procedure with pictures / drawings is mandatory to be able to duplicate always the same things to avoid any delta in the results.

Writing a procedure is not a punishment as that an usual  engineering document , no need to write an encyclopaedia too but something as a tool able to be improved ..

Regarding repeatability , before to claim that a result is GOOD , this results shall be demonstrated by a repeatability test , in various industries , test as Anova R&R are very common .

Whatever you are or not under Quality management system , these rules are valid and useful .

Regards
OS

[alex-kv]

Hello,

I've tried to produce a calibration procedure. It is available as an attachment.
So far, I have only processed the 100 mV range. I still need to figure out how to integrate the measurements into the mendip_discovery Excel file.

Alex

[Overspeed]

Hello

2 interesting documents

one Korean calibration method
one Fluke '' how to measure power coefficient '' this document include clear schematics

I will post longer today

Regards
OS

[alex-kv]

Hello

Thank you Overspeed. It seems you have a well-stocked library of very interesting documents 

I updated the procedure

Alex

[alex-kv]

When applying the procedure, I wonder if the quality of the calibration could be improved by using the statistical functions of the devices and taking an average of 50 measurements. The average will span 500 PLCs 2500 PLCs for each DMM. Since it's not possible to obtain a stable value exactly as indicated by the calibration certificate of the 'standard' DMM, I calculate the gain coefficients of the positive and negative values from the standard readings and then apply them to the UUT readings.

I updated the procedure accordingly.

EDIT: corrected number of PLCs

[alex-kv]

Here is a practical example of the calculation steps in the procedure, given that the calibrated values of the standard DMM for 100 mV and -100 mV are 100.0012 and -100.0008 respectively.

DMM7510 AVG 50RDGS 100mV = 100.00161 mV
DMM7510 AVG 50RDGS -100mV = -100.00128 mV
K2000 AVG 50RDGS 100mV = 100.0031 mV
K2000 AVG 50RDGS -100mV = -100.0031 mV

Poscoeff = 100.0012 / 100.00161
Negcoeff = -100.0008 / -100.00128
[ EDIT:
Poscoeff = 0.9999959
Negcoeff = 0.9999952
]

"Measured value" for the UUT's cal cert:
100.0031 * 0.9999959 = 100.0027 mV
-100.0031 * 0.9999952 = -100,0026 mV

Have I done things right?

Alex

[Overspeed]

Hello

When you run a calibration your procedure shall match the exact way / method of measurement you will use to compare '' apple with apple '' OR you run the standard calibration procedure as provided by the manufacturer .

If you chase ultimate accuracy ( precision ... ) and if you select an average measurement on 50 measurements OK why not , but you shall use the same for your daily measurement.

I never use the last digits of a instrument IF I need a real / better value I switch to a 7.5 or a 8.5 digits instrument . With low to very low voltage as shunt output voltage that also a solution to use a low noise preamp .

Regards
OS

[Kleinstein]

One should also include a zero measurement - not abosultely necessary to check the gain, but it would help to detect problems with linearity. Once set up the extra reading does not take that much time.

[Overspeed]

Hello

a doc on shunt tempco vs resistance change

Regarding the calibration I agree with Kleinstein on linearity , best is to try a 3 point calibration 0 V mid range so 5 V and full range 10 V if that a 10 V range

That also possible to limit reduce the calibration range to be closer to the measurement need as for example if you plan to measure between 0 and 5 V that better to run 0 V 2.5 V and 5 V calibration rather than  the full range as 10 V 

Regards
OS

[cscaccetti]

Did a scope expansion for DC energy measure. So not only dealing with just DCI measure, but DCV, time, simultaneity of measure, and more. You are going to have so much fun.

Get a danisense DCCT, can't beat it. Like 10 ppm, in series with your shunt, and compare. The danisense guys, were some of the best from LEM. I've tested both the ultrastabs and the DS series, both are incredibly good, the danisense edges out the LEM IMO. But likely, given the level of your equipment, it wont matter, so get what ever is cheaper. They also have great acd/dc properties, so could open up your lab to ac measure later as well.

As for the thermals of the shunt, you either go one of two ways:
You energize to current level, hold, and stabilize temp. What I have found with this is that, shunts can take a significant amount of time to stabilize to a level of very low variability. Like over 30 minutes. Then doing that for a number of test points, make this method incredibly time consuming.

Or either swap connections, or shut on off, and negate thermal EMFs. I do something like 5 second measure off, 5 second measure on. 5 second measure off. Average before and after and subtract from measure. 5 repeats. Point being quick measures to mitigate self-heating. high dTemp/dTime kills ability to temp correct, and will throw erroneous results.

Also looking at your pictures, need better metal matching. The owner of Ohm-labs, Jay has great material, and is a good source of knowledge when it comes to all things resistance and testing. Sadly he is mostly retired, so he may not be as available to chat as he used to.

For the thermal characterization, you need a more homogenized environment, like an air bath. But you can go more junkyard approach, with a box and a fan and temp sensor. If you are curious to see what type of equipment it takes to do this work, what I typically do is research who the players are in the test equipment field, then I look for the ISO17025 scope, and then find the measure, and measure range to find the calibrating equipment.

There is so much more, but my brain's quite scrambly from not sleeping. Relatively new father (1y.o.), much harder than uncertainty budgets.

[Overspeed]

Hello

An interesting 100 Amp current circuit

Regards
OS

[Hydron]

Hello

An interesting 100 Amp current circuit

Regards
OS

Interesting that they are running the IGBT in it's linear region (I was under the impression this was a good way to make the smoke come out) - I guess they are being conservative enough (or just lucky) to get away with it?

[Overspeed]

Hello

Yes I agree thermal effect is real concern for this circuit but nothing say they stay ''ON'' a long time and they can also have thermal control

Question how are managed the static relay ?

Regards
OS