As mentioned in the last unit, you can either choose to include fast diagnostics or both
fast and slow diagnostics.
As the name suggests, including slow diagnostics will cause the Propagate step of the simulation
to take longer to complete.
Fast diagnostics will return local gain and loss, and the forward and backward propagating
coefficients, while including slow diagnostics will also return the tangential field discontinuities
at each interface.
In the next slides, we will consider error diagnostic results for the polarization converter
example that we looked at in the mode convergence sweep unit with 25 modes in each cell.
When looking at the gain and loss results from the local diagnostics, first make sure
that the energy conservation method is set to "none", and the CVCS subcell method is
not being used since the CVCS subcell method also enforces energy conservation.
Enforcing energy conservation can improve the simulation accuracy, but only when the
simulation is already close to conserving energy.
When running this test, it's best to disable energy conservation so you can check if the
simulation is close to conserving energy.
Typically if local gain and loss is less than about 0.1%, then this is considered close
to conserving energy, and you can apply energy conservation and the CVCS subcell method again.
Otherwise, you may need to increase the number of modes or cells used until the gain and
loss falls below 0.1%.
Here, in both cases, with 5 or 19 cells, the gain and loss are low enough for energy conservation
and the CVCS subcell method to be used, although you can see that the loss is reduced by more
than an order of magnitude over the taper region of the device when the number of cells
is increased to 19.
Checking the field discontinuity values deltaE_1, deltaE_2, deltaH_1, and deltaH_2 from the
local diagnostics results can show you where more modes are needed.
Make sure that the CVCS subcell method is not being used.
If you see a peak in the field discontinuity values at a particular interface, the number
of modes should be increased in the cells on either side of the interface.
In this case, you can see that the highest field discontuities occur in the region near
the end of the taper.
Coefficients can give you a good idea of where power is going.
For example, if you are designing an adiabatic taper, but the transmission is not 100%, you
can plot the forward propagating coefficients to see which regions need to be stretched.
You should see that after a certain distance, power will get coupled into higher order modes,
so this is the region that needs to be stretched to prevent loss.
For the case of this taper, looking at the absolute value of the forward coefficients
to plot the fraction of power travelling in each mode in the forward x-direction, we can
see that the power is carried by the fundamental mode as the value for mode 1 is close to 1,
and the value for other modes is negligible over the full length of the device, so this
taper is adiabatic.
Looking at the backward propagating coefficients, we can see from the color bar scale that the
reflected power is negligible.