Feedback Loop Design
Feedback compares a measurement to a target and drives the difference toward zero; good design balances responsiveness against stability and noise.
The Idea of Feedback
A feedback loop continuously measures a controlled quantity, compares it to a desired setpoint, and adjusts an actuator to reduce the error. This closed loop lets a system reject disturbances and track targets without needing a perfect model of the plant. Feedback is the foundation of automatic control, from thermostats to plasma-position control.
Open Versus Closed Loop
Open-loop control commands the actuator from a model alone, without measuring the result; it is simple but cannot correct for disturbances or model error. Closed-loop control uses the measurement to correct itself, trading added complexity and the risk of instability for accuracy and disturbance rejection. Feedforward can be combined with feedback: a model predicts most of the needed action, and feedback trims the residual error.
The Core Trade-Off
Higher loop gain makes the system respond faster and reject disturbances more strongly, but too much gain, especially combined with delay, causes oscillation or instability. The art of loop design is maximizing responsiveness while keeping adequate stability margin. Delay is the enemy of both goals, which is why latency and jitter in the digital implementation directly limit achievable performance.
- Bandwidth: how fast the loop can track and reject, limited by delay and sampling
- Stability margin: how much delay or gain error the loop tolerates before oscillating
- Steady-state error: residual offset, reduced by integral action
- Noise sensitivity: high gain amplifies measurement noise into the actuator
From Continuous to Digital
Classical control is formulated in continuous time, but a digital controller acts at discrete instants. The sample rate, computation delay, and hold behavior all modify the loop, so a design that is stable on paper must be verified in its discrete-time form. In safety-critical settings, the loop is designed with margin against the worst-case timing the real-time system can guarantee, not the typical timing, so that a slow cycle cannot tip a healthy loop into instability.