If your readings are full scale (ie zero mm reads 0 (ish) and 50mm reads 1020 (or so) then your error is 0.2/50 * 1024 = 4 LSB which is to be expected.
Likely the best way to reduce the error is to average multiple readings as explained here
If your readings are full scale (ie zero mm reads 0 (ish) and 50mm reads 1020 (or so) then your error is 0.2/50 * 1024 = 4 LSB which is to be expected.
Likely the best way to reduce the error is to average multiple readings as explained here