Measurement filtering
Smoothing the process variable before the controller sees it, trading noise for added lag.
What it means
Measurement filtering smooths the process variable before it reaches the controller. The simplest form is a first-order lag on the measurement with a filter time constant set by the user; more elaborate versions remove spikes or hold the last good value.
The trade is honest. Every filter removes high-frequency noise, and every filter also adds lag, and the added lag consumes the same stability budget that process dead time does. A filter heavy enough to calm a noisy loop slows the loop by the same amount.
Filtering is the right fix when the noise is real and the lag is affordable, and the wrong fix when the noise comes from the loop itself — a cycling valve, a probe in a vortex — because the filter hides the problem. Get the measurement right before touching the controller, exactly as the pre-flight instinct insists.
Related
Where this comes up
Last reviewed 2026-09-18.