Why One Cooling Slot Can Undo a Mission Computer's Shielding

Rand Simulation — Applications Engineering AI
Aerospace & defense · Ansys HFSS · 10 min read

AI disclosure: RandSim Labs is an experimental AI-driven engineering simulation platform. Content on this site, including simulations, analyses, figures, and written materials, may be generated or assisted by AI using licensed Ansys tools. AI-generated content may contain errors and is provided for educational, informational, and demonstration purposes only. Users should independently verify all results before relying on them for engineering, design, manufacturing, safety, or other production decisions.

Flying around the mission computer, cut open at its cooling slot, while the solved field pulses through its 2.3 GHz cycle with all five module sources driven. Inside: the card cage bathing in its own standing-wave interference. Above the lid: the leak breathing through the 80 × 3 mm slot — at this frequency the enclosure radiates more than the same electronics with no box at all. Solved in Ansys HFSS.

The result: we put the same noisy electronics inside five versions of the same aluminum avionics enclosure and solved each one's shielding effectiveness in Ansys HFSS — the decibel ratio between what escapes with the box and without it. Welded shut, the box leaks nothing the solver can measure. Cut one 80 mm cooling slot in the lid and the shielding collapses — about 32 dB left at the GPS frequency, and at the enclosure's own cavity resonances the box becomes an amplifier, radiating up to 14 dB more than bare electronics. Replace that slot with eight round holes of the same total open area and the shielding never drops below 33 dB anywhere, with 79 dB at GPS. Add a 12 mm deep ribbed chimney over the slot and you are back above 100 dB. Vent shape decides the leak — not vent area. Then we filled the box with its card cage and gave every module its own transmitter — processor, power supply, Ethernet, RF comms, radar interface — with the GPS receiver as the victim: at the GPS frequency, the loudest neighbor couples 60 dB (nearly a thousand times) harder into the receiver slot than the quietest, and every source shouts through the same cavity resonance.

The Box Is the EMC Design

A multi-core mission computer is a broadband radio transmitter that happens to also run software. Processor clocks and their harmonics, DDR memory strobes, gigabit Ethernet, and switching power supplies fill the spectrum from megahertz to several gigahertz. One detail from the design-review folklore is worth repeating: the 63rd harmonic of a standard 25.000 MHz Ethernet crystal lands at 1,575.000 MHz — 420 kHz from the center of GPS L1, inside the main lobe of the signal a receiver is trying to hear at around −158 dBW, some 20 dB below thermal noise. That fragility is why DO-160's radiated-emission limits carve protected notches around the GNSS bands, about 10 dB deeper than the neighboring limit for standard cabin equipment and 15 dB deeper for equipment in view of the antennas.

Between all that digital noise and the aircraft's radios stands one part: the aluminum box. Solid aluminum is effectively opaque here — the skin depth at 1 GHz is 2.6 µm against a 3 mm wall. Real enclosures leak where the metal is interrupted: vents, seams, and connector penetrations. So the interesting engineering question is not "does shielding work" but "what does each opening cost" — and that is a question you can put numbers on.

Five Versions of the Same Computer, One Number Each: Shielding Effectiveness

We modeled a generic half-ATR-style conduction-cooled chassis — 124 × 194 × 320 mm, 3 mm aluminum walls, the classic finned flanks and 38999-style front panel — and placed a small broadband dipole inside as a stand-in for the noisy electronics. Five HFSS solves, 0.8–3 GHz: the bare source with no enclosure (the reference), the box welded shut, the box with one 80 × 3 mm cooling slot in the lid, the box with eight 6 mm round holes whose open area matches the slot's, and the slot again with a ribbed below-cutoff chimney over it — the home-made cousin of a honeycomb EMI vent. Shielding effectiveness is the honest ratio used in the enclosure-shielding literature: the field radiated with the enclosure versus the same source with no enclosure at all, frequency by frequency.

Shielding effectiveness versus frequency for the slot, hole-array, and chimney variants, with the DO-160 protected GNSS band shaded — What the enclosure is worth, by vent design. The 80 mm slot (red) decays from 57 dB toward its half-wave resonance and goes *negative* at cavity resonances. Eight round holes with the same open area (blue) never drop below 33 dB. The ribbed chimney (purple) holds 60–127 dB. The welded box leaks exactly nothing in the solve — its real-world limit is seams and gaskets, which is why they get so much attention in avionics packaging.

One Slot Is an Antenna

A narrow slot in a conducting wall is a slot antenna — the Babinet complement of a dipole. Below resonance it leaks in proportion to its longest dimension; at the half-wave frequency, f = c/2L — 1.875 GHz for our 80 mm slot — it radiates about as well as a purpose-built antenna of the same size. The solve shows exactly that collapse: 57 dB of shielding at 0.8 GHz eroding to roughly 20 dB through the resonance region. It also shows what the hand formula cannot: the slot does not act alone. It couples to the resonant modes of the cavity behind it, and the worst point lands at 2.30 GHz — a slot–cavity hybrid — where shielding reaches minus 13.7 dB. At that frequency the "shielded" computer out-radiates the same electronics sitting in free air, the box working as a resonator that collects the interior field and feeds it to the slot. That counterintuitive sign flip is a documented behavior of enclosures near cavity resonance in the shielding literature, and it is precisely the kind of thing that only shows up in a full-wave solve.

The mission computer with the solved electric field erupting through the lid cooling slot — The leak from outside: solved |E| at 2.3 GHz erupting through the 3 mm slot and washing across the lid — the slot radiating like the antenna it has become.

Cutaway of the enclosure at the slot centerline showing the standing-wave interference pattern inside the cavity and the leak jet through the slot — The box cut open at the slot centerline, with the solved |E| field on the cut plane. Inside: the cavity's standing-wave interference pattern at 2.3 GHz — the enclosure resonating like a microwave oven. At the lid line: the field funneling through the 3 mm slot, locally the strongest field anywhere in the model — the slot acting as the antenna feed. Outside: the escaping wavefronts, with dark interference nulls arcing through the radiated field.

Peak radiated field versus frequency for bare electronics and the three vented enclosures — The same data as raw emissions. Where the red slot curve crosses *above* the gray bare-electronics line, the enclosure is amplifying its contents. The hole array and chimney keep the radiated field 50–150 dB down across the band, including through the shaded DO-160 protected GNSS window around GPS L1.

Same Open Area, Very Different Leak

The eight-hole variant is the free lunch of EMC design. Its open area matches the slot's within a few percent — the airflow argument is a wash — but each 6 mm hole is electrically tiny, and small-aperture leakage falls with the sixth power of aperture size. The penalty for using several holes grows only as 10 log₁₀N. Net: at GPS L1 the hole field shields 79 dB against the slot's 32 dB, and at the slot's resonance the gap is 68 dB. Even at the nasty 2.25–2.3 GHz cavity mode, the holes hold 33 dB where the slot went negative. And when the airflow budget genuinely demands the full slot cross-section, depth does what diameter cannot: the 12 mm ribbed chimney turns each cell into a waveguide operating far below cutoff, whose attenuation the classic 27.3 · depth/width rule puts at +36 dB — the solve delivers that and more, holding above 100 dB through the GNSS band. Commercial honeycomb vent panels are this exact trick, mass-produced.

Then We Filled the Box: Five Sources and a Victim

A shielding number answers what escapes. The other half of the original question is what happens between the modules locked inside together — so we populated the enclosure: a backplane and four 3U cards on standard 25.4 mm pitch, and five independent broadband transmitters placed where the real offenders live. The processor sits mid-card on the SBC; the Ethernet PHY at the SBC's front edge; the RF comms module on card three; the power electronics and the radar interface mounted with their radiating structures turned 90° — different positions and different polarizations. The GPS receiver becomes what it really is in this drama: the victim — a field probe at its module position just behind the front-panel antenna feedthrough. One 232,000-tetrahedron HFSS solve carries all five ports at once, and because the physics is linear, each source can be switched on alone afterward without re-solving.

Field at the GPS receiver module versus frequency, one curve per source — Source-by-source coupling to the GPS receiver module, per watt driven. The hierarchy is brutal: the RF comms module reaches the receiver at 8.9 V/m at GPS L1 — roughly 60 dB (a factor of ~950) above the power supply — and *every* source spikes together on the shared cavity modes. The resonance is the highway; proximity and polarization decide who gets to drive on it.

Peak exterior radiated field versus frequency, one curve per source — The same five sources ranked by what escapes the slot. The vertically-polarized sources (processor, Ethernet, RF module) drive the slot efficiently; the power supply and radar interface, oriented across the slot's polarization, radiate 30–40 dB less — the aperture only amplifies the neighbors that speak its polarization.

What This Covers — and What a Full Campaign Adds

The community request named the whole EMI/EMC problem: interference between processors, power electronics, Ethernet, RF modules, GPS receivers, and radar interfaces inside a compact enclosure, plus shielding effectiveness, PCB stack-up, cable routing, and enclosure design. This study now covers the box-level levers end to end: enclosure design and shielding effectiveness priced across four vent designs, and module-to-module interference solved directly — five sources at their card positions, the GPS receiver as the measured victim, a full coupling hierarchy out of one solve. The two remaining levers are their own solves in the same toolchain, and we have run each on this blog: PCB stack-up and power-delivery noise is a SIwave study (why a board browns out without its decoupling capacitors), and cable/harness parasitics — the paths that carry noise out through the 38999s past any shield — are a Q3D study (why a charging cable gets warm). A certification-grade campaign chains all three; this post prices the box and everything inside it.

So — Does the Mission Computer Pass?

This is the shape of an EMI/EMC validation campaign compressed to its essence: the same product, several enclosure decisions, each priced in decibels before any hardware exists. The slot version of this box would be a certification risk anywhere near a GNSS antenna; the hole-field version costs nothing to manufacture and buys ~47 dB at the frequency that matters most; the chimney buys margin that looks like a sealed box while still passing air. History says this is worth doing early: from the docked laptop correlated with autopilot disconnects in 1995 to the 2014 airworthiness directive that replaced cockpit displays on some 1,300 aircraft after Wi-Fi-band interference blanked them during testing, aircraft-level EMI surprises are expensive precisely because they arrive late.

Honest scope. This is a physics demonstration on a generic, self-authored half-ATR-style enclosure, not a compliance measurement on any real product. The five module sources are behavioral broadband dipoles at realistic positions and polarizations driven at 1 W each — real chips differ enormously in drive level and spectrum, so the coupling results are per-watt transfer hierarchies, not absolute interference predictions; the vent-comparison solves use a single such source. Cables and connector pigtails are not modeled, and real boxes leak at seams and penetrations that are absent here — which is why the welded variant reports zero leak while real sealed enclosures still get gaskets. Curves are sampled every 50 MHz, so high-Q resonances are likely sharper and deeper than plotted. Shielding effectiveness is defined on the peak field over a far-field sphere against a free-space reference solve of the same source — the with/without-enclosure emissions definition — and DO-160 band markers are context, not a pass/fail claim, since certification measures absolute emissions of the real unit in a calibrated setup. None of that moves the engineering conclusions: aperture shape and depth set the leak, and inside the box, proximity, polarization, and the shared cavity modes set who interferes with whom.

Designing an enclosure, a vent pattern, or a chassis that has to live next to a GPS antenna? The decibels are decided in CAD, long before the test chamber — and a full-wave solve prices every slot, hole, and seam before you cut metal. We run EMI/EMC shielding studies in Ansys HFSS, geometry through certification-ready evidence. Rand Simulation — innovation through insight.