My application is a battery current measurement. The shunt is placed in the negative path and has a resistance of 250µOhm. Measured current is between -100A and +20A. Therefore the reference pin is fed of a 2/3U+ low impedance voltage source. The monitor itself is run from a 3V3 single rail supply. The tricky part is that the measured current is pretty noisy. Without filtering the monitors output is clipping to the supply rails and hence producing a variable offset error. I came up with two possible solutions: Increase the supply voltage to allow for a higher voltage swing or input filtering. Sadly changing the supply voltage is out of the question. So I added a filter network to the input as described in Figure 24 in the INA21x datasheet. Values are 100Ohm Rs and 22µF Cf. The output is now reasonablely smooth. Using lower series resistance would require absurdly high capacitance.
The Problem is now that the Gain has quadrupled?! I have checked a second PCB and the gain there was also around 380. After taking a closer look it is pretty sure that the root cause has to do with the input filter: Right befor the filter or rather the output of the current shunt the signal is noisy but ok. After the series resistors some voltage that is proportional to the monitors output adds to the input signal causing the high gain. For example: Current is 10A, shunt voltage is around 2,5mV and Ein+/- dif is 10mV. From this stage on the monitor works as designed. It takes the voltage differential amplifies it by 100 and outputs it.
I am aware that it is not recommended to use series resistors higher than 10 Ohms but there are enough examples of people succesfully doing just that with even higher resistance than I attempt to do.
