I noticed when assessing exposure times using any of the R_E_R_F models, often times the best exposure for larger features is not necessarily the best exposures for small ones.

For example, judging by the three exposures below, the shorter exposure sample on the right yields the best center target result but all of the smaller details are lost due to underexposure. The longer exposed tile on the left retains most of the smaller detail, but all of the larger features are overexposed.

I think a non-linear exposure scheme where smaller features get slightly longer exposure may work best. This of course extends the total exposure time so print duration will increase some, but if at the end it produces better results, then it may be worth it. Perhaps tolerance compensation may be another way, but that too is linear so if you apply a -.1mm compensation, anything smaller than .1mm gets obliterated so not sure if that would help either. What do you think of non-linear exposure?