5.3 Loop Tuning and Response Analysis

Key Takeaways

  • Use an approved test scope, operating limits and recovery plan. Recorded trends or simulation can supply evidence without perturbing an operating process.

  • Quarter-amplitude decay compares successive same-side peaks and is a historical criterion.

  • A self-regulating manual step can supply signed gain, delay and time constant for model-based PI tuning.

  • Classic ultimate-sensitivity values come from a valid proportional-only stability-boundary test. Measurement conditions and controller form matter.

  • Trend command, travel and measurement together. Sticking, windup and interaction can all cause cycling.

Last updated: November 2026

Loop Tuning and Response Analysis

Tuning changes how the controller reacts to a setpoint change or a process disturbance. Specify the required performance first: limits on overshoot, time to recover, allowed process variation and acceptable valve activity. Trend setpoint, PV, requested and applied output, valve travel when available, active limits, and relevant upstream variables on a common time base. A limit cycle is a repeated oscillation. It can arise from tuning, a nonlinear final element, or another loop's disturbance.

Separate the response from the test method

For the intended operating region, record the original settings and controller form, and confirm input scaling, action, signal quality, output limits, available travel and mode tracking. Use existing trends or a simulator if the proposed perturbation is unsuitable for the operating process. For an authorized field test, use the approved operating state, perturbation size, limits, stop criteria and restoration procedure. Coordinate with operations and respect mandatory protective functions. Under OSHA process safety management, covered processes have documented limits and operating procedures, and qualifying changes follow management of change.

Inspect both the response to a setpoint and the response to load where appropriate. A small overshoot may be acceptable in one loop while the same overshoot is unacceptable in another. Quarter-amplitude decay is a historical design criterion. It compares two successive peaks on the same side of the settled value. If they are 8 and 2 engineering units above that value, the decay ratio is 2/8=0.252/8=0.25. This measured ratio neither defines a universal target nor quantifies robustness by itself.

Read a reaction curve

On a self-regulating loop, an authorized small step in manual allows the process to reveal its response to a known output change. Keep other inputs steady enough to estimate the effect. Record the step, delay, shape and eventual steady change. In a first-order plus delay fit, the time constant is measured from the modeled onset to 63.2% of the final change. The observed response includes the installed final element and measurement dynamics. If the variable continues ramping, use an integrating model or another supported method. A clipped, noisy or disturbed trend may be unsuitable for applying a formula.

The original Ziegler–Nichols paper relates the steepest reaction-curve slope per unit output step and apparent delay to starting controller settings. A model-based alternative chooses a desired speed explicitly. For example, the Skogestad SIMC method starts from a first-order plus delay fit for PI control. Using positive gain magnitude and setting the action separately, it gives

Kc=τ∣Kp∣(τc+θ),Ti=min⁡{τ,4(τc+θ)}.K_c=\frac{\tau}{|K_p|(\tau_c+\theta)}, \qquad T_i=\min\{\tau,4(\tau_c+\theta)\}.

In this example, an output step of 5 percentage points produces a settled PV change of 10 input-span points. The fitted τ\tau is 30 s and θ\theta is 5 s. With a desired closed-loop time parameter τc=10\tau_c=10 s, ∣Kp∣=2|K_p|=2, Kc=30/[2(10+5)]=1.0K_c=30/[2(10+5)]=1.0, and Ti=min⁡(30,60)=30T_i=\min(30,60)=30 s, or 0.50 min per repeat. This is a calculated starting point for the stated model, with derivative disabled. A larger τc\tau_c in this rule reduces gain and typically yields a smoother response. Confirm the equation, units, limits and performance on the actual application.

Interpret ultimate gain and period on recorded data

In the classic closed-loop ultimate-sensitivity test, a proportional-only controller at the stability boundary of the model produces a sustained oscillation with unchanged amplitude. The gain magnitude there is KuK_u and the interval between same-side peaks is PuP_u. This is a different experiment from the manual reaction-curve test. Use lab, simulation or suitable existing records for the following calculation. For the classic PID settings, Kc=0.60KuK_c=0.60K_u, Ti=Pu/2T_i=P_u/2, and Td=Pu/8T_d=P_u/8. With Ku=4.0K_u=4.0 and Pu=8.0P_u=8.0 min, the values are 2.4, 4.0 min and 1.0 min. The classic reset setting is equivalent to two repeats per ultimate period. A flat-amplitude trace produced by output clipping or valve friction does not establish a valid KuK_u. This heuristic can give aggressive values and the applicable controller form must be confirmed before any use.

Diagnose before retuning

Observed behaviorDiagnostic check
Command moves while measured valve travel pauses and then jumpsCompare actuator loading and travel. Check air delivery, linkage and friction using the prescribed diagnostic procedure.
Command and travel follow each other while several loops cycle togetherCompare the timing of shared supply, load and upstream loops. Separate interaction from a local tuning problem.
Command is smooth but PV is noisy or delayedCheck the measurement and process effects before adding filter lag.
Command hits a limit and the integral term keeps increasingTrace the effective applied-output limits and the anti-windup or tracking path.

A noisy flow loop often uses PI. A temperature loop with significant lags can benefit from filtered derivative when the model and noise support it. A surge-level loop may allow more level variation to smooth downstream flow. Check the actual objective and observed process dynamics for any application. Compare recovery and output movement before and after an authorized change, then verify mode transitions and document the final settings.

Test Your Knowledge

A simulation uses a proportional-only controller. A verified sustained oscillation at the stability boundary gives ultimate gain Ku = 4.0 and same-side peak period Pu = 8.0 min. What are the classic Ziegler–Nichols PID starting values for Kc, Ti and Td?

A

2.0, 8.0 min, 2.0 min

B

1.8, 6.67 min, 0.0 min

C

3.2, 4.0 min, 1.0 min

D

2.4, 4.0 min, 1.0 min

Test Your Knowledge

An approved valve test shows a smooth changing command, actuator loading pressure that rises steadily, and a stem that remains at one position before breaking away in a jump. Which interpretation best fits these observations?

A

The transmitter’s thermal lag explains the pause in measured stem travel.

B

Stick-slip in the final element is indicated. Use valve diagnostics to locate the source of resistance.

C

The trend proves that the controller proportional gain is too high.

D

The trend proves that excess derivative gain is causing a setpoint kick.

Test Your Knowledge

A self-regulating process fit gives gain magnitude 2.0, a 30 s time constant and a 5 s delay. Use the SIMC first-order PI formula with an intended closed-loop time parameter of 25 s. What gain and integral time result?

A

Gain 2.0 and integral time 120 s.

B

Gain 0.5 and integral time 120 s.

C

Gain 0.5 and integral time 30 s.

D

Gain 1.0 and integral time 30 s.

Sections you finish are checked off in the contents.