Emissions and susceptibility are not the same test

A power supply designer optimizes a CE102 filter for conducted emissions — minimizing noise the supply pushes onto the line. The same filter is then assumed to protect against CS101 conducted susceptibility. It does not, and the reason is a resonance hiding inside the filter itself.

CS101 (MIL-STD-461H, retained from Rev G, standard released April 17, 2026) injects an audio-frequency voltage in series with the input power leads, 30 Hz to 150 kHz. The limit curve for a 28 VDC system requires the EUT to tolerate roughly 5 V RMS at the low end, ramping down with frequency. This is not noise riding on the line — it is a forcing function deliberately driven into your input filter.

⚠️The miss

CE102 is tested 10 kHz–10 MHz. CS101 is tested 30 Hz–150 kHz. The overlap is only 10 kHz–150 kHz, and below 10 kHz your CE102 filter has zero specified performance. Worse — a filter that is excellent at attenuating emissions can be a near-perfect amplifier of an injected CS101 tone if its resonance lands in band.

The physics: your filter is an underdamped LC tank

Every input filter is, at its core, an LC network with a resonant frequency:

f₀ = 1 / (2π·√(L·C))

Below f₀ the filter passes signal. At f₀, an underdamped filter exhibits a transfer-function peak with gain:

Q = (1/R)·√(L/C) → peak gain ≈ Q

A practical EMI filter with L = 100 µH and C = 10 µF resonates at:

f₀ = 1/(2π·√(100×10⁻⁶ · 10×10⁻⁶)) = 5.03 kHz

— comfortably below the CS101 band, which is correct. But designers chasing higher attenuation often shrink L and C, pushing f₀ upward. With L = 22 µH and C = 4.7 µF, f₀ = 15.6 kHz; with L = 10 µH and C = 2.2 µF, f₀ = 33.9 kHz — all squarely inside 30 Hz–150 kHz. At that resonance, a Q of 5–10 means a 14–20 dB gain applied to the injected CS101 tone. The 5 V test stimulus becomes 25–50 V at the regulator input.

The rule: place f₀ on purpose

Set the filter resonant frequency either below 50 kHz or above 200 kHz — never in the 50–200 kHz CS101-sensitive zone.

Below 50 kHz is preferred for emissions filters: large L and C, resonance well below the band, and any residual peaking lands where the CS101 limit is most forgiving.
Above 200 kHz suits compact designs but sacrifices low-frequency emissions attenuation — verify CE102 still passes.

Then damp it. An undamped placement is still risky if a temperature shift or component tolerance moves f₀. Add a series RC damping leg (R_d ≈ √(L/C), C_d ≈ 3–5× the filter C) to hold Q ≤ 2, capping resonant gain at ~6 dB.

Worked example

Design a 28 V input filter, CE102-compliant, CS101-safe:

Choose attenuation target: 40 dB at 150 kHz. A single LC stage gives 40 dB/decade above f₀; 150 kHz is 1.6 decades above f₀ if f₀ ≈ 3.8 kHz.
Pick f₀ = 4 kHz (well below 50 kHz). Choose C = 10 µF → L = 1/((2π·4000)²·10×10⁻⁶) = 158 µH.
Damp: R_d = √(158 µH/10 µF) = 3.97 Ω; C_d = 33 µF in series with R_d, across the filter capacitor.
Verify: Q drops from ~8 to ~1.8; resonant peak ≤ 5 dB; CS101 stimulus passes through with no amplification.

POCONS connection

When the CS101 forcing function couples around the filter — through the input cable's common-mode path rather than the differential pins — the cure is at the enclosure boundary. A POCONS board-level shield with a gasketed power-entry feature terminates the input harness shield 360° to the can wall and adds a defined CM-choke seat, removing the parasitic coupling path that even a perfectly placed filter cannot reach.

One thing

CE102 tests what leaves your box. CS101 tests what you can force into it. They are different physics, and a filter tuned for one can be a resonant amplifier for the other. Place f₀ deliberately — below 50 kHz or above 200 kHz — and damp it.