Current measurement from 0.1uA to 1A

[sandy.holmarc]

hello
is it possible to get high precision current measurement (0.1uA to 1A) with a  low value shunt resistor?.

[IanB]

hello
is it possible to get high precision current measurement (0.1uA to 1A) with a  low value shunt resistor?

From 0.1 µA to 1 A is a range of 10⁷ orders of magnitude.

If you want a precision of 0.1 µA it means you want to tell the difference between 0.9999999 A and 1.0000000 A

Have you thought about what you are asking?

[sandy.holmarc]

Hi IanB,
 I developed a circuit with shunt resitors 25K ( 0.1uA -0.1mA),250R (0.1mA-10mA),2.5R (10mA to 1A).I got result with these resistors .But i need a low value resistor to read all the current to use in a closed loop.Can you please tell me about any method to achieve this (0.1UA to 1A)measurement  .

[nsrmagazin]

No. Especially with a small value resistor.

[Gyro]

... is a range of 10⁷ orders of magnitude.

Thats a lot of orders of magnitude. 

[qno]

Have a good look at the uCurrent from Dave.

[Kalvin]

What is the required bandwidth? For slow signals you could use something that late Jim Williams has created: A voltage to frequency converter which has 8 decades of dynamic range:

https://www.analog.com/media/en/technical-documentation/application-notes/an14f.pdf

For faster signals you could use a 24-bit or 32-bit ADC converter, but the bandwidth would be few kHz at best and you will probably get maximum dynamic range around 100 dB depending on how well you can manage the noise issues and what voltage level you can get from the shunt resistor compared to the ADC input range. For example, if the shunt resistor is 0.1 ohms and the ADC input range is +/- 1V, the voltage across the shunt resistor will +/- 100 mV, and you will lose more than 3 bits of ADC range (ie. 24-bit ADC will provide only 20 bits at best). At the other end, the noise voltage of the ADC converter will limit the dynamic range.

Probably you could come up with some sort of fast autoranging system, which would select the most appropriate shunt resistor according to the instantaneous current.

If you want to an analog output, you could use a logarithmic output so that you can represent the low and high measurement values in sensible voltage range.

[spec]

Hi sandy.holmarc

Would this be any good?
100nA  : 5K
1uA      : 500R
10uA    : 50R
100uA  : 5R
1mA     : 500mR
10mA   : 50mR
100mA : 50mR
1A       : 50mR

[BrianHG]

... is a range of 10⁷ orders of magnitude.

Thats a lot of orders of magnitude. 

Yes, that numbers of orders of magnitude actually exceeds the capability of all the matter in our solar system by a long shot, even going from a fentoamp through all the current produced if all our sun's energy output was turned into electrical current all at once.

[IanB]

... is a range of 10⁷ orders of magnitude.

Thats a lot of orders of magnitude. 

Yes, that numbers of orders of magnitude actually exceeds the capability of all the matter in our solar system by a long shot, even going from a fentoamp through all the current produced if all our sun's energy output was turned into electrical current all at once.

Damn, such literal minded people around here   

[BrianHG]

... is a range of 10⁷ orders of magnitude.

Thats a lot of orders of magnitude. 

Yes, that numbers of orders of magnitude actually exceeds the capability of all the matter in our solar system by a long shot, even going from a fentoamp through all the current produced if all our sun's energy output was turned into electrical current all at once.

Damn, such literal minded people around here   

Come on, were having fun with a little at the 'expanse' of your little scale.
What you have actually really done is gone from the fento-amp to the same amount of current produced by around 1000 class 1-a supernovas all going off all simultaneously at once... 

[coppercone2]

I think thermals will destroy your calculations. Find how many PPM sstability you need first and then look at how the resistor is going to react to cooling and heating to those ranges.

[splin]

A few questions:
1) What is the maximum voltage drop you can tolerate at 1A?

2) Is .1uA the minimum resolution you need? If not what resolution do you need when measuring a .1uA current?

3) What accuracy do you hope to achieve on the 1A and .1uA ranges?

4) What bandwidth do you need? Are you measuring DC, steady AC or dynamic signals such as the current draw of a microcontroller system switching from full power to sleep mode? If the latter, how fast does it need to track the signal?

[sandy.holmarc]

A few questions:
1) What is the maximum voltage drop you can tolerate at 1A?

2) Is .1uA the minimum resolution you need? If not what resolution do you need when measuring a .1uA current?

3) What accuracy do you hope to achieve on the 1A and .1uA ranges?

4) What bandwidth do you need? Are you measuring DC, steady AC or dynamic signals such as the current draw of a microcontroller system switching from full power to sleep mode? If the latter, how fast does it need to track the signal?

Hi splin,
Thank you very much for the assistance.
1.  maximum voltage drop  at 1A is 1V(my application needs 1V to -1V voltage range)
2.  0.1uA  is the minimum range ,i need 10nA resolution if possible
3   I can tolerate up to 10nA error current
4   I am measuring a DC voltage ,output range of -1V to 1V.
any help would be greatly appreciated.


 

[mvs]

1.  maximum voltage drop  at 1A is 1V(my application needs 1V to -1V voltage range)
2.  0.1uA  is the minimum range ,i need 10nA resolution if possible
3   I can tolerate up to 10nA error current
4   I am measuring a DC voltage ,output range of -1V to 1V.

It is getting better and better... 1 Ohm shunt, 10nV max error.
Best opamps available on the market have offset error of about 1-10µV. Thermoelectric effect may add also some µV to error.
You do not get away with single shunt, you need to switch ranges.

[Marco]

There's no commercial solution which will deliver that offset voltage, even manual zero'ing won't help since the offset will drift more than that for a single degree Celsius.

You could certainly make your own chopper with far lower offset and bandwidth, but that's a whole lot of bother. Easier to just have a second higher resistance shunt and clamp the voltage across it.

[splin]

Hi splin,
Thank you very much for the assistance.
1.  maximum voltage drop  at 1A is 1V(my application needs 1V to -1V voltage range)

Ok, so a 1 ohm shunt.

2.  0.1uA  is the minimum range ,i need 10nA resolution if possible

Well it's possible but very challenging. 10nV is .01ppm of 1V; an 8 1/2 digit DVMs such as the HP/Agilent 3458A has a resolution of 10nV on the 1V scale and has a noise specification of .01ppm rms if it is set to measure for at least 2 seconds per reading. .01ppm rms is 10nV rms or approx 60nV peak to peak. You would need to average plenty of readings to get down to 10nV peak to peak noise - if in fact you can, as averaging suffers diminishing returns because of 1/f noise..

Noise wise, you have the advantage over a DVM as the source resistance is very low (1 ohm) so you can use a very low noise bipolar amplifier. One of the lowest noise opamps is the LT1028 with 35nV pp noise (0.1 to 10Hz bandwidth). Put 4 in parallel and in theory you should halve that 17nVpp. Reduce the bandwidth to say 1Hz will reduce the noise further but 1/f noise will limit the improvement.

Problem is that these bipolar opamps have offsets of around 10uV which drifts at .8uV/C. You can trim out the offset but the offset also drifts with time. To keep the drift due to temperature to less than 10nV you would need to maintain the opamp temperature within 0.0125C - not impossible but not trivial. You can get zero drift and chopper opamps which significantly reduce the offset errors but are rather noisier - 100nV pp or more. You could make your own chopper amp with low noise bipolar transistors to achieve very low noise, but again not a trivial exercise.

Thermal EMFs, generated at all the connections from the shunt to the amplifier will be an even bigger problem and could be several uV per degree C.  The best way to eliminate these is to reverse the current through the shunt and subtract the new reading from the previous thus cancelling out the thermal EMFs and the opamp offset. If you are in control of the current that you are measuring, that may be possible.

Alternatively, assuming you are measuring the currents with a ADC, it may be easier to stop, or divert the current away from the current shunt so that you can measure the shunt voltage with zero current which will give you the thermal EMF and offset voltages which you can subtract from subsequent readings. You would need to repeat this periodically because of temperature changes and other drifts.

   I can tolerate up to 10nA error current

I wonder if you are confusing resolution and accuracy? Measuring 0.1uA to an accuracy of 10nV is only 10% but measuring 1A to 10nA accuracy is .01ppm which is nigh on impossible. The 3458A can resolve 10n on the 1V scale but the accuracy is around 2ppm at best with a newly calibrated meter - 200 times worse than your requirement.

Realistically, getting a 1 ohm shunt resistor with a tolerance better than 100ppm is going to be expensive and it will likely drift by 100pm or more each year. Also with 1A, the shunt will be dissipating 1W which could raise the temperature of the shunt by several 10s of degrees, that's another 50 to 100ppm of error assuming the shunt has a TC of 5ppm/C. So the error from the shunt alone could be 300ppm or more.

4   I am measuring a DC voltage ,output range of -1V to 1V.

How quickly do you need to be able to measure the current? Can you tell us something about the application as it would help us understand your requirements better?

[David Hess]

A 1 ohm shunt would return 100 nanovolts at 0.1 microamp.  That sort of sensitivity and resolution is barely feasible with chopper stabilized amplifiers but thermocouple effects and noise will be major problems.

There is a way to remove the thermocouple effects, offset drift, and flicker noise; place the shunt inside of an h-bridge and chop the entire shunt!  But most loads and applications will not support this type of operation.

A generally better way would be to use multiple shunts in series with the higher resistance shunts clamped.  An alternative to this is to place the shunts on the drain side of an FET allowing low (practically zero) output impedance with higher resistance shunts and easier clamping.

[duak]

I have been toying with the idea of a tester that needs a large dynamic range current measurent.

I've worked out a two range I to V converter. (see rough sketch)

The upper section handles the 1 mA to 1 A range with a Kelvin sense resistor and an INA.

The lower section is a current to voltage converter using an opamp.  U2 will keep its - input at essentially 0 V (virtual ground) as long as its output has not limited.  When the input current increases, U2 will limit, the - input voltage will rise and one of the schottky diodes will conduct current.

A three range converter could probably be designed by replicating the lower section in series with the other two, but with a smaller value feedback resistor to reduce sensitivity.  However it would add another 300 mV to the voltage drop.  Design details would also have to consider the bias and leakage currents from the less sensitive ranges affecting the most sensitive range.

FWIW,

[Nauris]

What about 1N4007 or some such in place of the sense resistor? Should easily give the required dynamic range, though the response is more logarithmic than linear. 

[Kleinstein]

The logarithmic curve would not be that bad. A problem is that the diode voltage is also temperature dependent. With nonlinear self heating this can get quite tricky is good accuracy is required.

The two or 3 range version is probably the better way.

[Kalvin]

The logarithmic curve would not be that bad. A problem is that the diode voltage is also temperature dependent. With nonlinear self heating this can get quite tricky is good accuracy is required.

The two or 3 range version is probably the better way.

EDN has a Design Idea for logarithmic current measurement circuit using carefully selected diodes:
https://www.edn.com/design/test-and-measurement/4443319/Multi-decade-current-monitor-the-epitome-of-simplicity

I have no idea how well this will work in real life with real components and/or how hard it would be to find a suitable diode pair. But very interesting idea anyway.

[sandy.holmarc]

How quickly do you need to be able to measure the current? Can you tell us something about the application as it would help us understand your requirements better?
[/quote]

Hi splin,
Thank you very much for the reply.I would like to develop a VI measurement circuit for the solar cell experiment.Measurement time under 3 minute is fine .I need accuracy of 10nA only in uA range measurement ,uA accuracy in the mA range measurement  and mA range accuracy in the A range measurement .

[Marco]

Given how often this topic come up, I think Dave should do an auto-ranging multi-output uCurrent.

Have 3 shunts in parallel, with the lower value shunts gated by a bidirectional switch (back to back MOSFETs). When the absolute voltage across the previous shunt goes over a threshold, start smoothly switching in the next shunt.

[Habropoda]

It seems like the CurrentRanger could handle this:
https://lowpowerlab.com/guide/currentranger/

[Marco]

Nice, so it exists already. Looking at their schematic I guess a bidirectional switch isn't really necessary since the shunt voltage is too low for the body diodes to carry significant current ... should have thought of that.

[beanflying]

And almost like he is a mind reader Andreas Spiess does a review on both (for low currents) Not a great photo Mr Jones 

https://youtu.be/HmXfyLyN38c

[David Hess]

What about 1N4007 or some such in place of the sense resistor? Should easily give the required dynamic range, though the response is more logarithmic than linear.

The logarithmic response is actually an advantage for wide dynamic range but at high currents, the diode's series resistance and temperature compensation become large error terms.  This circuit type would be more suitable for current measurements between nanoamps and milliamps.