The Design Brief #008 — CE102 Frequency Planning: Engineering Switcher Harmonics Out of the 150 kHz–30 MHz Failure Zone

Where CE102 failures live

Across military power-supply qualification programs, conducted-emissions failures are not uniformly distributed. Roughly 73% cluster between 150 kHz and 30 MHz — and the cause is structural, not coincidental. That band is where switching-regulator fundamentals and their low-order harmonics naturally land, and it is where the CE102 limit curve is most demanding.

The fix is not "more filter." It is frequency planning: deliberately placing the switching fundamental and its first three harmonics outside the worst of the band before you commit the schematic.

The physics: harmonic content and the limit curve

A switching converter produces a comb of harmonics at n·f_switch. The amplitude envelope follows the trapezoidal-waveform spectrum: flat to the first knee f₁ = 1/(π·t_d), then −20 dB/decade, then −40 dB/decade past the rise-time knee f₂ = 1/(π·t_r). The first few harmonics carry the most energy and are the hardest to filter, because filter attenuation (40 dB/decade for a 2-pole LC) is least effective close to the fundamental.

MIL-STD-461H CE102 (released April 17, 2026) tests power-lead conducted emissions 10 kHz–10 MHz, with the limit curve stepping downward through the 150 kHz–30 MHz region. If f_switch = 500 kHz, the fundamental and harmonics at 1.0, 1.5, 2.0 MHz all sit deep in the punitive zone, and your filter must deliver 50–70 dB of attenuation starting just above the fundamental — a hard, bulky, expensive filter.

The rule: harmonics below 100 kHz or fundamental above 16.7 MHz

Keep the primary switching frequency and its first three harmonics out of the failure zone:

f_switch × 3 < 100 kHz → f_switch < 33.3 kHz (audio-frequency switching; the low-frequency escape)

OR

f_switch > 16.7 MHz → third harmonic > 50 MHz (high-frequency escape, suits GaN)

Either placement moves the three highest-energy harmonics to where the limit is relaxed and where a modest filter is highly effective. The measured benefit is roughly 15 dB of filter design budget — meaning a smaller, lighter, cheaper filter achieves the same compliance margin.

The 8 dB receiver-correlation gap

Pre-compliance results often disagree with the accredited-lab result by a meaningful margin. A documented contributor: up to 8 dB of difference at 1 GHz between a CISPR 16-1-1-compliant EMI receiver and a field-deployed or lower-grade receiver, arising from detector bandwidth, detector type (peak vs. quasi-peak vs. average), and measurement-receiver accuracy class. Build this into your margin budget — if your pre-compliance scan shows only 5 dB of margin, you do not have 5 dB; you may have −3 dB at the accredited lab.

Worked frequency plan

Design a 28 V → 5 V converter, 30 W, CE102-compliant, minimum filter mass:

1. Reject the mid-band default (500 kHz–2 MHz).
2. Choose the high-frequency escape: f_switch = 18 MHz with a GaN half-bridge. Harmonics at 36, 54, 72 MHz — third harmonic at 54 MHz, well above the punitive zone.
3. Filter target: now only ~25 dB needed at 18 MHz instead of 60 dB at 500 kHz. A single compact LC stage (f₀ ≈ 1.8 MHz, damped) suffices.
4. Verify the receiver gap: design for 12 dB pre-compliance margin so the accredited-lab result holds ≥ 4 dB even with an 8 dB correlation swing.
5. Cross-check CS101: confirm the input filter's f₀ (1.8 MHz) sits above the CS101 sensitive zone (f₀ > 200 kHz ✓).

POCONS connection

An 18 MHz GaN switcher trades a low-frequency filtering problem for a high-frequency radiated one — the fast edges push spectral content toward 1 GHz. Pair the frequency plan with a POCONS board-level shield over the GaN stage so the radiated cost of the high-frequency escape is contained at the source.

One thing

73% of CE102 failures are in one band because that is where unplanned harmonics fall. Plan the fundamental — third harmonic below 100 kHz, or fundamental above 16.7 MHz — and you buy 15 dB of filter budget before drawing a single component.