Back to all posts
Control Systems7 min read·November 1, 2026

Why Does a PID Control Loop Keep Oscillating?

TB
ThynkBlox Automation Team
Automation Engineering

A PID control loop can oscillate because its tuning does not suit the process, but measurement noise, sticking valves, process delays and changing operating conditions can produce similar symptoms. Start by comparing the setpoint, measured value, controller output and actual actuator response on a shared timeline. Change tuning only after checking whether the measurement is trustworthy and the equipment follows its commands.

What should you establish before investigating?

Define what “oscillating” means for this loop. Is the measured temperature repeatedly crossing its target? Is a valve moving constantly while pressure stays steady? Is the complaint based on a live display, a historian trend or a physical observation?

Record when the behaviour started, which products and operating loads are affected, and whether it followed maintenance, a recipe change or a software change. Describe the operational consequence: rejected product, unstable throughput, equipment wear or repeated operator intervention.

Agree the investigation limits before testing. Identify who owns the process, who can authorise changes and which conditions require stopping. Any safety-related test design must involve a qualified assessor. Do not use this guide as authority to bypass protection, force an output or change an operating mode.

Which trends should you capture?

Request a trend that includes several complete oscillations and enough preceding history to show how they began. Keep the original data alongside any annotated screenshots.

Signal or contextWhat to captureDiagnostic question
SetpointRequested value and any ramped or selected valueIs the target itself moving?
Process valueValue used by the controller, quality status and raw measurement where availableIs the controller seeing the same signal as the display?
Controller outputCalculated output, applied command and output limitsIs the requested correction reaching the actuator?
Actuator feedbackActual valve travel or drive speed, where availableDoes the equipment follow the command?
Control stateAutomatic, manual, cascade, override and tracking statesWho currently controls the output?
DisturbancesRelevant flow, pressure, load and upstream conditionsDoes another event precede the oscillation?
Operating contextProduct, recipe, production stage and equipment configurationDoes the behaviour depend on operating conditions?

Check timestamps and the recording method before interpreting which signal moved first. Sparse sampling or historian deadbands can conceal movement and distort the apparent sequence.

The guides to choosing plant-data sampling intervals and aligning timestamps across plant systems cover those recording decisions.

Could measurement noise be driving the output?

Compare the measurement entering the controller with the value shown to operators. Record any filtering, scaling or averaging between them.

Look for rapid measurement fluctuations accompanied by rapid output changes. Check whether the fluctuations coincide with equipment switching, vibration, cleaning or other identifiable events. Treat these as investigation leads, not proof of an electrical fault.

Inspect the measurement chain: sensor condition, installation, connections, signal quality and configured range. Use an independent measurement only where it is suitable for the process and can be obtained safely.

Derivative action can amplify rapid measurement changes, while added filtering can introduce delay. Increasing a filter may therefore make a trend look calmer without resolving the underlying problem. The PID chapter of Feedback Systems, chapter 11 (Åström and Murray) covers derivative filtering and its trade-offs.

Document the signal path before changing either filtering or tuning.

Is the valve or actuator following its command?

Compare controller output with actual movement. A changing output is a command; it is not evidence that the valve moved.

Look for periods when the command changes but position remains still, followed by a sudden movement. Also check whether behaviour changes when movement reverses. Treat such patterns as a reason to inspect the valve, positioner and linkage for mechanical causes; a trend alone does not establish the fault.

Review available maintenance evidence for the actuator, positioner, linkage and supporting supplies. Check the meaning of the feedback signal: a command echoed back through software is not an independent measurement of travel.

Even genuine position feedback does not prove that flow changes as intended, so check the process response as well.

Resolve confirmed actuator faults before using tuning changes to compensate for them.

Could process delay or operating conditions explain the cycling?

Ask how long it takes for an output change to produce a visible process response. Distinguish that delay from the time the process then takes to approach its new condition.

Transport distance, mixing, heat transfer and measurement location can affect the response seen by the controller. A controller that keeps correcting before the effect becomes visible can contribute to overshoot and repeated reversal.

Compare episodes at different production rates, loads and equipment configurations. Record whether the same tuning is stable in one condition and unstable in another. If behaviour depends on the operating condition, record that before judging the tuning.

Also inspect interacting loops and external commands. If an upstream pressure cycle consistently precedes a downstream flow cycle, investigate that relationship before assuming the downstream controller caused it.

How do you distinguish tuning from an instrument fault?

Use several pieces of evidence together. No single waveform is a reliable diagnosis.

Observed patternWhat to investigate firstEvidence still needed
Measurement jumps without a corresponding process observationMeasurement chain or signal processingSuitable independent measurement and instrument checks
Output changes while actuator position holds, then jumpsActuator behaviourTravel feedback and maintenance inspection
Setpoint or upstream conditions cycle firstExternal command or disturbanceCommand ownership and related process trends
Output stays at a limit while error persistsCapacity, constraints and controller handling of limitsApplied limits, process demand and controller configuration
Measurement is credible, actuator follows, and cycling persists under steady conditionsTuning and process dynamicsVerified configuration and an authorised test plan

An output limit deserves particular attention. Integral action can accumulate while an actuator cannot deliver further correction; the resulting recovery depends on the controller's anti-windup and tracking behaviour. Feedback Systems, chapter 11 (Åström and Murray) describes integrator windup and common anti-windup schemes.

Check the actual controller implementation, including action direction, parameter units, execution interval and limit handling. Do not assume that tuning values transfer directly between different PID implementations: the same controller can be written with gains or with integral and derivative times, and products differ in how they implement it. Check the controller's own documentation.

When is a tuning change justified?

Consider tuning after the measurement, actuator response, operating context and controller configuration have been checked.

Before changing anything, record:

  • The current settings and a recoverable configuration.
  • The operating condition for the comparison.
  • The approved change and person responsible.
  • The permitted process deviation and stop conditions.
  • The observations required to judge the result.
  • The rollback decision and procedure.

Define success in process terms. The aim might be stable product temperature within an agreed operating band, without unacceptable actuator movement or repeated output limiting.

Make a controlled comparison and retain the evidence. Any safety-related test design must involve a qualified assessor. Record accepted changes through the plant's control-software change process.

What common mistakes should you avoid?

Changing several things together. Adjusting tuning, filtering and output limits in one attempt makes the result difficult to attribute.

Treating every repeated waveform as poor tuning. Repetition identifies a pattern, not its cause.

Trusting a smooth display without checking its source. Establish what the controller receives and what the historian records.

Testing outside representative conditions. State which loads and production stages the evidence covers.

Assuming manual operation proves the diagnosis. Changing mode alters the control arrangement. Any proposed mode change needs an authorised test plan; it is not a universal troubleshooting step.

Frequently asked questions

Should I reduce proportional gain immediately?

First establish that the measurement and actuator are credible. A gain change may alter the symptoms without resolving an instrument or mechanical fault.

Can I diagnose the loop without valve-position feedback?

You can identify suspects, but confidence is lower because command and movement cannot be compared directly. State that limitation and agree what safe inspection or additional measurement would close it.

What should the investigation handover contain?

Keep the annotated trends, operating conditions, instrument and actuator findings, controller settings, authorised changes and verification results. Name any unresolved uncertainty and its owner. The handover should make clear why the diagnosis was accepted and what evidence would require reconsidering it.

Next step

Want to apply this to your plant?

Every project starts with a fixed-fee assessment, and you can stop after the report.