This video is taken from the varFDTD Learning Track on Ansys Innovation Courses.
Transcript
We can compare the simulation results with the results from a 3D FDTD simulation of a
same device to verify the accuracy of the results.
The 3D FDTD simulation results that are shown are from the example linked below this video.
Here is the power spectrum from the drop port, which is the T result from the drop monitor
using the varFDTD simulation that we set up, plotted together with the same result from
the 3D FDTD simulation.
You can see that the height and wavelength of the peaks in the spectrum are shifted,
however the free spectral range is consistent between the two.
The differences are because there is some vertical coupling out from the device at the
coupling regions between the straight waveguide and ring.
The varFDTD method assumes that there is no vertical coupling which causes it to overestimate
the coupling efficiency of the device and it also introduces some error in the effective
index of the coupled mode between the straight waveguide and ring.
The difference in the effective index from the 3D case can be thought of as being equivalent
to a small change in the gap distance or coupling length which causes the peaks in the spectrum
to shift.
Since the 3D FDTD simulation accounts for loss in the vertical direction, the peaks
from the spectrum simulated using 3D FDTD have a lower amplitude.
The free spectral range, or spacing between the peaks, is the same between the two methods
since this value depends on the group index and path length which do not suffer from approximations.
It turns out that the free spectral range is typically the main quantity that the component
designer would be interested in extracting from the simulation, since small differences
in the fabrication of each device will lead to much larger variations in the resonant
frequencies and Q factor between the individual devices after fabrication compared to the
differences that we see in the results between the varFDTD and 3D FDTD.
We can also extract the S-parameters of the ring resonator by using mode expansion monitors
to get the amount of forward and backward propagating power in the fundamental TE mode
at the input and output ports.
We will show how to set up mode expansion monitors for this purpose in a later section
of this course.
The plots on the left show the real part of the S parameters using varFDTD and 3D FDTD,
and the plots on the right show the phase of the S parameters.
You can see they exhibit the same trends.
Although there is still a shift in frequency as we saw with the power spectrum, the shift
is not important to the design since the resonant frequency is very sensitive to the fabrication.
This table shows the simulation memory and time to get the simulation results using the
same computer.
As you can see, the simulation memory requirement is more than 15 times lower, and the simulation
time is more than 100 times faster when using varFDTD compared to 3D FDTD.
In summary, although some properties of the device, such as the coupling coefficient between
the straight waveguide and ring may suffer from the assumptions of the varFDTD method,
other results like the free spectral range are comparable to 3D FDTD.
The trends in the device performance due to varying the design parameters will also be
the same as for 3D FDTD simulations, with a much smaller simulation time.
The ring resonator is a particular case where the effect of the approximations of the varFDTD
method can be seen clearly.
For planar devices that don't include any components that introduce loss in the vertical
direction, for example ridge waveguide tapers, MMIs, or star couplers for arrayed waveguide
gratings, the simulation accuracy is much closer between varFDTD and 3D FDTD.
Due to the lower simulation time, varFDTD can be useful for determining the effect of
varying specific design parameters, and for optimization of the design by running many
simulations to narrow down the range of design parameters which give optimal device performance.
3D FDTD simulations or EME simulations which are more time consuming can then be used to
verify the results and get final high accuracy results.