This is a note I wrote some years ago during the AO-40 project. If people think it's useful, perhaps I could post it to Wikipedia.
Alan N1AL
-------------------------------------------------------------------
Our local PC board technical expert here at HP has done an extensive study on power dissipation of surface-mount parts. While his report is based on non-space applications, there are a number of conclusions that are very interesting.
I won't include the entire report, partly because it is much too long, and partly because it is HP proprietary. But here are a few highlights.
He found surprisingly little discussion of these issues in the literature. So he performed a series of experiments using an Inframetrics 760 infared camera system to measure component temperatures. All measurements were taken on similar-sized PC boards mounted horizontally with no significant airflow.
One main conclusion: "Part ratings by the vendors have little meaning by themselves since trace width and part density have a huge impact on heat dissipation."
For example, what is the allowable power dissipation of an 0603 resistor? The answer is somewhere between 12 mw and 526 mw, depending on layout!
This is based on a 60 degC temperature rise. (Most resistors have a 125 degC max case temperature spec.) Temperature rise is directly proportional to power dissipation.
The 12 mw number was measured on 112 parts packed into a 0.75 square inch area and all dissipating the same power. The 526 mw measurement was on a single part soldered between two half-board-sized ground planes and tied through many vias to another ground plane on the back side of the board.
Manufacturers rate 1206-case resistors at 125 mw, versus 62.5 mw for 0603-case resistors, a 2:1 ratio. Actually, for most reasonable trace widths, an isolated 0603 resistor can dissipate roughly 80% as much as a 1206. When parts are packed densely together, power dissipation is limited by the maximum watts per square inch. Note that I used a 30 degC temperature rise for the following table:
0603 1206 Isolated Resistor: Large ground plane: 263 mw 403 mw 0.060 inch traces: 170 mw 199 mw 0.040 inch traces: 148 mw 177 mw 0.012 inch traces: 106 mw 128 mw 0.005 inch traces: 77 mw 100 mw 0.0025 bond wires: 55 mw 79 mw High-density part layout, 0.75 square inch (4.84 cm^2) area: Number of parts 112 32 Power per part 6 mw 20 mw Total power 672 mw 640 mw Number of parts 56 16 Power per part 11.5 mw 43 mw Total power 644 mw 688 mw Number of parts 20 8 Power per part 31 mw 80 mw Total power 620 mw 640 mw
Since the power dissipation depends so strongly on trace width, then clearly most of the heat must be conducted, not radiated, on SMT resistors (and other parts). Even in the thin-trace case, much of the heat is conducted to, and radiated from, the PC board. You can see that on the infared photos: the PC board surrounding the part is quite hot.
Thermal resistance of isolated SOT-23 transistors was very similar to 0603 resistors up to 0.040" line widths. SOIC-8 voltage regulators had about 1/2 the thermal resistance of SOT-23.
The above numbers are probably conservative, since they are based on still air, even though most earth-based applications have forced-air cooling or at least natural convection (vertical PC board). The rule of thumb I generally use is 1 watt per square inch (155 mw/cm^2), which gives an average temperature rise of around 35 degC in still air.
I have been told that for space applications, 15 mw/cm^2 is a more appropriate limit. Assuming radiation cooling is only 1/10 as efficient as convection cooling, then that seems a reasonable spec.
Conclusion:
I have heard several rules of thumb on what percentage you should derate component power specifications for space applications. For through-hole devices, such a rule of thumb probably makes sense. THD devices dissipate most of their heat from the component body, and relatively little heat is conducted out the leads. On earth, most of that heat is conducted to the air; only a little is radiated. In space, radiation is the only heat-dissipating mechanism, so the power must be derated by a large factor.
But SMT devices are cooled mainly by conduction out the leads to the copper traces on the PC board, so they are not directly affected by the lack of air. It seems to me that as long as I keep within the 15 mw/cm^2 limit on the PC board, that allowable power dissipation of individual components should be governed by trace width, per the above table. If I have a hot component, I'll connect it to lots of copper and make sure no other hot components are nearby.
<END>