Feedback in a system is the process of routing a portion of the output signal back to the input to regulate, stabilize, or modify the system's overall behavior. In electrical and electronic circuits, this mechanism fundamentally alters a system's gain, bandwidth, input/output impedance, and distortion characteristics. Think of it like a thermostat in your house: the heater doesn't just run blindly; it measures the room's actual temperature (the output) and compares it to your dial setting (the input) to decide whether to keep heating or shut off. Without feedback, an operational amplifier is essentially a comparator that slams to its voltage rails at the slightest microvolt difference; with feedback, it becomes a precision, predictable building block.
The Core Mechanics: Negative vs. Positive Feedback
When we talk about what feedback changes in a real circuit, we are usually talking about the trade-off between raw gain and system stability. By sacrificing some of an amplifier's massive open-loop gain, negative feedback drastically reduces Total Harmonic Distortion (THD), flattens the frequency response, and makes the circuit's behavior dependent on passive, highly stable components (like resistors) rather than the temperature-sensitive silicon inside the chip. Conversely, positive feedback reinforces the output, driving the system into saturation or sustained oscillation.
| Application / Circuit | Feedback Polarity | Sensing Network | Primary Effect on Output |
|---|---|---|---|
| Non-Inverting Op-Amp | Negative | Resistor Voltage Divider | Stabilizes voltage gain, lowers output impedance, reduces THD to < 0.01% |
| Wien Bridge Oscillator | Positive (with amplitude limiting) | RC Series/Parallel Network | Sustains continuous sine wave oscillation at a precise target frequency |
| Buck Converter (e.g., LM2596) | Negative | Resistor Divider to FB Pin | Regulates Vout to exact target (e.g., 5.0V) despite input voltage or load transients |
| Schmitt Trigger (Comparator) | Positive | Resistor to Non-Inverting Input | Adds hysteresis (deadband) to prevent output chatter when crossing noisy thresholds |
| Phase-Locked Loop (PLL) | Negative | Phase Detector + VCO | Locks output frequency and phase to an external reference clock signal |
For a deeper dive into the mathematics of loop gain and stability margins, the Electronics Tutorials guide on Op-Amp Feedback provides excellent Bode plot breakdowns, while Texas Instruments' Op-Amp resources offer practical application notes on compensating feedback loops to prevent phase margin collapse.
Worked Numeric Examples: Designing Feedback Networks
Let's look at two distinct bench scenarios where you must calculate feedback components using real values.
Scenario 1: Op-Amp Closed-Loop Gain
You are building a sensor interface using an LM358 op-amp. Your sensor outputs a maximum of 0.25V, but your Arduino's ADC needs a 3.0V full-scale reading. You need a non-inverting amplifier with a gain of exactly 12.
The formula for non-inverting gain is:
Av = 1 + (Rf / Ri)
- Target Gain (Av) = 12
- Choose a standard Ri = 1.0 kΩ (to keep input impedance reasonable)
- 12 = 1 + (Rf / 1000)
- 11 = Rf / 1000 → Rf = 11.0 kΩ
What this changes in the real circuit: The LM358 has a typical open-loop gain of 100,000. If you fed 0.25V directly into it without feedback, the output would instantly slam to the positive rail (~3.5V on a 5V supply). By adding the 11kΩ and 1kΩ resistors, the feedback network forces the inverting input to match the non-inverting input (the "virtual short" principle). Even if the chip heats up and its internal open-loop gain drops to 50,000, the closed-loop output remains locked at 3.0V (0.25V × 12) with an error of mere microvolts.
Scenario 2: Switch-Mode Power Supply (SMPS) FB Pin
You are designing a custom buck converter using the ubiquitous LM2596, which has an internal reference voltage (Vref) of 1.23V. You need a rock-solid 5.0V output rail to power a Raspberry Pi.
The feedback formula for this regulator is:
Vout = Vref × (1 + R1 / R2)
- Target Vout = 5.0V
- Vref = 1.23V
- Choose R2 = 1.0 kΩ (standard practice for the lower divider leg)
- 5.0 = 1.23 × (1 + R1 / 1000)
- 4.065 = 1 + (R1 / 1000)
- 3.065 = R1 / 1000 → R1 = 3065 Ω
Since 3065 Ω isn't a standard E96 resistor value, you select the nearest 1% tolerance resistor: 3.09 kΩ.
Recalculating the actual output: Vout = 1.23 × (1 + 3090 / 1000) = 5.03V. This 30mV deviation is well within the Raspberry Pi's acceptable 5V ±5% tolerance. The feedback loop continuously adjusts the internal MOSFET's PWM duty cycle to maintain this 5.03V at the output capacitors, regardless of whether the Pi is idling at 0.5A or pulling 2.5A under load.
Where You Meet This In Practice
Bench Reality Check: When troubleshooting a dead power supply, the first thing I check after the main switching FET is the feedback resistor divider. If the upper resistor drifts high due to thermal stress, the controller thinks the output is too low and drives the duty cycle to 100%, blowing the output capacitors. Always measure feedback resistors in-circuit (power off) if a supply is over-volting.
Beyond textbook op-amps, feedback is the invisible backbone of modern power and control systems:
- Switch-Mode Power Supplies (SMPS): Every buck, boost, or flyback converter uses a feedback pin (usually labeled FB or COMP). An optocoupler or auxiliary winding routes the secondary side voltage back to the primary side controller to maintain regulation across an isolation barrier.
- Audio Amplifiers: Class AB and Class D audio amplifiers use global negative feedback. Without it, the output impedance of the amplifier would be high, failing to "grip" the speaker cone and resulting in muddy, uncontrolled bass (poor damping factor).
- Motor Speed Controllers (ESCs): Sensorless brushless DC (BLDC) motor controllers rely on back-EMF feedback. By measuring the voltage induced in the undriven stator winding, the microcontroller determines the rotor's exact physical position to time the next commutation step.
- Phase-Locked Loops (PLLs): Used in RF synthesizers and microcontroller clock generation, PLLs use a phase detector to compare a divided-down VCO output against a crystal reference, feeding the error voltage back to lock the frequencies together.
Common Confusions: Feedforward and Hysteresis
When studying control theory, people commonly confuse standard negative feedback with two related but distinct concepts: feedforward and hysteresis.
Feedback vs. Feedforward
Feedback is reactive. It waits for an error to occur at the output, measures it, and corrects it. If a heavy load is suddenly applied to a power supply, the voltage sags, the feedback loop detects the sag, and increases the duty cycle to recover.
Feedforward is proactive. It measures a disturbance at the input before it affects the output. In a line-regulated power supply, if the input AC voltage suddenly drops, a feedforward circuit instantly increases the PWM duty cycle before the output DC voltage has a chance to sag. High-end server power supplies use both: feedforward for immediate input transient response, and feedback for precise output regulation.
Negative Feedback vs. Hysteresis (Positive Feedback)
Beginners often assume all feedback is meant to stabilize a system at a single point. However, hysteresis deliberately injects positive feedback to create a deadband. Consider a comparator monitoring a noisy 12V battery to trigger a low-voltage alarm at 11.0V. With pure negative feedback (or no feedback), electrical noise from a motor starting might cause the input to rapidly bounce between 10.99V and 11.01V, causing the alarm to chatter violently. By adding a high-value positive feedback resistor from the output to the non-inverting input, you create a Schmitt trigger. The threshold to turn on might be 11.0V, but the threshold to turn off shifts to 11.2V. This 0.2V hysteresis window completely ignores the noise, ensuring a clean, single transition.
