# The control loop: the heart of automation > Measure, compare, correct: the control loop is the basic mechanism that keeps a quantity (temperature, pressure, flow) at its setpoint, automatically and continuously. - Source : https://industryhub.site/en/learn/automation/boucle-de-regulation - Track : automation - Level : Discovery - Reading time : 7 min - Language : English ## In brief A control loop measures a quantity, compares it to a setpoint, and acts to correct the deviation. Sensor, controller, actuator: this trio, often driven by a PID controller, is the elementary gesture of automation. ## In one sentence A control loop **keeps a physical quantity, all by itself**, at the value you ask for, by continuously correcting deviations. It is the basic gesture that every control system in a plant repeats thousands of times a second. ## The three organs in a circle A loop links three organs in a closed circle: the **sensor** measures the actual quantity, the **controller** compares it to the **setpoint** and computes the deviation, the **actuator** acts to reduce that deviation. The result changes the quantity, which the sensor measures again — the loop is *closed*. As long as the deviation exists, the system corrects. This is **feedback control** (*closed loop*). ## Open loop, closed loop - In **open loop**, you command without checking the result: you set heating to 50% and hope. If a window opens, no one corrects. - In **closed loop**, you measure continuously and adapt: if the temperature drops, the controller raises the heating by itself. The closed loop is what makes a process **robust** against disturbances. ## The PID controller The most widespread controller is the **PID** (Proportional-Integral-Derivative). From the deviation $e(t) = \text{setpoint} - \text{measurement}$, it computes the command: $$u(t) = K_p\, e(t) + K_i \int_0^{t} e(\tau)\,d\tau + K_d\, \frac{d\,e(t)}{dt}$$ Three reactions add up: - **Proportional ($K_p$)**: the larger the deviation, the harder the correction. Immediate, but leaves a small residual error. - **Integral ($K_i$)**: accumulates the deviation over time to erase that residual error. A patient reaction. - **Derivative ($K_d$)**: anticipates by looking at the rate of the deviation. Damps the jolts. Tuning a PID means balancing **speed** and **stability**: you typically aim for an overshoot below **10–20%** and a fast return to setpoint, with no sustained oscillation. Methods like **Ziegler-Nichols** or *lambda tuning* give a first set of gains, then refined on the real process. A badly tuned PID oscillates; well tuned, it holds tight. ## Where you find them Loops run everywhere: in a controller ([PLC](/en/glossary#term-plc)), a distributed control system ([DCS](/en/glossary#term-dcs)), a simple thermostat. A large process unit has **hundreds to thousands** of them, interacting — hence the challenge of decoupling and anti-windup. A key safety point: a control loop **drives** the process, but it does not **protect** it. Protection belongs to a separate system, the safety instrumented system ([from risk to SIL](/en/learn/safety/du-risque-au-sil)). You never entrust safety to the loop that drives.