Closed-loop

What is closed-loop control?

Closed-loop control, also called feedback control, a feedback loop or a closed control loop, is automatic control that continuously measures the actual outcome of a process, compares it with the desired value (the Setpoint) and adjusts the control signal so that the difference becomes smaller. Because the controller sees the effect of its own action in the next measurement, it also corrects disturbances nobody anticipated, such as a changing load, wear or a cold outdoor temperature. Closed-loop control is therefore the foundation of industrial automation, from the temperature in a fermenter to the position of a robot arm. The opposite is open-loop control, which issues a command without checking the result.

Closed-loop = measure → compare → correct → measure again.

🧠 How does closed-loop control work?

A closed control loop runs through the same cycle over and over:

1. A Sensor or transmitter measures the process variable, for example temperature, pressure, level, flow or speed.
2. That measured value, the Process Value (PV), is compared with the Setpoint (SP).
3. A control algorithm in a controller calculates a new control signal from the error (e = SP − PV). In the process industries this is, in the vast majority of cases, a PID controller.
4. That control signal, the Manipulated Variable (MV), goes to an Actuator: a control valve, a pump or a motor driven by a variable frequency drive.
5. The process responds, the sensor measures again and the cycle repeats, in a PLC or DCS with a fixed cycle time.

The word feedback describes exactly that last part: the output is fed back to the input. In automation, feedback loop, feedback control and closed-loop control therefore mean the same thing; “feedback loop” emphasises the loop itself, “closed-loop” the fact that the loop is closed. Closed-loop control can be extended with a feedforward element that compensates known disturbances in advance, or with cascade control, in which a fast inner loop controls an intermediate variable.

🔄 What is the difference between closed-loop and open-loop control?

The difference lies not in the hardware but in the flow of information: a stepper motor normally runs open-loop, but becomes closed-loop as soon as an encoder reports the actual position back. The drive can then detect and correct missed steps, which makes a closed-loop stepper motor more reliable than a classic open-loop one.

⚖️ What is the difference between negative and positive feedback?

Almost every industrial control loop uses negative feedback: the controller counteracts the measured error. If the temperature is too high, the heating is turned down; if the pressure is too low, the valve opens further. That counteraction is what makes the system stable. The principle is old: the centrifugal governor that James Watt fitted to his steam engines from 1788 onwards was a mechanical negative feedback on engine speed, and in 1927 Harold Black of Bell Labs conceived the negative feedback amplifier, which sharply reduced distortion.

With positive feedback, the output reinforces the deviation instead. A small change grows larger and larger until the system hits a limit. That is sometimes intended, but in a control loop it is usually a fault.

Two practical situations make the distinction concrete:

• Wrong controller action - every controller has a direct acting or reverse acting setting. In a heating loop the output must fall when the temperature rises (reverse). If that setting is wrong, negative feedback turns into positive feedback: the controller opens the valve even further and the process variable runs to its maximum within a short time. Commissioning should therefore always include a check of the controller action.
• Deliberate memory - a relay with a self-holding contact uses positive feedback: after a short start pulse the relay keeps itself energised until a stop contact breaks the loop. This is how the classic start/stop circuit of a motor works, including when it is written in ladder logic in a PLC.

Negative feedback can also behave like positive feedback in practice. Because of lags and dead time, the correction arrives exactly half a period late at certain frequencies. If the gain is high enough at that point, the controller amplifies the swing instead of damping it, and the loop starts to oscillate.

🔧 When is a closed-loop controller well tuned?

Feedback is powerful but not free. An overly aggressive controller reacts too strongly to an error, overshoots and can make a stable process oscillate. Dead time in particular (the delay between an action and its measurable effect, as with a long pipe or an analyser) makes a loop prone to instability. Control engineers therefore judge a loop against a few standard criteria:

A practical rule of thumb is that sampling must be well ahead of the process dynamics: the controller’s cycle time is typically around a tenth or less of the dominant time constant, or the update rate at least ten times the loop bandwidth. That is why flow loops in a DCS often run at cycles of a fraction of a second, while temperature control of a large furnace can manage with seconds. How to tune the P, I and D actions, with methods such as Ziegler–Nichols (1942) or lambda tuning, is covered under PID.

🏭 Where is closed-loop control used?

Closed-loop control is found in almost every installation, but the speed varies enormously by application:

A servo drive is itself a textbook example of stacked closed-loop control: a position loop sets the setpoint of a speed loop, which in turn sets the setpoint of a very fast current-control loop that controls the motor current. For its Lexium 32 servo drive, for example, Schneider Electric quotes a cycle time of 62.5 µs for the current and speed loops and 250 µs for the position loop. On more complex process units, a layer of MPC or APC sits on top: the manipulated variables of such a multivariable controller are usually the setpoints of the existing PID loops.

Feedback is everywhere outside classic process control too: when dosing and filling tanks, a weight or flow measurement trims the filling valve; on multi-axis machines a slave axis follows the master axis through position feedback; and in HVAC systems control loops keep heating and cooling in balance.

Closed-loop systems work closely with SCADA and the HMI: the operator chooses the setpoint and the control mode (automatic or manual), and the historian records PV, SP and MV for trending and analysis.

📊 How well do control loops perform in practice?

On paper a closed-loop system regulates itself, but in real plants the picture is often disappointing. Around 2000, Honeywell analysed the performance of about 26,000 PID controllers in the continuous process industries. Only one third performed acceptably or excellently; 22% were fair, 10% poor and 36% were effectively running open loop: in manual mode, or with an output stuck at a limit for more than 30% of the time. In a separate Honeywell survey of 11,600 controllers across eighteen sites, the median share of PID per site was above 97%. According to the same source, a refinery, chemical plant or paper mill typically has between 500 and 5,000 control loops.

A practical approach to improving the health of your control loops:

1. Take stock of the control mode - which loops are in manual, and why? A loop that is permanently in manual is in effect an open-loop process with an operator as the controller.
2. Check the actuator first - sticking valves (stiction), a wrongly sized valve or an actuator running against its limit are more often the cause than the tuning.
3. Measure with historian data - use PV, SP and MV from the historian to find oscillating and sluggish loops.
4. Retune - only then adjust the PID parameters, preferably with a model-based method.
5. Record changes - document parameters and changes so that the next shift knows what was adjusted.

🔐 How do you secure a closed-loop control system?

A closed control loop blindly trusts its own measurement. That makes the integrity of measured values and setpoints the critical point:

• Manipulated measurements - if an attacker falsifies the PV (spoofing), the loop dutifully controls towards a false reality. Stuxnet showed operators normal values while it drove the frequency of centrifuge motors alternately to 1,410 Hz and 2 Hz.
• Manipulated setpoints - whoever can change the SP controls the process. The Oldsmar incident (Florida, 2021), where the sodium hydroxide setpoint went from 100 to 11,100 ppm and an operator immediately reverted it, showed how weakly remote access and setpoint limits are sometimes managed. An operator error later became the most likely explanation, but the lesson stands.
• Altered control logic - changes to PLC code or PID parameters via a compromised engineering station are hard to spot.

Measures in line with IEC 62443 and general cybersecurity practice:

• Hard limits in the controller itself: minimum and maximum for SP and MV, plus a maximum rate of change.
• Redundant or independent measurements and plausibility checks, so that a manipulated sensor stands out.
• Anomaly detection and logging to a SIEM, paying attention to unusual setpoint changes and mode switches.
• Monitored data integrity of PLC programs and control parameters, with periodic comparison against an approved baseline.

An important distinction: a closed-loop controller is not a safety function. Under IEC 61511, a basic process control system (BPCS) may be credited as a protection layer for a risk reduction factor of at most 10. Larger risks require an independent SIS, often built on a safety PLC, that intervenes when the control loop fails or is manipulated. This is also where security and safety meet.

❓ Frequently asked questions

What is an example of closed-loop control?

A room thermostat in a central heating system is a classic closed-loop example: it measures the room temperature, compares it with the set value and switches the boiler on or off. In industry, PID temperature control of an oven or fermenter works on the same closed-loop principle, only continuously and more accurately.

Is closed-loop control always better than open-loop control?

No, closed-loop control is more accurate and corrects disturbances, but it needs a measurement, a controller and good tuning. For simple, predictable tasks such as a timed filling cycle, open-loop control is cheaper and sufficient, and a poorly tuned closed-loop controller can actually start to oscillate.

Why does a closed-loop controller oscillate?

A closed-loop controller oscillates when it reacts too strongly or too late relative to the process dynamics, often because of excessive gain, a lot of dead time or a sticking control valve. The fix is usually to check the actuator first and then retune the controller parameters more calmly.

How many control loops does a plant have?

A refinery, paper mill or chemical plant typically has between 500 and 5,000 closed-loop control loops, according to research by Honeywell. Almost all of those loops use a PID algorithm; where advanced control such as MPC is applied, it adjusts the setpoints of these PID loops.

Is a feedback loop the same as closed-loop control?

Yes, in automation a feedback loop, feedback control and closed-loop control are three names for the same principle: the output is measured and fed back to the controller. Industrial control loops almost always use negative feedback, which reduces deviations; positive feedback amplifies deviations and is only used deliberately, for example in a self-holding relay contact.

Is closed-loop control a safety measure?

No, closed-loop control keeps the process within its normal operating limits, but under IEC 61511 it is no substitute for a safety system. Catching hazardous situations requires an independent Safety Instrumented System that intervenes when the control loop fails.

📌 In summary

Closed-loop control measures the result of a process, compares it with the setpoint and corrects continuously. This makes processes stable, accurate and resilient to disturbances, from process loops in a DCS to servo axes that control in microseconds. In practice closed-loop control needs maintenance: many loops run in manual or are poorly tuned. Because the loop blindly trusts measured values and setpoints, limits, integrity checks and an independent SIS belong in every critical design.

Closed-loop control is control with feedback: the foundation of stable, accurate industrial automation, provided measurements, setpoints and tuning can be trusted.