Feedback Control

Feedback control is a control strategy that continuously compares a desired value with the measured output of a system. The controller acts on the difference, called the error signal, so the plant or process moves closer to the setpoint. Because the output is measured and reused, feedback control is a form of closed-loop control.

In electrical and control systems engineering, feedback is used whenever the real output may drift because of changing load, disturbances, model uncertainty, component tolerances, or operating conditions. A motor speed controller, for example, does not simply apply a fixed voltage and hope the motor spins at the right speed. It measures speed, compares it to the command, and adjusts the drive signal when the load changes.

A basic feedback control system starts with a reference input or setpoint. The measured output is subtracted from that reference to form an error signal. The controller converts the error into a control signal, the plant responds, the sensor measures the output, and the measurement returns to the summing junction.

Main Parts of a Feedback Control Loop

Continuous-Time and Digital Feedback

Feedback control can be implemented with analog circuits, continuous-time models, digital controllers, programmable logic, embedded software, or computer-based control systems. The same loop idea still applies, but digital feedback also introduces sampling rate, quantization, computation delay, and discrete-time stability considerations.

Most engineering control loops use negative feedback because the measured output is subtracted from the reference. If the output rises above the setpoint, the error changes in a direction that reduces the output. If the output falls below the setpoint, the controller increases the command to bring it back.

The difference between feedback, open-loop, and feedforward control is how the controller uses information. Open-loop control applies a command without checking the result. Feedback control corrects based on measured output. Feedforward control acts on a known or measured disturbance before the output fully changes.

Feedback Control vs Feedforward Control

Feedback control is reactive because it corrects after the output measurement shows error. Feedforward control is predictive because it adjusts the input based on a known disturbance or command change before the output fully responds. Many high-performance systems use both: feedforward handles predictable effects, while feedback removes the remaining error.

A feedback control transfer function shows how the reference input becomes the output after the loop is closed. For a standard negative feedback system with forward-path transfer function \(G(s)\) and feedback-path transfer function \(H(s)\), a common closed-loop form is:

This equation is useful because the denominator contains the loop behavior. The term \(1+G(s)H(s)\) influences sensitivity, bandwidth, disturbance rejection, and stability. If that denominator produces unstable closed-loop poles, the system can oscillate or diverge even if the individual controller and plant blocks look reasonable by themselves.

The equation is a compact model, not a complete design check. Engineers still review pole locations, time response, frequency response, sensor behavior, actuator limits, and robustness before trusting a feedback design.

Disturbance rejection is one of the main reasons engineers use feedback control. When a load, environmental change, or process disturbance moves the output away from the setpoint, the sensor detects the output change, the error signal changes, and the controller drives the plant back toward the target.

Example: Motor Speed Under Changing Load

Consider a motor drive commanded to hold a fixed shaft speed. If the mechanical load suddenly increases, the motor slows down. A speed sensor detects the lower output speed, the error signal increases, and the controller commands more torque or current until the speed recovers.

Why Feedback Correction Is Not Instant

Feedback reacts after the measured output changes, so the response always depends on sensor delay, controller computation, actuator speed, and plant dynamics. This is why real feedback systems may show overshoot, undershoot, or settling time even when the final error is small.

Feedback can improve accuracy, but it also changes the dynamics of the system. A well-tuned loop moves the output toward the setpoint without excessive overshoot. A poorly tuned loop may oscillate. A loop with too much gain, too much phase lag, too much delay, or the wrong feedback sign can become unstable.

The Gain Tradeoff

Increasing loop gain often reduces steady-state error and improves disturbance rejection at low frequencies. However, higher gain can also increase overshoot, amplify noise, excite unmodeled dynamics, or reduce stability margin. The engineering goal is not maximum gain; it is enough gain to meet performance requirements while keeping the loop robust.

Delay and Phase Lag Matter

Time delay is especially dangerous in feedback systems because the controller may react to old information. If the measured output arrives late enough, the controller can add energy in the wrong phase of the response, turning correction into oscillation.

Engineers rarely judge a feedback control system by a single result. A loop may reach the setpoint eventually but still overshoot too far, settle too slowly, amplify noise, or become fragile when operating conditions change. The metrics below help connect the diagram to real performance.

Feedback control appears anywhere engineers need a system to hold, track, regulate, or stabilize a variable despite disturbances. It is used across electrical, mechanical, aerospace, process, robotics, automotive, biomedical, and power systems applications.

Textbook feedback diagrams often assume a clean measurement, a linear plant, unlimited actuator authority, and a controller that updates instantly. Real systems rarely behave that cleanly. Sensors are filtered, actuators saturate, signals are sampled, mechanical systems have backlash and friction, and plants may change with operating point.

Experienced engineers look for the non-ideal pieces of the loop before tuning the controller. A noisy sensor may require filtering, but filtering adds delay. A larger controller gain may reduce tracking error, but it can also excite flexible modes, switching ripple, or high-frequency measurement noise. A stable simulation may still fail if the model left out the slowest thermal lag or the fastest actuator limit.

The simple feedback control model becomes less reliable when the real system violates the assumptions behind the loop diagram or transfer function. This does not make feedback useless, but it changes the design problem from simple error correction to robustness, modeling, filtering, and safety review.

• Wrong feedback sign: The measured output reinforces the error instead of reducing it, creating positive feedback behavior.
• Large time delay: The controller reacts to old output information and may correct in the wrong direction for the current state.
• Actuator saturation: The controller requests more action than the actuator can deliver, increasing overshoot or recovery time.
• Sensor noise: The controller reacts to measurement noise instead of true output behavior, especially when derivative action or high gain is used.
• Unmodeled dynamics: Flexible modes, resonance, backlash, transport lag, switching behavior, or nonlinear friction can appear outside the simplified model.
• Operating point changes: A loop tuned for one speed, load, temperature, or pressure may perform poorly at another condition.

Feedback control is powerful, but many control problems come from applying the concept too casually. The common mistakes below are especially important because they can make a loop look correct on a block diagram while behaving poorly in hardware or simulation.

• Assuming more gain is always better: Higher gain may reduce error, but it can also reduce phase margin and create oscillation.
• Ignoring the sensor: A slow, noisy, or poorly calibrated sensor can dominate the loop behavior.
• Treating feedback as disturbance prevention: Feedback usually corrects after the output moves; it does not always prevent the first deviation.
• Using open-loop intuition for closed-loop behavior: Closing the loop changes poles, bandwidth, sensitivity, and transient response.
• Checking final value but not transient response: A system can eventually reach the setpoint while still overshooting too far or settling too slowly.
• Forgetting actuator limits: Saturation can cause sluggish recovery, overshoot, integral windup, or loss of control authority.
• Assuming feedback always means PID: PID control is common, but feedback can also use lead-lag, state-space, model-based, adaptive, digital, or custom controller structures.