Change the size of the step. If a small change behaves perfectly and a large one overshoots badly, the problem is actuator saturation, not tuning.
Windup is a large-signal effect. A small step never demands more than the actuator can give, so the integrator never accumulates anything it cannot justify and the loop behaves exactly as tuned. A large step pins the actuator, the error persists for reasons the controller cannot fix, and the integral climbs anyway — storing a correction that has to be paid back by negative error after the process reaches setpoint. A genuinely mistuned loop overshoots proportionally at every step size.
The test needs the actuator to actually be reachable. If the setpoint is beyond what the plant can produce at all, the output stays pinned for ever, the integral grows without bound, and no amount of anti-windup helps because there is no correct integral value to settle at — that is a sizing fault wearing a control fault’s clothes. Dead time also produces step-size-independent overshoot that looks similar on a trend and is a different problem entirely.
Estimate with the rule, then check it against the calculator that models it properly.
Open Integral Windup & Anti-Windup →Change the size of the step. If a small change behaves perfectly and a large one overshoots badly, the problem is actuator saturation, not tuning. Windup is a large-signal effect.