User guide

Getting Started with CapSense Document No. 001-64846 Rev. *A 37
3.4 Software Filtering
Software filters are one of the techniques for dealing with high levels of system noise. Table 3-3 lists the types of
filters that have been found useful for CapSense.
Table 3-3. CapSense Filter Types
Type Description Application
Average Finite impulse response filter (no feedback) with
equally weighted coefficients
Periodic noise from power supplies
IIR Infinite impulse response filter (feedback) with a step
response similar to an RC filter
High frequency white noise (1/f noise)
Median Nonlinear filter that computes median input value
from a buffer of size N
Noise spikes from motors and switching power supplies
Jitter Nonlinear filter that limits current input based on
previous input
Noise from thick overlay (SNR < 5:1), especially useful for
slider centroid data
Event-Based Nonlinear filter that causes a predefined response to
a pattern observed in the sensor data
Commonly used to block generation or posting of non-
existent events
Rule-Based Nonlinear filter that causes a predefined response to
a pattern observed in the sensor data
Commonly used during normal operation of the touch
surface to respond to special scenarios such as accidental
multi-button selection
3.4.1 Average Filter
An average filter is a Finite Impulse Response filter (FIR) with equal-weighted coefficients. Average filters work well
with periodic noise. Periodic noise is attenuated by spacing the samples out over one noise cycle. Sample spacing is
not critical. For example, power line noise can be anywhere from 50 Hz to 60 Hz. Without adjusting the sampling rate,
the average filter works equally well for 50-Hz and 60-Hz noise. Figure 3-11 shows a sample rate that is synchronized
with a simple periodic waveform. There is no feedback path in this filter.
Figure 3-11. Synchronized Sample Rate
The general equation for an average filter is:

[

]
=


(
[

]
+ 
[
  1
]
+  + 
[
   + 1
]
) Equation 12
Figure 3-12 and Figure 3-13 on page 38 illustrate the results of using an average filter on real CapSense data using
the 16-sample filter equation:

[

]
=


(
[

]
+ 
[
  1
]
+  + 
[
  15
]
) Equation 13