INA169NA

Alex,

After examining your system, I think you are driving the INA169 too close to the zero mark to achieve reasonable results. The recommended V_sense of this part is 100mV. By figure 4, at room temperature, you are closer to 7-9% error, on an incredibly sharp slope. The slope of the error being introduced at the 10mV range is extreme, and any deviations caused by external factors are contributing to the inaccuracy of the part on the bottom end. A way to check this would be to take the inductor off your board and inject a test voltage into the part more in the flat region of the curve of figure 4 (somewhere around 25-50mV differential), and see if this mitigates the error of your system. There is a 5% tolerance of the DC resistance of the inductor, but it seems like you've already taken that into account. You also need to look at the tolerance of the 100K resistor setting your gain on the output end, as this will also contribute error at the output. If this experiment provides more accurate results, it would correspond to finding an inductor with a higher DC coil resistance to increase the value of V_sense into that flat region of the curve.

Alternatively, you could look at our parts LMP8640 and LMP8645. The LMP8640 is a socket equivalent part that can be driven more closely to zero without increasing error as you approach zero. It is available in 100 V/V. The LMP8645 has the ability to adjust the gain via resistor as the INA169 does, but requires a 6th pin.