How to calibrate current shunt measurement device

[Spikee]

Hello,

Together with a friend I'm working on a current shunt measurement device.
Which ranges form 1uA to 4A or so. We are especially interested in the accuracy in the low portion of the measurement range.
0 - 150mA (especially the low - mid uA range is important)
150 - 500 mA range (if possible).

We use high quality resistors, 24b adc, dual LTC6655 1.25V 2ppm vrefs (to lower noise) etc.

The first thought was using some good spec transfer standards but these are not available locally.

Equipment that I already have:
Rigol DM3058 (out of cal for a long time)
Fluke 87 III
DMMCheck Plus


Buying a new fancy Keithley SMU or current source / sink is too expensive at this moment.
I could buy my own transfer standards (dual - quad LTC6655 1.25V 2ppm , with some 0.2ppm; custom value; high accuracy laser trimmed vishay precision resistor) which is basically a dummyload.
And get it calibrated / characterized at a test lab here somewhere in Shenzhen.

Or buy an olden goldie from taoboa and get that calibrated / characterized locally.
Here is a list of some of the things that I found:

Yokogawa    2553 - ~1000 rmb
Yokogawa   2554 - ~1300 rmb
Yokogawa   7651   - ~1000 - 2600 rmb
Keithley   6220   - ~6800 rmb
Advantest   TR6141 ~2680 rmb
Advantest   TR6143 ~1880 - 2500 rmb
Advantest    R6141 ~1700 rmb

The idea is that the setup can also be used during the small batch assembly / testing.
To calibrate the circuits.

I have not done calibration / characterization like this before so any help or tips are appreciated.

[ap]

The problem with anything built by yourself is how do you define, in a formally clean way, the drift (time, temperature). This is not easy, and in some cases may be impossible, or expensive to find out. You could buy hermetic resistors and use their spec for drift. All others are harder. So the most efficient (cost wise) way therefore really is, in your case to buy used gear and have it calibrated. Seems you can get it at reasonable prices.

[Spikee]

Can anyone recommend one of these machines ?
I can't seem to find the spec of the Advantest TR6141

[ap]

It really depends on what your exact requirements are (resolution/accuracy/functions...). And also your budget. Let the group know and there will be lots of ideas I am sure.

[Spikee]

DC current source and or sink.
Rough specs:
100nA -> 100uA <0.25%
100uA -> 1mA <0.035%
1mA -> 100mA <0.025%

If possible also get this range: But from what I can see this will add a lot of cost.
100mA -> 2A  <0.025%

These specs are fairly low if I see some of the specifications of these devices. The plan is that the device I'm designing will meat these specs.
Most likely it is going to perform much better.

For future (small scale) production GPIB or any other means of remote current setting will be useful.

Room is always air conditioned ; temp / humidity logging can be implemented.

[alm]

Those are pretty tight specs. The specifications for example the 1 year Yokogawa 7651 (0.03%), Advantest TR6143 (0.05%) or Keithley 6220 (0.05% on most ranges) are not even exceeding that, never mind getting a decent test uncertainty ratio (TUR). You may be able to mitigate that by doing your own characterization and doing more frequent calibrations, or adding an external ammeter to monitor, but that also adds to the costs.

Even the $$$ Keithley 2450 SMU will only barely meet those specs on most ranges, never mind getting a TUR that allows you to verify a device under test matches those specs with any degree of certainty.

[Spikee]

Those were short term specs I listed. Over a longer term this can be less tight.
Once used in small batch production it can be send over for cal a few times per year if needed.  It is not too bad of an price here.

In terms of the ranges I need this seems to do :
Advantest TR6143 Can't find much regarding the stability / accuracy over time

[alm]

So you mean short term stability rather than accuracy? Even on a 90 day cal cycle the accuracy will only be marginally better than the 1 year accuracy for most instruments. Short term stability it easier, except that not all manufacturers will specify short term stability, let alone guarantee it. I do not think a cal lab will certify short term stability either.

[Spikee]

Clearly I have no good idea regarding these kind of specs.

So lets reset the specs and look again:
DC current source and or sink.
Rough specs:
100nA -> 100uA <0.5%
100uA -> 1mA <0.5%
1mA -> 100mA <0.5%

If possible also get this range: But from what I can see this will add a lot of cost.
100mA -> 2A  <0.5%

Accuracy over lets say 180 days / one year.

[alm]

If budget is an issue (sounds like it is), then you could look for a Keithley 220 (the predecessor to the 6220). No idea about the availability in China, but it was the standard current source for years. Price on eBay US is usually around $300-$400. Should be able to comfortably meet your specs (up to 100 mA). Keep in mind that most of them use two-lug triax connectors that are somewhat uncommon (connectors are available, and converters are available for $$$, as everything triax).

If you want from 100 nA (which means you want a resolution down to maybe 10 pA) to 2A, then it might be cheaper to buy two instruments rather than buy one expensive instrument (like Keithley 2450) that can do it all. One instrument that might be affordableish ($1000?) is the Keithley 238 SMU. It does from 1 A full scale down to 1 nA full scale (100 fA resolution), and will measure voltage for free . Obviously it is older and clunkier than a fancy new SMU.

Parts and repair are probably no longer available, but calibration should be no problem. I have no information (beyond publicly available specs) on the Advantest/Yokogawa units.

Do not forget to budget for cables. Current sources that go down to the nA or lower will often have triaxial connectors (two lug for older or three lug for newer instruments). These are not compatible with BNC. Triax cables, connectors and adapters are all fairly expensive (around $50 at Digikey). Even used you can still easily pay $100 for a single Triax cable. There are some threads on this forum about cheaper sources for cables and connectors.

[Alex Nikitin]

The Keithley 263 can source  up to 20mA full scale (0.035% 1 year), 2mA to 2uA FS (0.025% 1 year), 200nA - 2nA FS (0.07% 1 year) and lower ranges with reduced accuracy down to 2pA FS (0.5% +10fA 1 year).

Cheers

Alex

[Spikee]

Those options are not available / affordable at this moment.
Lets say the 100nA range is gone. Will that change much in terms of options ?

Looking at the local sources the Advantest gear seems to be the only affordable thing locally

[Spikee]

I also found a advantest R6240A for about 720 USD. This seems to have a useful range and is somewhat modern

[alm]

Is approximately 0.83% accuracy for 100 µA (assuming output voltage < 2 V) sufficient? One day stability is approximately half that. This is something that would combine well with a lower current unit like the Keithley 220 to get a decent accuracy from nA to A (or something similar from a Japanese brand).

As for other options, the Keithley 224 is an option (does not go as low as the 220). As would the Keithley 236 or 237 (up to 100 mA). If you do not need programmability, then older analog sources like the Keithley 225/227 or HP 6177/6181/6184 might work. The latter can do remote voltage programming, but you would obviously need to verify the set current with a DMM. Sorry, no specific recommendation for Japanese equipment, I rarely see them offered on eBay in Europe/North America. I am trying to suggest as many models as possible in the hope that one is available and affordable in your region.

[Spikee]

Thank you for all the help.
1mA to 100mA models seem quite available:
Advantest R6142
Advantest R6144 (on the more expensive side)
Yokogawa 2553
Yokogawa 2554 (on the more expensive side)
Advantest R6145 (free electron has one, seems high spec)
Keithley 228
Yokogawa 7651
Keithley 6220

Having a bag with money would be so much easier 

As an alternative would it be possible to just get a fairly high spec multimeter,
get that calibrated and use that as the correct value. And just make my own constant current sink / source which is calibrated to the meter ?
Since multi-meters a much easier and cheaper to get.

[Spikee]

If I'm not mistaken my Rigol DM3058 can measure with the following accuracy with a new cal:

200uA range 0.055% of reading  + 0.005% of full range; 1 Year 23?± 5?

So 1uA would have the following accuracy: (1uA*0.055%) + (200uA *0.005%)  = +- 1.00500e-8 A
Which seems way too high for a device of this class. Or is this just not the full picture ?
http://www.batronix.com/pdf/Rigol/Datasheet/DM3058_Datasheet.pdf

This seems to fit the requirements that I need to calibrate / check something.

[alm]

Near the bottom of the range, the 0.005% of range contributes a lot to the tolerance. Pretty much all of it. So accuracy at 1 µA is only 1.06%.

Yes, a meter with sufficient accuracy can be used to monitor a source. Of course the short-term stability and noise should still be low enough, and it makes the procedure more complicated (send value X to current source, measure current as Y from DMM, send value x+dx to source, check if DMM reading is now right).

[Spikee]

In the case that a DMM is used for the current verification it is not that important
that the 1uA is spot on. The only thing that is important is that the measured current incl. accuracy is good enough for my application.

I also saw the fluke 8845A and that should be 0.075% accurate over 1 year in the 100uA range.

I have to think about what is the best way to tackle this problem.

[Spikee]

For now I am getting my DM3058 calibrated. It is only out of cal for 4 years.
It is going to be adjusted and calibrated in China by Rigol. They use a Fluke 5520.
The cost is 87 USD