high-side current sensing is much more accurate, without ground-loop issues.
Not true. Ground-loop issues are avoided by using a differential measurement of the voltage across the current-sense shunt resistor, whether that resistor is on the positive or negative rail.
In a typical circuit where everything is referenced to the negative supply rail, high-side sensing has two additional sources of error:
- For a given common mode rejection ratio, more of the noise present on the positive supply rail remains unrejected, compared to the much lower level noise on the negative rail
- The transfer of an analog signal from the positive side to the negative reference introduces inaccuracies
Therefore, as far as accuracy goes, sensing the negative rail current is much better. The only reason we don't do it all the time is if we cannot because the power supply and the powered circuit must use the same ground.