The Design Brief #002 — Surviving CS114 Bulk Cable Injection: Common-Mode Physics from 10 kHz to 200 MHz

The failure mode nobody budgets for

A defense-grade DC/DC converter passes CE102 conducted emissions with 8 dB of margin, sails through RE102, then fails CS114 bulk cable injection at exactly 27.12 MHz with the unit latching off. This is not a filter problem in the differential sense — it is a common-mode coupling problem, and the two obey different physics.

CS114 (now governed by MIL-STD-461H, Table I, released April 17, 2026, superseding Rev G after eleven years) drives a calibrated bulk current onto the entire harness through an injection probe. The requirement is to inject power that produces a forward current equal to the limit curve — for Navy and Army interior equipment, that is 107 dBµA (≈ 0.22 A) from 10 kHz to 200 MHz. The equipment under test sees this as a common-mode (CM) excitation referenced to the ground plane, not as a pin-to-pin differential signal. Your π-filter, sized for differential-mode insertion loss, is nearly transparent to it.

The physics: cable as a transmission line, not a wire

The reason failures cluster at 27 MHz and 108 MHz is geometric resonance. A cable harness over a ground plane is a transmission line. When its electrical length approaches an odd multiple of λ/4, the CM input impedance seen at the connector swings toward an extreme — a short or an open — and the injected current concentrates there.

For a 2 m harness:

• λ/4 at 27 MHz ≈ 2.77 m → first quarter-wave resonance
• 3λ/4 at 108 MHz ≈ 2.08 m → third resonance, near-exact match

At resonance the CM current that actually penetrates the enclosure is governed by the cable's transfer impedance Z_T (Ω/m). Coupled voltage into the victim circuit is:

V_coupled = Z_T · I_CM · ℓ

A braided shield with 80% optical coverage has Z_T rising from ~1 mΩ/m at 1 MHz to 50–200 mΩ/m above 30 MHz as the braid's porpoising inductance dominates. At 108 MHz with 0.22 A injected, a poorly terminated shield delivers tens of millivolts of CM noise straight onto a logic reference — enough to trip a reset comparator.

The solution: 40 dB CM rejection, engineered across the band

Target ≥ 40 dB common-mode rejection above 10 MHz for any harness run exceeding 3 m. Three independent mechanisms, stacked:

1. Common-mode chokes at the connector. A bifilar CM choke presents high impedance to CM current while passing differential signal. Specify a core with ≥ 300 Ω at the problem frequency — for 27–108 MHz, Fair-Rite 31-material toroids deliver 250–350 Ω per pass. Two passes through the core quadruples impedance (Z ∝ N²).
2. Ferrite cable sleeves at λ/8 intervals. Distributing impedance along the harness damps the standing wave before it can build. At 100 MHz, λ/8 ≈ 37.5 cm.
3. 360° shield termination. Removes the pigtail inductance entirely and keeps Z_T low where it matters.

For CS116 damped oscillatory transients, your power supply's PSRR must exceed 60 dB through the 10 kHz–100 MHz ringing band, or the transient walks straight onto the rail.

Field case: the 27 MHz latch-up

A vehicle power-distribution module failed CS114 at 27 MHz. Root cause: a 2.1 m harness in third-quarter-wave resonance, shield pigtailed 30 mm at both ends. The fix was not more filtering — it was a POCONS custom board-level shield with an integrated 360° gasketed entry that terminated the harness braid directly to the can wall, plus a 31-material choke at the connector. CM rejection went from 14 dB to 43 dB at 27 MHz. Pass, with 9 dB margin.

When the coupling path is through the enclosure boundary, the cure belongs at the boundary. A shield with a properly engineered cable-entry feature solves a CS114 failure that no amount of in-line filtering will.

One thing

40 dB of common-mode rejection at 27 MHz is the line between a one-pass qualification and a six-figure redesign cycle. Measure CM impedance — not differential insertion loss — before you book the chamber.