FM Card

The FM card is used to instruct the Feko solver to calculate the solution using accelerated methods, for example, using the multilevel fast multipole method (MLFMM) or adaptive cross-approximation (ACA). An option is available to apply compression to looped plane wave sources.

On the Solve/Run tab, in the Solution settings group, click the  MLFMM (FM) icon.



Figure 1. The FM- Fast method settings (acceleration) dialog.

Parameters:

None
Neither the MLFMM nor the ACA is activated.
Use multilevel fast multipole method (MLFMM)
The MLFMM is used instead of the standard method of moments (MoM) to calculate the solution on all structures.
Use adaptive cross-approximation (ACA)
The ACA method is used to calculate the solution. This method approximates the MoM impedance matrix by constructing a sparse H-matrix (only a few selected elements are computed). This method is applicable to low frequency problems or when using a special Green’s function.
Activate additional stabilisation
Activate additional stabilisation for the MLFMM to address models with severe convergence problems.
Box size at finest level
The MLFMM is based on a hierarchical tree-based grouping algorithm, and depending on the frequency and the model dimensions Feko automatically determines the number of levels in this tree and the size of the boxes at the finest level. It is also recommended that this default box size of 0.23λ is kept. When there is no convergence in the MLFMM, then advanced users might try to slightly increase or decrease this box size by setting it manually (the input is in terms of the wavelength).
Field calculation methods
The Solver determines automatically which calculation method to use for the fastest field calculation.
Enable fast far field calculation
If selected, the MLFMM method is included in the available options for calculating the far field (default). If not selected, the Solver uses either traditional integration or other fast methods to calculate the far field.
Enable fast near field calculation
If selected, the MLFMM method is included in the available options for calculating the near field (default). If not selected, the Solver uses either traditional integration or other fast methods to calculate the near field.
Characteristic basis function method (CBFM)
Enable CBFM
Enable characteristic basis function methods (CBFM) to reduce the number of unknowns (which reduces the memory and runtime).
MoM-based CBFM
Use the MoM solution method to generate the CBFMs. Solution is more accurate, but at the cost of an increase in runtime and memory.
PO-based CBF
Use the PO solution method to generate the CBFMs. Solution is less accurate, but with a shortened runtime and a decrease in memory.
Block size in wavelegths
Size of the blocks in which the characteristic basis functions (CBFs) are defined. A larger size requires more processing time, but yields a higher reduction of the number of unknowns. The box size can be a value between 0.5 and 2, where 1 is the recommended size.
MLFMM level to create CBFM blocks
It defines the MLFMM level that will be used to create the boxing. These boxes will determine the size of the blocks in which the characteristic basis functions (CBFs) are defined where the finest level is 1.
Threshold
Parameter used to discard less significant CBFMs, between 0 and 1. A higher value discards more CBFMs, with a better reduction of the number of unknowns. The accuracy may be affected for high threshold values. The typical threshold value is about 5e-4.
Compression for looped plane wave sources

Compression of a plane wave loop over direction of incidence can be used to speed up calculations over multiple directions of incidence (for example, when calculating RCS).

Auto
The Solver determines automatically if compression should be used for the model (if the method is likely to speed up the solution).
Enable compression
Enable compression for looped plane wave sources in the model.
Disable compression
Do not use compression for looped plane wave sources in the model.

MLFMM

Note: The MLFMM is not supported in conjunction with the multilayer Green's function.

Advantages of adaptive cross-approximation (ACA)

  • ACA is done on the matrix level.
  • No frequency breakdown like MLFMM.
  • Can also be used with the multilayer Green's function.
  • Direct solution.