Terrain height is only accounted for in determining the effective base station and mobile end-point heights, and the path clearance over (or under) one or two predominant terrain peaks. The far-field diffraction over these peak(s) results in one or two "lumps" of added loss.

The Cost, Hata and other empirical models cannot be used within the near-field of elevated base stations or terrain obstacles; the near/far boundary for a street-level mobile being about one (1) kilometer for each 100 feet of antenna height. The spatial resolution of such model calculations is at best about 200 meters, even with 30-meter data.

Other "gross" lumps of loss are added to account for one of several categories of "clutter" (vegetation, water, and man-made objects).

Predictions from these empirical models do not depend on the incremental effects of either specular reflection or diffraction performed sequentially in a chained calculation along the path. Thus, all fine-grain variations of the signal are integrated out. The rms values of these variations (prediction error) are typically 9-14 dB in a suburban morphology setting with the mean matched to within one (1) dB.

Lee's model adds the benefit of combining both specular (ground-reflected) and direct rays along the path. Thus, Lee's model may be more accurate than others in the foreground of the station, but averages the reflections over greater distances. Again, the end points, terrain obstacles, and reflection points must be in the far-fields of one another. The spatial resolution is similar to the other empirical models.

None of these models can accurately predict PCS coverage and overlap of small cells/sectors sufficient for CDMA capacity estimation and the design of isolation between cells/sectors in varying terrain and vegetation.

Athena (the CRC-PREDICT^TM code) separately integrates the Huygens source field in height from the ground (or top of the clutter) to infinity for the previous point to describe the direct and ground-reflected contributions to the new Huygens source field in height at the current point. The chained calculation continues sequentially along radial paths from the transmitter, adding radials at increasing distance. The radial-defined points are then interpolated into a point grid. The technique has no inherent resolution limit, with radial spacing of less than 10 meters being possible over sharply-defined clutter objects. Thus a quasi-exact diffraction calculation is performed in two dimensions.

The novelty of the CRC-PREDICT^TM approach is in the efficiency of making the infinite integrations, point-by-point. The result is radial spatial resolution equivalent to the terrain database, with angular resolution capability equivalent to 50 meters at a range of 10 kilometers in finely-defined terrain.