Model Tuning Procedure

1. Set your planning tools correctly, i.e. Equipment, Antennae masks, output powers, etc...
2. Focus on one region, which contains the major clutter types: dense urba, suburban, rural...etc...
3. Examine your map data and preferably you should get map data with high resolution. Set the inital clutter attenuations as they are very important.
4. Run a simulation and produce a coverage plot.
5. Go on drive tests in this area.
6. Visualize the drive tests and run a report to see the delta between the planned and measured data. Chech the error report as well, (RMS...)
7. Edit some of the parameters in the propagation model and some of the clutter attenuations.
8. Run a coverage plot again.
9. Go on drive tests, visualize and prdouce comparison report.
10. if you fit in less then 5dB error from the simulation, you have done well. If not...continue working.

Model tuning is very very demanding in terms of knowledge and experience, but you must strive to do it. This is the way. Also I am again asking to refer to the other topics on this. There are very good detailed explanations.

1. ALL models work as more or less linear functions against the log distance. When tuning a model, you need at least a whole decade of distance included in your drive test (e.g. 100m to 1km, or 300m to 3km), and your drive test points density must be spread linearly against the log distance. This requires as many points you can gather near to the base station, and only a selected few on different clutters at far side.
2. Always perform Lee transform (decimation) prior to model tuning to exclude Rayleigh from equation.
3. Validity of a model is checked by filtering out a group of points that have some distinct feature, e.g. LOS, and see if your average difference (measured against model) remains zero. If not - try harder.
4. RMS errors smaller than 7 dB are fine.
5. Do it by hand, and tune a single parameter at one time, observe RMS error average difference, and if possible observe a difference distribution plot - if you observe two or more distinctive "hills" there - your cartography may be wrong or your clutters are not selected appropriately.
6. Take time.

A final report of COST-231 can be downloaded at: http://www.lx.it.pt/cost231/final_report.htm
Mr. Coreia usually makes a book out of a final report, and sells it while it is hot. This one is not that hot any more, and it is a perfect starting point for serious researcher. Unfortunately OFDM is not included.

About the Lee criterion and how to collect and decimate drive test data see this: http://whitepapers.zdnet.co.uk/0,1000000651,260090593p,00.htm and it requires a free registration

[get this widget]