Optimization tricks. Share your experience!

  • 15 April 2020
  • 8 replies
  • 900 views

Userlevel 6
Badge +2

Optimization is extremely important in a computer-aided design. However, it's never as easy as just a click on a button. We usually need to carefully set up the variable and merit function. While there are no a general rules to follow for all systems, there are still some useful trick to follow for most of common cases.



Here I'm sharing some from my experience. If you put yours in reply, I will also update this thread.:)


[Michael Cheng]


1. Keep this in mind: Always check and consider whether a variable is really required to be a variable.

* If the variable won't changing system performance much, turn off it first. You can turn it on at final stage for a fine tune. A typical case is the thickness of a lens. In many cases, they contribute much less than those AIR thicknesses in the system.

* Before careful that, if you have a redundant variable, for example Radius on an isolated STOP, the optimizer may be confused and just cannot work well.

* If during optimization, a variable just tend to reach its limit defined in merit function, just set it to the limit value and turn off variable. A typical case is you may find that optimization want a lens to be as thick as possible and always reach the limitation you set up in merit function.



2. In merit function, if a boundary cannot be exceeded, do not put high weight. Instead, shrink or swell the boundary a little. For example, if the central thickness of the lens should be smaller than 2 mm and you find the optimization always give 2.01 mm. Try to shrink the limitation to 1.99 instead of set a Weight to 1000 or more. A too high Weight make the operand too important and optimizer may find the others are just not important and it's very slow or even not working in improving other operands.



[Heng Li]


* If you're designing an imaging system, always give Contrast Optimization a try! Systems using Contrast Optimization ususally have a faster converging speed. 

* If you're optimizing a non-sequential illumination system or an imaging system with very big aberrations, Orthogonal Descent usually works better. 




8 replies

Userlevel 1
*`If you're designing an imaging system, always give Contrast Optimization a try! Systems using Contrast Optimization ususally have a faster converging speed. 

* If you're optimizing a non-sequential illumination system or an imaging system with very big aberrations, Orthogonal Descent usually works better. 

Hi Michael Cheng,


 


Your tips are really helpful. I have some following questions related to optimization. Can you please go through those?1


(1) Which operand can be used to set the chief ray incident on a particular point on the image plane?

(2) Is there anyway to fix the variable limit of the lens to vary in that particular limit? Let's say, radius is variable and how this we can limit in a particular range?


Best and Regards,

Faheem Ahmad

Userlevel 6
Badge +2

Hello Faheem,


Thank you for reply! Both are good questions and please see my answers below.


(1) Probably the easiest way is to define the Field with Real Image Height. There is no way to know where the chief ray would land on image plane if the Field Type is not Real Image Height.


(2) Usually the first method to try is to increase the boundary opearnd's weighting in merit function. However, sometimes this may not work or requires too high weighting that even affect the optimization stability. In case increasing weight doesn't work, I would suggest you just set the limitation to be smaller or larger. For example, assuming you want the Radius to be smaller than 100 in merit function, however you find, after optimization, the resulted Radius is always slightly larger, like '100.1'. And it doesn't work if you set the Weight of the operand to even 100 either. In this case, I would just set the limitation to 99.9. If that doesn't work, try 99.8. 


 


Hope this works for you!


 


Michael


 

 


Thank you Michael for your helpful response, it is working.


Best,

Faheem

Userlevel 3
Badge

If you are designing a hand-held optics, in terms of tolerancing and manufacturing. it is recommended to select diameter-to-thickness ratio of 8:1

In ZOS we can define those constraints using MNDT and MXDT operands by giving an interval

6< Diameter to thicknes ratio of the lens <10

 

 

 

Userlevel 3
Badge

The  Z-factor (Karow factor) describes the ability of a lens which leads automatically centering itself between bell clamps during centering in the fabrication and testing processes.

 

The Karow factor (Z factor) is given by 

D1 and D2 are clear apertures of the lens. R1 and R2 are the radii of the lens surfaces.

 

In ZOS we can define those using the operands below;

 

 

 

 

 

You can have a look at the webpage for further details

https://www.photonics.com/Articles/Nuances_in_Optical_Design_for_Manufacturing/a64048

Userlevel 3
Badge

The  Z-factor (Karow factor) describes the ability of a lens which leads automatically centering itself between bell clamps) during centering in the fabrication and testing processes.

 

The Karow factor (Z factor) is given by 

D1 and D2 are clear apertures of the lens. R1 and R2 are the radii of the lens surfaces.

 

In ZOS we can define those using the operands below;

 

 

 

 

You can have a look at the webpage for durther details

https://www.photonics.com/Articles/Nuances_in_Optical_Design_for_Manufacturing/a64048

to mantain the lens ability to self-center between two bell chucks, Karow (Z-factor) is recommended to be greater than 0.56

Userlevel 3
Badge

 

Hello Everyone,

Optimization is sometimes extremely painstaking in complex designs. I want to share one of my experience;

 

Decreasing the F/# or field of an existing design can lead to ray failures during lens evaluation and optimization. Sometimes we can get warning from OS as; total internal reflection (TIR) or ray intercept failures (where rays physically miss a surface). The solution is to gradually step by step walk up to the desired aperture or field in small increments, and accordingly optimize the lens at each step.

Reply