This presentation, "The Magic of Feedback and Analog Circuit Design" by Jans Anders, comprehensively explores the fundamental principles and diverse applications of both negative and positive feedback in analog electronic circuit design. The talk begins with a historical overview, tracing negative feedback from Harold Black's 1928 patent for distortion reduction in early telephone amplifiers to Edwin Armstrong's 1913 pioneering use of positive feedback for gain enhancement and selectivity in radio. Anders details how negative feedback desensitizes gain, extends bandwidth (creating inherent trade-offs), and effectively reduces noise in systems like Phase-Locked Loops (PLLs) and Low Dropout Regulators (LDOs), and significantly lowers distortion in various amplifier configurations, including those employing local feedback like source degeneration. The critical aspect of stability in negative feedback systems is thoroughly covered using Nyquist criteria, gain and phase margins, and a simplified slope-based stability rule. Finally, positive feedback is discussed for its utility in achieving high gain, designing oscillators, comparators, and bistable circuits, exemplified by a differential pair with cross-coupled loads, emphasizing the delicate balance required to harness its benefits without succumbing to instability.
The Magic of Feedback and Analog Circuit Design
0:00 Introduction & Outline: Jans Anders introduces the presentation, outlining topics such as classical feedback with op-amps, local feedback loops, feedback's role in reducing noise and distortion, stability considerations, and the applications of positive feedback.
0:12 Historical Context of Negative Feedback: The concept of negative feedback, patented by Harold Black in 1928, was developed to address critical challenges in long-distance telephone communication, including reducing distortion, desensitizing gain to manufacturing tolerances, extending bandwidth, and improving noise performance in early tube amplifiers.
0:59 Historical Context of Positive Feedback: Positive feedback predates negative feedback, with Edwin Armstrong's 1913 invention of the regenerative amplifier (as a student) providing much-needed gain and sensitivity for wireless telegraphy at a time when amplifier circuits offered zero gain. It also improved selectivity and enabled continuous wave detection, crucial for local oscillators.
The core idea involves a high loop gain (beta multiplied by the amplifier's open-loop gain 'A'), which forces the error signal at the input of the amplifier to be near zero.
This leads to a closed-loop transfer function (output/input) that is approximately 1/beta, making the overall system gain largely independent of the potentially variable open-loop gain 'A'.
Key takeaways: This "virtual ground" property and gain desensitization are fundamental to designing stable and precise analog circuits.
0:08:11 Negative Feedback Trade-offs: Gain vs. Bandwidth: Negative feedback inherently introduces a trade-off: as the amount of feedback (1 + beta*A) increases, the DC gain of the amplifier decreases, but its 3dB bandwidth extends by the same factor, keeping the gain-bandwidth product constant.
0:10:18 Gain Desensitization Explained: Using an inverting op-amp as an example, the presentation shows how feedback makes the closed-loop gain depend solely on the precise ratios of passive components (e.g., resistors R1 and R2) rather than the highly variable open-loop gain of the op-amp. This results in a very precise and accurate gain, as passive component ratios can be matched with high accuracy.
0:14:34 Local Feedback Loops for Performance Improvement:
Cascode Current Mirror (0:16:27): By introducing an additional transistor, local negative feedback significantly increases the output resistance of the current mirror by a factor related to the transistor's self-gain (GM*RDS).
Whitler Current Source (0:19:00): Similar to the cascode, this circuit uses a source resistor (RS) to create local negative feedback, dramatically improving the output resistance by a factor of approximately RSGM2RDS2.
Key takeaway: Local feedback, like global feedback, always involves a trade-off, where improving one parameter (e.g., output resistance) often comes at the cost of another.
Phase-Locked Loops (PLLs) (0:21:35): Feedback is leveraged to suppress the intrinsic noise of a Voltage-Controlled Oscillator (VCO). The loop effectively transfers the stable, low-noise characteristics of a reference signal (scaled by a frequency divider 'N') to the output, cleaning up the noisy oscillator.
Low Dropout Regulators (LDOs) (0:23:55): In LDOs, feedback is crucial for stabilizing fluctuating input voltages (e.g., from a degrading battery). The loop greatly suppresses battery variations while ensuring the output voltage is a precise, scaled version of a stable, low-noise reference voltage (e.g., a band gap reference).
Feedback significantly reduces amplifier distortion by decreasing the effective input signal applied to the internal nonlinearities of the amplifier. This reduction is often super-linear, meaning a small reduction in input can lead to a large reduction in distortion.
Non-inverting amplifier (0:29:01): Simulations demonstrate that the third-order distortion (HD3) is reduced proportionally to the loop gain.
Common Source Amplifier with Source Degeneration (0:32:47): Local negative feedback through a source resistor degrades the linear gain (GM) but dramatically improves the amplifier's linearity and reduces second-order distortion (HD2) by linearizing the transistor's behavior.
A critical concern is avoiding instability, which occurs when the loop gain (beta*A) becomes -1, leading to sustained oscillations.
Stability is assessed using the Nyquist criterion, focusing on gain margin (how much gain is below 0dB when phase is 180°) and phase margin (how much phase is above -180° when gain is 0dB).
A simplified stability criterion involves plotting the open-loop gain and 1/beta on a Bode plot: the system is stable if the slope difference at their intersection is not less than -20dB/decade.
Frequency-dependent feedback networks (e.g., due to parasitic capacitances) can introduce additional phase shift, potentially leading to instability (e.g., a -40dB/decade slope difference).
Historically used to enhance gain and sensitivity, positive feedback is still employed in modern CMOS designs for these purposes, as well as for creating oscillators, comparators, and bistable circuits (like the "latch" shown in a previous talk).
Differential Pair with Cross-coupled Loads (0:47:52): This circuit uses positive feedback to significantly increase the amplifier's output resistance and thus its gain. However, exceeding a certain gain threshold (where GM2*RL approaches 1) causes instability, leading to latching or oscillations.
0:52:00 Conclusion: Feedback is an indispensable tool for engineers. Negative feedback offers significant performance improvements (bandwidth, gain stability, noise, distortion) through controlled trade-offs, requiring careful attention to both small-signal and large-signal stability assessments. Positive feedback can enhance gain and sensitivity but demands precise design to prevent instability. The presentation concludes by noting that all real-world circuits are inherently nonlinear, inviting further exploration into nonlinear control.