Hi,

Let me use LTspice to illustrate why the original posters circuit doesn't work very well in practice.

**Simplified Model**If I build a model for the shunt part of the circuit and the voltage dividers that bring the shunt signals into the common mode range of the amplifier I get:

And the circuit has the desired output 1V per Amp of load current.

**Component Tolerances**As described by other contributors, the accuracy of the voltage dividers is very important in this circuit. If the two dividers are not the same a voltage will be appear between the two outputs.

In LTspice we can use Monte Carlo analysis to implement component tolerance. The format is:

{mc (nominal value, tolerance)}

{mc(10K,0.01)} this is a 10K 1% resistor.

The distribution is rectangular. That is equal probability of getting any value in the tolerance window.

.step param X 0 9 1

This causes the simulation to run 10 times, from 0 to 9 step 1

The results are:

If I increase the number of runs to 100 I get:

This is like building the circuit 100 times with 1% resistors, randomly selected.

IC manufacturers can do circuits like this, because the can trim and match the resistors.

Check LT1999

Link:

https://www.analog.com/en/products/lt1999.htmlI have attached the LTspice models.

Regards,

Jay_Diddy_B