[Post #26/38] Negative Feedback — Properties and Benefits

CONTENT_START [HERO_HERE: A high-level schematic showing an amplifier with a feedback path and the comparison node.]

📘 Microelectronic Circuits Series — Post #26/38 — 12.1-12.5 (Key)

Negative feedback is the cornerstone of precision analog design. Without it, the inherent variability of integrated circuit manufacturing would make reliable amplifiers impossible, as we would be at the mercy of wildly fluctuating transistor betas and threshold voltages.

This chapter is the "holy grail" for system-level stability and accuracy. It is the primary tool used by IC designers to trade raw, unpredictable gain for highly stable, controlled performance. Mastering these concepts is non-negotiable for anyone designing op-amps, LDOs, or data converters.

1. Overview & Background — Why this matters

Imagine steering a car on a highway while blindfolded, relying only on a friend in the passenger seat calling out your deviation from the center line. If you start drifting right, the friend tells you to steer left. This correction process—comparing your actual output (position) to your desired input (lane center)—is the essence of negative feedback.

Figure 1. Basic negative-feedback block diagram (A, β, summing point)

In electronics, we suffer from the "drift" of manufacturing tolerances. An open-loop amplifier made in a 180-nm process might have a gain of 1000 V/V one day and 500 V/V the next due to process-voltage-temperature (PVT) variations. By taking a sample of the output and subtracting it from the input, we force the amplifier to "correct" its output until the error is minimized.

This is precisely how the TI OPA350 or the error amplifier in an Apple M-series SoC's power management unit achieves performance that is independent of the silicon’s specific batch variability. We effectively "waste" gain to buy predictability.

[DIAGRAM_1_HERE: A block diagram showing the forward amplifier block A and a feedback block β, with a summing junction at the input.]

2. How it Works (Physical & Circuit Principles)

At the circuit level, negative feedback works by creating a "loop." The input signal V_s is fed into a summing node where the feedback signal V_f (derived from the output V_o) is subtracted from it. The difference, called the error signal, is amplified by the gain A. Because the signal is subtracted, the output is forced to track the input scaled by the feedback factor β.

If the loop gain (T = Aβ) is very large, the output V_o becomes almost entirely dependent on the feedback network components (often precision resistors) rather than the amplifier itself. This is why a simple LM741 can provide a rock-solid gain of 10 V/V even if its internal transistors change characteristics significantly due to heat.

where A is the open-loop gain, β is the feedback factor, and A_f is the closed-loop gain.

💡 Intuition: If A approaches infinity, the denominator 1 + Aβ is dominated by Aβ, and the A terms cancel out, leaving A_f ≈ 1/β. The system "ignores" the amplifier and listens only to the passive network in the feedback path.

3. Key Design Equations

This shows that sensitivity to gain variations (dA/A) is reduced by the factor 1 + Aβ, which is known as the desensitivity factor.

Figure 2. Gain desensitization (open vs closed-loop gain variation)

where ω_H is the open-loop 3-dB bandwidth and ω_H,f is the extended closed-loop bandwidth, demonstrating the classic gain-bandwidth tradeoff.

where D is the nonlinear distortion of the open-loop amplifier and D_f is the reduced distortion, showing how feedback "straightens out" the transfer curve.

4. Worked Numerical Example — Calculate it yourself

Consider an op-amp with an open-loop gain A = 10,000 V/V and a 3-dB bandwidth f_H = 100 Hz. We apply negative feedback with a factor β = 0.1.

1. Closed-loop gain: A_f = 10,000 / (1 + 10,000 · 0.1) = 10,000 / 1001 ≈ 9.99 V/V. Note how close this is to 1/β = 10.

2. New Bandwidth: f_H,f = 100 Hz · (1 + 1000) = 100.1 kHz. We traded a massive amount of gain for a 1000-fold increase in bandwidth.

3. Sensitivity: If A drops by 10% (from 10,000 to 9,000), the change in A_f is only 10% / 1001 ≈ 0.01%. The system is now extremely robust.

[DIAGRAM_2_HERE: Plot showing the open-loop and closed-loop frequency response on a Bode plot, illustrating the corner frequency extension.]

5. Design Considerations & Trade-offs

• Stability: Adding too much feedback (increasing Aβ) can lead to phase shifts that turn negative feedback into positive feedback at high frequencies, causing oscillation.
• I/O Impedance: Series-input connections increase input impedance, while shunt-output connections decrease output impedance; always match the feedback topology to the load requirements.
• Noise: Feedback does not improve the Signal-to-Noise Ratio (SNR) of the amplifier stage itself. While it reduces gain, it attenuates the internal noise and the signal by the same factor, leaving the noise floor relative to the signal unchanged.
• Power: Achieving high A requires higher supply current (I_DD). Designers must balance the need for high Aβ (for linearity) against the power budget.

6. Where it Shows Up in Practice

In the TI INA240 current-sense amplifier, precision feedback is used to define the gain precisely to 20 V/V, 50 V/V, etc., ensuring that current readings don't drift as the chip heats up during high-current monitoring. Similarly, in an LDO (Low-Dropout Regulator), the feedback loop samples the output voltage and compares it against a bandgap reference (typically 1.25 V) to ensure the output stays regulated at, say, 3.3 V, regardless of load current changes.

7. Common Pitfalls & Debugging Tips

• ⚠️ Phase Margin Issues: If you see ringing on a step response, you have pushed the loop gain too hard near the unity-gain frequency; increase the phase margin by adding a compensation capacitor.
• ⚠️ Miscalculating β: Remember that β is the fraction of the output *returned* to the input. If you have a voltage divider R₁ and R₂ in the feedback path, β is usually R₁ / (R₁ + R₂)—don't just assume it's the component values themselves.

8. Exam & Interview Hot Spots

• 💡 "How does feedback affect input impedance?" (Answer: It depends on the connection—series feedback increases it, shunt feedback decreases it.)
• 💡 "Can feedback reduce noise?" (Answer: No, it reduces gain; it does not change the Signal-to-Noise Ratio of the amplifier itself.)
• 💡 "Define Loop Gain (T)." (Answer: T = Aβ; it is the total gain around the loop when the loop is broken.)

9. Key Takeaways

• Negative feedback trades high, unstable gain for low, stable, predictable gain.
• The closed-loop gain is defined by A_f = A / (1 + Aβ).
• Feedback extends bandwidth by the desensitivity factor (1 + Aβ).
• Nonlinear distortion is reduced by (1 + Aβ), linearizing the amplifier.
• The gain-bandwidth product (GBW) remains constant for a fixed amplifier design.

Educational content only. Always verify with datasheets and SPICE simulation before production design.