LED Thermal Management Guide
Junction Temperature, Heatsink Sizing & Lifetime
1. Introduction — An LED Is a Heat Pump, Not a Light Bulb
A high-power LED converts roughly 70–80% of its input electrical power into heat, not light. That heat must flow from the semiconductor junction to the ambient, and every thermal resistance in between pushes the junction temperature up. Junction temperature (Tj) is the single most important number in LED reliability: it drives luminous flux (light output), forward voltage, color shift, and above all lifetime (L70 — hours to 70% lumen maintenance). This guide walks the full thermal chain: the resistor network model (Rth → Rth), a worked junction-temperature calculation, how to size a heatsink from a thermal budget, PCB thermal design (aluminum-core boards and copper), constant-current driving and its thermal coupling, and the derating curves you must honor.
2. The Thermal Resistor Model
Tj = Ta + Pheat · (Rth jc + Rth c-s + Rth s-a)
Where:
– Tj: junction temperature (°C)
– Ta: ambient temperature (°C)
– Pheat: dissipated thermal power ≈ Vf·If − optical power ≈ 0.75·Vf·If (typical)
– Rth jc: junction→case (internal, from datasheet, e.g., 1.5–3 °C/W for a 1 W class, lower for larger die)
– Rth c-s: case→sink (TIM / solder / mounting, e.g., 0.2–1 °C/W)
– Rth s-a: sink→ambient (the heatsink, the design variable, e.g., 5–15 °C/W)
The lower Tj, the more light and the longer life.
2.1 Why Tj matters
- Luminous flux: flux drops ~5–10% per +25 °C of junction temperature for many packages (temperature coefficient is negative, roughly −0.2%/°C).
- Forward voltage: Vf falls with temperature (−1 to −2 mV/°C); with a constant current source this shifts operating point and changes power slightly.
- Color shift: chromaticity drifts with Tj, especially phosphor-converted white LEDs — a thermal runaway of color in enclosed luminaires.
- Lifetime: LM-80 data and the L70 criterion show lifetime halves for roughly every +10 °C of junction temperature in the typical operating range.
3. Worked Example — Junction Temperature & Lifetime
Step 1 — electrical power: Pel = Vf·If = 3.2 × 0.7 = 2.24 W
Step 2 — thermal power (assume 75% heat): Pheat = 0.75 × 2.24 = 1.68 W
Step 3 — total thermal resistance: Rth tot = 1.8 + 0.3 + 8 = 10.1 °C/W
Step 4 — junction temperature:
Tj = 25 + 1.68 × 10.1 = 25 + 16.97 ≈ 42 °C
Lifetime estimate: the LED datasheet gives L70 ≈ 50,000 h at Tj = 85 °C, with lifetime halving per +10 °C above a reference. From 42 °C to 85 °C the temperature difference is 43 °C ≈ 4.3 “ten-degree steps”; extrapolating: L70(Tj = 42) ≈ 50,000 × 2^(4.3) ≈ 50,000 × 19.7 ≈ 985,000 h — vastly more than the fixture’s practical life. At 42 °C the lifetime is effectively infinite for this design. Now repeat with a too-small heatsink: Rth s-a = 20 °C/W → Rtot = 22.1, Tj = 25 + 1.68×22.1 ≈ 62 °C → L70 ≈ 50,000 × 2^(2.3) ≈ 246,000 h. Still fine. But raise ambient to 50 °C (outdoor fixture, enclosed): Tj = 50 + 1.68×22.1 ≈ 87 °C → L70 ≈ 50,000 × 2^(−0.2) ≈ 43,500 h — now the fixture is limited by thermal design, and at 95 °C+ it drops to ~25,000 h. The lesson: heatsink and ambient matter as much as the LED’s own Rth jc.
4. Heatsink Sizing
4.1 Thermal budget method
1. Compute Pheat = 0.75 · Vf · If.
2. Allowable rise: ΔT = Tj,max − Ta,max − Pheat·Rth jc − Pheat·Rth c-s.
3. Required sink resistance: Rth s-a, req ≤ ΔT / Pheat.
Example: Pheat = 3 W, Rth jc = 2.0, Rth c-s = 0.5 →
ΔT = 85 − 45 − 3×2.0 − 3×0.5 = 85 − 45 − 6 − 1.5 = 32.5 °C
Rth s-a, req ≤ 32.5 / 3 = 10.8 °C/W (a modest extruded fin heatsink in free air).
Rule of thumb: a natural-convection extruded sink gives roughly 5–15 °C/W per 1–2 W dissipated per 10 cm of finned length; fans cut Rth s-a by 3–5× but add noise, power, and failure modes.
4.2 Heatsink best practices
- Orient fins vertically for natural convection; horizontal fins trap hot air and increase Rth s-a significantly.
- Don’t block airflow in enclosures — vent the housing, or the “heatsink” radiates into a sealed hot box and Rth s-a balloons.
- Use a good TIM — thermal pad vs. paste vs. phase-change: the interface is often the biggest improvable Rth term; compress pads to spec, use the right thickness.
- Match expansion coefficients for big COBs — CTE mismatch between Al sink and ceramic COB can crack solder; use appropriate mounting and not over-torque.
- Consider anodized aluminum — improves emissivity for radiation and corrosion resistance.
5. PCB Thermal Design
| Board type | Thermal conductivity | Use for | Trade-offs |
|---|---|---|---|
| FR-4 (2 oz Cu) | ~0.3 W/mK (laminate), Cu ~385 | Low-power LEDs, THT | Poor lateral spreading; needs thick Cu + vias |
| Aluminum-core (MCPCB) | Dielectric 1–3 W/mK, Al ~150 | Mid/high-power, 0.5–3 W | Not reworkable with hot air easily; single-sided |
| Copper-core / IMS with vias | Very high | High-power COBs, >3 W | Cost, weight |
- Spread heat with copper: a large copper pour under the LED pads spreads heat laterally before it hits the dielectric — the copper area is often the limiting factor, not the dielectric.
- Thermal vias: a grid of small vias (0.3 mm) with thermal relief under the LED to the opposite copper plane dramatically lowers Rth c-s; fill or tent them per assembly method.
- Keep the LED’s thermal pad well-connected: the exposed pad on SMD LEDs must be soldered with proper stencil aperture and adequate solder volume — a dry joint is a thermal open circuit.
- Aluminum-core dielectric thickness: thinner dielectric = lower Rth but higher cost and weaker isolation (careful with mains-isolated designs — isolation requirements may force a thicker dielectric and thus higher Rth).
6. Constant-Current Driving & Thermal Coupling
LEDs are current-driven devices: a small forward-voltage change (driven by temperature) causes a large current change if driven by voltage. A constant-current driver is mandatory for consistent light output and to protect the LED from thermal runaway:
- Constant current (CC): the driver regulates If, so Pheat ≈ Vf·If stays roughly constant regardless of Vf drift with temperature — no positive feedback runaway.
- Driver loss adds heat: a linear CC driver dissipates (Vin − VLED)·I — that heat is on the board too and must be included in the total thermal budget. A switching driver is ~85–92% efficient but adds EMI and a heatsink of its own.
- Positive thermal feedback in voltage drive: Vf falls with temperature → with a fixed voltage supply, current rises → more heat → lower Vf… a runaway. This is why you never drive LEDs directly from a voltage rail without series resistance or a CC stage.
- Thermal derating of the driver: many driver ICs include temperature-based current fold-back; verify the driver’s own Tj stays within limits at the worst case.
7. Derating Curves
Datasheet derating curves give the maximum allowed forward current vs. ambient temperature (or vs. Tsolder point) for the package. Read them properly:
- The curve typically shows a flat region (thermal resistance low enough to carry the current) that falls off after a threshold ambient — that threshold is where the package can no longer keep Tj under its max (often 120–150 °C).
- Apply the derating curve for the worst-case ambient, not the nominal. An indoor LED at 25 °C may run at 100%; the same part in a 60 °C outdoor enclosure must be derated to a fraction of the current.
- Derating is per board: the datasheet curve assumes a reference board (specific Cu area and thickness). Your layout with less copper gives a worse effective Rth and thus a worse derating curve — measure your own board’s Tsolder point to confirm.
8. Common Mistakes
- Sizing the heatsink for the nominal ambient instead of the worst-case enclosed ambient — the classic “works in the lab, overheats in the field.”
- Ignoring the driver’s heat — the driver sits on the same board, its dissipation adds to the local temperature the LED sees.
- Thermal pad not soldered / dry joint — an open thermal path under an SMD LED turns a 2 °C/W path into 20+ °C/W.
- Fins horizontal / enclosure sealed — natural convection stops, Rth s-a roughly doubles, Tj rises to a level that halves lifetime.
- Voltage driving instead of current driving — thermal runaway risk and uneven brightness across parallel LEDs.
- Thin FR-4 with no vias for a mid-power LED — the board itself is the bottleneck; use MCPCB or thick copper + via arrays.
- Not measuring Tsolder point in the prototype — thermal models drift; a thermocouple on the pad confirms the design before production.
9. Frequently Asked Questions
Q1. Why does my LED get hot even though it’s “efficient”?
Electrical efficiency only converts ~20–30% of input power to light; 70–80% becomes heat. A “high-efficiency” LED still dissipates most of its electrical input as heat — the heat must be moved, not hoped away.
Q2. What Tj should I design to?
For long lifetime, target Tj ≤ 85 °C at the worst-case ambient. Many quality LEDs are rated to 120–150 °C max junction, but running there cuts L70 to a fraction of the datasheet number.
Q3. Aluminum-core vs. FR-4 with thermal vias?
For 1 W+ LEDs, MCPCB is usually the right call: the aluminum layer spreads heat far better than vias in FR-4. For tiny LEDs or very low power, thick-copper FR-4 with a dense via grid can suffice and costs less.
Q4. Is a fan acceptable for LED cooling?
Fans give excellent Rth s-a but add noise, power draw, and a moving failure point. For outdoor or sealed products, passive (convection) is strongly preferred; use active cooling only when the heat density demands it.
Q5. How do I measure my board’s real thermal performance?
Mount a fine thermocouple to the LED’s solder point (or use the package’s Tsolder point measurement point), run at max drive in the worst-case enclosure, record the steady-state temperature, and back-calculate Tj = Tsolder + Pheat·Rth js from the datasheet.
Q6. My white LED shifts color when it warms up — is that normal?
Yes. Phosphor-converted white LEDs shift chromaticity with junction temperature (blue shift and/or CCT change). If color stability is critical (e.g., display or lighting), design the thermal budget to keep Tj narrow, or use color feedback.