Ahmedonics

Electronics

Heat sink and junction temperature calculator

The heat sink thermal resistance a transistor, MOSFET, regulator or diode needs from its dissipation, the ambient in the enclosure and a design junction temperature; the junction, case and sink temperatures with the sink you have; the same device with no sink; and the power each can take.

Device, interface and heat sink
For a linear regulator (Vin − Vout) × I; for a MOSFET I² × RDS(on) at the hot temperature plus switching loss; for a diode Vf × I. From the datasheet or a measurement.
The absolute maximum is 150–175 °C; design 30–50 °C below it. Reliability roughly halves for every 10 °C, as a rule of thumb.
The air the heat sink actually sees. A closed panel in a Pakistani summer sits 10–20 °C above the room.
From the datasheet. TO-220 0.5–2, TO-247 0.3–1, D²PAK 0.5–2; small-die parts such as a 78xx regulator in TO-220 are higher, about 3–5. A TO-92 has no case path worth using (about 80 to ambient).
TO-220 with thermal grease 0.3–0.5; silicone insulating pad 0.8–1.5; mica plus grease 0.5–1; bare metal with nothing 1–2. Set 0 for a device soldered to a copper pour.
From the sink datasheet at natural convection. Small clip-on TO-220 sinks 15–25; a 50 × 50 × 25 mm finned sink 5–8; a 100 mm extrusion 1.5–3. Forced air divides these by 2–4.
Datasheet figure for the bare package in free air: TO-220 about 62, TO-247 about 40, D²PAK on 1 in² of copper about 40, TO-92 about 160. Only for the "no sink" comparison.

Temperatures

Enter your values and press Calculate.

What the calculator does

Every watt lost in a transistor, a MOSFET, a linear regulator or a rectifier diode has to get out through the package to the air, and on the way it raises the temperature of the silicon above the air by a fixed number of degrees per watt. That number is the sum of three thermal resistances in series: from the junction to the case, across the interface between the case and the heat sink, and from the heat sink to the surrounding air. The calculator takes the power, the temperature of the air inside the enclosure and the junction temperature you are prepared to run at, and gives the heat sink thermal resistance you must beat. It then takes the sink you have and works out the junction, case and sink temperatures it produces, the power it could take at the design temperature, and for comparison what the same device does with no heat sink at all.

It is the check to run before bolting a 7805 to a panel, before choosing a MOSFET for a motor driver or a solar charge controller, before deciding whether a fan is needed in a sealed box, and after a device has failed and you want to know whether it was heat.

Formula

Heat flow obeys the same rule as current flow: a temperature difference drives a heat flow through a thermal resistance, exactly as a voltage drives a current through an electrical one. Watts play the part of amperes, degrees of volts, and °C/W of ohms, and resistances in series add.

Rθja (with sink) = Rθjc + Rθcs + Rθsa

Tj = Ta + P × (Rθjc + Rθcs + Rθsa) Tcase = Tj − P × Rθjc Tsink = Tcase − P × Rθcs

Heat sink needed: Rθsa ≤ (Tj,design − Ta) ÷ P − Rθjc − Rθcs

Power the sink allows: Pmax = (Tj,design − Ta) ÷ (Rθjc + Rθcs + Rθsa)
No sink: Tj = Ta + P × Rθja Pmax = (Tj,design − Ta) ÷ Rθja

where P is the power dissipated in the device in watts, Rθjc the junction-to-case resistance from its datasheet, Rθcs the resistance of the mounting interface, Rθsa the heat sink's resistance to the surrounding air, and Rθja the datasheet junction-to-ambient figure for the bare package. All temperatures are in °C and all resistances in °C/W (the same thing as K/W). If the required Rθsa comes out at zero or below, the junction-to-case and interface resistances alone use up the whole temperature budget and no heat sink can rescue the design.

Worked example

A MOSFET in a TO-220 dissipating 15 W in a motor driver, mounted with thermal grease on a finned heat sink rated 2.5 °C/W, inside a panel whose air reaches 45 °C, with a design junction temperature of 110 °C. Datasheet Rθjc = 1.0 °C/W; grease interface 0.5 °C/W; bare-package Rθja = 62 °C/W:

  1. Temperature budget from junction to air: 110 − 45 = 65 °C. Spread over 15 W that is 65 ÷ 15 = 4.33 °C/W for the whole chain.
  2. Heat sink needed: 4.33 − 1.0 − 0.5 = 2.83 °C/W or better. The 2.5 °C/W sink qualifies.
  3. With that sink the chain is 1.0 + 0.5 + 2.5 = 4.0 °C/W, so the junction sits 15 × 4.0 = 60 °C above the air, at 105 °C: 5 °C inside the target.
  4. Working back along the chain: the case (the tab) is 105 − 15 × 1.0 = 90 °C and the heat sink surface under it is 90 − 15 × 0.5 = 82.5 °C. That is the temperature a thermocouple on the sink next to the device should read at full load, and a figure to check against on the bench.
  5. The most this sink allows at 110 °C is 65 ÷ 4.0 = 16.25 W.
  6. Without a heat sink, Tj = 45 + 15 × 62 = 975 °C by the formula: the device would fail in seconds. The bare package can pass only 65 ÷ 62 = 1.05 W at 110 °C in 45 °C air.

Change the interface to a silicone insulating pad at 1.2 °C/W (because the tab is at drain potential and the sink is earthed) and the same sink puts the junction at 45 + 15 × 4.7 = 115.5 °C, above the target: the pad has cost 10.5 °C. Either a better sink, or mica plus grease, or an insulated-tab package puts it back.

Typical thermal resistances

ElementTypical Rθ, °C/WNotes
TO-220 power MOSFET or transistor, junction to case0.5–2From the datasheet; a large die is lower. Small-die parts in the same package (78xx regulators, small transistors) are 3–5.
TO-247, junction to case0.3–1Bigger tab, bigger die.
D²PAK (TO-263), junction to case0.5–2The tab is soldered to the board; the "sink" is then the copper pour and Rθcs is zero.
TO-92, junction to ambient~160No usable case path; the leads carry most of the heat. Keep dissipation under about 0.3 W.
TO-220 bare in free air, junction to ambient~62JEDEC-style test board, still air. Only for the no-sink comparison.
TO-247 bare in free air, junction to ambient~40Same caveats.
Interface: thermal grease, bare tab to sink0.3–0.5A thin smear, screw torqued to the datasheet figure (typically 0.5–0.8 N·m for M3 on a TO-220). Thick grease is worse than thin.
Interface: mica washer plus grease0.5–1Insulates the tab to a few kV; grease both sides.
Interface: silicone or ceramic-filled insulating pad0.8–1.5Convenient and clean; the price is a degree per watt.
Interface: bare metal, no compound1–2Air in the surface roughness; avoid.
Clip-on or bent-sheet TO-220 heat sink15–25Enough for 1–3 W.
Finned sink about 50 × 50 × 25 mm5–8Vertical fins, free air, roughly 5–10 W at a 40–60 °C rise.
Extruded sink 100 mm long, 40–60 mm high fins1.5–3Vertical fins, 30–50 mm of clear air above and below.
Any of the above with a fan÷ 2 to ÷ 4Depends on the airflow through the fins; read the sink's forced-convection curve at your air speed.

All figures are typical, taken from device datasheets and heat sink catalogues, and are for orientation only; the datasheet of the actual part and the actual sink govern. Natural-convection Rθsa figures are usually quoted at a 75 °C rise and vertical fins in free air; at a smaller rise, with horizontal fins or in an enclosure they are worse.

Assumptions and limitations

  • Steady state only. The calculation assumes the power has been on long enough for every temperature to settle, which for a heat sink takes minutes. For pulses, surge currents and short-duty loads the junction sees the transient thermal impedance Zθ(t), which is far lower than Rθjc for pulses shorter than the package's thermal time constant; use the Zθ curves in the datasheet for that case. The steady-state answer is conservative for pulses and correct for continuous loads.
  • The sink datasheet figure has conditions. Rθsa is measured with the fins vertical, in free air with 30–50 mm of clearance above and below, usually at a 75 °C rise. Horizontal fins, a sink lying on a chassis floor, or a sink in an enclosure with no air path can add a third to double the resistance. Forced-air figures are for a stated air velocity through the fins, not for a fan somewhere in the box.
  • The ambient is the local air, not the room. Inside a closed panel the air is heated by everything in it; 45–55 °C inside a box in a 35 °C room is normal in Pakistan in June. Measure it or estimate it from the total dissipation and the enclosure's own thermal resistance to the outside air.
  • Several devices on one sink share it. The sink's rise above ambient is set by the total power on it. Add the powers to find Tsink, then take each device up from there through its own Rθcs and Rθjc.
  • Rθcs depends on how it is mounted. Flatness of the sink, thickness of the grease, torque on the screw and whether the tab is bent all move it. The figures in the table assume a flat, clean, properly torqued joint.
  • Rθja is not a design number. The junction-to-ambient figure for a bare package comes from a JEDEC test board (JESD51-2 and JESD51-7) of a size and copper area you probably do not have, and it changes with the board. It is on the page only to show what "no heat sink" means; do not design to it.
  • Losses must be found first. The calculator takes the dissipated power as given. For a MOSFET remember that RDS(on) at 110 °C is typically 1.5–2 times the 25 °C datasheet figure, so the loss at temperature is larger than the cold calculation, and that switching loss is separate; for a linear regulator include the quiescent current at the input voltage; for a diode use Vf at the actual current and temperature.
  • No altitude or dust derating. Air is thinner at altitude (a fan moves less mass of air at Quetta's 1 700 m than at Karachi) and a fin gap clogged with dust conducts far less; neither is modelled.

Frequently asked questions

How hot is too hot?

The datasheet's absolute maximum junction temperature, 150 or 175 °C for most silicon, is where the manufacturer stops guaranteeing anything, not a place to design to. The usual rule of thumb is that the failure rate of a semiconductor roughly doubles for every 10 °C, so a design at 110 °C lasts something like sixteen times longer than one at 150 °C, and the derating curves in the datasheet reduce the allowed current with case temperature anyway. Most industrial designs aim for 100–125 °C at the worst-case ambient; the default here is 110 °C.

Do I need a fan?

When the required Rθsa comes out below about 1–2 °C/W and the sink has to fit in a small box, yes. A natural-convection sink at 1 °C/W is a 150–200 mm extrusion with free air on both sides, which rarely fits; a 60 mm fan on a sink a third that size does the same job. The price is a moving part that fails, a filter that clogs and dust that arrives anyway, so the design must survive the fan stopping: an over-temperature switch on the sink that trips the load is cheap insurance.

Thermal grease or a pad?

Grease, if the tab can be at the heat sink's potential: it is the lowest resistance (0.3–0.5 °C/W on a TO-220) and cheap. A silicone pad is cleaner and quicker on a production line and insulates, at 0.8–1.5 °C/W. Phase-change pads and graphite sheets sit between. Whatever it is, apply it thinly: the compound fills the roughness of two surfaces that are already nearly touching, and a thick layer is an insulator.

What if the tab is live?

The tab of a TO-220 is the collector or drain, so on a mains-side or bridge-connected device it carries voltage, and a sink bolted to the chassis must be insulated from it. The options are an insulating pad or mica washer with an insulating bush on the screw, which costs 0.5–1 °C/W; an isolated-tab package (TO-220F, TO-247 with full-pack insulation), which has a higher Rθjc built in, typically 2–4 °C/W; or floating the whole heat sink at the tab's potential and keeping fingers off it. Put the penalty into Rθcs or Rθjc and re-run the numbers.

Can I use the chassis or the enclosure as the heat sink?

Yes, and for a few watts it is the best sink there is: a steel or aluminium panel of a few hundred square centimetres is a few °C/W to the room. The caveats are that paint and powder coat are insulators on the contact face and must be removed under the device, that a thin steel sheet spreads heat poorly so the area that counts is within a few centimetres of the device, and that an outdoor enclosure in the sun is itself hot. Bolt through a flat, bare area with grease and treat the panel's measured rise as the Rθsa.

How do I measure the junction temperature?

You cannot reach it, so infer it. Put a fine thermocouple on the tab as close to the die as the package allows (a small hole drilled in the sink under the tab, filled with grease, is the standard method) and add P × Rθjc from the datasheet. Or use the diode method from JESD51-1: the forward voltage of the body diode or base–emitter junction falls by about 2 mV/°C at a fixed small sense current, so calibrate it in an oven, run the device at load, switch briefly to the sense current and read the voltage. An infrared thermometer on the package reads the surface, which is cooler than the junction, and misreads shiny metal badly unless a patch of matte tape sets the emissivity.

References

  • JEDEC JESD51 series: JESD51-1 (electrical test method), JESD51-2A (natural convection test environment), JESD51-7 (high effective thermal conductivity test board) — how Rθja and Rθjc are defined and measured, and why Rθja is a comparison figure
  • IEC 60747-1, Semiconductor devices — Part 1: General — ratings, thermal resistance and transient thermal impedance definitions for discrete devices
  • P. Horowitz and W. Hill, The Art of Electronics, 3rd ed., Cambridge University Press, 2015 — §9.4, heat and power design: thermal resistance, heat sinks and derating
  • Manufacturer thermal design application notes, for example ON Semiconductor AN-1040 (Mounting Considerations for Power Semiconductors) and Infineon's thermal design guides — interface resistances, mounting torque and insulating washers
  • Heat sink manufacturers' catalogue data (Aavid Thermalloy, Fischer Elektronik and others) — Rθsa against length, temperature rise and air velocity for extruded and clip-on sinks

Last reviewed 2026-09-20.