How to Improve Convergence for the FDTD
The finite difference time domain (FDTD) is a solution method that may fail to converge under certain conditions. Several model or solution settings are presented that could improve the model's convergence behaviour.
Adding a 50 Ohm Load to the Feed Port
FDTD solutions, in general, converge faster when there are losses in the model. A 50 Ohm load can be added to the feed port but will affect the impedance (input reflection coefficient), far fields (gain only) and also the near fields.
In POSTFEKO the effect of this load can be removed from the impedance by selecting the Subtract loading check box in the result palette. The gain is lower when using a load but the directivity is unaffected.
Increasing the Total Time Interval
The maximum total time interval is the propagation time required for the time signal to pass through the domain. If the total time interval is too short, the signal has not yet propagated through the domain. Increase the total time interval to ensure the signal has propagated through the domain and therefore improving convergence.
Using Double Precision
The Solver uses single precision by default- a single byte is used to store a complex number.
Including Losses in the Dielectric
The FDTD shows slower convergence for lossless models, therefore including the dielectric loss could improve convergence.
Setting the Wire Radius to the Intrinsic Value
When a wire port is present in the model, convergence varies according to the wire thickness. Select the Use intrinsic wire radius check box to determine automatically the wire radius best suited for the voxel mesh.
Changing the Feed
If it practical for the specific model, change the feed. If a wire port is used, change to an edge port, and conversely.