Deadband and hysteresis are nonlinearities that make a control valve respond differently than its command signal says it should, and both are common causes of the repeating oscillation known as a limit cycle in process control loops.
Deadband is a range of controller output change that produces no valve movement at all. When the signal reverses direction, it must cross this dead zone before the stem moves the new way. It is usually mechanical play, or backlash, in the linkage between actuator and valve: loose pins, worn gear teeth, or coupling clearance. On a signature test plotting commanded position against actual stem position, deadband is the flat, unresponsive segment right after a direction change.
Hysteresis is different: not a zone of no motion, but a path-dependent offset. The valve moves as the signal changes, but its position for a given signal value depends on whether the signal is increasing or decreasing. It shows up as a gap between the "opening curve" and "closing curve" on the same test. Hysteresis is mostly a friction effect, from packing drag, seal friction, and bearing friction, so the valve lags the signal by an amount that flips sign when direction reverses. Most valves show both effects together, and diagnostics often report "dead band plus hysteresis" as one combined band because they are hard to separate in field data.
Stiction (static friction) is the resistance that must be overcome before a stationary valve starts moving again, and static friction is higher than the dynamic friction that follows once it is moving. That gap produces stick-slip: the controller's integral action ramps its output while the valve sits still, error builds, and once accumulated force exceeds stiction the valve breaks free and moves abruptly, overshooting because force built up while it was stuck. It then sticks again and the cycle repeats. Stiction is generally quantified as a percentage of valve travel or signal span: 2 percent stiction means the signal must change by roughly 2 percent of its range before the valve breaks free. Common causes include under-lubricated or over-tightened packing, seal swelling, stem corrosion or scale, and worn guide bushings.
Backlash is physical clearance in a mechanism, most often in valve linkages, gear trains, and shaft couplings. Gear teeth are made with intentional clearance so they do not jam from tolerance stack-up or thermal expansion, but that clearance means the teeth or linkage pins must travel across the gap before they re-engage when drive direction reverses. Until then, the actuator moves while the valve does not, which is deadband's mechanical origin. Worn linkages and a loose positioner feedback arm increase backlash over time, so deadband tends to worsen gradually with age.
A limit cycle is a sustained, self-repeating oscillation: it does not decay or grow, it settles into a stable pattern of roughly constant amplitude and period. The mechanism in a PI or PID loop acting on a sticky or backlash-affected valve is well documented:
The result is a recognizable signature: a roughly square-wave process variable paired with a saw-tooth controller output, which engineers use to tell a sticky valve apart from a controller that is simply tuned too aggressively.
Oscillation from valve nonlinearities is not cosmetic. A cycling process variable increases product variability, pushing more output toward specification limits and raising rejects. Repeated stick-slip motion accelerates wear on packing, seats, and linkages, so a sticky valve tends to get stickier if left uncorrected, and cycling loops propagate disturbance downstream since one loop's process variable often feeds another loop as an input. Because deadband, hysteresis, and stiction all break the assumption that a signal produces a proportional, repeatable, direction-independent response, they undermine the basic premise PID tuning depends on. That is why retuning alone rarely fixes a sticky valve, the valve, packing, or linkage usually needs mechanical attention.
Catching this kind of degradation early is the same condition-monitoring problem as bearing wear and other friction-related failure modes, which follow a similar slow-drift pattern and are easier to catch from trend data than a single spot check. Fabrico reads machine condition and OEE directly from the line, with computer vision that catches what sensors miss, and auto-routes a work order the moment it detects a developing loss. It is EU-built with EU data residency and is ISO 27001, ISO 20000-1, and ISO 9001 certified. Book a Fabrico demo.
For loops already cycling, a valve signature or step-test comparing commanded and actual position is the standard first move, the same way techs rely on quantified checks like shaft alignment tolerances or vibration severity bands rather than a guess.
No. Deadband is a range where the signal changes but the valve does not move, mainly from mechanical backlash. Hysteresis is a directional lag where the valve does move but its position depends on the direction of signal travel, mainly from friction. Diagnostics often report them together because they are hard to separate from routine trend data.
Rarely on its own. Stiction and backlash are valve-side mechanical problems, not tuning problems. Detuning the controller can shrink the visible amplitude of the cycle, but it does not remove the underlying friction and it sacrifices responsiveness. The more durable fix is mechanical: packing adjustment, lubrication, or linkage repair.
As a percentage of valve travel or of the control signal span, based on how far the signal must change before a stationary valve breaks free. Valve signature testing, comparing commanded signal to actual stem position through a full stroke and back, is the standard way to quantify it.
A process variable that oscillates in a roughly square-wave pattern together with a controller output that ramps in a saw-tooth pattern, one of the standard ways engineers spot a valve mechanical problem versus aggressive tuning.
Connexion
Blog
