PCB Controlled Impedance: What the Fab Actually Controls and What You Assume
Ordering controlled impedance from a PCB fab sets a target. The actual impedance depends on dielectric constant variation, copper weight, and trace width tolerance.
Analog and digital circuit analysis, PCB layout, and electronic component selection
USB 2.0 differential pairs tolerate modest impedance variation. USB 3.0 SuperSpeed does not. The length matching, impedance control, and via stub rules that differ between them.
Ordering controlled impedance from a PCB fab sets a target. The actual impedance depends on dielectric constant variation, copper weight, and trace width tolerance.
Adding a fifth layer for a dedicated analog ground plane solved a noise floor problem that three different layout iterations couldn't fix. The trade-off analysis that led there.
Each ESD protection technology has a different clamping voltage, capacitance, and energy rating. The wrong choice for an industrial interface kills the IC it was meant to protect.
Off-the-shelf transformers rarely match your flyback or forward converter spec exactly. How to read the datasheet parasitics and what to measure on the bench to verify fit.
Each current sensing topology has a sweet spot defined by frequency, isolation requirements, and acceptable power loss. A bench comparison at 20 A steady-state and 50 A pulse.
Type II compensators work for most voltage-mode buck converters. Type III becomes necessary when the LC double pole and ESR zero interact in specific ways. A worked example.
Conformal coating protects against humidity and contamination. Designing for it requires planning clearances around connectors, test points, and adjustable components before the first gerbers.
SPICE says your decoupling works. The spectrum analyzer says otherwise. A board that taught me why placement matters more than value above 100 MHz.
The hot loop in a synchronous buck converter is small but its current is fast. A ground plane split under that loop radiates efficiently. Before-and-after measurements from a real board.
Voltage noise density is the headline number but 1/f noise corner, input bias current noise, and supply rejection matter more in many instrumentation front-ends.
LDOs are simple and quiet. Switching regulators are efficient. The crossover point in power budget depends on dropout voltage, quiescent current, and load profile — not just efficiency.