When signal measurement data is imported into Overture for visualization and analysis no filtering is applied. For prediction model tuning this is not ideal, so the Prediction Comparison component has filter functions that it applies to the input data.

Filter Functionality

All of the filtering functionality is collected together into the Preprocessing section shown here

Bin Distance

Measurements that are closer together than this distance will be aggregated into a single average measurement. This is to filter out the effect of fast-fading on the signals, but it is also because the tolerance of other input data, such as terrain or building locations, is unlikely to be accurate enough to support modeling very small distances. Therefore, there is no utility in over-tuning the prediction model to an unecessary resolution.

Min & Max Distance

Measurements very close to the transmitter may be distored or may involve modes of propagation that the model is not designed to account for. Therefore near-by measurements should be clipped out of the data set, especially when the reference transmitter has a complex or partially obstructed location.

Measurements very far from the transmitter may in fact be due to other transmission sources, and if this has not been ruled out by a prepartory drive without the transmitter, a sensible maximum clipping distance should be applied.

Min & Max Signals

Very high signals can swamp the scanning reciever and be innaccurate, so unless the sensitiviy of the receiving equipment is well specified, very high signals should be eliminated.

When the scanner is nearly out of range of the reference transmitter, the background noise will dominate the reported signal. If signals close to this noise floor (up to 3 dB above) are not filtered out, the automatic tuning algorithm will try and force the model to match this impossibly flat signal profile.

Note that some scanners reserve special values to indicate error conditions such as "no signal detected", and that unless these have been filtered prior to being read by Overture, they will be intepreted as extreme signal values. Clipping to a safe range can eliminate these problem results.

Also note that the when automatically tuning predictions, the goal is to get closest fit to all of the measured data points. This means that an unusually high number of measurments at a particular signal level can dominate the tuning process leading to a poorer fit in other areas. If this happens at the nominal cell edge signal power level, the result can be poorly defined cell edges. In this scenario it is better to set agressive thresholds to filter out all but the signals near the cell edge power level.

Max Angle From Boresite

If a directional antenna transmitted the reference signal, this property can be used to filter out measurements that are not close to the boresite. This is particularly useful when the pattern is inaccurate outside of the main beamwidth as is often the case. In fact some antenna patterns do not even include a back pattern because the front-to-back ratio is so high it is not useful for normal planning activities.

Setting this parameter to zero (the default) disables angular filtering.