Two numbers, one division
Watt density is the power you put into a square centimetre of heated area. You need the power at nominal voltage (P = V² / R) and the heated area — the printed resistor footprint, not the outer dimensions of the plate. Divide one by the other and you have W/cm². Every other decision in the thermal design is a consequence of that number. Take a 40 × 20 mm chip rated 24 V, 60 W, with the resistor covering 32 × 16 mm. The heated area is 5.12 cm², so the watt density is 11.7 W/cm². The same part rated 100 W would sit at 19.5 W/cm²: same ceramic, same geometry, nearly twice the thermal stress.
What the alumina can actually conduct
Heat crosses the plate by conduction, q″ = k · ΔT / t. With k = 20 W/m·K for 96% alumina and a fired thickness of 0.635 mm, every kelvin of difference across the plate carries 3.15 W/cm². So 45 W/cm² needs only about 14 K across the ceramic. That is the part most people get wrong: the ceramic is not the bottleneck. At 45 W/cm² the plate is barely working; something else is limiting you.
Then why does 45 W/cm² behave like a ceiling?
Because three other limits arrive first. The resistor must stay below the temperature where the tungsten–alumina interface starts to change, and that temperature is fixed — a higher watt density means a hotter element for the same surface temperature. The interface to your load, whether air, a metal block or a graphite pad, usually dominates the total thermal resistance, and poor contact makes the element run hot while the surface stays cool. And the gradient across the plate, however small, is a stress: 14 K held for a minute is harmless, 14 K applied in 200 ms is a shock.
Duty cycle and mounting change the answer more than the ceramic does
These ranges are starting points from industry practice, not guarantees. The number that matters is the one your thermal model produces for your mounting.
- Free convection in still air — 3–8 W/cm², surface temperature is the limit
- Clamped to a metal block with a compliant interface — 15–30 W/cm²
- Flowing liquid on the other side — 30–50 W/cm²
- Intermittent duty under 30% — 45–60 W/cm² is reachable
Lowering watt density without losing power
Four levers, cheapest first:
- Improve the interface — contact resistance is usually the cheapest fix and the most often skipped
- Increase the footprint — a meander or serpentine pattern buys area inside the same outline
- Split the circuit — zone the resistor and drive the zones in sequence
- Thicken the plate — more spreading, lower surface flux, slower response
From power to trace geometry
Once power and area are fixed, the resistor follows. R = V² / P gives the hot resistance. The paste datasheet gives sheet resistance in ohms per square, and the printed meander then has to deliver R = R_sheet × N × (L / W), where N is the number of squares. A 60 W, 24 V part needs 9.6 Ω; at 20 mΩ/sq that is 480 squares, for example 24 passes of 20 squares each. The width that fits those squares into your footprint is what sets the real watt density — which is why changing the voltage on an existing drawing is never free.
The border does not heat
Every plate carries an unheated margin: the distance from the resistor edge to the plate edge, plus the termination area. On a small chip that margin can be 25–35% of the outline. Calculating watt density on the outline flatters the design by exactly that much, and it is one of the most common specification errors we see. Put the resistor footprint on the drawing, not just the outline, and say which area the number refers to.
Uniform flux is not uniform temperature
Edges lose heat on three sides and run cooler; the centre of a large plate can run 30–60 K hotter than the edge at the same flux. A process that needs ±5 K across the surface needs a graded resistor pattern — denser at the edges — or a spreader plate. Ask for the predicted profile, not the average.
The measurement nobody does
Watt density is a design number; surface temperature is the measured one. Put a thermocouple on the plate and another on the load, run to steady state at nominal voltage, and record both. If the plate sits 60 K above the load, fix the interface before asking for a lower watt density. The heater is doing its job; the mounting is not.
Specifications
Questions buyers ask
Intermittent duty, yes. Continuous in still air, no. The failure is not instant — the element runs hot, resistance drifts, and the part dies at 2,000 hours instead of 20,000.
The resistor footprint. Using the outer dimensions flatters the number by 30–50% on a typical chip with a border.
No. For a fixed power, watt density depends only on power and area. Voltage changes the resistance and the trace geometry, not the flux.
Lower voltage cuts power by V², so it does cut watt density — and it cuts the temperature the part can reach by the same logic. If you need the temperature and not the flux, add area; do not drop voltage.
Not the number, but it changes surface temperature at the same flux: a thicker plate spreads heat better and runs cooler at the surface, at the cost of a slower response.
Send us your geometry
Outline, voltage, target power and mounting method are enough for us to check the watt density before you commit to a drawing.