Natural vs Forced Convection — Calculating Whether You Need That Fan

Half the fans we see in customer designs are there because somebody on the team said "it gets warm, add a fan." The other half are there because the team did the calculation and the fan was unavoidable. The line between the two camps is a one-page thermal budget — usually about twenty minutes of work — and it pays for itself the moment you delete a moving part from a sealed enclosure.

The thermal budget answers three questions:

• How much heat is the electronics dumping inside the box? Sum the dissipation from every IC, regulator, and high-current trace at worst-case load. Include the LDOs — a 5V-to-3.3V LDO at 1A is dumping 1.7W of heat, and we have seen four of them on one board.
• What is the ambient the box has to survive? Indoor commercial is 25–35°C. Industrial cabinet is often 50–55°C internal. Outdoor enclosure in direct sun on a black powder-coat finish can hit 65–70°C surface before the electronics turn on.
• What is the junction-temperature ceiling on the hottest component? Usually the SoC or the switching regulator. Industrial-grade silicon caps out at 105°C junction; AEC-Q100 grade 2 sits at 125°C; aggressive consumer parts derate hard above 85°C.

"A natural-convection box that runs at 12W internal in a 35°C ambient is almost always feasible. The same box at 30W in a 55°C ambient is almost never feasible without forced air, regardless of how clever the heatsinking is." — Pioneer Horizon thermal lab

With those three numbers in hand, the convection question stops being an opinion and becomes a calculation. The rest of this article walks the calculation, then compares it to CFD on a representative enclosure.

For a sealed enclosure cooled by natural convection only, the steady-state heat-transfer equation is:

Q = h × A × ΔT

where Q is the dissipated power in watts, h is the heat-transfer coefficient (W/m²·K), A is the effective surface area in m², and ΔT is the temperature delta between the surface and the ambient. For a typical industrial box in still air, h falls in the range 5–10 W/m²·K. We use h = 7 W/m²·K as a working number for vertically-oriented painted aluminium and 5 W/m²·K for plastic.

Worked example

Take our reference 200×150×80mm enclosure. Surface area is roughly 0.118 m² (sum of six faces; subtract 10% if it is panel-mounted and one face is against an unventilated surface). Assume aluminium, h = 7. Allowable rise ΔT = 20°C (so a 35°C ambient lands the surface at 55°C, and the internal air-to-surface gradient pushes the hottest component to ~80°C junction with a reasonable thermal interface).

Q_max = 7 × 0.118 × 20 = 16.5 W.

That is your sealed-enclosure budget for natural convection on this geometry, at this ambient, with this surface emissivity. Above 16.5W internal dissipation, the surface temperature exceeds 55°C and the internal hot-spots start to climb.

What moves the number

• Surface finish: bare polished aluminium has emissivity ~0.05 (catastrophic for radiation); anodised black hits 0.85. Anodising the box can add 20–30% to the natural-convection budget for free.
• Orientation: a horizontal top plate convects ~25% better than a vertical wall; a horizontal bottom plate is nearly useless.
• Internal layout: a hot component bolted to the enclosure wall via a copper thermal pad uses the entire wall as a heatsink. The same component floating on the PCB radiates only into trapped internal air.

For most industrial designs under 15W dissipation in a 35°C ambient, natural convection is achievable with discipline. Above 25W in the same ambient, you are almost always pushing air.

A customer brought us a 220×160×90mm sealed industrial gateway dissipating 22W at peak. The original design had a 40mm axial fan moving 12 CFM through inlet and outlet louvres. They asked whether the fan was actually doing useful work, given the audible noise complaint from their pilot customers. We ran two CFD models in Simcenter Flotherm and validated each on a benchtop with thermocouples on the hottest three components.

Case 1 — sealed, natural convection only

Internal air temperature stabilised at 71°C in a 35°C ambient. The SoC junction reached 96°C — within absolute spec (105°C ceiling) but with only 9°C of headroom. The DC-DC inductor hit 87°C. Acceptable for a continuous-load product, marginal for one that has to survive a 45°C ambient excursion.

Case 2 — forced convection at 12 CFM

Internal air dropped to 47°C. SoC junction sat at 71°C. Inductor at 62°C. Lots of headroom, but the fan was the dominant noise source and the inlet filter clogged at the 18-month service interval, dropping flow to 7 CFM and pushing the SoC back into the high 80s.

Case 3 — sealed with internal heat-spreader

We added a 3mm aluminium plate thermally coupling the SoC and DC-DC to the rear enclosure wall, anodised the enclosure black, and removed the fan and louvres entirely. Internal air settled at 58°C. SoC junction reached 78°C. Inductor at 69°C. Better than case 2 after filter clogging, with no moving parts, no ingress path, and no audible noise. Bill-of-materials change: +₹140 per unit for the heat-spreader, -₹220 per unit for the fan + filter + grilles.

Net: removing the fan paid for the spreader plus saved ₹80/unit, eliminated the warranty failure mode at the 18-month service interval, and let the customer move from IP54 to IP67 on the same enclosure. The lesson is not "fans are bad" — the lesson is "the surface-area calculation tells you whether you have a choice, and most teams haven't run it."

If the calculation says you cannot make natural convection work, the next question is how much air you actually need. Under-spec the fan and the box overheats; over-spec it and you pay in noise, dust ingress, and bearing-life warranty claims.

The airflow equation

Required CFM ≈ (1.76 × Q × ΔT⁻¹) where Q is in watts and ΔT is the allowable air-temperature rise from inlet to outlet. For Q = 30W and ΔT = 12°C, you need ~4.4 CFM of useful airflow through the box.

The catch: useful airflow is not rated airflow

Fan datasheets quote free-air CFM. By the time you pull air through inlet louvres, an EMI mesh, the PCB obstructions, and out through outlet grilles, system back-pressure typically drops actual airflow to 30–50% of the free-air rating. We size to a 2.2× safety factor over the airflow equation result as a default.

Practical sizing rules we hold to

1. Use a fan rated at least 2.2× the calculated CFM, ideally with a pressure-volume curve that crosses the system impedance line in the flat part of the curve, not the knee.
2. Prefer 60mm or 80mm fans over 40mm — bigger fans move the same air at lower RPM, which means lower noise and longer bearing life. A 60mm at 1800 RPM lasts roughly twice as long as a 40mm at 3500 RPM.
3. Use ball-bearing fans for any product expected to live beyond 3 years. Sleeve-bearing fans cost less but quote MTBF at 25°C — at 50°C the practical life is one-third of the headline number.
4. Always provide a tach output and monitor it in firmware. A fan that has stalled in the field but still draws current is the classic silent thermal failure.

Filter strategy

If you fit an inlet filter, design assuming it is 50% clogged at your stated service interval. That clogging factor is what should drive your CFM safety margin, not the free-air rating.

This is the decision tree we walk a customer through in the first thermal review. It is deliberately conservative — when in doubt, lean toward natural convection because moving parts are warranty liabilities.

Step 1 — compute Q_max for natural convection

Q_max = h × A × ΔT with h = 5–7 W/m²·K, A from the enclosure CAD, ΔT from the ambient spec and the component derating limits. If actual dissipation ≤ 0.85 × Q_max, natural convection is the answer. Stop here.

Step 2 — improve the surface

If you are at 0.85–1.3 × Q_max, you may still be able to land natural convection by:

• Switching to anodised black or matte powder-coat finish (emissivity bump).
• Adding internal heat-spreader plates to couple hot components to the enclosure wall (effective h boost).
• Adding external fin geometry — extruded aluminium with integral fins doubles the effective surface area for ~₹40–60/unit at scale.

Step 3 — accept forced convection if needed

If after step 2 you are still above 1.3 × Q_max, fit a fan. Size it per the airflow equation with 2.2× margin, prefer a larger fan at lower RPM, monitor RPM in firmware, and design the inlet filter for 50% clogged at end-of-service.

Step 4 — re-check ingress protection

A sealed natural-convection enclosure can easily hit IP67. A fan-cooled enclosure rarely exceeds IP54 without exotic membrane filters or pressurised cabinets. If the product needs IP65+ outdoor, step 4 routes you back to step 2 and you redesign the thermal interface rather than fight a hopeless gasket battle. See our companion article on IP67 vs IP68 gasket geometry for that decision.

If you would like us to run this tree on your own design — including a Flotherm CFD pass before EVT — share your mechanical CAD and dissipation budget and we will return a sizing report within a week.