Sub-Resolution Assist Features (SRAFs) have been used extensively to improve the process latitude for isolated and semi-isolated features in conjunction with off-axis illumination. These SRAFs have typically been inserted based upon rules which assign a global SRAF size and proximity to target shapes. Additional rules govern the relationship of assist features to one another, and for random logic contact layers, the overall ruleset can become rather complex. It has been shown that model-based placement of SRAFs for contact layers can result in better worst-case process window than that obtained with rules, and various approaches have been applied to affect such placement. The model comprehends the specific illumination being used, and places assist features according to that model in the optimum location for each contact hole. This paper examines the impact of various illumination schemes on model-based SRAF placement, and compares the resulting process windows. Both standard illumination schemes and more elaborate pixel-based illumination pupil fills are considered.