This video is taken from the FDTD Learning Track on Ansys Innovation Courses.
Links
- Understanding analysis and layout modes
- Parameter sweeps, Optimization and Monte Carlo analysis
- Convergence testing for the FDTD solver
Transcript
The typical design workflow involves multiple iterations of simulations where after setting
up the simulation, running the simulation, and analyzing the results, changes are made
to the design based on the results, the simulation can then be run again, the results analyzed
and so on until the desired performance of the device is achieved.
There is also a built-in parameter sweep, optimization, S-parameter sweep, and Monte
Carlo analysis tool in the graphical user interface in the Optimizations and Sweeps
window.
These tools can also be used in the design process to sweep a parameter of the design
and observe trends in the results, or find the optimal values for a set of parameters
that optimize a figure of merit.
We won't cover how to set up these tasks in detail here, but you can find the details
and tutorials for using these tools in the related links below.
Some general tips for the design process are to start with a relatively coarse simulation
mesh for initial simulations.
Using a coarse mesh allows the simulations to run quickly which is desirable since you
often need to run many iterations to narrow down the range of design parameters that result
in an optimal design.
Then, convergence testing can be done to obtain high accuracy results as a final step.
Convergence testing is a process of varying settings that can have an effect of the numerical
accuracy of simulations, such as the mesh step size in order to quantify the level of
numerical error to make sure that numerical error in the simulation results is at an acceptable
level.
Convergence testing will be covered in detail in one of the following units.
Let's go through an example workflow for designing the optimal shape of a taper.
Here the taper width is described by an equation which includes a variable m that we want to
vary in the design.
We want to find the value of m which results in maximum transmission through the taper.
We can start by using a coarse mesh and sweeping over a large range of values of m to get a
rough idea of the range where the transmission is maximized.
This allows us to narrow down the range of m to sweep to get the final results.
To make sure the results will be accurate, we can perform convergence testing on a single
design to determine the settings which result in the required level of accuracy.
We can then use these higher accuracy settings to run a second sweep over a smaller range
of m values that we identified from the first sweep to get the final value of m that maximizes
transmission through the taper.
Since using the higher accuracy settings will require more computation time and memory,
it's more efficient to first run the simulations with a coarse mesh to narrow down the range
of m values compared to using the high accuracy settings to run all of the simulations over the
larger range.
We can save time by putting some consideration into the design workflow before starting to
run simulations.