THE DESIGN BRIEF

Issue #006 · Thursday, April 9, 2026 · Engineering Intel

Stop using the λ/20 rule blindly. At 1 GHz, that's 15mm — but your via inductance already kills you at 3mm. I've been measuring actual impedance vs. distance on 0402 ceramics, and the sweet spot is tighter than you think.

Real measurement: 2.2 μF X7R at 1mm from pin = 0.8Ω impedance. Same cap at 5mm = 4.2Ω. Your target should be <1.5Ω for decent supply noise rejection above 500 MHz. Use this: Z_total ≈ ESL×2πf + trace_inductance×distance×2πf. Assume 0.5 nH/mm for microstrip trace inductance.

The kicker: switching noise couples back through that same path. Every millimeter of separation between your decoupling cap and IC costs you 3-4 dB of conducted emission performance above 800 MHz.

• • MIL-STD-461G CS114 failures: 80% trace back to ground plane discontinuities >0.5mm
• • IEC 60601-1-2:2021 Amendment 1 tightened radiated limits by 6 dB for 80-1000 MHz band
• • CISPR 25 Class 5 automotive: saw three failures this month from 48V rail noise coupling through CAN transceivers
• • FCC Part 15 Subpart B: measuring 12 dB margin minimum for Class A equipment to survive production variations

• • CISPR 32 revision targeting data center switch-mode supplies — draft limits are 10 dB tighter
• • SAE J1113 automotive immunity updates for 800V EV architectures expected Q3 2026

Have you tested the 1mm decoupling placement rule against conducted emissions? Reply with your CS101/CE101 delta measurements — I'm collecting field data.

Sources: 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 · 9 · 10