PIDSnap

PID loop oscillating

An oscillating loop is giving you more information than most people take from it. The shape of the oscillation, its period, and whether it decays all point at different causes, and reading them costs nothing.

The three questions worth answering, in order: does it decay, how long is one cycle, and does it survive putting the loop in manual. Between them they narrow the field to one or two candidates before you touch a single constant.

An over-tuned loop hunting after a setpoint change Trend of an over-tuned PID loop: the process variable hunts for several cycles after a setpoint change before settling. high low PV SP Output PIDSnap time
Trend of an over-tuned PID loop: the process variable hunts for several cycles after a setpoint change before settling. Four or more peaks that shrink slowly. The output mirrors the process variable and swings just as hard. The oscillation decays, which is what separates a tuning problem from a mechanical limit cycle — stiction cycles do not decay.

Does the oscillation decay?

A decaying oscillation — each peak smaller than the last — is a tuning signature. The loop is over-correcting but the correction is still stabilising, so it settles eventually. This is fixable from the faceplate.

A constant-amplitude oscillation that never decays is a limit cycle, and limit cycles come from nonlinearities, not from tuning. Stiction, backlash, and a valve hitting a travel limit all produce them. Tuning changes the period and does not touch the amplitude.

A growing oscillation means the loop is genuinely unstable. Put it in manual now and reduce the proportional action before putting it back.

How long is one cycle?

The period tells you roughly where in the loop the problem lives. A period of a few seconds on a flow loop is the loop's own dynamics — tuning territory. A period of many minutes on the same loop is not: nothing in a flow loop is that slow, so something external is driving it.

Compare the period against the loop's natural response time, which you can measure by stepping the output in manual and timing how long the process variable takes to settle. An oscillation much slower than that is coming from outside.

What to change first

If it decays and the period matches the loop's own dynamics, reduce the proportional action and observe. If it decays but the period is several times the loop's response time, the reset is too fast — slow it down before touching gain.

A rule that saves a lot of wasted effort: change one thing at a time and capture a trend after each change. Two changes at once and you have learned nothing about either.

What causes it, most likely first

Proportional action too high

most common tuning can fix this

The controller responds to error more strongly than the process can absorb, overshoots, and then over-corrects in the other direction.

How to confirm it
Decaying oscillation with a period close to the loop's natural response time. Output swings as hard as the process variable.
What to do
Reduce proportional action and observe the overshoot reduce.
An over-tuned loop hunting after a setpoint change Trend of an over-tuned PID loop: the process variable hunts for several cycles after a setpoint change before settling. high low PV SP Output PIDSnap time
Trend of an over-tuned PID loop: the process variable hunts for several cycles after a setpoint change before settling.

Reset too fast for the dead time

most common tuning can fix this

Integral action repeating faster than the process can respond. It keeps adding correction while the last one is still in transit.

How to confirm it
Slow decaying cycle, period several times the loop's response time, output drifting steadily in one direction then the other.
What to do
Slow the reset substantially before adjusting anything else.

A nonlinearity in the final element

common tuning cannot fix this

Stiction, backlash or a valve sitting against a limit. These produce cycles that do not decay because the nonlinearity is the same size regardless of how small the error gets.

How to confirm it
Constant amplitude across many cycles. Stops in manual.
What to do
Deal with the valve. Tuning will not remove the cycle.
Valve stiction producing a limit cycle Trend showing valve stiction: the controller output ramps smoothly while the process variable moves in abrupt steps, producing a limit cycle that does not decay. high low PV SP Output PIDSnap time
Trend showing valve stiction: the controller output ramps smoothly while the process variable moves in abrupt steps, producing a limit cycle that does not decay.

Derivative acting on a noisy measurement

common tuning can fix this

Derivative amplifies rate of change, and noise has a large rate of change even when the process is steady.

How to confirm it
High-frequency activity on the output that mirrors noise on the process variable, with the output noisier than the measurement.
What to do
Set rate to zero. On flow, pressure and speed loops it should almost always be zero anyway.
Derivative action amplifying measurement noise Trend showing noise amplification: a modest amount of noise on the process variable appears on the controller output magnified many times over. high low PV SP Output PIDSnap time
Trend showing noise amplification: a modest amount of noise on the process variable appears on the controller output magnified many times over.

Questions that come up

Is quarter amplitude damping a good target?

It is a traditional one and it is more aggressive than most operating plants want. Quarter amplitude damping leaves a loop close to the edge — a modest change in process gain, which happens whenever load changes, can tip it into a much less comfortable place. Aim for a response that settles inside two peaks.

Why does the loop only oscillate at certain production rates?

Because the process gain changes with load, and a tuning set that suits one end of the range is too aggressive at the other. This is very common on flow loops with equal-percentage trim, and on any loop with a nonlinear process.

Related

Last reviewed 2026-08-01.

Run a free PIDSnap check on this loop

Two photos, thirty seconds, no signup. The pre-flight check finds the structural problems that no amount of tuning will fix — before you touch the controller.

Start a free check