This is an interesting case. With the stop on its own surface, there is no power between the object and the stop and so no pupil aberrations. With the stop on a surface with power, there is power and hence aberrations.

I would turn on ray-aiming in the example where the stop is the actual lens surface. Remember the object is supposed to radiate into all angles, and so overfill the stop. If you want the limiting ray to be the one that hits the embedded stop, ray-aiming should be on.

Both cases are correct, it just depends on which is the physically correct one for your system.

Mark

Hey @Jeff.Wilde & @Mark.Bennahmias 

To continue the conversation a bit, except for “speed” (which isn’t really a modern concern) and Ray Aiming failures (which Zemax is actively addressing):

Is there a time when Ray Aiming shouldn’t used?

Zemax historically has not set Ray Aiming on by default because of speed & failure modes, but if those two considerations are taken off the table, is there a scenario when no Ray Aiming is better than Paraxial Ray Aiming?

Also, what’s a more physically accurate representation of the Exit Pupil...a dummy surface with (virtual) propagation or the actual curved surface itself?  Would Ray Aiming on a curved surface (without a dummy surface) be more physically accurate because you’re mapping the Marginal Ray to the edge of the physical aperture itself and not to a dummy surface?

I’ve seen all combination of files (ray aiming on/off, paraxial vs real RA, dummy surface vs actual surface for the Stop, etc) and it would be nice if the “Zemax Community Brain Trust” could have an agreed upon recommendation for where the Stop should go.  I know there was a discussion about 8 months ago but I think it stopped before a meaningful conclusion:

Are you choosing a physical stop? | Zemax Community

I personally think the only difference between the “virtual propagation” and the “embedded” model is where the Chief Ray is...I also personally think that beyond 3rd order wavefront values which require Chief/Marginal rays, you would be using more robust Centroid-based metrics for your analysis, so the more simple & intuitive virtual propagation is both “good enough” and accurate.  This “virtual propagation” method is more realistic since the Exit Pupil is the image of the Stop from the detector plane and the “image of the Stop” can’t know an “embedded” surface..  I also think you should always use Paraxial Ray Aiming (as long as it doesn’t fail).

Thanks Dirk! If there’s any power between the stop surface and the object, ray aiming should always be turned on.

Personally I think it should always be on and the user should not be able to turn it off. The case of zero power between the object and stop is so fast that having ray aiming on does not affect things anyway, but could easily be trapped in the code anyway. From the perspective of a 21st century user rather than someone banging away with a 386 or 486 processor and 1 MB of RAM, this switch is otiose now.

But rather than wait for Ansys Zemax to change things, just start OS up, turn ray-aiming on in the file that auto-loads and then just save it as lens.zmx. Never be bothered by 1990-era workarounds again!