I have a query
During optimization process Zemax varies defined variables and tries to minimize the merit function (optimization).
But still I dont know upto what decimal places Zemax varies the defined variables and tries to achieve the better solution.
For eg: Suppose I have defined a variable of 'Radius of Curvature' having value 100mm. During optimization process upto what decimal
places Zemax varies this value and tries to find the next best solution.
Is this trying like new radius of curvature of 100.000001mm and computing the merit function, then next radius of curvature of
100.000002mm and computing new merit fuction etc..
Since my practical limitation is 0.01mm (10 micron) or 0.001mm (1micron), so whether I can tell Zemax to vary the given variables in
minimum step size of 0.01mm or 0.001mm only, not finer than this. More fine computation beyond practical limit is wastage of computation energy
and time.
In short whether I have control over given variable to tell Zemax upto what decimal place the variable is allowed to vary during trying for
next best solution(optimization)?
[Unwanted minimum step size is just wastage of computing time, so my practical limit(tolerance) is nearly 1 micron (0.001mm) only]
Please help me with the above query.
Thanks and Regards
--Satheesha.
Optimization: How to put limit on the number of Decimals to the variables?
+3
If you use Automatic cycles, OS will keep optimizing until the difference in merit function betwen cycles is something like 10^-8 of the starting value. So it's not the variables, its the MF that controls the number of cycles.
Your best best is to just run 5 cycles or some number rather than using Automatic.
OS uses double precision floats for all variables.
- Mark
Thanks Sir,
As you said OS uses double precision floats for all variables.
But from Google search I found that smallest double precision floats number is ±10^(-323) [Please correct me if I am wrong].
Isnt it a very small change in variables to find next minimum merit function?
[One question also in my mind is that why OpticStudio uses floats as provided from OS.. instead cant it be use floats limit as per own requirement]
In other words, I strongly recommend OpticStudio Scientists and Engineers to add a feature where we choose decimal places upto
which Zemax allowed vary the variables to find next best merit function during the optimization, not finer than this limit.
I guess its more efficient way of using the processing time.
Please clarify.. Thanks Mark..
Regards
+1
Hi Satheesha,
I'm pretty sure that there is a difference between the smallest number a chip can represent and the smallest difference between numbers that it can work on. In other words, if that 10^-323 is used with much larger magnitude numbers for math purposes, it will be rounded out entirely.
But what Mark said is the essential lesson for the optimization. Typically a double precision number is around 16 decimal places, but the optimizer goes to about 8. It does not try for higher precision. And more importantly, it is the merit function that is calculated to this precision, not the value of any given variable. One of the reasons for having different weights on operands is to account for some of them returning values of drastically different magnitudes and trying to compare them. I'm not a Zemax developer so I can only make reasonable guesses here, but the optimization routines are based on continuous mathematical functions, so trying to impose a discreteness to the internal calculations, like the step sizes for physical quantities, might be a more complicated matter than we realize.
+3
The best way to approach granular values in the LDE is to optimize and then set the values in the LDE to the desired precision. This is what we would do prior to tolerancing, for example.
For radii, I would suggest running the test plate fitting tool, even if you are not using that manufacturer (but especially if so). This will set all radii to sensible values taken froma manufactuere's tooling list. For thicknesses, when you run the design lockdown tool you can specify the precision of the thickness values.
Optimization is a continuous process, but manufacture is not so it's best to keep an optimize-versus-tolerancing mindset here.
- Mark
Hi Mark Sir and Kevin,
Thanks very much for your responses.. However I will go through once with Zemax Help manaul about the details of Optimzation process, then I will come back to you.
Thanks
--Satheesha.
Reply
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
Need more help?
To Chinese users:
Do not provide any information or data that is restricted by applicable law, including by the People’s Republic of China’s Cybersecurity and Data Security Laws ( e.g., Important Data, National Core Data, etc.).
不要提供任何受适用法律，包括中华人民共和国的网络安全和数据安全法限制的信息或数据（如重要数据、国家核心数据等）。