What Is Hysteresis?
When the Answer Depends on Direction

Back to Blog

Intentional Hysteresis in Switching

Deliberate hysteresis is what keeps an on/off device from chattering. A tank pump that starts at 80% level and stops at 60% has 20 points of hysteresis: the turn-on and turn-off thresholds are different by design. Because the switching points differ by direction, a level jittering around a single value cannot rapidly cycle the pump - it must travel the full gap before the state can flip back.

The same idea protects alarms (an alarm clears at a lower value than it set), thermostats, and any comparator-driven logic. This intentional hysteresis is closely related to deadband, and in everyday plant language the two words are used almost interchangeably, though hysteresis specifically refers to the direction-dependent switching.

Unwanted Mechanical Hysteresis

The other face of hysteresis is a defect. In a control valve, mechanical slack, packing friction, and linkage backlash mean that when the controller reverses direction, the valve does not begin moving until the signal has changed by some amount - the actuator has to take up the lost motion first. This shows up as a gap between the up-going and down-going stroke curves and degrades control: the loop cannot make small, precise corrections, and it can even drive limit-cycle oscillation.

Sensors and mechanical gauges show hysteresis too - a Bourdon tube or a float linkage may read slightly differently on rising versus falling pressure or level. It is quantified during calibration as the maximum difference between readings taken while increasing and decreasing the input across the range. Valve positioners and good instrument selection minimize it, which is why hysteresis is a standard line item on instrument accuracy specifications.

Measuring Hysteresis with a Step Test

The standard way to quantify hysteresis in a valve or instrument is a slow bidirectional traverse. Put the loop in manual, step the signal upward in small equal increments across the range of interest, and record the response at each step after it settles. Then reverse and step back down through the same values. Plot both traverses against the input: a device with hysteresis shows two distinct curves, and the maximum gap between them, expressed as a percentage of span, is the hysteresis figure. On a transmitter the same idea appears in bench calibration - each test point is approached once from below and once from above, and the difference is recorded.

As a symbolic example, suppose the up-going traverse reaches a given stem position at a command of X percent, but on the way down the stem does not return to that position until the command has fallen to X minus d percent. That gap d is the lost motion the controller must burn through on every reversal. A loop that reverses often - level or pressure control hovering around a fixed setpoint - crosses the gap constantly and pays for it in limit cycling. Choose step sizes smaller than the gap you expect to find, or the traverse will jump straight across it and hide it.

1. Place the loop in manual and let the process settle.
2. Step the output up in small equal increments, recording the settled response at each.
3. Step back down through the same values and record again.
4. Plot both traverses; the widest gap between them is the hysteresis.
5. Repeat the traverse once to confirm you are seeing repeatable slack, not drift or noise.

Sizing Intentional Hysteresis

There is no universal number for how much switching hysteresis to use, but there is a universal method. The gap must be wider than the noise on the signal, or noise alone will bounce the state across the thresholds. It must be wide enough that the equipment gets an acceptable rest between cycles - a pump or compressor has a minimum practical cycle interval per the manufacturer's guidance. And it must be narrow enough that the process stays inside its acceptable band at both ends of the swing. Those three constraints usually leave a comfortable window, and where they conflict, the equipment-protection constraint should win the argument.

The same reasoning applies to alarms: the clearing threshold sits far enough below the set threshold that a measurement hovering at the limit produces one alarm, not a stream of them. Time-based conditioning stacks on top of the gap, and in practice the combination of alarm deadband and on-delay settings is what tames a chattering point for good. Document the chosen gap and the reasoning next to the setpoint itself, so the next engineer who finds the thresholds does not tighten them back up and reintroduce the cycling you designed out.

Hysteresis, Stiction, and Backlash: Three Different Defects

Field crews tend to bundle every reversal problem into one word, but the three mechanisms leave different fingerprints. Backlash is pure lost motion: on reversal the actuator moves while the stem does not, then motion resumes cleanly - the mechanism described under valve backlash in control loops. Stiction is different: the stem sticks while actuator force builds, then breaks free and jumps past where it should have stopped. On a trend, valve stiction tends to produce a square-shaped oscillation in the process variable, while backlash and distributed friction produce a smoother limit cycle.

The distinction matters because the fixes differ. Backlash points at linkages, couplings, and worn pins. Stiction points at packing friction, often aggravated by an over-tightened gland. Distributed friction hysteresis responds well to a positioner, which closes a local loop around stem position and drives through the friction band. Diagnose before anyone opens a toolbox: a step test in manual, or a valve signature where diagnostics are available, tells you which mechanism you actually have.

Frequently Asked Questions