Page 356 - Integrated Wireless Propagation Models

334 C h a p t e r F i v e

a
FIGURE 5.6.2.2.3 Data model for FDTD simulation. (A color version of this figure is av i l able at
www.mhprofessional.comjiwpm. )

t
o
can o the iterations within one time step o gain all the E and H values f all the cells
d
or grids. Since the iteration is processing one by one in time steps, we can reach any
time's state of the whole system after repeating the same procedure. This is the key part
of the FDTD procedure. The initial E and H are equal to zero when the procedure starts
from the very beginning. Therefore, we may not need to start at the very beginning. If
we have knowledge of the past and current state of the system at any time point, we can
start from that time point to predict the future state of the system. The iterations make
up the procedure of the FDTD.
Figure 5.6.2.2.3 shows the data model for the FDTD calculation obtained from a
picocell. Based on the input data, the FDTD can calculate the expected values for each
point. Again, this depends on how much detail and on the granularity to derive the final
value. Figure 5.6.2.2.4 shows electromagnetic field simulated using the FDTD technique.
The electric field is strong at the corner of the picocell, as shown in Fig. 5.6.2.2.4.

5.6.2.3 3 D Ray-Tracing and FDTD Models versus the Lee Model
In this case, the 3D ray-tracing and FDTD models were implemented in the same sce­
narios for comparison with the Lee Model. The room is assumed to have a uniform
dielectric environment, and the simulation was done for just one room case-the second
1
floor of north wing at Republic Polytechnic (RP) in Singapore (see Fig. 5.3.2. . 2).
The deviation is the difference between the predicted and the measured values, as
1
shown in Fig. 5.6.2.3. . As shown, the ray tracing (with and without furniture) and Lee
models gave a decent prediction in this scenario. However, the FDTD model did not
show its performance to be as good as that of the other two models. The main possible
reason is that the 1 x 1 m-square grid is too large to be accurate for the FDTD model in
this case. Furthermore, with the consideration of furniture, the gird should be even
smaller. This is also the reason that applying the FDTD model in a furnished room here
gives the worst performance. However, as we mentioned earlier, if we put more com­
putational resources into the FDTD model, the results should be better.