Skip to main content

Improvements in Propagation Models

This month, I'll look at an ITU propagation model that could replace Longley-Rice and some major updates to the popular open-source SPLAT! propagation software.


While discussing a single frequency network mobile DTV demonstration we were developing for this year, Nivia Walker in the Dish technology group introduced me to the ITU-R P.1812 path-specific propagation prediction method for point-to-area terrestrial services in the VHF and UHF bands. Like Longley-Rice, P.1812 is based on physical rather than statistical/empirical models.

In my August 6, 2008, column, "Propagation Models in Perspective," I described the software errors Sid Shumate found in the current implementation of Longley-Rice used to determine DTV station coverage and interference in the United States. It is possible that the P.1812 model could be a used in place of Longley-Rice for more accurate coverage and interference predictions. P.1812 takes into account the following model elements:

  • • line-of-sight
  • • diffraction (smooth-Earth, irregular terrain and sub-path cases)
  • • tropospheric scatter
  • • anomalous propagation (ducting and layer reflection/refraction)
  • • height-gain variation in clutter
  • • location variability
  • • building entry losses

Fig. 1 shows a portion of a test plot of coverage from KNBC-DT and a DTS transmitter on Table Mountain in Angeles National Forest. Diffraction loss is calculated using a method based on the Deygout construction for a maximum of three edges as described in ITU-R P.452. Recommendation P.1812 notes "other diffraction methods are under consideration now that will lead to an update of this section."

Terminal clutter losses are calculated when the receiver or transmitter antenna (the model is symmetrical) is located below the height representative of the ground cover surrounding the transmitter or receiver. P.1812 offers recommendations for receivers with antennas below clutter height in urban or suburban environments, for mobile systems with omnidirectional antennas at car-top height, receivers with rooftop antennas near the clutter height, and for receivers in rural areas.

Building entry loss is specified for three different frequencies along with a standard deviation. At 200 MHz, the loss is 9 dB with a standard deviation of 3 dB. At 600 MHz, loss increases to 11 dB with a standard deviation of 6 dB. The same numbers apply at 1.5 GHz. The recommendation states these values may have to be updated when more experimental data becomes available.

I'll have more on Recommendation ITU-R P.1812 after I've had a chance to work through the calculations on some specific paths where I have measured data. ICS telecom nG is the only propagation software package I've found with P.1812. More information on it is available at Recommendation ITU-R P.1812 can be obtained at A Google search on ITU-R P.1812 will reveal work on updates.


I've been beta-testing Version 1.3.0 of John Magliacane's excellent SPLAT propagation software. On April 11, he released the final version on If you are not familiar with SPLAT, you'll find links to my previous articles on it at

In addition to a number of bug fixes, Version 1.3.0 introduces SPLAT! HD. SPLAT! HD allows 1 second (about 30 meter) resolution on Longley-Rice coverage plots and path studies using the shuttle radar terrain mapping (SRTM) mission data files available from a NASA FTP site. The main advantage of using SRTM data is that it is based on radar reflections, so buildings that reflect the signal, as well as forests, are included in the terrain height. The disadvantage, of course, is that receive antennas are based on the height above terrain. When using SRTM terrain data, the FCC 9.1 meter antenna height will give misleading results except in areas with no buildings or forests. I've found using an antenna height of 3 to 4 meters works well.

SPLAT! HD will produce huge image and data files. An area plotted with 1 second terrain data has nine times as many points as one plotted using 3 second data. Fortunately, SPLAT-1.3.0 has a new compile script that makes it easy to set the size of the area to be studied for both standard and HD studies. When using 1 second resolution, each degree plotted will create a 3600x3600 pixel image. With a two degree square, the image size grows to 7200x7200 pixels. Some graphics programs may have trouble with files this size. I've had good luck with GIMP.

SPLAT 1.3.0 allows for plotting received signal power in dBm in addition to field strength in dBµV/m and path loss in dB. I noticed SPLAT returned levels about 3 dB higher than those predicted using RadioSoft's ComStudy program to plot field strength. This may be due to the use of EIRP (effective isotropic radiated power) in the program. John and I have been going back and forth on the discrepancy and unfortunately at this time I can't say which program is accurate. To make SPLAT output closer to that of ComStudy, I reduced the ERP by 2.15 dB, the difference in gain between a dipole and an isotropic antenna. If readers have an opportunity to compare the PPM image files and the text output from SPLAT with any other Longley-Rice based propagation software I'd be interested in hearing about any differences in predicted signal levels.

One of SPLAT's limitations is it doesn't provide a way to do interference studies between multiple transmitters at different locations. Using the text output data from SPLAT, I've now found a way to accomplish this with a few simple PERL scripts. The picture on p.21 shows a portion of a test plot of coverage from KNBC-DT and a DTS transmitter on Table Mountain in Angeles National Forest near Wrightwood, Calif. Areas where the signal strength from the two transmitters is within 15 dB are shown in orange. The blank lines are due to errors in mapping the SPLAT output to an X-Y grid.


Using PERL (, I've been able to convert the text output available from SPLAT into data files that can be sorted, combined and plotted. I've posted PERL programs at They should be considered a "work in progress"—I'm still fixing bugs and some may consider the command line programs "user hostile." However, PERL programs are written in plain text and compiled at run time, making modifications easy. Programs are small, so it is easy to have multiple programs to do different studies.

SPLAT generates plots by a calculating paths along a radial swept around the transmitter site. This speeds calculation, but it isn't easy to combine output from multiple sites. Data has to be mapped into a grid to compare multiple sites. Because the signal plot is a circle and the map is a square, not all points will have data. Some of the grid cells will have more than one point in them while adjacent cells will have none. I'm still working on the best way to deal with this.

For now, you can use the program to convert the output to a list of x, y, field strength data then run the program to create a PNG file and associated kml file for display in Google Earth. Modify the file to change the colors in the plot. The xy_data.txt output can be sorted with the Linux "sort" program for combining with other studies using PERL once the grid mapping problem is solved.

Let me know how you are using SPLAT! If you have improved PERL routines for converting the SPLAT text output into an x-y grid please share them. Look for updates on the Web site.

E-mail Doug Lung

Doug Lung is one of America's foremost authorities on broadcast RF technology. He has been with NBC since 1985 and is currently vice president of broadcast technology for NBC/Telemundo stations.