Numerical Modeling of Indoor Propagation Using FDTD Method with Spatial Averaging

The error in the local mean magnitude of the electric
field (E-field), due to the numerical anisotropy, obtained by the
finite-difference time-domain (FDTD) method is investigated. The
spatial averaging is applied over a cube. In order to quantify
the error, the numerical results are compared with theoretical
and measured ones. The comparison between the FDTD method
and theory is conducted for two empty rooms with perfect
electric conductor (PEC) walls at 3 and 5 GHz. It is found
that averaging over a cube with side length of 3.3 wavelengths
(λ₀) ensures a good matching between the local mean magnitude
of the FDTD and theoretical E-field - maximum error below
23%, 95th percentile of the error below 6% and correlation
above 0.83. Measurements over a cube at 3 GHz in empty
and office environments are performed. The difference between
the averaged numerical and measured magnitude of the E-field
decreases with increasing the averaging stencil. For empty room
the maximum error in the local mean FDTD results is 46% and
for office scenario is 49% if the cube side length is 0.5λ₀.