This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

INA 122 reference input

Other Parts Discussed in Thread: INA122

Hello everyone,

I am designing a coulomb counter for usage in low power wireless sensor networks.

To measure the voltage differential across a shunt resistor I am using a INA122 and feeding the output to a 22bit ADC. This whole is supplied by 5V.

My question is about the reference input of the INA122.

Does the reference input have the same limitations as the common mode input voltage range as the inputs of the INA122?

I would like to offset the output voltage of the INA122 by the ADC reference voltage (4.096V). This is done so I will not run into problems that at very low power consumption (very low differential voltages) the INA will not reach the ground all the way. This difference is only a few mV but this swamps out voltage I try to measure.

Thank you in advance for any answers.

  • The reference voltage is a totally different animal than the common mode input range.

    You need to apply a reference voltage that will always keep the output of the INA122 out of the ground rail.  If you look at the following drawing that shows the equations applicable to the INA122 nodes, you can see that the output voltage is a function of  "Vdiff" times your selected gain.  It is important to note, however, that the internal offset voltage of the INA122 will be seen as part of Vdiff, so to calculate the minimum reference voltage needed to provide an input to a 4.096V ADC you need to know your mimimum sensor input (this includes any sensor offsets) and the minimum VOS of the INA122.  Let's say as an example that your sensor outputs voltages from -1mV to 10mV.  You need to treat your minimum input signal as 1mV to the minimum VOS = -1mV -500uV = -1.5mV.  In a gain of 400, your output will vary between ~4V and ~ -.6V.  Since the output swing limitation of the INA122 is 100mV from the rail, you will need to apply AT LEAST 700mV (i.e. 600mV  + 100mV) to the reference pin.  Ideally, you are best in giving yourself some head room on this and setting your reference pin to .8-1V.  The issue with this is that now your maximum signal will hit your ADC input MAX and/or your power supply rail, so to compensate you will have to scale down your gain to a value that will not hit the top rail under your full scale sensor output.  A good comprimise might be to set your gain to 230 and your reference to .5V to given you an output range of 155mV to 3.95V.

    In the case where your sensor ouputs a symmetric bipolar signal (i.e. =+/-5mV), in this case the reference is ideally placed in the center of your desired output range, which in this case would be +2.045V.

  • Matt said:
    The reference voltage is a totally different animal than the common mode input range.

    + good explanation

    Thank you Matt for the explanation.

    I had another idea that might be possible. What if I set the REF voltage to the max voltage of the ADC (4.096V) and switch around the + and - across my sensor. My sensor only delivers positive voltages from 0V to 40mV. I am using a gain of approximate 100. This means, 0V across my sensor then the INA122 would output 4.096V. If I would have 1mV across my sensor , with a gain of 100 resulting in 100mV subtracted from my ref voltage instead of added. This would create an output voltage of 3.996V. My scale is reversed this way but not a problem at all as I just correct this in my microcontroller.

    Would this be a feasible solution?

  • This is actually a pretty clever idea; however, a good design should always have some margin from the input and output limits to prevent clipping.  In your case, if you do not care about inputs below 1mV this should work.  If need resolution down to 0V this will not work as your lower voltages might get clipped on the inputs of the ADC.  This will be exacerbated if you have a positive offset with respect to your input voltages.  In this case you run into the same issue again of having to create a divided down reference voltage to ensure the output does not clip.

    -Matt