</table>
Instead of running a single-frequency SBR analysis, this time we ran a frequency sweep of the propagation scene over the frequency range [1.5GHz, 2.5GHz]. A frequency step of 100MHz was chosen, hence, a total of 11 frequency samples. The geometrical optic (GO) part of the simulation is the same for all frequency samples except for the fact that the reflection coefficients are at the building specular points are frequency-dependent. Diffraction coefficients are frequency-dependent, too. [[EM.Terrano]] first determines all the optical paths in the scene using the k-d tree algorithms. Then, it calculates the reflection and diffraction coefficients at each frequency sample and computes the field components and received power at the location of each receiver. As a result, [[EM.Terrano]]'s frequency sweeps are extremely fast. The figures below show the received power coverage maps of the propagation scene with a rotated horizontal Yagi-Uda array at three different frequencies: f = 1.5GHz, 2GHz and 2.5GHz. The color-coded intensity scale is identical for all three plots. As you would have expected, the maximum received power level drops as the operating frequency is increased.
<table>
</tr>
</table>
Â
The table below shows the computation time of different stages of the SBR sweep simulation:
Â
{| class="wikitable"
|-
| Simulation Step
| Computation Time
|-
| Facet Mesh Generation
| 80s
|-
| SBR Ray Tracing
| 160s
|-
| Frequency Computations
| 90s
|-
| Total
| 330s
|}
== References ==