If you need to measure the "zero offset" values, the only way to accurately accomplish this is to tie the (+) and (-) inputs together for each channel and measure the result. I have duplicate SCXI terminal blocks for all of the channels I need to measure offsets and I install this "shorting" block at the beginning of each calibration period. I store these results and subtract them during post-run processing. The raw data is always acquired and stored. I remove the zero offsets ONLY during data processing in order to have "traceable" data acquisition.
P.S. If you are measuring ONLY 40 micro-volts of offset, then I congratulate you. Your data system is very clean. As a point-of-reference, the 6052E series (16-bit) boards have a "best" data resolution of 15.3 micro-vol
ts ( 2*0.5 / 2^16 ) for the input range of (+/-) 0.5V. I really doubt that you will be able to achieve better results than you currently measure. It is physically impossible to ever achieve a "perfect" zero value.