Hello user5304320,

To answer your first question you should understand that inside the INA219 and INA3221 there are physical resistances from the IN- pin to the GND pin. For the INA219 this resistance is 320kΩ and for the INA3221 this resistance is 670kΩ. These resistors are to measure VBUS and both of these values are in the respective datasheets. There is nothing you can do to this resistance, it will always be there.

These internal resistors induce unbalanced input bias currents. For INA219, while IB+ will remain around 20µA, the IB- will be 20µA + VCM/320kΩ = 51.25µA assuming your bus voltage is 10V. Now usually this is not a problem when sensing current, but once you insert external filter resistors on the inputs Rin+ = 10Ω and Rin-=10Ω, then you will induce unbalanced voltage drops across each resistor, which will then increase your offset error as seen with 313µV.

Now if you follow the last link on the last post, you will see the possible solution you could implement. This solution would be to add an external 320kΩ resistor from IN+ pin to GND. Once you do this then both IN+ and IN- will equal same current which will be very beneficial and improve accuracy if you have implemented an external input filter to the INA219. See below.

Hope this helps,

Peter Iliya

Current Sensing Applications

Hey user5304320,

The internal resistor for IN- to GND (320kΩ for INA219 and 679kΩ for INA3221) are there to only sense the bus voltage. These devices are actually power monitors because they measure current and voltage at same time and produce a power measurement.

We don't add a balancing internal shunt resistor at IN+ to GND because it usually is not necessary. The imbalance in input bias current IN+ and IN- is only really a problem when resistors are placed on the input pins to the device. Once there is input resistors then the imbalance in current will turn into an imbalance of voltage drops resulting in a larger offset. Keep in mind that will adding a balancing resistor will improve offset, the internal resistor from IN- to GND can vary conservatively +-30%.

Hope this helps,
Peter Iliya
Current Sensing Applications

The internal Rin- resistors to ground sense the bus voltage because they draw current from the bus. This current equals VBUS/Rin-. The internal ADC will measure the voltage drop across Rin- (320kΩ or 670kΩ) to determine the bus voltage.

The input Rfilter is recommended in INA3221 datasheet for two reasons:
1. The sense current contains transients at harmonic of 500kHz.
2. If VCM (common mode voltage) will change very quickly (large dV/dt) and there are no electrlytic shunt capacitors from IN+/IN- to ground.

We list reason 1 because we do not want input signals to interact with the sampling frequency of internal ADC.

We list reason 2 because continously switching the VCM very quickly from 0V to 26V will overtime damage the internal ESD cell structures. This situation would occur in the in-line current sensing of a H-bridge controlling a motor. This is why we describe this part only for high or low side sensing, not in-line sensing where the VCM can be pulled very quickly from ground to a high voltage repeatedly. For an in-line application we recommend the INA240 because we manufactured this device with PWM rejection internal circuit. The IN240 uses different ESD cell structures that are made to withstand fast-switching VCM. The INA3221 was not.

Anyway, if your system cannot avoid fast switching of VCM and you cannot insert electrolytic capacitors at the input to slow down the VCM the INA3221 sees, then you can resort to adding the Rfilter. The Rfilter helps because it limits the current sunk into the device input pins and ESD cells during a high dVCM/dt event.

A third reason you would add the Rfilter is if you expect large VCM transients greater than 26V at the inputs. The datasheet suggests using TVS diodes or tranzorbs to clamp the input voltage and divert current, but a Rfilter network would help limit the current and thus power needed by TVS diodes.

If you end up needed a Rfilter network for whatever reason, you could consider the INA190 which has a much lower input current of 2 nA. This would allow you to use a much high input resistor (e.g., 1kΩ) without sacrificing much on accuracy.

Here is some educational material on protecting device inputs:
www.ti.com/.../tidu473.pdf
www.ti.com/.../slyt492.pdf
www.ti.com/.../slla305.pdf
e2e.ti.com/.../700891
e2e.ti.com/.../699662
e2e.ti.com/.../700877

Hope this helps
Peter Iliya