Setting Up an Optimization Analysis
Use optimization to vary predefined variables in the nominal design to search for the solution that best satisfies a set of user-defined goals or cost functions. Optimetrics modifies the variable values until the minimum is reached with acceptable accuracy.
- You can define more than one optimization analysis setup per design.
- You can create an Optimization setup before defining variables, but all variables must be defined before you start the Optimization analysis.
- Once you have created an optimization analysis setup, you can copy and paste it, then make changes to the copy, rather than redoing the whole process for minor changes.
To provide a broad range of capability, Optimetrics incorporates the following types of numerical optimizers:
- Sequential Nonlinear Programming (Gradient) (SNLP)
- Merit-based Sequential Quadratic Programming (Gradient) (MBSQ)
- Sequential Mixed Integer Nonlinear Programming (Gradient and Discrete) (SMINLP)
- Quasi-Newton (Gradient)
- Pattern Search (Search-based)
- Genetic Algorithm (Random search)
- MATLAB
- Screening (Search based)
- Multi-Objective Genetic Algorithm
- Non-linear Programming by Quadratic Lagrangian (Gradient)
- Mixed-Integer Sequential Quadratic Programming (Gradient and Discrete)
- Adaptive Multiple Objective (Gradient)
- Adaptive Single Objective (Gradient)
Click the links above to view the setup procedure for each optimizer. Options for the analysis are listed in the table.
You can also use these optional optimization solution setup options:
- Modify the starting variable value.
- Modify the minimum and maximum values of variables that will be optimized.
- Exclude variables from optimization.
- Modify the values of fixed variables that are not being optimized.
- Set the minimum and maximum step size between solved design variations (For the quasi-Newton and Patterns Search optimizers, Variables tab).
- Set the minimum and maximum focus size. (For the SNLP and SMINLP optimizers, Variables tab).
- Set Linear constraints.
- Request that Optimetrics solve a parametric sweep before an optimization analysis.
- Request that Optimetrics solve a parametric sweep during an optimization analysis.
- Automatically update optimized variables to the optimal values during an optimization or after an optimization analysis is completed.
- Change the norm used for the cost function calculation (Advanced Option)
- Set HPC and Analysis Options to select or create an analysis configuration.
- Enable a fast calculation-update algorithm to speed up Optimetrics and report updates during Optimetrics analyses. By default, the fast calculation-update option is enabled whenever it is applicable.
Sweeping or using a complex variable is not allowed in any optimetrics setup, including optimization, statistical, sensitivity, and tuning setups.
Improper or undefined simulation setups will cause errors during Optimetrics Analysis. To verify the analysis setups, select Analysis > Analyze in the Project Manager pane to analyze the nominal circuit and review the messages in the Twin Builder Message Manager pane prior to running the Optimetrics Analysis.