INA293: Realistic Measurement Range/Accuracy

Dajon,

The short answer is no. There's a bit to unpack here, but while the INA293 is capable of handling common mode voltages up to 100V, no single device is going to capable of handling this range of measurement on its own.

Before we even get into the total error requirement, the first issue here is the sheer magnitude of current range the customer wants to measure here. If I were to use a G=20 device, then the maximum full scale range would dictate that Rshunt be 50mΩ, as .05Ω*5A*20=5V. For this same shunt, at the lower bound of 5nA, the sense voltage is 5E-9*0.05=250pV. The offset of the INA193A1 is 150uV, so at the lower end of this design, assuming only error from offset and nothing else, the error is literally 1,000,000%.

By the same token, if I were to try to design to the lower end of the FSR, I would need to achieve a >10mV output to satisfy the swing to GND limitation. Using a G=200 device, to satisfy this, I calculate you would need a 10kΩ resistor to achieve this. With this resistor, not paying attention to the bias currents creating additional offsets and errors, at the upper end of the measurement now of 5A, our output voltage is saturated (calculated Vout = 10^7 V), and you are measuring 50,000V in the shunt, for a power dissipation of 250kW. Keep in mind that this design path only guarantees my device to be operating linearly, and I haven't looked at the error implications of this lower sense voltage.

My point is that both of these scenarios are completely unrealistic due to the wide range of current being demanded here. The only realistic way to do this is to break the range down into realistic blocks that can then be designed to individually and stitched together. This is often done with an analog front end that is switched onto the same device, or several devices that each handle their own individual section of the current range.

Finally, regarding error, 0.1% error is going to be a challenging metric to hit, even at the full scale range, let alone on the reading, and we would need to clarify which error the customer desires. eFSR is typically a bit more forgiving than reading error, but there's more to unpack here. Typically, the gain error of the INA293 is .02%, while the worst case is 0.15%. This means that there will exist devices lot to lot that will not satisfy the 0.1% error criterion anywhere on the curve, let alone at lower voltages. This is also not taking any additional error sources into account.