Heat spreader or fins in a sealed enclosure: sizing the thermal path

Short answer: at 10 W you need a spreader block, and fins inside only slow you down. Still air inside a sealed box moves heat at about 5-8 W/m2-K, so internal fins multiply a weak coefficient. A greased 20 x 20 mm aluminum block from chip to wall carries the same load at about 0.5 K/W, a 5 C rise. The wall then does the rejecting outside: 0.06 m2 of painted aluminum at 30 C ambient rises about 15 C, which holds your chip near 50 C against an 80 C limit. Each step below carries its arithmetic, the numbers that move it, and the checks that prove the build.

What numbers do you need before you size the path?

You need six numbers: the real heat load in watts, the worst ambient, the skin material and its area, the chip-to-wall gap, the chip's temperature limit, and how steady the load stays through a working day. Get them by measurement, not from the nameplate. A board sold at 10 W often averages 3-4 W at its true duty, and your metal should be sized for the peak you will sustain around the clock. A bolted block also carries thermal mass, so 20 g of aluminum absorbs a two-minute spike without a blip. Steady state is what kills the chip, and a small sealed box reaches it in about 40-60 minutes.

Do fins inside a sealed enclosure actually help?

For one warm chip at the example load, no: internal fins enlarge the air path, which still runs at 5-8 W/m2-K in still air, while a bolted block reaches the wall for about a tenth of the temperature difference the air path needs. Fins earn their area only where air moves, and inside a sealed box the air moves by natural convection alone. Run the air-path arithmetic for the example: the full load into all 0.06 m2 of inside surface at 6 W/m2-K is 0.36 W/K, so the internal air sits 28 C above the wall before the chip-to-air step adds its own 10-20 C. Carrying that much heat at a 10 C rise by air alone takes about 0.17 m2 of finned surface, nearly three times your box's whole skin.

The honest cases for fins inside are these: heat sources scattered across several boards that cannot each touch a wall, a gap too long for a block, or a load you have already moved to the wall and now need to spread. Where one chip dominates, conduction cooling a single-board computer with a solid path beats every internal fin array, and conduction cooling a single-board computer is covered with its own numbers. If you do build internal fins, leave 8-12 mm of air between them; tighter spacing merges the boundary layers and the fins stop working.

How much heat will the outside wall reject?

Count on about 6 W/m2-K of natural convection from a warm vertical wall, plus roughly another 6 W/m2-K of radiation once the surface is painted or anodized: 0.06 m2 of painted surface then rejects the full load at a 15 C rise. The two mechanisms add. Convection holds near 6 W/m2-K on a vertical face and drops toward 2-3 W/m2-K on a face pointing down, while radiation behaves like a coefficient of about 6 times emissivity. Bright bare aluminum has an emissivity near 0.05, so it radiates almost nothing; paint or an anodized finish takes emissivity to 0.8-0.9 and nearly doubles the rejection. The convection and radiation loss formulas for metal surfaces are the same physics behind these tables.

Treat the enclosure rating and the thermal design as separate problems. A box rated NEMA 4X / IP66 by the manufacturer (ours are rated this way and are not UL listed) is proven against hose water, and it says nothing about watts. Your heat numbers come from the arithmetic above; conductive cooling through an enclosure wall walks the same method end to end.

Aluminum or stainless wall for a conduction build?

Aluminum, by a wide margin: heat entering a 1 mm sheet spreads over a distance of about sqrt(k t / h), which works out near 160 mm [6 in] for 6061 aluminum and 50 mm [2 in] for 304 stainless at about 16 W/m-K. With aluminum, the whole wall joins the heat path. With stainless, metal more than 50 mm from the block stays at ambient, and your rejecting area collapses to a patch. The fix on a washdown job is a spreader plate: 3-5 mm of aluminum, 80-100 mm square, between block and wall, bare on both faces. It costs a few millimetres of depth and restores the wall as a sink. You can order either metal to your dimensions from custom enclosures, aluminum included.

How big should the heat spreader block be?

For the example build across a 14 mm [0.55 in] gap, a 20 x 20 mm aluminum block carries the heat with about 2 C of rise in the metal itself, and its two greased joints add 3-4 C, so the chip-to-wall path holds near 0.5 K/W. Conduction resistance is R = L / (k A). For 6061 aluminum, k is about 167 W/m-K, so a 20 x 20 mm bar (A = 0.0004 m2) bridging 14 mm gives 0.014 / (167 x 0.0004) = 0.21 K/W, or 2.1 C at 10 W. Each greased joint adds about 0.16 K/W, from a 0.05 mm film of 0.8 W/m-K grease over the 4 cm2 face, and there are two of them.

The joints, not the metal, own the budget. A dry bolted joint runs 1-2.5 K/W and would eat 10-25 C at this load on its own. From the same formula: a 20 x 20 mm 6061 block stays under 0.5 K/W for gaps up to about 33 mm, and a 30 x 30 mm bar stretches that to 75 mm. Extra thickness beyond the gap adds spike-absorbing mass rather than conduction. This is the same approach behind using a sealed enclosure as a heat sink, with the wall playing the fin.

When does copper or a heat pipe pay?

Copper at about 390 W/m-K against 167 for 6061 halves the block's resistance, but that block was 2 C of a 20 C budget, so copper saves you about 1 C; a heat pipe pays once the gap passes 50 mm or the load passes 30 W. Copper also weighs 8.9 g/cm3 against aluminum's 2.7, which loads your mounting hardware on a wall-hung or moving box. A heat pipe behaves like a bar of several thousand W/m-K, tens of times copper, but it needs its evaporator at the hot end and good pads at both interfaces, so treat it as a 30 W-and-up part. For sinking one hot module, heat sink thermal resistance in sealed enclosures and the module to cold plate thermal path carry the details. Past about 50 W, stop adding metal and choose among fans, air conditioners and heat exchangers.

How do you build the conduction path step by step?

Build it in six steps, from the measurement to the clearance check, and your finished path lands within 1-2 C of the arithmetic above. Lock out the feeder and confirm absence of voltage before your hand goes inside a live panel.

1. Measure the load you will truly sustain, on a bench supply over an hour; for the example build, that is 10 W flat.
2. Choose the shortest chip-to-wall run in the layout, 14 mm in the example, and mark that face. On a custom box, ask for a machined flat pad there, and lay the mounting holes out in the cutout designer along with your glands and display opening; the box is delivered in 10 business days, counted for custom work from your approval of the drawing.
3. Cut the block from 6061 bar, 20 x 20 mm for gaps to 33 mm, and skim both faces flat on fine paper.
4. Spread a pea-sized dab of grease to a 0.05 mm film on both faces, then clamp with M3 hardware at about 0.6 N-m [5 lb-in], tightening corners in sequence.
5. Hold the block without new holes through the wall: a bracket off the backplate, or a 0.1 mm bond line of 1.5 W/m-K thermal adhesive. A sloppy 0.5 mm bond line adds 0.8 K/W and gives back 8 C at this load.
6. Keep 2 mm [0.08 in] of clearance between the hot block and any gasket, lid seam or cable insulation, and leave at least 50 mm [2 in] of free air outside the rejecting faces.

How do you check the finished path?

Run the box at full load for 2-3 hours, since a small box settles in 40-60 minutes and the longer run proves steady state, then read two numbers: the wall rise over ambient and the chip rise over the wall. For the example build the budget closes like this:

Chip temperature = 30 C ambient + 15 C wall + 5 C path = 50 C, with 30 C of margin to the 80 C limit. Read your two rises against that table and the fault names itself. A wall running hot points outside: paint it, give it area and free air. A chip running hot over a cool wall points at the joints: strip, re-grease, re-torque. Both on target with the chip throttling means your load is bigger than the bench number. The load-to-cooling arithmetic behind these steps is in the enclosure heat calculation guide.

Which mistakes cost you the degrees you calculated?

Four mistakes eat the margin in a working build: a dry or painted joint under the block, a bond line left thick, a wall starved of outside air, and an outdoor box that forgets the sun. A 0.05 mm paint film under the block adds 0.6 K/W, or 6 C at full load, so machine the wall pad bare or order the box with the pad machined. Mount the box with its big faces vertical and 50 mm [2 in] clear, because a face pointing down convects at 2-3 W/m2-K instead of 6. Sun adds 300-800 W/m2 to an outdoor wall and swamps the example budget in a small box; read solar heat load on an outdoor enclosure before you mount one facing south. At a 45-50 C site ambient, the same arithmetic with the smaller window points you to sealed enclosure hot ambient cooling options instead of more metal.

The decision by load, then: up to 10 W, a greased aluminum block and a painted wall; 10-30 W, the same block plus a painted external fin set or the next size of box; past 30 W, or a gap past 75 mm, heat pipes or a cold plate to the wall; heat spread across many boards, internal fins at 8-12 mm spacing or a pad per board to the backplate.

Enclosures for this job

Catalog sizes that suit the job this article answers, picked from its subject; open one for its full specifications.

Questions