Feedback Control Systems

Learning Objectives

By the end of this page, you will be able to:

• Identify the six components of a feedback control loop: process, sensor, controller, actuator, feedback path, and forward path.
• Distinguish positive feedback from negative feedback and explain why negative feedback is the basis of nearly all regulating control systems.
• Define gain, sensitivity, and phase margin, and explain how each relates to stability.
• Classify proportional, integral, derivative, and PID control by what each term corrects for.
• Explain why time delay and nonlinearity threaten the stability of a feedback loop.

Quick Answer

Feedback control is the technique of continuously measuring a system's actual output and feeding that measurement back to adjust the input, so the system tracks a desired value even as conditions change. It matters because it is what lets systems reject disturbances they were never explicitly designed for — a cruise control doesn't need to know about every hill in advance, it just keeps comparing actual speed to target speed and correcting. The two flavors are negative feedback, which reduces the error and stabilizes the system (used in almost all regulating control), and positive feedback, which amplifies deviations and is used deliberately in things like oscillators, but destabilizes a system if it appears where negative feedback was intended.

Anatomy of a Feedback Loop

A feedback control system is built from six recurring parts, and being able to point to each one in an unfamiliar diagram is a core exam skill.

• Process (plant) — the system being controlled, e.g., a room's temperature, a motor's speed.
• Sensor — measures the current state of the process (a thermistor, tachometer, encoder).
• Controller — compares the sensor reading to the setpoint and computes a control signal (often a PID controller).
• Actuator — carries out the controller's decision (a heating element, a motor drive).
• Feedback path — carries the measured output back to the controller for comparison.
• Forward path — carries the controller's decision to the actuator.

Negative vs Positive Feedback

Definition: Negative feedback subtracts the measured output from the setpoint, so any deviation from the target produces a correction that opposes (reduces) that deviation. Positive feedback adds the measured output to the input, so any deviation is reinforced (amplified) rather than corrected.

Explanation: In a negative-feedback voltage regulator, if the output voltage rises above target, the controller reduces the drive signal — pushing the output back down. In a positive-feedback oscillator circuit, a small signal is deliberately amplified and fed back in-phase so it grows into a sustained oscillation rather than settling.

Example: A public-address system where a microphone picks up sound from its own speaker is an accidental positive-feedback loop — the signal grows until you hear the familiar squeal.

Real-world example: A voltage regulator chip uses negative feedback to hold its output steady despite changes in input voltage or load current; an LC oscillator circuit deliberately uses positive feedback at a specific frequency to sustain oscillation.

Why it matters: Negative feedback is what makes almost every regulating control system (thermostats, cruise control, voltage regulators) stable and self-correcting. Positive feedback is genuinely useful, but only in circuits designed to grow a signal (oscillators, Schmitt triggers) — if it shows up unintentionally in a regulating loop (e.g., due to wiring or sign errors), the system will diverge or oscillate uncontrollably instead of settling.

Common misunderstanding: Students assume "positive feedback" is always good because "positive" sounds favorable. In control engineering it specifically means "reinforces deviation," which is destabilizing in a regulator — the word describes the sign of the loop, not a value judgment.

Gain, Sensitivity, and Phase Margin

Definition: Gain is how much the output changes per unit change in input. Sensitivity measures how much the output changes when one of the system's internal parameters (not the input) shifts slightly. Phase margin measures, in frequency-domain terms, how far a system's phase shift is from the point that would cause instability.

Explanation: If a controller has high gain, a small error produces a large corrective action — fast, but risking overshoot. Sensitivity tells you how fragile the design is to real-world variation: a highly sensitive amplifier circuit might work perfectly with one resistor tolerance but misbehave with a slightly different one. Phase margin comes from frequency-response (Bode plot) analysis and is one of the standard numeric stability checks used in the next chapter, Control System Analysis.

Example: A control loop with a phase margin of 45° is generally considered to have a reasonable stability buffer; a phase margin near 0° means the system is on the edge of oscillating uncontrollably.

Real-world example: Op-amp feedback amplifier designers explicitly check phase margin in datasheets to guarantee the amplifier won't oscillate when used with real-world capacitive loads.

Why it matters: These three quantities are the bridge between "the loop works in theory" and "the loop is robust when built with real, imperfect components." Phase margin in particular is one of the two standard margins (with gain margin) used throughout frequency-domain stability analysis.

Common misunderstanding: Students conflate "gain" (gain in general, e.g., a controller's proportional constant) with "gain margin" (a specific stability metric measured in decibels at the frequency where phase hits −180°). They are related but not the same thing — gain margin is how much extra gain the loop could tolerate before instability, not the gain itself.

Proportional, Integral, and Derivative Control

Definition: Proportional (P) control produces an output proportional to the current error. Integral (I) control accumulates past error over time to eliminate any steady, lingering offset. Derivative (D) control reacts to the rate of change of error to anticipate and dampen overshoot. PID control combines all three.

Explanation: A thermostat using pure proportional control might settle 1–2°C below setpoint forever (called steady-state error) because the heater output naturally shrinks to nearly zero as error shrinks to zero. Adding an integral term keeps "remembering" that small persistent error and keeps nudging the output up until the error is fully eliminated. Adding a derivative term looks at how fast the temperature is rising and eases off the heater in advance, reducing overshoot.

Example: Consider a system with error $e(t)$; the PID control signal is $u(t) = K_p e(t) + K_i \int_0^t e(\tau)\,d\tau + K_d \frac{de(t)}{dt}$ With only $K_p$ nonzero, the controller is pure proportional control.

Real-world example: Cruise control in most cars uses PID logic: proportional action for the immediate speed gap, integral action to eliminate any lasting speed offset on a long hill, and derivative action to smooth out the throttle response.

Why it matters: PID is the single most widely used control algorithm in industry (estimated in over 90% of industrial control loops) precisely because these three simple terms cover the three most common performance problems: slow response, steady-state error, and overshoot.

Common misunderstanding: Students think adding all three terms (P, I, and D) is always "the best" choice. In practice, the derivative term amplifies sensor noise (since it reacts to fast changes), so many real systems use only PI control, and simple on/off loads sometimes use pure P or even bang-bang (on/off) control instead.

Visual Learning

Trace this diagram for negative feedback: the feedback signal is subtracted at the summing junction. If it were added instead, the loop would become positive feedback and the same PID logic would drive the system away from the setpoint rather than toward it.

Real-World Applications

• Industrial automation — temperature control in chemical reactors, speed regulation on production lines.
• Aerospace — guidance and navigation loops that continuously correct heading and altitude.
• Medical devices — insulin pumps and ventilators that adjust dosing/pressure based on continuous sensor feedback.
• Automotive — anti-lock braking systems (ABS) and electronic throttle control both rely on tight negative-feedback loops.
• Consumer electronics — audio equalizers and op-amp based circuits use feedback to control gain precisely.

Key Terms

Term	Definition
Feedback path	The route by which the measured output is returned to the controller for comparison.
Negative feedback	Feedback that opposes and reduces deviation from the setpoint; the basis of stable regulation.
Positive feedback	Feedback that reinforces and amplifies deviation; used deliberately in oscillators.
Gain	How much a system's output changes per unit change in its input or error signal.
Sensitivity	How much the output changes in response to a small change in an internal system parameter.
Phase margin	A frequency-domain measure of how far a system is from the instability threshold.
Proportional (P) control	Control action proportional to the current error.
Integral (I) control	Control action based on the accumulated (past) error, used to eliminate steady-state error.
Derivative (D) control	Control action based on the rate of change of error, used to reduce overshoot.

Common Mistakes

Misconception 1: "Positive feedback is better because it amplifies the effect." Why it's wrong: In a regulating system, amplifying deviation is exactly the opposite of what's needed — it drives the output further from the setpoint instead of correcting it. Correct understanding: Negative feedback is what regulating systems need because it counteracts deviation; positive feedback is reserved for applications that specifically want growth or sustained oscillation, like oscillator circuits.

Misconception 2: "A PID controller with a derivative term is always more accurate than one without." Why it's wrong: The derivative term differentiates the error signal, which sharply amplifies high-frequency sensor noise — in a noisy real-world sensor, this can make performance worse, not better. Correct understanding: Many practical systems intentionally use PI control (omitting D) or filter the derivative term, trading a bit of overshoot reduction for much better noise immunity.

Misconception 3: "Integral control eliminates error instantly." Why it's wrong: Integral action accumulates error over time — it needs time to build up enough correction to close a steady-state gap, and if tuned too aggressively it can cause slow oscillations ("integral windup"). Correct understanding: Integral control eliminates steady-state error eventually, not instantly, and must be tuned carefully to avoid overshoot from accumulated error building up faster than the system can respond.

Comparison and Connections

Feedback Type	Effect on Error	Typical Use
Negative feedback	Reduces/opposes error	Regulating systems: thermostats, cruise control, voltage regulators
Positive feedback	Amplifies/reinforces error	Oscillators, Schmitt triggers, deliberately unstable/switching circuits

Control Term	Reacts To	Fixes	Risk if Overused
Proportional (P)	Current error	Slow response	Steady-state error remains
Integral (I)	Accumulated past error	Steady-state error	Overshoot, slow oscillation ("windup")
Derivative (D)	Rate of change of error	Overshoot	Amplifies sensor noise

Practice Questions

Recall

1. Name the six components of a feedback control loop. Answer guidance: Process, sensor, controller, actuator, feedback path, forward path.
2. What is the difference between negative and positive feedback? Answer guidance: Negative feedback subtracts the output from the setpoint to reduce error; positive feedback adds it, reinforcing and amplifying deviation.

Understanding

3. Explain why a PID controller with only a proportional term typically leaves a steady-state error. Answer guidance: Proportional output shrinks as error shrinks, so near the setpoint the corrective action becomes very small — often too small to fully close the gap — leaving a residual offset that only integral action can remove.
4. Why does the derivative term in a PID controller amplify sensor noise? Answer guidance: Derivative action responds to the rate of change of the error signal, and random noise fluctuates quickly, producing large rate-of-change values even though the underlying signal hasn't meaningfully changed.

Application

5. A public-address system starts to squeal when the microphone is placed too close to the speaker. Identify the feedback type involved and explain the mechanism. Answer guidance: This is unintended positive feedback — sound from the speaker re-enters the microphone in-phase and gets re-amplified repeatedly, growing into an audible squeal instead of settling.
6. A cruise control system settles at a speed slightly below the driver's setpoint on a long uphill grade. Which PID term should be increased to fix this, and why? Answer guidance: The integral term, because the lingering, constant offset is a steady-state error, and integral action is specifically designed to accumulate and eliminate exactly that kind of persistent error.

Analysis

7. Compare a system with a phase margin of 60° to one with a phase margin of 5°. Which is more robust to component variation, and why? Answer guidance: The 60° phase margin system is far more robust — it has a large buffer before the loop's phase reaches the −180° threshold that causes instability, so small real-world variations in components are unlikely to destabilize it, whereas the 5° margin system is nearly on the edge of oscillating.
8. A junior engineer proposes always using full PID control on every feedback loop "to be safe." Evaluate this proposal for a simple on/off refrigerator compressor control. Answer guidance: Full PID is unnecessary and often impractical for a system that can only switch a compressor on or off — such systems typically use simple hysteresis (on/off with a deadband) rather than continuous PID output, since the actuator itself cannot apply a continuously variable signal.

FAQ

Q1: Why is negative feedback used far more often than positive feedback in control systems? Because most control applications need to hold a value steady against disturbances, which requires the error-reducing behavior only negative feedback provides. Positive feedback is reserved for special cases like oscillators where growth or fast switching is the goal.

Q2: Does every feedback control system need all three PID terms? No. Many real systems use only P, or PI (omitting the noise-sensitive derivative term), depending on the process dynamics and how much steady-state error and overshoot can be tolerated.

Q3: What's the difference between gain and gain margin? Gain is simply how much a controller's output changes per unit of error. Gain margin is a specific stability metric — how much additional gain the loop could tolerate before oscillation — measured from frequency-response analysis.

Q4: Can time delay in a feedback loop cause instability even with sensible gain? Yes. Delay adds extra phase lag as frequency increases, which can eat into the phase margin and destabilize a loop that would otherwise be stable — this is a key theme of the Control System Analysis chapter.

Q5: How does this chapter connect to PID controller design later in the unit? This chapter introduces P, I, and D conceptually; the Control System Design chapter goes further into how to actually choose (tune) the $K_p$, $K_i$, $K_d$ gains for a specific plant.

Quick Revision

• Feedback loop = process + sensor + controller + actuator + feedback path + forward path.
• Negative feedback opposes/reduces error → stabilizes regulating systems.
• Positive feedback reinforces/amplifies error → used deliberately in oscillators, destabilizing elsewhere.
• Gain = output change per unit input/error change; sensitivity = output change per unit internal-parameter change.
• Phase margin = how far the system's phase is from the −180° instability threshold; larger is more robust.
• P control reacts to current error but leaves steady-state error.
• I control eliminates steady-state error by accumulating past error, but risks slow oscillation if overtuned.
• D control reduces overshoot by reacting to the error's rate of change, but amplifies noise.
• PID combines all three and is the most common industrial control algorithm.
• Time delay and nonlinearity are the two biggest real-world threats to feedback-loop stability.

Related Topics

Prerequisites: Introduction to Control Systems (open-loop vs closed-loop, setpoints, gain).

Related Topics: Control System Analysis (formal stability tools: Routh-Hurwitz, Bode plots, gain/phase margin); Control System Design (tuning PID gains for a specific plant).

Next Topics: Control System Analysis — the mathematical tools used to formally check the stability and performance ideas introduced qualitatively here.