How to calibrate DMM with 10mA? DIY

[001]

How to calibrate DMM with 10mA?
Actually I need for current source of 10mA (at least +-0.01% 100ppm)
Can I build simple voltage controlled source around common +-0.01% Zener and opamp?
Thanx for Yours opinions

[try]

How 001,

How to calibrate DMM with 10mA?
Actually I need for current source of 10mA (at least +-0.01% 100ppm)
Can I build simple voltage controlled source around common +-0.01% Zener and opamp?
Thanx for Yours opinions

Yes, you can.
With P = I^2*R and P = 1e-4*R in your case, you have some margin on resistor selection.
You can keep the power low and reduce the change in resistor value caused by self-heating.

[Brak]

Would a current sink do?



There's a lot of TL431 variants around, one might be accurate enough for you.  They're cheap and you'd only need one precision resistor.

Brak

[001]

Would a current sink do?



There's a lot of TL431 variants around, one might be accurate enough for you.  They're cheap and you'd only need one precision resistor.

Brak

Thanx!

Can I use cheap china ad584 voltage standard in Yours manner?

[Brak]

I don't think so.

Brak

[Alex Nikitin]

Would a current sink do?



There's a lot of TL431 variants around, one might be accurate enough for you.  They're cheap and you'd only need one precision resistor.

Brak

Thanx!

Can I use cheap china ad584 voltage standard in Yours manner?

With a bipolar transistor you'll get 1/beta (so, ~0.5%) error due to the base current component, plus some error due to the TL431 input current and the reference accuracy is 0.4% ? So, "precision" in this case means ~1% error  .

Cheers

Alex

[SvanGool]

This one is better: http://www.ti.com/lit/an/sbva001/sbva001.pdf

[Brak]

“The best way to get the right answer on the Internet is not to ask a question, but to post the wrong answer”,  Ward Cunningham.

https://www.todayifoundout.com/index.php/2014/09/cunningham-cunninghams-law/

Brak

[Andreas]

“The best way to get the right answer on the Internet is not to ask a question, but to post the wrong answer”,  Ward Cunningham.

Doesnt work when it comes to "precision".

Question: how do you reliably check the 10mA 0.01% ?
My 6.5 digit DMMs are specced much worse.

with best regards

Andreas

[Wolfgang]

... and mine, too (Keysight 34465A at 500pm 1 Year spec). What you need is 3458A which has about 25ppm. Not even a 34470A will meet the 100ppm you want.
So, it can be done - but not for free. 

[ogden]

Actually I need for current source of 10mA (at least +-0.01% 100ppm)

If you (have to) ask how to build +-0.01% 100ppm source of anything, you are not educated enough to comprehend answer

[2N3055]

Would a current sink do?



There's a lot of TL431 variants around, one might be accurate enough for you.  They're cheap and you'd only need one precision resistor.

Brak

Thanx!

Can I use cheap china ad584 voltage standard in Yours manner?

With a bipolar transistor you'll get 1/beta (so, ~0.5%) error due to the base current component, plus some error due to the TL431 input current and the reference accuracy is 0.4% ? So, "precision" in this case means ~1% error  .

Cheers

Alex

Actually TL431 is the one regulating 2.5V on sense resistor, including collector and base current. TL431 ref input current is on order of 2-3 uA. That is static error. But ref input current will change less than 1uA over full temp range, so while not perfect, regulation will be easily 0.1 % , or better stability. Initial accuracy is a different thing, that will not be stellar, but can be calibrated out.
Regards,

[Wolfgang]

... tempco and long term stability wont be stellar, either. I am sceptic about this approach, and for cal you need a 8 1/2 digit multimeter anyway. I think you should try it this way:

- a very good voltage reference (LM399)
- a precision resistive divider made from 0.01% resistors and a *small* trim range
- a high precision Op Amp with extremely low offset, drift and tempco.
- a MOSFET in the driver so there is no base current
- proper compensation to avoid oscillations
- a 100ppm sensing resistor to measure the current
- all this with a very well filtered power supply (batteries are best)
- in a shielded box, if you can with constant temperature.

This could work out with 100ppm, if everything is done properly. Hacking metrology is pointless, because there are so many pitfalls to drop in.

[001]

Well
I must use at least x10 tolerance source to calibrate something, isn`t it?
But how to check 0,1% ampermeter if 0,01% current source is incredible? 

[e61_phil]

Question: how do you reliably check the 10mA 0.01% ?
My 6.5 digit DMMs are specced much worse.

I would use a 100R resistor and measure the voltage.

One can buy a 100R resistor with 50ppm tolerance (Vishay Z210 for example) or measure it to 8ppm with a Fluke 8508A.
The input current of the meter (30pA max for a 34401A for example) doesn't matter at 10mA (3ppb)

1year spec for 1V is 47ppm (34401A)

Therefore, you end up around 70ppm. And in reality much better.

Or did I miss a point?

Edit: Even if you use your 3478A in the 3V range you end up at ~95ppm with the 1year specs.
The TUR is only 1:1 in this case...

[2N3055]

... tempco and long term stability wont be stellar, either. I am sceptic about this approach, and for cal you need a 8 1/2 digit multimeter anyway. I think you should try it this way:

- a very good voltage reference (LM399)
- a precision resistive divider made from 0.01% resistors and a *small* trim range
- a high precision Op Amp with extremely low offset, drift and tempco.
- a MOSFET in the driver so there is no base current
- proper compensation to avoid oscillations
- a 100ppm sensing resistor to measure the current
- all this with a very well filtered power supply (batteries are best)
- in a shielded box, if you can with constant temperature.

This could work out with 100ppm, if everything is done properly. Hacking metrology is pointless, because there are so many pitfalls to drop in.

I wasn't saying it was 10 ppm circuit, but merely pointed out that it wasn't transistor doing regulation and other factual errors. It is not very good circuit but much better than 1%.

Also,when you start pushing ppms, it will be thing like tempco of sensing resistor and such that will be problematic...

Member Dr. Frank some time ago shared his design of precision current box. It is much better than 100 ppm. There is no magic, but clever design by somebody that knows a lot about topic...
Problem is that it needs precision resistors, that each cost more than meters OP is trying to calibrate...

When chasing ppm levels, there is a point where it isn't more complicated anymore, just needs really expensive components..
Regards,

Dr. Frank's design:

https://www.eevblog.com/forum/metrology/recycling-of-precision-current-source-noise-reduction-for-low-burden-shunts/msg1432410/#msg1432410

[e61_phil]

... tempco and long term stability wont be stellar, either. I am sceptic about this approach, and for cal you need a 8 1/2 digit multimeter anyway. I think you should try it this way:

- a very good voltage reference (LM399)
- a precision resistive divider made from 0.01% resistors and a *small* trim range
- a high precision Op Amp with extremely low offset, drift and tempco.
- a MOSFET in the driver so there is no base current
- proper compensation to avoid oscillations
- a 100ppm sensing resistor to measure the current
- all this with a very well filtered power supply (batteries are best)
- in a shielded box, if you can with constant temperature.

This could work out with 100ppm, if everything is done properly. Hacking metrology is pointless, because there are so many pitfalls to drop in.

I wouldn't go for a current source as a reference. Instead I would build a very stable current source and use a reference (a good shunt or another known multimeter) in series with the meter under test. This way you don't need a long term stable voltage reference, shunt, dividers and so on in your source. Many things become easier.

I started something similar a couple of days ago for 1A. I attached a first result. I measured the current with a Burster 1240-1 Shunt connected to my Keithley 182.

[2N3055]

Well
I must use at least x10 tolerance source to calibrate something, isn`t it?
But how to check 0,1% ampermeter if 0,01% current source is incredible? 

Little bedtime reading...

[Wolfgang]

Its not incredible. It is *difficult*. When you look at industry calibrators from Fluke or others, it can be done, of course, but for a lot of money.