Hi Petar,

The error that you are getting is due to the 3rd reflection inside the Prizma.STEP object, namely the roof angle on the prism. When a ray hits a surface, the surface itself must be differentiable at that local point so the normal vector can be calculated; this means the surface must be smooth and cannot have an abrupt angle like a roof line.

In sequential or mixed-mode, almost all analyses rely on the chief ray, which goes directly down the main optical axis of the system.  When this chief ray hits the roof exactly at the roofline, the normal cannot be calculated and a ray trace error occurs.  There are a few ways to avoid this:

1. Use pure non-sequential mode.  NSC does not rely on chief or marginal rays for analysis so the ray tracer can throw away ray errors and still provide accurate information.  With that said, non-sequential mode does not provide wavefront analysis, only ray-based geometric analysis.
2. Shift your roof-prism slightly off the optical axis so the chief ray can be traced.  This will allow sequential mode to calculate the exit pupil & wavefront maps for diffraction analysis, but you will have a slight shift for your wavefront, akin to a slightly tilted element.  This might be a small enough error for your calculation.

You can always ignore these errors by clicking on the Ignore Trace Errors in the NSCE, but if you can’t trace the chief ray in sequential then many analyses will not be valid.

OK, I see. Thank you for this exhaustive explanation. I moved the prism 1 micron to the left and it solved the problem. The result is good enough.

I also tried filetting and chamfering the rooftop, but this also caused artefacts.