Signal Filtering
Signal filtering is obtaining a stable reading by suppressing, in hardware or in software, the unwanted fluctuation on a noisy analog or digital input. Every filter introduces delay: the more you suppress the noise, the later the system responds to a real change.
In practice two layers are used together.
- Hardware filter: a simple low-pass structure made of a series resistor and a capacitor placed at the input pin attenuates high frequency interference before it reaches the converter. The cut-off frequency is chosen above the real rate of change of the quantity to be measured.
- Software filter: a moving average, an exponentially weighted average, or taking the median of a few samples. The median filter is more effective than the average at suppressing occasional spike values.
In a control application the balance is struck like this: on a speed setting potentiometer, a few hundred milliseconds of smoothing is not noticed by the user and makes the setting stable. By contrast, the same delay is unacceptable in overcurrent protection; the protection path is kept separate and fast. That is why the filtering decision is made signal by signal, and not settled with a single value across the whole board.
The context in which this term is used: Our Services