Everytime TouchLib completes a sequence of measurements, it converts the series of measurements into a single sample. The way this is done depends on the type of filter that is used.

Currently, three types of filters are implemented: an averaging filter, a slewrate limiting filter and a median filter. The filter type can be set with the property enum FilterType tlSensors.data[<n>].filterType.

filterTypeAverage

This is a simple averaging (or more accurately: integrating) filter. It is useful if there are no (common mode) noise sources and a very sensitive sensor is required (for example for a distance sensor).

Advantages: no reduction in signal strength.

Disadvantages: sensitive to spikes and common mode noise.

Example: tlSensors.data[0].filterType = TLStruct::filterTypeAverage;

filterTypeSlewrateLimiter

This is a simple slewrate limiting filter. It is useful if your processor has very little memory (such as ATmega-based boards) and the sensor experiences (common mode) noise sources or spikes.

Advantages: relatively robust against spikes and (common mode) noise.

Disadvantages: some reduction in signal strength. If a spike occurs during first measurement of a new sample sequence, an incorrect value could be produced. This filter therefore requires good debounce filtering (long timings for state transitions).

Example: tlSensors.data[0].filterType = TLStruct::filterTypeSlewrateLimiter;

filterTypeMedian

This is a median filter. It is useful if your processor has plenty of memory (such as Teensy or Particle boards) and the sensor experiences (common mode) noise sources or spikes.

Advantages: no reduction in signal strength, not sensitive to spikes.

Disadvantages: requires a lot of memory and is therefore not available on ATmega boards.

Example: tlSensors.data[0].filterType = TLStruct::filterTypeMedian;