Controller output plus every major published tuning method: Ziegler-Nichols, Cohen-Coon, IMC/Lambda, Tyreus-Luyben, Chien-Hrones-Reswick, SIMC (Skogestad), a relay-feedback auto-tune estimator, and a practical loop-type quick reference.
A PID controller drives a process toward its setpoint using three terms acting on the error (setpoint minus measured value): Proportional response reacts to the size of the current error, Integral response accumulates error over time to eliminate the steady-state offset that proportional-only control leaves behind, and Derivative response reacts to how fast the error is changing, damping overshoot. Most industrial loops actually run PI (no derivative) — derivative amplifies measurement noise, which is a problem on noisy signals like flow, and is only worth the trade-off on slower, cleaner loops like temperature.
Frequently Asked Questions
Why do most flow and level loops skip the derivative term?
Flow and level measurements are often noisy (turbulence, splashing). Derivative action amplifies that noise into large, erratic controller output swings — for these loops the added noise usually costs more than the derivative term’s stability benefit.
What is integral windup?
When the controller output saturates (valve fully open, for example) but error is still present, a naive integral term keeps accumulating anyway — then overshoots badly once the process finally responds, because the integral has to unwind first. Real controllers use anti-windup logic to stop the integral term accumulating while saturated.
What are common PID tuning methods?
Ziegler-Nichols (open-loop reaction curve or closed-loop ultimate-gain methods) is the classic starting point; many modern loops are tuned by trial adjustment against the process’s known dynamics, or by auto-tuning features built into the DCS/PLC.
Why does my loop oscillate even though it eventually settles?
That is usually too much proportional or integral gain for the process’s actual dynamics — the controller is reacting faster than the process can physically respond, overshooting the correction each time. Reducing gain (or increasing TMS/reset time) usually settles it, at the cost of slower response.