Keeping Your Cool: Heat, Vents, Fans and Airflow in an RF Shielded Test Enclosure

Every so often a customer calls us with the same story. They put their device in one of our chambers, closed the door, ran their test — and an hour later the inside of the box was noticeably warm, or the device throttled back, or a temperature-sensitive instrument drifted. The chamber was doing exactly what it was designed to do: it sealed the device off from the outside world. The trouble is that "the outside world" includes the cool air of your lab.

An RF shielded test enclosure is, thermally speaking, a picnic cooler. It's a sealed, gasketed box — and to add insult to injury, the LS-30 absorbing foam that lines the walls to keep your RF measurements clean is a very respectable thermal insulator. The same open-cell polyurethane structure that soaks up microwaves also does a fine job of keeping heat from reaching the aluminum walls. Whatever power your device dissipates has to go somewhere, and in a shielded chamber, "somewhere" is almost entirely air moving through the vents.

This paper explains the whole business: why chambers heat up, why the wattage printed on your device's label is usually the wrong number to design around, how the honeycomb vents pass air without passing RF, how much airflow it takes to hold the temperature rise to a given limit, and why two fans on one chamber don't necessarily give you twice the cooling. We'll keep the math down to one honest rule of thumb in the main body; the full derivation lives in the appendix for those who want to check our arithmetic. As usual, we'll use real JRE chambers as examples — the JRE 1714, the JRE 2525, and the big JRE 3036, including its new six-vent high-airflow version, which moves enough air to cool a fifty-phone device farm — with real fan numbers, including the derating nobody's spec sheet likes to talk about.

Start with a law you can't negotiate with: essentially every watt your device draws from its power source ends up as heat inside the chamber. A 40-watt device is a 40-watt space heater. (The sliver of energy that leaves as RF down a coax cable or as data down a filtered line is a rounding error — even a "high power" 1-watt transmitter is only exporting 1 watt.)

In open air on your bench, that heat wanders off on its own — room air drifts past the device, warm air rises away, and the whole room is the heatsink. Close the chamber door and you've cut off that escape route. Now there are only three exits, and two of them are lousy:

Exit 1: conduction through the walls. Heat can soak through to the aluminum shell and radiate off the outside. But the LS-30 foam lining stands in the way — foam is mostly trapped air, and trapped air is what every insulation product on earth is made of. Some heat gets through (the foam isn't magic, and the aluminum floor under the DUT conducts a little), but it's a slow, weak path. Don't count on it for more than a handful of watts.

Exit 2: passive convection through the vents. Warm air is lighter than cool air, so if the chamber has a low vent and a high vent, warm air drifts out the top and pulls cool air in the bottom all by itself — a tiny chimney. This works, and it's free, but the driving pressure is feeble (we put a number on it in Appendix A.3 — it's about a thousand times weaker than a fan). Passive venting comfortably handles a device dissipating a few watts up to maybe 15–25 watts, depending on the chamber's vent placement.

Exit 3: forced air — a fan. This is the workhorse, and it's what the rest of this paper is about. A fan pushes lab air through one honeycomb vent, across your device, and out another vent, carrying the heat with it. How much heat it carries is a simple function of how much air actually moves — which, as we'll see, is not the number printed on the fan.

Figure 1 shows the whole picture in cross-section: heat rising off the DUT, foam-insulated walls, cool air entering the low vent and warm air leaving the high one.

One more definition before we go on, because it matters. When we talk about "temperature rise," we mean the rise of the air inside the chamber above your lab's room temperature. Your device then runs its own additional rise above that internal air temperature, just as it would on the open bench. Cooling the chamber air well doesn't make your device run cold — it makes the device run the same amount above 30 °C chamber air that it would have run above 25 °C room air. The chamber's job is to keep that internal ambient from climbing.

Here's the single most common sizing mistake we see, and it errs in the expensive direction. A customer reads the label on their device — "Input: 12 VDC, 10 A" — multiplies, and concludes they have 120 watts to remove. Sometimes they read the wall-adapter rating instead, which is padded even further. Then they call us asking for airflow numbers that would keep a toaster comfortable.

That nameplate figure is the maximum the device is permitted to draw — a safety and regulatory number, sized for the worst instant of the device's life: every radio transmitting, every peripheral powered, the battery fast-charging, the processor flat out, all at once, plus margin. It is almost never what the device draws sitting in your chamber running your test.

Real devices have a duty cycle. A radio transmits in bursts and listens in between. A processor sprints and idles. A charger tapers off. Figure 4 shows the shape of the thing: a device with a 120 W nameplate, spiking briefly at startup, pulsing up to 65 W during transmit bursts, and idling near 22 W — for an honest average dissipation around 40 W. And it's the average that heats the chamber. Air has enough thermal mass that a few-second burst just doesn't move the needle; the chamber temperature settles according to the long-term average power, not the peaks. Designing your cooling for 120 W when the truth is 40 W means buying three times the airflow you need — or, worse, concluding a chamber "can't be cooled" when it can.

So don't read the label — measure the draw. It's cheap and quick:

• DC-powered device: run it from a bench supply through the chamber's filtered DC interface and read volts × amps right off the supply's display while your actual test runs. Watch it for a few minutes and eyeball the average.
• AC-powered device: put a $25 plug-in power meter (the classic "Kill A Watt" style) between the wall and the device's supply, run the real test, and read the watts. Most of these meters will happily show you an averaged figure.
• Battery-powered device under charge: the heat inside the chamber is the charger's input power minus the energy actually banked in the battery — but for sizing purposes, just use the measured wall draw during charge. It's conservative and it's easy.

Measure during the test you'll actually run, because the duty cycle is the test. A throughput soak test dissipates a lot more than a receive-sensitivity test on the very same device.

A real example of how far apart the two numbers can sit. A customer building a mobile device farm — fifty phones on charge in one chamber — budgeted 30 W per phone, the fast-charge maximum: 1,500 W. But a phone doesn't fast-charge for twelve hours; it bulk-charges hard for twenty or thirty minutes and then falls off a cliff, settling near 8 W once the pack is topped off. The honest sustained load for the whole farm came out around 425 W — about a quarter of the nameplate arithmetic. The 1,500 W figure wasn't useless: it's the right number for sizing the power feed and for a worst-case ceiling check. It was just the wrong number to design the sustained cooling around.

One heat source people forget while they're adding up device labels: the power conversion gear itself. USB hubs, PoE injectors, and power supplies mounted inside the chamber all burn their own conversion losses as heat — figure roughly 10% of the power passing through them. A hub whose datasheet says "0.12 W" is quoting its idle draw; when that hub is pushing a kilowatt out to phones, it's dissipating an extra hundred watts of its own. If the supply lives inside the chamber, its losses are your problem.

At first glance a vent in a shielded wall sounds like a contradiction — didn't we spend two other papers (this one and this one) preaching that every opening leaks?

The trick is the shape of the opening. Our vents are built from a thick panel of small hexagonal metal cells — a honeycomb. Each little cell is a short piece of waveguide operating below cutoff: a metal tube so narrow that radio waves below its cutoff frequency physically cannot propagate through it — the wave dies off exponentially inside the tube instead of traveling. Make the tube a few times longer than it is wide and the attenuation is enormous. Air molecules, of course, couldn't care less about waveguide cutoff and stroll right through.

The result is a panel that flows air freely while holding better than 100 dB of shielding at 1 GHz and better than 95 dB at 10 GHz. And here's a property worth tucking away: below cutoff, that attenuation actually gets stronger as you go down in frequency. Keep that in your pocket — it becomes important when we talk about mounting fans.

The honeycomb does exact one toll for its services: flow restriction. Pushing air through hundreds of narrow cells takes pressure, and that's the root of the fan derating we'll get to shortly.

Here it is — the entire thermal design of a ventilated chamber in one line:

CFM ≈ 1.76 × watts ÷ allowable rise in °C (or 3.16 × watts ÷ rise in °F)

CFM is cubic feet per minute of air actually moving through the chamber, watts is the measured average dissipation inside, and the rise is how far above room temperature you'll allow the chamber air to climb. The formula is nothing exotic — it falls straight out of the heat capacity of air, and Appendix A.1 derives it in four lines. Air just isn't a very dense fluid, so it takes a fair volume of it to cart away a watt.

Our standard design target is a 10 °C rise. That's comfortable for nearly all electronics — a 25 °C lab gives you 35 °C chamber air, no worse than a warm summer day — and it keeps the airflow requirement civilized. If your payload is touchier (precision oscillators, calibrated instruments), design for 5 °C and expect to need double the airflow. If your DUT is rated for a hot environment anyway, 15 °C buys you slack in the other direction.

A few landmark points, at a 10 °C rise:

Figure 2 plots the full picture — required airflow vs. watts for 5, 10, and 15 °C rises — with our actual fan options marked on it so you can see at a glance what a given fan setup can carry.

Now for the fine print that catches nearly everyone.

Every fan CFM figure in our catalog — and in everybody else's — is a free-air rating: the flow the fan moves with nothing in its way, blowing from open room into open room. Bolt that same fan onto a honeycomb vent feeding a foam-lined box that exhausts through another honeycomb, and the restriction pushes back. The fan slides down its performance curve, trading flow for pressure. As a practical rule, figure on delivering about half the free-air rating through a chamber's vent path. Sometimes you'll do a bit better; plan on half and be pleasantly surprised.

Here's the JRE fan lineup with both numbers side by side (delivered @ 10 °C uses the rule-of-thumb formula):

Two more rules about combining fans, and these are the ones that bite. A borrowed electrical analogy makes both of them stick: airflow is like current, static pressure is like voltage.

Rule 1 — Fans in series (push-pull) add pressure, not flow. Put one fan pushing air in the intake vent and a second pulling it out the exhaust, and you have two fans on one flow path — sources in series. Their pressures add, but the flow is bounded by the path: two 31 CFM fans in push-pull deliver roughly 31 CFM of through-flow, at roughly double the pressure. That extra pressure genuinely helps — it's what muscles air through the honeycomb restriction and keeps the delivered flow up near the rating — but it does not double your airflow. In steady state every bit of air that comes in must go back out; it's all one air stream. If your cooling budget was built on the idea that fans add up, recheck it before anything else.

Rule 2 — Parallel paths do add flow. Fans ganged side by side, facing the same way, are current sources in parallel: flows add at the same pressure. Two independent intake-to-exhaust paths — which requires four vents — move twice the air of one, and three paths (six vents) move three times. That's the whole difference between chambers in the next section.

Figure 3 draws both cases side by side.

And one shopping tip that surprises people: axial fans come in two very different families — high flow at low static pressure, and lower flow at high static pressure. A honeycomb vent is a restriction, and a chamber packed with equipment is another one, so a high-pressure fan will outperform a glamorous high-CFM low-pressure fan through that path by a wide margin, even though the low-pressure one looks better on paper. Our super-high-flow 120 mm unit is the working example: it's a 38 mm-thick, 4,500 RPM ball-bearing industrial fan rated to nearly an inch of water column of static pressure — built specifically to push against backpressure, so it holds much closer to its rated flow in a real restricted path. A thin, whisper-quiet fan with the same headline CFM would stall against the honeycomb and deliver a fraction of it. When comparing fans, read the pressure-flow curve, not the headline number.

The price of pressure is noise. That industrial fan runs about 62 dBA — and six of them on one chamber lands around 70 dBA at the box. That's vacuum-cleaner territory: fine in a lab or server room, miserable beside somebody's desk. High-pressure, high-RPM fans are inherently loud, so decide where the chamber will live before choosing the fan. Stepping down one fan grade often trades a few degrees of rise for a dramatically quieter room — a real trade-off worth making deliberately rather than discovering on install day.

Model numbers, for reference: the first two digits are the approximate outside width in inches, the second two the depth — a 1714 is about 17" × 14". Full details on the selection guide.

JRE 1714 — small box, surprisingly good lungs. The 1714 (15.75" × 12.75" × 9" inside) carries four vents: two low on the front, two high on the back. That geometry is no accident. With no fans at all, it forms a natural chimney — cool air in low at the front, warm air out high at the back — and passive convection alone will keep a 10–20 W payload happy.

Add fans and the four vents pay off again: two low intakes and two high exhausts make two parallel flow paths. The 1714's vents are 60 mm, so it takes our 60 mm 23 CFM fans. With all four fitted, that's roughly 23 CFM delivered (two paths × ~11.5 CFM each), call it 130 W of cooling at a 10 °C rise. That's a healthy amount of dissipation for a box you can carry under one arm.

Example: a Wi-Fi access point with a "54 W PoE" label, measured at 30 W average during a throughput soak. Passive vents alone would run warmer than we'd like; one fan (~11.5 CFM delivered → ~65 W capacity) covers it about twice over. Done.

JRE 2525 — big box, small lungs. The 2525 is our largest portable chamber (23.5" × 23.5" × 18" inside), and we'll say it plainly: it's the weakest ventilator in the line. It has two vents — one low on one side, one high on the other. Two vents means exactly one flow path, and one flow path means one fan's worth of delivered air, no matter how many fans you bolt on (Rule 1 above).

The ceiling for a 2525 is therefore the biggest fan that fits the vent: the 84 CFM high-flow 80 mm unit, delivering ~42 CFM through the path — about 240 W at a 10 °C rise. There is no larger fan and there is no third vent. Two vents is two vents.

Example — a real case from our inbox. A research lab housed a set of sensitive digitizer scopes in a 2525, fitted with two standard ~31 CFM fans in push-pull, and asked us for 150 CFM. Two things needed untangling. First, their two fans were delivering ~15 CFM, not 60 — series fans don't add (they'd assumed they did). Second, 150 CFM at a 10 °C rise corresponds to about 850 W of dissipation — and when we asked for the measured draw rather than the sum of the nameplates, the requirement came down substantially. Upgrading the exhaust position to the 84 CFM high-flow fan tripled their delivered airflow in the same chamber. The lesson runs both directions: check what your fans actually deliver, and check what your equipment actually dissipates, before concluding a chamber can't do the job.

JRE 3036 — when you genuinely need big airflow. If the arithmetic demands more than ~240 W of heat removal, the answer is the 3036. The standard build carries four 80 mm vents — two intakes, two exhausts, two parallel paths. With high-flow 84 CFM fans that's roughly 84 CFM delivered, or about 475 W at a 10 °C rise in a reasonably open chamber. (Pack the chamber wall-to-wall with racks and equipment and the airflow pays a second toll beyond the honeycomb — figure less.)

And for the truly big loads, there's now a six-vent version. Recently made a standard build option — no custom engineering charges — the high-airflow 3036 carries six 120 mm vents: three low across the front, three high across the back. That's three parallel push-pull paths, placed so cool air enters low at the front where equipment sits, sweeps the floor, and exits high at the rear — working with the rising heat instead of fighting it. With the three 120 mm fan grades, the six-vent 3036 delivers (derated by half, as always):

That top row is kilowatt-and-a-half territory — full device-farm scale — from a chamber that still holds its full shielding spec, because every one of those six openings is a proper waveguide-below-cutoff honeycomb, fabricated and final-tested by us.

Example — a real device-farm build. A customer sized a fifty-phone test farm for a 3036: phones on fast charge, four big USB hubs, two Wi-Fi access points, and a switch. Nameplate arithmetic said ~1,650 W; adding the hubs' ~10% conversion losses put the true worst-case ceiling near 1,800 W. But the sustained twelve-hour load — once the phones' bulk charge tapered to trickle — measured out around 425 W. On the six-vent 3036 with super-high-flow fans (~307 CFM delivered), that's about a 2 °C rise at the real load and only ~10 °C even at the full ceiling: in a 25 °C lab, roughly 27 °C inside during normal operation and the mid-30s at absolute worst case. Lithium batteries don't even clear their throat at those temperatures. The same customer had previously run a similar farm in a chamber with standard 80 mm venting — a ~35 °C-plus rise at that load — and cooked a cage of phones into bloated batteries. Same physics, opposite outcomes: the difference was nothing but delivered CFM. (Bloating cells are a fire hazard, not a nuisance — if your payload is batteries, size the cooling with margin, not just adequacy.)

Sidebar: big watts size the power entry, too. The same measured wattage that sizes your airflow also sizes your filtered power feed. Our PEM-1 filtered power entry module is rated 10 A / 1,200 W at 120 V — so an 1,800 W ceiling is out of reach for one module, and even two would run near 70% of rating for twelve-hour stretches. A filter running cool is a filter that lasts: for that farm we specified three PEM-1s, each on its own wall circuit, each loafing along under half its rating — with the side benefit that a tripped breaker takes down a third of the farm instead of half. If your load is up in the kilowatts, plan the power entry and the cooling from the same measured number.

The lineup at a glance (10 °C rise, honeycomb derating included):

This is the good question, and it has a clean two-part answer.

Radiated noise through the vent: not a problem. The honeycomb's attenuation is a property of the honeycomb's own geometry — it does not care what's sitting on the other side of it. It provides its 100 dB whether there's a fan in front of it, behind it, or nothing at all. Better still, remember that pocketed fact from earlier: below cutoff, the attenuation gets stronger at lower frequencies — and a brushless fan motor's commutation noise lives in the kHz-to-low-MHz range, exactly where the honeycomb is at its very best. Radiated fan noise coming through the mesh is a non-issue.

Conducted noise on the wiring: this is the real hazard. A fan is a conducted noise source before it's a radiated one, and its power leads are conductors. If a fan lead penetrates the chamber wall, shares a ground path with anything running inside, or simply lies alongside your signal cabling on the way in, the honeycomb does nothing for you — you've built the classic unfiltered-conductor leak our filtered-I/O paper warns about. This is precisely why JRE fans mount entirely outside the shielded boundary and run from an independent filtered external supply that never touches the I/O plate and never touches your DUT's power rails. Nothing new penetrates the wall, and the fan's electrical life stays outside the fence.

If you mount a third-party fan, follow the same rules: fan and all wiring stay completely outside the shield; nothing new penetrates the wall; and if you must drill mounting holes, fill every one with a metallic fastener — stainless preferred, seated tight — so the hole is metal, not air.

• Measure, don't read. Actual average watts, during the actual test. The label is a legal maximum, not a thermal fact.
• Intake low, exhaust high. Warm air wants to rise; arrange the flow so gravity helps instead of fights.
• Don't smother the vents. A DUT, foam block, or cable bundle parked against the inside of a vent chokes the path. Leave breathing room on both faces, outside too.
• Give the air a route past the DUT. Airflow only removes heat from what it touches. Position the device in the stream between intake and exhaust, not in a dead corner.
• Halve the fan rating in your head. Free-air CFM is an optimistic stranger; delivered CFM is your friend.
• Don't stack fans on one path expecting double flow. Series buys pressure, parallel buys flow.
• Check it with a thermometer — but respect the shield. Simplest of all: an infrared thermometer or thermal viewer. Pop the door at the end of a run and shoot the DUT and interior surfaces — those hold their heat for well over a minute, so a quick read is honest even though the air escapes immediately. For a record over time, a $10 min/max or data-logging thermometer sitting inside the chamber logs while you run and tells the story when the door opens. Two tempting shortcuts to avoid: a wireless thermometer won't work — if the chamber is doing its job, that signal isn't going anywhere — and pinching a probe wire through the door gasket breaks the very seal you're testing inside of. If you truly need a live readout with the door shut, run a thermistor or thermocouple through a filtered terminal interface like any other low-frequency line, and it crosses the wall legitimately.
• Duty-cycle your test, not just your device. If the chamber trends warm during an all-day soak, a brief door-open between runs resets the clock.

A shielded chamber isolates your device from RF — and, as a side effect, from the lab's cool air. The foam that keeps your measurements honest also insulates the walls, so nearly all the heat must leave as warm air through the honeycomb vents, which pass air freely while blocking RF by better than 100 dB. The sizing rule fits on a sticky note: CFM ≈ 1.76 × watts ÷ °C of rise, using measured average watts — not the nameplate, which reflects a worst-case duty cycle your test probably never reaches (a farm of fast-charging phones is a quarter of its label once the packs top off — and don't forget the ~10% the hubs and supplies burn on their own). Derate every fan's free-air rating by about half for the honeycomb restriction; remember that push-pull fans on one path add pressure but not flow, while parallel vent paths genuinely multiply it; favor high-static-pressure fans against the honeycomb, with eyes open about their noise; and keep all fan wiring outside the shield. Do that arithmetic honestly and a 1714 cools about 130 W with its four 60 mm fans, a 2525 cools 240 W, a standard 3036 cools 475 W — and the six-vent high-airflow 3036 carries over 1,700 W, enough to keep a fifty-phone farm within a couple of degrees of room temperature.

Questions about a hot payload, or a configuration that doesn't fit the standard recipes? Contact us — sizing this correctly the first time is a ten-minute phone call. More application notes live on our resources page.

A.1 Where 1.76 comes from. Moving air carries heat according to:

P = ṁ · c_p · ΔT

where P is the heat carried away (watts), ṁ is the mass flow of air (kg/s), c_p is air's specific heat (≈ 1005 J/kg·°C), and ΔT is how much the air warms between intake and exhaust — which, in steady state, is the chamber's temperature rise.

Mass flow is density × volume flow: ṁ = ρ · V̇, with ρ ≈ 1.2 kg/m³ for sea-level air near room temperature. Solving for volume flow:

V̇ = P / (ρ · c_p · ΔT) = P / (1206 · ΔT) [m³/s]

Converting to cubic feet per minute (1 m³/s = 2119 CFM):

CFM = 2119 · P / (1206 · ΔT) ≈ 1.76 · P / ΔT(°C)

For Fahrenheit, ΔT°F = 1.8 · ΔT°C, giving CFM ≈ 3.16 · P / ΔT(°F). That's the whole derivation — the constant is just the heat capacity of air dressed up in mixed units.

A.2 Altitude and temperature corrections. The 1.76 assumes sea-level air density. Thin air carries less heat per cubic foot, so at altitude, scale the required CFM up by the density ratio: about +4% per 1000 ft. A lab in Denver (5280 ft) needs roughly 20% more airflow for the same watts and rise. Hot intake air matters the same way, but the effect is small (≈ 3% per 10 °C of intake temperature) — the bigger issue with a hot lab is simply that "room + 10 °C" starts from a worse place.

A.3 Why passive convection is feeble: the stack effect. With a low vent and a high vent separated by height h, warm chamber air creates a natural draft pressure:

ΔP = ρ · g · h · (ΔT / T)

For a 2525-class chamber — h ≈ 0.35 m, ΔT = 10 °C, T ≈ 295 K — that's ΔP ≈ 1.2 × 9.81 × 0.35 × (10/295) ≈ 0.14 pascals. A typical 80 mm fan develops static pressure in the range of 20–40 Pa: a couple of hundred times more. Squeezing flow through two restrictive honeycomb panels on 0.14 Pa yields only a few CFM — which, by the formula in A.1, carries a couple dozen watts at most with a generous rise. That's why passive venting is fine for small payloads and hopeless for big ones.

A.4 Fan curves, system curves, and where "half" comes from. A fan's datasheet curve runs from free-air flow (maximum CFM, zero pressure) down to stalled (zero CFM, maximum static pressure). The chamber — two honeycomb panels, the foam-lined interior, whatever clutter the air must dodge — presents a system curve: pressure drop rising roughly with the square of flow. The fan operates where the two curves cross. For typical axial fans against a two-honeycomb path, that crossing lands near 40–60% of free-air flow, which is where our plan-on-half rule comes from. It also explains the fan-selection tip in the main text: a fan whose curve stays flat (holds pressure) as flow rises keeps its operating point high against a restriction, while a fan with a steep, droopy curve — however grand its free-air number — collapses to a trickle. And it explains why a second series fan helps some: adding the two pressure curves shifts the crossing point up the system curve — more flow, but nowhere near double.