Solved

Optimization alternative to RELI for relative illumination.

  • 28 January 2020
  • 2 replies
  • 201 views

I notice that adding RELI to my merit functions significantly slows down the optimization speed. Is there an alternative for relative illumination optimization that is more computationally efficient? My application is general imaging lens design/optimization. 



Other than optimization speed, RELI seems to work fine, this is just a minor gripe that I realize might have a clever solution. 



Thanks!
icon

Best answer by Mark.Nicholson 28 January 2020, 23:21

View original

2 replies

Userlevel 7
Badge +3
Hi Sean,



RELI is a fairly big calculation, so I can see why it impacts speed. I'd suggest using it with an OPGT/OPLT operand so that you constrain it to be within a range that is acceptable, rather than targeting a specific number. That way the optimizer won't spend ages trying to achieve a specific RELI value to 64-bit accuracy. Something like:







will ensure the RELI is >90% without specifying an actual target value. It also means that as long as you get the RELI you want, it doesn't affect the rest of the optimization.



Keep the sampling as small as you can, too.



Last, the Effective f/# EFNO is used in this calculation, and it may be faster to use it and just save RELI for a final system check.



HTH,



- Mark
Thanks Mark,



OPGT does help significantly, I've been using that as a best practice for a while. It's one of the best tips I've gotten out of the Zemax training courses. 



I will try EFNO as an indirect check. Thank you for the tip!



-Sean

 

Reply