Photon Engineering has developed and integrated an Optimization feature into FRED. With a heightened awareness of the flexibility required to address a general optimization problem, our developers have conceived and implemented a methodology of optimization with an equally general framework able to adapt to the wide variety of configurations encountered in modern optical technology. FRED's Optimization feature has the ability to treat virtually any quantity of an optical model as a variable. Performance criteria are specified through built-in or User-defined merit functions and applied with Simplex (multi-variable) or 1-D (single variable) algorithms. The user has an option to accept the optimized configuration or return to the initial values of all variables involved.
This feature can be accessed by selecting Optimize > Define / Edit from the menu.
Please see the following help topics: •Optimize - Variables •Optimize - Merit Function Aberrations •Optimize - Method •Optimize - Output Results
The optimization method controls allow the user to explicitly set how the optimization will proceed.
Once the Variable(s), Merit function(s) and Optimization Method are configured, proceed to the optimization step by selecting Optimize > Optimize from the menu.
Results from the optimization process will be printed to the output window.
When the optimization cycle completes, the model parameters will be set to their optimum values obtained during the final iteration step and any traced rays are deleted. Any subsequent analysis of the current system configuration requires that source rays be re-traced.
If so desired, model parameters can be restored to their starting values before the previous Optimization cycle using the Restore Previous Starting Values option. Select Optimize > Restore Previous Starting Values from the menu.
Approach to Optimizing CAD Geometry It is incumbent upon the user to carefully evaluate the nominal configuration of their optical system with an eye toward expediency when planning and designing an optimization routine. The design of reflectors serves as one example where alternative approaches should be explored. In many cases, the general shape of a reflector is initially conceived and its packaging issues explored using a CAD tool. When these CAD models are imported into FRED, they are represented exclusively as NURBs which are a parametric representation of arbitrary order often containing copious numbers of control points and weight factors. Large numbers of control points necessarily inhibit the ability to optimize a surface due to the sheer number of variables involved. Reducing NURB order during import, while affecting the accuracy of surface shape, will decrease the complexity of objects by reducing the number of control points thereby making optimization a more tractable problem. An additional benefit to NURB order reduction is increased raytrace speed, an extremely beneficial attribute in illumination problems that require large raysets to achieve adequate SNR.
When requested, FRED prints values of the variables and n+1 merit functions where n is the number of merit functions (see Simplex Optimization). The tag "<---variables printed" indicates which of the merit functions is closest to the local minimum.
Noise in general optimization routines can be a serious issue and sabotage convergence. Make sure that the number of rays traced minimizes the statistical noise in your merit function. If the system does not converge, try tracing a factor of 10 more rays.
All currently active sources will be used in the optimization raytrace. Any sources referenced by the user aberration scripts must be traceable or the optimization will fail (with the message "cannot compute merit function"). This limitation extends to referenced nodes as well.
Analysis Surfaces are required as part of the Merit Function definition. However, unless they are used to evaluate one of the "Energy" quantities (Irradiance, Intensity, Energy Density, Color Image), their sole purpose is to provide access to Ray Filters during the optimization process.
Optimizing Ball Lens Coupler - Optimizes the position of a ball lens to affect maximum power onto fiber face LED Color Optimization - Optimizes the powers of three LEDs to produce a specific color (chromaticity coordinates)
|