This video challenges the conventional wisdom of PCB decoupling by demonstrating that the common strategy of using multiple, different-valued capacitors in parallel (e.g., 100pF, 1nF, 10nF, 100nF) can be detrimental to power supply integrity.
Using RLC capacitor models, simulations, and real-world VNA measurements, the presenter shows that this standard approach creates unintended parallel resonant circuits. These resonances occur because capacitors behave as inductors above their self-resonant frequency (SRF), leading to high-impedance peaks that cause supply noise and ringing. The video also debunks the myth that smaller value capacitors are inherently better for high-frequency decoupling.
Two superior strategies are proposed and verified: first, using multiple capacitors of the same large value (e.g., four 100nF capacitors) to eliminate parallel resonances and increase local energy storage. Second, for critical applications, adding small series resistors to each capacitor to aggressively damp the network, creating an exceptionally smooth impedance profile across a wide frequency range.
Rethinking PCB Decoupling: From Flawed Tradition to Optimized Design
0:00 The Common Decoupling Myth: The video begins by challenging the widespread engineering practice of using a range of capacitor values (e.g., 100pF to 100nF) for decoupling, suggesting it can create more noise than it solves.
0:27 Understanding Real Capacitors: A real-world capacitor is best represented as a series RLC circuit. Crucially, above its self-resonant frequency (SRF), a capacitor's impedance becomes inductive and increases with frequency.
3:10 The Flawed Strategy Analyzed: The traditional method of paralleling different capacitor values is simulated and measured. The results show multiple high-impedance peaks in the frequency response.
5:00 The Root of the Problem: Parallel Resonance: These unwanted peaks are caused by parallel LC resonances. They form between a larger capacitor operating in its inductive region and a smaller capacitor still in its capacitive region, leading to supply ringing and noise.
5:35 Placement Theory Debunked: Contrary to popular belief, placing the smallest value capacitor closest to an IC pin for RF performance is often incorrect. For a given package size, larger value capacitors (e.g., 100nF) frequently have lower series resistance (ESR) and inductance (ESL), making them better for high-frequency decoupling.
7:45 A Better Solution: Use Identical Values: A superior strategy is to use multiple capacitors of the same, larger value (e.g., four 100nF capacitors). This eliminates parallel resonances because all capacitors switch from capacitive to inductive behavior at the same frequency. This also improves local energy storage and reduces the overall Q factor.
8:34 Verification of the Improved Method: Measurements of the "identical value" strategy confirm the absence of high-impedance peaks, resulting in a much smoother and lower impedance profile compared to the traditional method.
11:02 The Ultimate Strategy for Critical Designs: For the most noise-sensitive applications, the video proposes adding a small series resistor (e.g., 1 ohm) to each capacitor. This technique intentionally "damps" the network, "smashing the Q factor" to prevent any possible resonance, even those caused by PCB trace inductance.
12:22 Final Comparison & Key Takeaway: The final comparison shows a dramatically improved, flat impedance curve for the damped strategy. The key takeaway is that using multiple identical, larger-value capacitors is a more robust and effective decoupling method that also simplifies the bill of materials.