A compensation circuit senses the delta difference between d+ & d- and then uses a DAC  to compensate d+ & d- until it has 101010 patterns at the comparator output (conventional digital compensation circuit).

The drawback of this circuit is that it does not know which compensation codes is the closest to the  zero difference (assume no glitch) since the analog value cannot be stored.

is there a way to know which code is the best code? the present code or the previous code?

Quote:
Forget what is closest to the Actual Value. This is the funda of Digital Calibration.
You set your resolution acc. to max. error you can tolerate.
After that, you just have to accept one or two bit error depending on your scheme.
Resolving / Rounding off final code requires Delta/2 comparison, which again means higher resolution either in analog OR Digital cktry --> Hardware Overhead.