How to determine the number of cells to use over a taper



When simulating a tapered waveguide or fiber using the EME solver, the taper will be discretized along the propagation direction using a specified number of cells in each cell group.

One way to make sure that you are using enough cells is to perform convergence testing by increasing the number of cells and re-running the simulation until the results converge, as demonstrated in this example:

Another tool that can be used is the error diagnostics which are returns results in the EME solver region object after running the simulation and performing the EME propagate step. The error diagnostics can return information such as the amount of field discontinuity, and power gain and loss between neighboring cells. More information about the EME error diagnostic results as well as an example showing how to use the results can be found on this page:


Will the above suggestion applicable to an S-band waveguide?

I understand that there is a spot size converter example that shows the idea of using the EMEsweep, where you can sweep over the taper length for s-parameters in almost no extra computational cost. My understanding is that you keep the modes calculated in each cell, and the propagate them for a longer distance to run the sweep. My question: Is this sweeping approach valid for an S-band too? I want examine the transmission (ie, s-parameter) as a function of s-bend length.


Yes, the same approach using convergence testing, and error diagnostics can be used to determine the number of cells to use over a bend, or any other portion of the structure where the structure cross section varies continuously along the x-direction.

You can also use the EME propagation sweep tool to sweep the length of the cell group region, however, for a bend it will not preserve the width of the waveguide since it uniformly shrinks or stretches the x span of each cell while keeping the same device cross section in each cell as the propagation distance is varied, so the device shape will be warped as illustrated in the following image: