Thermal resistance is the single number that connects electrical power to die temperature. It is also the number most often quoted without its reference point, which makes the figure close to meaningless. A 3 K/W rating and a 12 K/W rating can both be correct for the same part.
What Rth Actually Means
Thermal resistance is a temperature difference divided by a heat flow, in kelvin per watt. Multiply it by the thermal power passing through the path and you get the temperature rise across that path:
Delta T = Rth x Pd
For an LED, the power that becomes heat is forward current times forward voltage, less the optical power that leaves as light. A blue die inside a white package converts roughly 40% to 50% of input power to light, so a nominal 3W part might pass 1.7W to 1.8W as heat. Using the full 3W in the calculation overstates the rise by more than half.
The Three Definitions
Every Rth figure is quoted between two reference points, and the choice of points decides the value.
| Symbol | Path | Typical high-power value | Notes |
|---|---|---|---|
| Rth(j-s) | junction to solder point | 2 to 8 K/W | Most common for SMD and ceramic packages |
| Rth(j-c) | junction to case | 3 to 10 K/W | Metal-can and star-board parts |
| Rth(j-a) | junction to ambient | Board and heatsink dependent | Valid only with the stated board |
Rth(j-s) is the most transferable of the three, because it stops at the package boundary and leaves the board design to the engineer. Rth(j-a) includes the PCB, the thermal vias, the heatsink and the airflow, so it changes with every layout change.
Why the Test Board Matters
Thermal testing follows the JEDEC 51 family. A part mounted on a 25mm by 25mm FR-4 board with a small copper pad and the same part on a metal-core board with a large copper area will show Rth(j-a) values that differ by a factor of three or more. Both are honest measurements of a defined system, and neither describes the customer’s design.
When a datasheet lists Rth(j-a), look for the board description next to it. If there is no board description, treat the figure as a marketing number and work from Rth(j-s) instead.
From Rth to Junction Temperature
A worked example with a 3535 ceramic part. Forward current 1000mA, forward voltage 3.2V, so input power is 3.2W. With 40% leaving as light, thermal power is about 1.9W. Solder point temperature measured at 85C. Package Rth(j-s) is 3.0 K/W.
Tj = 85C + 3.0 K/W x 1.9W = 85C + 5.7C = 90.7C
That sits below the usual 115C or 125C maximum for the die, with some margin. Raise the drive current to 1500mA, and thermal power rises to roughly 2.7W, which puts Tj at about 93C on the same solder point. Note that the solder point itself would rise too in a real fixture, because the same heatsink now carries more heat.
Why the Margin Is Not Optional
Junction temperature drives three failure paths at once. It accelerates lumen depreciation, it shifts chromaticity as the phosphor and die emission move, and it speeds intermetallic growth at the die attach and bond pads. An LM-80 report at 105C against one at 55C shows the size of the effect on lumen maintenance alone.
A part held at 85C may hold its bin for years. The same part pushed to 120C can drift out of the bin in a few thousand hours and lose several percent of output in the first 1,000.
What to Ask Before Designing In
- Rth(j-s) with the measurement method stated.
- Maximum junction temperature, not just maximum case temperature.
- Die attach material, because solder and sintered silver behave differently at high temperature.
- Thermal derating curve for current against ambient, if the supplier publishes one.
Key Takeaways
Rth is only meaningful with its two reference points. Rth(j-s) travels between designs; Rth(j-a) does not. Always subtract the optical power before computing the thermal load, and compare the resulting junction temperature against the die maximum with real fixture conditions, not bench conditions.
Related reading
