Gaussian quadrature is used in optical design optimization to efficiently calculate integrals using only a few carefully chosen sample points, rather than uniformly sampling all points. This allows optimizations to be performed using orders of magnitude fewer ray traces. The sample points used in Zemax are the intersections of rings and arms on the pupil. The number of rings used depends on the order of the significant aberrations in the design. Gaussian quadrature has been shown to greatly speed up optimizations for systems with annular pupils or wide wavelength ranges.
Gaussian quadrature is used in optical design optimization to efficiently calculate integrals using only a few carefully chosen sample points, rather than uniformly sampling all points. This allows optimizations to be performed using orders of magnitude fewer ray traces. The sample points used in Zemax are the intersections of rings and arms on the pupil. The number of rings used depends on the order of the significant aberrations in the design. Gaussian quadrature has been shown to greatly speed up optimizations for systems with annular pupils or wide wavelength ranges.
Gaussian quadrature is used in optical design optimization to efficiently calculate integrals using only a few carefully chosen sample points, rather than uniformly sampling all points. This allows optimizations to be performed using orders of magnitude fewer ray traces. The sample points used in Zemax are the intersections of rings and arms on the pupil. The number of rings used depends on the order of the significant aberrations in the design. Gaussian quadrature has been shown to greatly speed up optimizations for systems with annular pupils or wide wavelength ranges.
Gaussian quadrature is used in optical design optimization to efficiently calculate integrals using only a few carefully chosen sample points, rather than uniformly sampling all points. This allows optimizations to be performed using orders of magnitude fewer ray traces. The sample points used in Zemax are the intersections of rings and arms on the pupil. The number of rings used depends on the order of the significant aberrations in the design. Gaussian quadrature has been shown to greatly speed up optimizations for systems with annular pupils or wide wavelength ranges.
When you tell a code like Zemax to optimize your design, it launches into a computationally intensive process of changing variables and assessing whether the system is improved or not. Each assessment potentially involves a lot of calculations, so efficient procedures are important.
Previously we looked at these patterns of rays launched into the pupil from a given object point for generating spot diagrams.
Uniform random square triangular polar
Now we consider how to launch FEWER rays into the pupil from a given object point during the more computationally intensive optimization process …
J. Geary, Introduction to Lens Design
with practical Zemax examples, Figs. 7.2, 7.4 1 Optical Design (S15) Joseph A. Shaw – Montana State University
Why not use uniform sampling for optimization?
Here’s an illustration of why uniform sampling is simply not practical in the computationally intensive procedure of optimizing even a simple lens system.
• If 10 sfcs, 100 wavelengths, 100 field pts, 1000 rays per wl & field pt 108 rays per system • Need to trace 1024 systems × 108 rays/system = 1032 ray-surfaces
If your computer can compute 109 ray-surfaces/second, you need 1023 seconds…
3 × 1015 Years!!!
Adapted from a talk given at MSU in
2011 by Brian Bauman from Livermore National Laboratory. 2 Optical Design (S15) Joseph A. Shaw – Montana State University
Gaussian Quadrature Efficient optimization requires an efficient way to integrate over a range of pupil locations, wave- lengths, field angles, etc. Optical design uses the Gaussian quadrature numerical method to efficiently calculate integrals as a weighted sum at just a few carefully chosen points (“nodes”).
1 𝑛
𝐼= 𝜔 𝑥 𝑓 𝑥 𝑑𝑥 ≈ 𝜔𝑖 𝑓 𝑥𝑖 −1 𝑖=1
• 𝜔 𝑥 is a weighting function and 𝑓 𝑥 is function being integrated
• Uses knowledge of maximum order of integrand • Exact result for polynomial integrand of order 2𝑛 − 1 using 𝑛 nodes • Orders of magnitude fewer evaluation points than uniform sampling
Actual interval converted to [-1,1] using “change of interval” equations
See, e.g. http://en.wikipedia.org/wiki/Gaussian_quadrature
Key papers: 1. G. W. Forbes, “Optical system assessment for design: numerical ray tracing in the Gaussian pupil,” J. Opt. Soc. Am. A 5(11), 1943-1956 (1988). 2. B. J. Bauman and H. Xiao, “Gaussian Quadrature for optical design with non-circular pupils and fields, and broad wavelength ranges,” Proc. SPIE 7652, 76522S (2010). Adapted from a talk given at MSU in 2011 by Brian Bauman from Livermore National Laboratory 3 Optical Design (S15) Joseph A. Shaw – Montana State University
Gaussian Quadrature on circular pupil
Gaussian quadrature applied to the variable 𝜌2 over the interval [0,1] yields the radii that are used in the Zemax default merit function.
𝜌= 𝐿𝑖 ′ = 0.3357, 0.7071, 0.9420
Adapted from a talk given at MSU in 2011 by Brian Bauman from Livermore National Laboratory 4 Optical Design (S15) Joseph A. Shaw – Montana State University
Rings and Arms
Zemax refers to the pupil Gaussian quadrature nodes as intersection points of rings and arms. The intersection points of these rings and arms mark the pupil points through which rays are traced for optimization calculations.
Arm 6 Arm 1
Arm 5 Arm 2
Arm 3 Arm 4
How many rings should you use? … It depends on the order of your significant aberrations. Using (2n-1) rings will provide optimal results for ray aberrations of order n. If in doubt, you can increase the number of rings until the result stops changing by more than ~1%.
5 Optical Design (S15) Joseph A. Shaw – Montana State University
Example of GQ used for annular pupils
Here is an example of using GQ to speed up optimization for optical systems with annular pupils. This shows that several orders of magnitude fewer rays are required to achieve comparable rms spot size error for the Large Synoptic Survey Telescope (8.4 m primary mirror!)…
http://www.lsst.org/files/docs/LSSToverview.pdf 6 Optical Design (S15) Joseph A. Shaw – Montana State University
Example of GQ used for wide wavelength range
Three-glass apochromatic lens designed for visible through short-wave infrared
B. J. Bauman and H. Xiao, “Gaussian Quadrature for optical design with non-circular pupils and fields, and broad wavelength ranges,” Proc. SPIE 7652, 6522S (2010). 7 Optical Design (S15) Joseph A. Shaw – Montana State University
Merit Function
The merit function in Zemax is defined as a weighted sum of the differences between target values and current values of design parameters that are designated as variables.
2 𝑖 𝑊𝑖 𝑉𝑖 −𝑇𝑖 2 MF = , 𝑖 𝑊𝑖
where 𝑊𝑖 is the weight, 𝑉𝑖 is the current value, and 𝑇𝑖 is the target value.
8 Optical Design (S15) Joseph A. Shaw – Montana State University
Remember: Many Possible Solutions
If you consider your design space as a surface, there can be many local maxima and minima. This means that an optimization algorithm rarely finds the best solution at first. Sometimes it is necessary to “beat on” the design to force it into a new local minimum.
Solar Vision Leveraging Automated Detection and Quantification of Solar Panels in Urban Areas To Promote Renewable Energy Adoption and Achieve Sustainable Development Goals
Solar Vision Leveraging Automated Detection and Quantification of Solar Panels in Urban Areas To Promote Renewable Energy Adoption and Achieve Sustainable Development Goals
