Power Supply Heat Sinks: Sizing for Efficiency and Life
Short answer: Size a power supply heat sink by working backwards from the junction temperature you refuse to exceed. For a typical 150 W switch-mode supply running 88% efficient, that is roughly 18 W of waste heat to remove. Targeting a 40 °C rise above a 45 °C ambient, you need a sink-to-ambient thermal resistance near 1.1 °C/W in natural convection, or about 0.35 °C/W with 200 LFM forced air. Undersize it and every 10 °C of extra junction temperature roughly halves electrolytic capacitor and semiconductor life. BQUQ machines and extrudes these heat sinks in Dongguan to ±0.005 mm on CNC interfaces and quotes in 12 working hours.
Why heat sink sizing decides power supply life, not just temperature
A power supply is a chain of thermally stressed components: switching MOSFETs or IGBTs, output rectifiers, magnetics, and — the usual first casualty — aluminium electrolytic capacitors. Every one of them has a rated life quoted at a reference core or junction temperature, and every one derates as temperature climbs.
The industry rule of thumb is brutal and worth memorising: for aluminium electrolytics, every 10 °C reduction in core temperature roughly doubles expected life. For silicon, Arrhenius-type degradation gives a similar order of magnitude. That means a supply that would last 50,000 hours at a 65 °C capacitor core might reach only 12,000 hours at 85 °C. The heat sink is not a cosmetic accessory bolted to the secondary side; it is the component that sets the temperature floor for the whole assembly.
Sizing errors usually come from three places:
- Using the semiconductor's rated power dissipation instead of the actual loss at your operating point.
- Ignoring the thermal path resistance between the device case and the heat sink base — interface material, mounting pressure, and flatness.
- Assuming the heat sink's catalogue rating applies unchanged inside an enclosure with restricted airflow.
Get those three right and the rest is arithmetic.
How do you calculate the heat load a power supply heat sink must remove?
Start with losses, not with the heat sink.
For a switch-mode supply, total loss is approximately:
P_loss = P_out × (1/η − 1)
So a 150 W output at 88% efficiency loses about 20.5 W. In practice, split that loss by stage:
| Loss source | Typical share of total loss | Notes |
|---|---|---|
| Primary switching devices | 35–50% | Conduction + switching, rises with frequency |
| Output rectifiers | 20–35% | Schottky or synchronous FETs |
| Magnetics (transformer, inductors) | 10–20% | Core + copper loss |
| Snubbers, gate drive, control | 5–10% | Often ignored, rarely negligible |
| Input rectifier / PFC | 5–15% | Only if PFC stage present |
Only the portion that reaches the heat sink matters for sizing it. Devices mounted directly to the sink — MOSFETs in TO-247 or TO-220, rectifiers in TO-247AC, IGBT modules — transfer most of their loss there. Magnetics and capacitors usually cool by convection and radiation to the enclosure, not through the sink.
A worked example
Take a 150 W, 88% efficient supply, 45 °C internal ambient, and a target junction temperature of 110 °C for a MOSFET with a 1.5 °C/W junction-to-case resistance dissipating 6 W.
- Allowable case temperature: 110 − (6 × 1.5) = 101 °C
- Allowable rise from ambient to case: 101 − 45 = 56 °C
- Interface resistance (thermal pad, 0.4 °C/W typical): 6 × 0.4 = 2.4 °C
- Allowable sink rise: 56 − 2.4 = 53.6 °C
- Required sink-to-ambient resistance: 53.6 / 6 ≈ 8.9 °C/W for that device alone
That number looks generous because it is for one device. Add the rectifiers, the PFC switch, and the shared base plate, and the aggregate requirement tightens fast. Sum the losses that land on the sink, then divide the allowable rise by that total. That single figure — total watts divided into allowable temperature rise — is the number you take to a heat sink supplier.
What thermal resistance does a power supply heat sink actually need?
The governing equation is simple and unforgiving:
T_junction = T_ambient + P_total × (R_jc + R_interface + R_sink)
Rearranged for sizing:
R_sink = (T_junction,max − T_ambient) / P_total − R_jc − R_interface
Everything hinges on R_sink, the heat sink's thermal resistance to ambient, which depends on geometry, material, surface finish, and — most of all — airflow.
| Cooling mode | Typical R_sink for a 100 × 100 × 40 mm extruded sink | Practical limits |
|---|---|---|
| Natural convection, vertical | 1.5–2.5 °C/W | Needs open fins, 8–12 mm gaps, no dust clogging |
| Natural convection, horizontal | 2.5–4.0 °C/W | Avoid; hot air pools between fins |
| Forced air, 100 LFM | 0.7–1.1 °C/W | Small fan, low noise |
| Forced air, 200 LFM | 0.4–0.6 °C/W | Standard 40 mm axial fan |
| Forced air, 400 LFM | 0.25–0.35 °C/W | Audible; often needs ducting |
| Liquid cold plate | 0.05–0.15 °C/W | Overkill for most supplies |
These are indicative ranges for extruded 6063-T5 aluminium with a natural or black anodised finish. Your actual value depends on fin count, fin thickness, base thickness, and how much of the sink is shadowed by other components.
Fin geometry rules that hold up in practice
- Fin gap: 8–12 mm for natural convection, 4–6 mm for forced air. Tighter than 4 mm in natural convection chokes the boundary layer and gains nothing.
- Fin height: useful up to about 40 mm in natural convection; beyond that, added surface area outruns the chimney effect.
- Base thickness: 5–8 mm spreads heat from a concentrated device footprint. Below 4 mm, spreading resistance dominates and the outer fins run cool while the junction cooks.
- Fin thickness: 1.2–2.0 mm is a good extrusion compromise between surface area and conduction along the fin.
If your device footprint is small relative to the base — a single TO-220 on a 100 mm plate — spreading resistance can add 20–40% to the effective sink resistance. Thicker base or an embedded copper slug fixes it. For genuinely concentrated, high-flux sources, a copper-base aluminium-fin design or a vapour chamber is the better answer.
Which heat sink construction suits which power supply?
Not every supply should get the same sink. The table below maps common topologies to construction.
| Supply type | Typical loss | Recommended heat sink | Why |
|---|---|---|---|
| 5–30 W wall adapter | 1–4 W | Stamped aluminium, adhesive-back or clip-on | Low cost, low flux, often no sink at all |
| 50–150 W enclosed SMPS | 8–25 W | Extruded aluminium, natural convection | Simple, no fan to fail, sealed against dust |
| 150–500 W industrial DIN rail | 20–60 W | Extruded with forced air, or bonded-fin | Higher flux needs more area per volume |
| 500 W–3 kW server / telecom | 50–200 W | Skived or bonded-fin with high-static fan, or cold plate | Density and airflow budget dominate |
| High-voltage IGBT inverter stage | 30–150 W per module | Black anodised extruded or CNC machined base with isolated interface | Creepage, flatness, and insulation matter as much as R_sink |
Two construction notes worth knowing:
Extrusions are the default for power supplies because they are cheap per kilogram, can be cut to any length, and integrate mounting rails, device slots, and screw bosses in one profile. BQUQ runs extrusion-based heat sink production alongside CNC finishing in the same Dongguan factory, so machining of device pockets, tapped holes, and interface flatness happens without a second supply chain.
Bonded-fin and skived assemblies beat extrusions when you need a fin density an extrusion die cannot produce — typically below 3 mm gaps or aspect ratios above 10:1. They cost more and add a thermal interface between fin and base, so specify them only when the airflow and flux genuinely demand it.
How do you validate a power supply heat sink before production?
Simulation gets you close; measurement closes the gap. A practical validation sequence:
1. Bench test the worst case. Run the supply at maximum load, minimum input voltage (highest input current), and maximum rated ambient. Minimum line is the case people forget.
2. Measure case temperature, not just sink temperature. A thermocouple on the heat sink base tells you the sink is working; a thermocouple on the device tab tells you whether the interface is working.
3. Soak for thermal equilibrium. Small supplies reach steady state in 20–30 minutes; larger magnetics can take over an hour.
4. Check the capacitors. Electrolytic can temperature is the life-limiting measurement in most supplies. If the can is 15 °C above the sink, airflow is bypassing the wrong components.
5. Test in the real enclosure. An open-frame bench result can be 30–50% optimistic versus a sealed metal box.
If you want to go deeper on airflow paths and ducting, the sibling article on forced airflow design covers fan selection, static pressure, and duct geometry in detail.
Interface material is not a rounding error
A dry joint between a TO-247 tab and an anodised aluminium base can add 1.0–1.5 °C/W. A good thermal pad brings that to 0.3–0.5 °C/W; grease and a clamped joint can reach 0.1–0.2 °C/W. On a 6 W device, that difference is 5–8 °C of junction temperature — enough to shift capacitor life by tens of thousands of hours. The thermal grease and pad selection guide walks through the trade-offs between pump-out, dielectric strength, and assembly cost.
Where does cost come from, and where can you cut it without losing life?
Heat sink cost in a power supply is driven by, in rough order: material mass, extrusion die amortisation, secondary machining, surface finish, and assembly labour.
| Cost lever | Typical saving | Risk to thermal performance |
|---|---|---|
| Reduce fin height by 20% | 8–15% material | R_sink up 10–20%; check margin |
| Switch from black anodise to mill finish | 5–10% | Radiation loss up to 10% at low airflow |
| Lengthen the profile instead of adding fins | Neutral to +5% | Usually safe if board space allows |
| Use a standard profile instead of custom die | 10–25% on low volumes | May not fit the enclosure |
| Machine fewer features, use clips | 5–15% | Clip pressure must be validated |
| Raise airflow instead of adding area | Fan cost + noise | Dust ingress, fan is a wear item |
The honest framing: material reduction is the fastest saving, but it eats thermal margin that you may need for the worst-case ambient. If you are already at 15 °C of junction margin, take the saving. If you are at 5 °C, spend the money. A broader treatment of these trade-offs lives in the heat sink cost reduction article.
Frequently Asked Questions
Q: How do I know if my power supply heat sink is big enough?
A: Measure the device case temperature at full load, worst-case ambient, and minimum input voltage after thermal equilibrium. If the case sits more than 15–20 °C below the device's rated maximum, you have margin. If it is within 10 °C, or the electrolytic capacitor cans run hot, the sink is undersized — add area, improve the interface, or increase airflow.
Q: Is extruded aluminium always the right choice for power supply heat sinks?
A: For most supplies under roughly 500 W, yes. Extrusions give the lowest cost per degree of thermal resistance, cut to any length, and can integrate mounting features in one profile. Above that, or where fin gaps must drop below 3 mm, bonded-fin, skived, or cold plate designs become competitive. The crossover depends on airflow and flux density, not on a fixed wattage.
Q: How much does surface finish affect heat sink performance?
A: Less than most people assume. Black anodising improves radiation by roughly 5–10% at low airflow and near zero in strong forced convection, where convection dominates. Its bigger practical value is corrosion resistance and consistent emissivity. For a fan-cooled supply, mill finish is often fine; for a sealed natural-convection enclosure, black anodise earns its cost.
Q: What is a reasonable thermal resistance target for a 150 W supply?
A: If roughly 18–20 W of loss lands on the sink and you allow a 40 °C rise, you need about 2.0–2.2 °C/W for the aggregate sink. Natural convection in a sealed enclosure rarely achieves that in a small footprint, so most 150 W designs use either a larger external sink or a low-noise fan targeting 0.5–0.8 °C/W.
Q: Can BQUQ supply heat sinks matched to a specific power supply design?
A: Yes. BQUQ runs CNC machining, extrusion finishing, stamping, and spring production in one ISO9001 factory in Dongguan. We machine device pockets to ±0.005 mm, tap mounting holes, apply finishes, and supply interface hardware. Send a drawing or a thermal target and we return a quote in 12 working hours, with flexible MOQ for prototype and pilot builds.
Related Resources
- About BQUQ and our Dongguan manufacturing footprint: /about/
- Full heat sink product range, including extruded and CNC machined options: /heat-sinks/
- Extruded aluminium heat sink profiles and custom die support: /extruded-heat-sinks/
- Industry trends in electronics thermal management: /industry-dynamics/
- Technical articles on heat sink design and manufacturing: /bquq-blog/
- Frequently asked questions about quoting, tolerances, and MOQ: /faq/
- Case studies from power electronics and industrial customers: /case/
- Contact the engineering team for a 12-hour quote: /contact/
Authored by the BQUQ Engineering Team. BQUQ (Dongguan) runs CNC machining (±0.005 mm), metal stamping, custom springs, and heat sink production in one ISO9001 factory. Source-direct from Dongguan, China — quote in 12 hours: sc@bquq.com | WhatsApp +86 13713157787 | www.bquq.com
