Estimación de coeficientes de convección: de la hoja de datos a la realidad
Short answer: plan on h ≈ 3–8 W/m²·K for natural convection from a heat sink, 10–30 W/m²·K for a gentle fan flow at 1–2 m/s, 30–80 W/m²·K for ducted airflow at 3–5 m/s, and above 100 W/m²·K only with high-velocity or impingement flow. The coefficient is not a material property: it depends on fin geometry, channel width, air velocity, orientation and temperature. Most design errors trace back to using one optimistic h for the whole sink, when the true value varies from fin root to tip and from channel to bypass air.
Every heat sink calculation leans on the convection coefficient h, and h is the least certain number in thermal design. It is also the most abused: datasheets quote it, spreadsheets assume it, and simulations spend thousands of cells computing it, yet a surprising share of thermal failures trace to a wrong h used with perfect math. This article is about estimating h honestly, from textbook ranges to the reality of a fin channel inside your enclosure.
The Ranges That Bound Reality
The convection coefficient describes how many watts leave each square meter of surface per degree of temperature difference. Its range across cooling methods spans more than three orders of magnitude, which is why the same heat load can need a palm-sized sink or a cabinet full of fins. Air is a poor heat-transfer fluid: its conductivity and density are low, so even vigorous air motion delivers far less than any liquid.
| Cooling mode | Air velocity | Typical h range | What it feels like |
|---|---|---|---|
| Natural convection, still air | ~0.1–0.5 m/s (buoyant) | 2–8 W/m²·K | Quiet, passive sink |
| Natural convection with radiation | — | +20–40% effective | Black anodized sinks |
| Gentle fan, open air | 1–2 m/s | 10–30 W/m²·K | Small axial fan, loose shroud |
| Ducted fan through fins | 2–4 m/s | 25–60 W/m²·K | Shrouded sink, real static pressure |
| High-velocity duct or impingement | 5–10 m/s | 60–150 W/m²·K | Server or blower applications |
| Liquid cold plate | — | 1,000–10,000 W/m²·K | Pumped coolant channels |
Use the middle of the range for planning and the low end for worst case. The single most common mistake is treating the optimistic end as the design point, which is how "the datasheet says this sink does 100 W" turns into a field failure at 60 W when the real fan only pushes 1 m/s of poorly ducted air.
Why Natural Convection h Is So Unstable
Natural convection has no fan to stabilize it, so everything drifts. The driving force is buoyancy: warm air rises off the fins, and h depends on the temperature difference itself, roughly h ∝ ΔT^0.25 for a vertical plate. Double the temperature rise and h grows by about 19%, which means the coefficient improves exactly when the sink is hotter — a feedback that flattens the temperature curve and breaks the assumption that Rth is constant.
Orientation changes natural convection more than most people expect. A vertical plate with free air on both sides achieves its highest coefficient, roughly 3–8 W/m²·K at typical electronics temperatures. The same plate horizontal, heated side up, loses 10–30%; horizontal with the hot side facing down, as when a module sits on top of a shelf-mounted sink, can lose 30–50% because the heated surface traps a stagnant air layer. Fin channels add another effect: narrow channels merge boundary layers and choke the buoyant flow, which is why natural-convection sinks need wide pitch. Our natural versus forced convection comparison shows the practical consequences for choosing the cooling architecture.
Forced Convection: Velocity Is Not the Whole Story
Forced airflow raises h, but only the air that actually moves through the fin channels counts. Three effects separate datasheet h from real h. Bypass: air takes the path of least resistance around the fin block, so a fan blowing at a sink in open air may push most of its flow past the fins, not through them; shrouding and ducting fix this. Entry effects: the first few centimeters of a fin channel have thin boundary layers and high local h, while downstream the flow develops and h falls; on short fin channels the whole channel behaves like an entry region, which is favorable but usually ignored in simple estimates. Turbulence: real fan flows are turbulent and mixing, which raises h well above the laminar correlations that textbook examples use.
| Air velocity through channel | Laminar-ish estimate | Realistic ducted value |
|---|---|---|
| 1 m/s | 10–15 W/m²·K | 10–25 W/m²·K |
| 2 m/s | 15–25 W/m²·K | 20–40 W/m²·K |
| 3 m/s | 20–35 W/m²·K | 30–60 W/m²·K |
| 5 m/s | 30–50 W/m²·K | 50–100 W/m²·K |
The gap between columns is engineering judgment, not precision, which is why serious designs measure. The measurement path — build the sink, run the fan, measure the airflow and the temperatures — converts these estimates into a real coefficient, as described in our heat sink thermal testing guide. Estimates size the project; measurements close it.
How to Use h Without Fooling Yourself
The discipline that saves most designs is to use h as a range, not a number, and to carry that range through the whole calculation. Run the thermal resistance budget twice: once with your optimistic h and once with the pessimistic one. If both answers fit your temperature budget, the design is robust and you can proceed to prototyping. If only the optimistic case fits, the design is a gamble dressed as an estimate, and it is cheaper to add fin area, airflow or a fan now than to rework after thermal testing.
Also apply h where it belongs. Fins near the base see the fastest air and the hottest metal; fin tips and the leeward half of the sink see degraded flow. Using one average h across the whole surface is acceptable for quick sizing only if the average is conservative. For natural convection, remember radiation: a black anodized sink at a 50 °C rise rejects roughly 20–40% of its heat by radiation, which behaves like a bonus h that pure convection numbers miss. That is one reason anodized machined heat sinks measurably outperform bare aluminum on passive applications even though anodizing barely changes conduction.
What a Good Supplier's Thermal Data Should Look Like
Supplier thermal data deserves the same scrutiny as the estimates in this article. A resistance number without conditions is not data: the quotation or datasheet should state the air velocity or fan configuration, the ambient, the heat source size, and whether the figure includes a TIM. Anything measured should say how, a wind tunnel at a stated face velocity, a fan-rig test in a stated shroud, or a natural-convection chamber test in a stated orientation. When two suppliers quote the same part, compare the conditions before comparing the numbers, because a sink quoted at 5 m/s ducted air will not behave the same in your 1 m/s enclosure.
Ask how the number was derived, measured or simulated. Measured values carry the ±5–10% uncertainty of the rig; simulated values carry the model's boundary-condition assumptions, often ±10–20%. Both are legitimate if labeled, and both are dangerous if not. A supplier who presents a simulation as a measurement, or omits the airflow entirely, is not necessarily dishonest, but the data cannot be used for your design, and the polite response is to ask for the test protocol.
The most useful question you can ask a heat sink factory is simpler than any of these: what do you need from me to size this correctly? The answer should include watts, source footprint, ambient, allowed rise or target temperature, and the airflow plan. A supplier who asks those questions is running the h-based calculation the way this article describes; a supplier who quotes from a photo is pricing metal. The difference shows up in the first thermal test, and it is far cheaper to discover it in the quotation stage than after tooling.
At BQUQ we quote heat sinks with the conditions stated: the airflow assumed, the resulting resistance, and the machining that makes the interface real, with critical faces held to ±0.005 mm. Because we machine, combine extrusions and finish in one Dongguan plant, the thermal data you get is tied to the part we actually ship. Send the watts, footprint, ambient and airflow to sc@bquq.com or WhatsApp +86 13713157787, and the quote will state its own assumptions within 12 working hours.
Frequently Asked Questions
Q: What is a typical convection coefficient for natural convection?
A: For vertical surfaces and heat sinks in still air, plan on h ≈ 3–8 W/m²·K at typical electronics temperatures, varying with temperature rise, size and orientation. Radiation adds effective cooling of 20–40% on dark, high-emissivity surfaces.
Q: What convection coefficient should I use for a fan-cooled heat sink?
A: It depends on air velocity through the fin channels: roughly 10–30 W/m²·K at 1–2 m/s, 25–60 W/m²·K at 2–4 m/s ducted flow, and higher with blowers or impingement. Use the low end of your range for a worst-case design.
Q: Why is my real heat sink worse than my h-based calculation?
A: Because the real airflow is lower than assumed, or bypassed the fins entirely. Fans deliver less flow through dense fin fields than their free-air rating, and open-air fan setups push much of the flow around the sink. Shroud the sink and measure the actual channel velocity.
Q: How do I get an accurate h for my specific sink?
A: Measure it: apply known power, run your real fan configuration, measure base temperature rise at steady state, and back-calculate h from the fin area and efficiency. That measured value, not the textbook number, is what belongs in your thermal model.
Q: Can BQUQ help turn these estimates into hardware?
A: Yes. Send your watts, airflow plan and temperature budget to sc@bquq.com or WhatsApp +86 13713157787, and we will machine a prototype heat sink to your geometry — extruded or CNC — within 12 working hours of quoting, so you can measure the real coefficient instead of estimating it.
Related Resources
- Natural vs Forced Convection Heat Sinks — choosing the architecture before estimating h.
- Heat Sink Thermal Resistance Calculation — the budget that turns h into a sink size.
- Heat Sink Sizing Calculation — carrying the h range through a full sizing example.
- Contact us — send your drawing for a quote within 12 working hours.
Authored by the BQUQ Engineering Team. BQUQ is an ISO9001-certified source factory in Dongguan, China, running CNC machining, metal stamping, custom springs, heat sink and collet lines under one roof. Send drawings to sc@bquq.com or WhatsApp +86 13713157787 for a quote within 12 working hours. www.bquq.com
