Current sensor

Thread Starter

Dollarday

Joined Jan 25, 2012
32
Hi All!

I need to design a high side current sensor thats able to measure 0-1A at 30V within an accuracy of 500uA and feed this value into an ADC of a microcontroller (6V input max)

I intend to use a current sensing chip (This takes the guesswork of trying to figure out CMRR and CMVR due to resistor inaccuracies etc. for a custom instrumentational OP-AMP).

The best current sensing chip I've found sofar for the application is the MAX4080 current sensing chip with a set gain of 60 from maxim with a 0.1ohm current sensing resistor. It has good Common Mode Rejection and the Common Mode Voltage range includes 0V - 30V.
(Output: 1A = 6V ; 500uA = 0.03V)

It has a gain accuracy of 0.1%.

The required gain accuracy (As far as I can calculate) is 0.05%
(0.05% of 1A is 500uA) and (As far as I can tell) this value is independant of the sense resistor used.

Any suggestions on chips that are more accurate or perhaps a way I can go around this inaccuracy?
 

wayneh

Joined Sep 9, 2010
17,496
Be careful about the term "accuracy". It's not the same as "precision", which is what I think you meant throughout your post. The accuracy of your strategy will depend on the degree to which you "know" the resistance of the current-sensing resistor. It cannot be better than that (except by luck, which rarely acts in our favor).

The limiting factor on precision may be your ADC, as you've calculated.
 

Thread Starter

Dollarday

Joined Jan 25, 2012
32
Hi Wayneh,

Thanks for blowing my mind... I always thought they meant the same thing, but I did a bit of reading:
Accuracy means getting a result that is close to the real answer. Precision means getting a similar result every time you try.

I agree, the resolution of my ADC will influence my precision. (However, with a resolution of 10-bits this is within "acceptable" margins)

My accuracy with regards to the resistor can also be kept within acceptable margins as I will ensure the temperature of the resistor will be within a certain temperature range. (Within that temperature range the resistance can be guarenteed within levels that will not influence my readings outside the range of 500uA.)

My problem remains the OP-AMP chip. (Gain accuracy of 0.1%)

I think my best aproach to increase accuracy would simply be to connect the sense resistor to a load and calibrate my circuit (via software) by measuring the voltage over the sense resistor with an oscilloscope?

If I repeat this test I should be able to determine if the precision is high enough as well.

Can I assume the oscillosope will give a correct value at a gain of 50? (The oscilloscopes pre-amplifier would need to have a common mode voltage range of 30V)

Otherwise, if you have any other suggestions that may help?
 

Thread Starter

Dollarday

Joined Jan 25, 2012
32
Sai, I think I mixed them up again

Sorry Wayneh, I think after taking a look at these pictures I won't mix the terms up again, thanx!
 

jwilk13

Joined Jun 15, 2011
228
I've used the ZXCT1083 from Diodes, Inc. before with success. It has adjustable gain, so some extra error is going to be introduced with the resistors used to set the gain. I'm pretty sure they make fixed gain versions as well, but you'll have to search around their website for those.
 

bretm

Joined Feb 6, 2012
152
10-bit ADC will only give you about 0.1% precision. You'll need to take four measurements and average them to get an extra bit of precision, to reach 0.05%.
 

ErnieM

Joined Apr 24, 2011
8,377
I'm presently using a Maxim MAX4080 chip to measure current via an A2D as you are. This device comes with a choice of a gain of 5, 20, or 60 so it can save a few parts.

When I design these things I am less concerned with the absolute accuracy of my components, all I care about is are they stable. I include a calibration routine in the software to determine the proper constant to fudge factor the reading back to reality.

The cal gets run once a year or so then the fudge factor gets applied to each reading. Works fine in our lab where all measurements need to be traceable to the National Bureau of Standards.
 

crutschow

Joined Mar 14, 2008
34,285
Hi Wayneh,

Thanks for blowing my mind... I always thought they meant the same thing, but I did a bit of reading:
Accuracy means getting a result that is close to the real answer. Precision means getting a similar result every time you try.

I agree, the resolution of my ADC will influence my precision. (However, with a resolution of 10-bits this is within "acceptable" margins)

My accuracy with regards to the resistor can also be kept within acceptable margins as I will ensure the temperature of the resistor will be within a certain temperature range. (Within that temperature range the resistance can be guarenteed within levels that will not influence my readings outside the range of 500uA.)

My problem remains the OP-AMP chip. (Gain accuracy of 0.1%)

I think my best aproach to increase accuracy would simply be to connect the sense resistor to a load and calibrate my circuit (via software) by measuring the voltage over the sense resistor with an oscilloscope?

If I repeat this test I should be able to determine if the precision is high enough as well.

Can I assume the oscillosope will give a correct value at a gain of 50? (The oscilloscopes pre-amplifier would need to have a common mode voltage range of 30V)

Otherwise, if you have any other suggestions that may help?
In school I learned that Accuracy is how accurate the measurement is as traceable to the NBS, and precision is the resolution (fineness) of the measurement (which can be much higher than the accuracy). Thus a 10-bit converter has a precision of 0.1% but its absolute accuracy may be much less, depending on its design and how it is calibrated.

You do not want to use an oscilloscope to calibrate your system since their accuracy is likely not much better than 1%. You need to use an accurate 5 to 6 digit voltmeter that has been calibrated within the year.
 

Thread Starter

Dollarday

Joined Jan 25, 2012
32
Thanx guys, I've decided on the max4080... It seems to be well within the required accuracy (I used a precision current source to test it).


Regards,
Hermann Havenstein
 
Top