Control valve hunting
Hunting is the word most people reach for when a loop will not sit still. The valve moves, the process variable swings past setpoint, the valve moves back, and the cycle repeats indefinitely. It is the single most common complaint raised against a control loop, and it is the one most often misdiagnosed.
The instinct is to re-tune. That is right perhaps a third of the time. The rest of the time the loop is hunting because of something the controller constants cannot reach — a sticking valve, an oversized trim, a reset time shorter than the loop's own dead time, or another loop pushing on the same process. Detuning a loop that hunts for a mechanical reason will slow the cycle down without removing it, and leaves you with a loop that is both cycling and sluggish.
The fastest way to split the field is the manual test, described below. It takes thirty seconds and it eliminates about half the possibilities.
Start with the manual test
Put the loop in manual and leave the output where it is. Then watch the process variable for a few minutes.
If the cycling stops, the loop itself is generating it. That points at the controller or at the valve responding to the controller — tuning, stiction, or backlash. If the cycling continues with the output frozen, the controller is not causing it, and you are looking at a disturbance coming in from somewhere else: an interacting loop, an upstream swing, or equipment cycling on and off.
This one test is worth more than any amount of staring at tuning constants, because the two branches it separates have almost no fixes in common.
Then look at whether the cycle decays
Capture a trend with the process variable and the controller output on the same time axis, and disturb the loop with a small setpoint change.
A tuning problem produces an oscillation that decays. Each peak is smaller than the last, even if it takes many cycles to die out. A mechanical limit cycle does not decay — the amplitude stays constant indefinitely, because the valve is stepping through the same stiction band every cycle regardless of how small the error gets.
That distinction is visible in the trend shape alone, which is why PIDSnap's diagnostic works from a photograph of the trend rather than from the numbers on the faceplate.
What causes it, most likely first
Valve stiction
most common tuning cannot fix this
Packing friction holds the stem still while the signal changes, then the stem breaks free and slips past the position that was asked for. The controller corrects, the stem sticks again, and the loop cycles. Repacking a valve slightly too tight is the usual origin, which is why this often appears shortly after maintenance.
- How to confirm it
- The process variable moves in abrupt steps while the output ramps smoothly — a sawtooth against a triangle. The cycle amplitude stays constant. In manual the cycling stops immediately.
- What to do
- Check packing friction and the positioner calibration. Compare actual valve travel against the signal if you have position feedback. Tuning can lengthen the cycle period but cannot remove it.
Reset faster than the loop's dead time
common tuning can fix this
Integral action keeps pushing while error persists. If it repeats faster than the process takes to answer, it is still pushing when the previous correction arrives, then has to unwind. That is a cycle by construction, and no gain setting will stop it.
- How to confirm it
- Compare the reset time against how long the process actually takes to start responding to an output change. If reset repeats faster than the dead time, this is your answer. PIDSnap's pre-flight check flags it from the loop profile alone, before any test.
- What to do
- Slow the reset down substantially, then reassess. This is the usual explanation for a loop that has hunted since the day it was commissioned.
Oversized valve
common tuning cannot fix this
A valve running near its seat delivers a large flow change for a small movement, so the process gain is at its highest exactly where the loop operates. Resolution is poor and any stiction is magnified.
- How to confirm it
- Check the normal valve position. Below about 20% open at normal load, the valve is oversized for the service.
- What to do
- Reduced trim is the real fix. Until then the loop needs detuning rather than tightening, because a high process gain wants a low controller gain.
Tuning that is simply too aggressive
common tuning can fix this
Too much gain, too much reset, or derivative on a loop that should not have it. The loop over-corrects, overshoots, over-corrects the other way.
- How to confirm it
- The oscillation decays after a disturbance rather than sustaining. The output swings as hard as the process variable does.
- What to do
- Back the proportional action off and observe. If derivative is switched on and this is a flow, pressure or speed loop, take it out first.
An interacting loop
less common tuning can fix this
Two loops sharing a process push against each other, each correcting for a disturbance the other created. Common on parallel flows into a common header, and on cascades where the inner loop is not fast enough.
- How to confirm it
- The cycling continues in manual. Trend the suspected partner loop over the same window and look for a matching period.
- What to do
- Detune one of the two so their response speeds separate, or fix the cascade architecture if the inner loop is the slower of the pair.
Questions that come up
Should I reduce gain or increase it to stop hunting?
It depends on which kind of hunting you have. An over-tuned loop needs less proportional action. A loop hunting because reset is too fast often needs the reset slowed while the gain stays where it is, and reducing gain alone can make that case worse. Establish which one you have from the trend before you change anything.
Why does the valve hunt in automatic but not in manual?
Because in manual the controller is not moving the output, so any cycle driven by the feedback path disappears. That points at the controller or at the valve's response to it, and rules out an external disturbance.
Can I just put a filter on the measurement?
You can, and it will make the trend look calmer. It will not make the loop control better — the filter adds lag, which limits how well the loop can ever perform, and it hides the evidence you need to find the real cause.
Related
Last reviewed 2026-08-01.