INA213: Gain error factor and input resister value

Daisuke,

The answer is both yes and no. The equations in section 7.4.1 are applicable, and there is even a simplified expression for the INA213:

However, these equations examine only the ideal scenario where the internal resistance is at the outlined value discussed in the datasheet section. The problem here is that you also need to take into consideration the potential tolerance in the internal resistances:

This is why we advocate the use of small resistances, as <10Ω's will have minimal bearings on all of these possibilities inside of this tolerance. Using a resistor value as large as 1k will introduce significant error into the measurement as even after you plan the operating point using this provided equation, the value of Rint inside this equation will swing by this tolerance. 

If the customer needs to use this type of topology, I would recommend looking at a device such as the INA186, which has a capacitively coupled front end, and current limiting resistors have less of an effect on the gain error. Check out this app note for additional information. 

Daisuke,

Kai is correct on all counts here. 

The 30% listed is the maximum deviation in these resistors. 

Regarding the temperature coefficients of the internal gain network, I've never looked "under the hood" to fully understand this on a transistor level, but I would speculate that the temperature coefficient of this gain network is correlated at least in part to the gain drift of the device, which is a worst case 10ppm. As Kai points out, keeping the external resistors at a value <10Ω will be far more effective in error mitigation here, as the resistors internally are trimmed to each other, which is how the device reaches the values of datasheet spec.