Hello,
I have to realize a high side current measurement with the following parameters:
Current to be measured: 0 ... 30mA
Accuracy: 0.1mA
Offset Voltage: 24V
I have found the INA226. This seems to be an interesting IC to solve the problem. Using a 2 Ohm shunt resistor would result in a maximum current of ca. 40mA. Is therefore a precission shunt resistor necessary or may the resistance value be compensated by calibration? Is the INA226 a good choice for this application or is there an IC that is better suited?
Best regards
Markus
Hello Markus,
Due to the INA226's very low initial input offset voltage and ability to handle a common-mode voltage of 24V, it is the best choice for this application.
However, a current range of 0-30mA is not strictly possible. This is because at minimum load (0A), the initial input offset voltage of the device will be the dominant signal. In other words, you need some load current to create a shunt voltage such that the initial input offset voltage becomes insignificant. Vos(max) of the INA226 is 10uV.
Concerning your accuracy requirement (0.1mA of 30mA), that equates to 0.33% error. It is important to distinguish this in terms of the load current. For example, the error due only to initial input offset voltage at maximum load is 100*(10u/(30m*2))=0.25%. As the load current decreases, the initial input offset voltage becomes a larger portion of the measurement. Ultimately we find that you'll violate your 0.33% error budget when Iload<=1.5mA.
There are additional errors to take into account, including but not limited to CMR & PSR. We have a series of articles that will help you with the calculations. Here is a link to the first of the four parts:
http://www.eetimes.com/design/industrial-control/4236026/A-Current-Sensing-Tutorial-Part-1--Fundamentals
Part 3 discusses accuracy.
Also, here is a link to a recent post that you may find useful.
View this post at http://e2e.ti.com/support/amplifiers/precision_amplifiers/f/14/t/183621.aspx
Regards,
Pete Semig
Precision Analog Applications
I forgot to address your question about the precision resistor/calibration. The calculations in the other post do not include the shunt resistor tolerance, though this is an additional error as discussed in the articles. Therefore you will require a precision shunt for this application. However, the decision to calibrate will be a cost vs. performance decision. Performing a calibration (or multiple calibrations) will include additional design & analysis in addition to the additional components. I recommend first calculating the accuracy assuming no calibration to see if such a solution is sufficient.
Hello Pete,
meanwhile I found the time to read your Current Sensing Tutorial. This was very helpful to me.
Actually I do have two questions:
1. In your calculation your result is 0,25%. When I calculate the equation the result is 0.017%. Is this correct?
2. Regarding Part III of the arcticle (Accuracy): When Vcm-pds equals to Vcm-sys the difference is 0. In this cas e(CMRR) would be 0%. Is this correct?
The same would be applicable for e(PSSR).
Best regards.
You're correct...the error due to a 10uV initial input offset voltage on a 60mV shunt voltage is 17m%. Sorry abou the miscalculation.
The answer to your second question is 'yes'.
meanwhile I ordered an INA226 EVM and I made some measurements. The results for current measurement are really good. Using the calibration register and verifying the results with a 6,5 digit multimeter shows that the accuracy is within +/-2uA for a range of 25mA.
Additionally I have to make a voltage measurement. The VBus-Pin of the INA226 could be used BUT the input current of the pin is to high.
How can I reduce the input current? I think of using e.g. a voltage follower prior to the VBus-Pin. Preferably the common mode range is the same of the INA 226. Is there an adequate component (instrumentation amplifier, op amp ...) for this application?
Good to hear!
You could buffer the voltage directly, but you would need a high voltage op amp and a sufficient supplies. Maybe you could just use large valued resistors to divide down the voltage, buffer using a lower-power op amp, then feed it into Vbus. I would beware of resistor tolerance/matching and temperature drift, though. Also, the values stored in the bus and power register will have to be recalculated in your microcontroller because the bus voltage measurement and power calculation will use the divided down voltage instead of actual bus voltage.
I hope this helps.