When deciding on the trace thickness required to carry a certain amount of current on a PCB, the answer depends on how much temperature rise you are willing to accept. This leads the designer into the difficult situation of trying to decide how much temperature rise is reasonable. Common rules of thumb are to allow no more than 5°C, 10°C or 20°C of temperature rise, depending on how conservative you want to be. These figures seem remarkably small compared to the maximum temperature rises of power transistors, ICs, power resistors, or other heat-dissipating components, which may be 60+°C. What is the reasoning behind these numbers?
Possible reasons I've thought of:
- Maximum temperature of PCB materials. For most FR4-type materials this is around 130°C. Even allowing for a very conservative ambient temperature (inside the chassic) of 65°C, this would still allow another 65°C temperature rise.
- Allowing for further temperature rises of components. If an SMT MOSFET was going to see a temperature rise of 80°C for example, you wouldn't want to start it out 40°C above ambient because of the temperature of the surrounding PCB. However, this seems far too situation-specific to make for a rule of thumb. In the case of a heat-sunk through-hole MOSFET, for example, the heat flow up the leads is a fraction of the heat flow out through the heat sink, so the PCB temperature should not be a major concern. Even with SMT parts, I could have a thin trace which dissapates a lot of heat for the majority of its length, but then widen that trace before it reaches the component.
- Thermal expansion of PCB materials. As the PCB heats up, the materials will expand. If different parts of the PCB are exposed to different amounts of heat, this could cause flexing of the board which could crack solder joints. However, given that PCBs are regularly exposed to higher temperature differentials than this due to power dissipation in the components mounted to them, this doesn't seem like the answer.
- Outdated standards. Perhaps the 5/10/20°C limits were thought up years ago and no longer apply to modern PCB materials, but everyone has kept on following them without thinking about it. For example, perhaps old phenolic board materials were less tolerant of heat than modern fiberglass.
To put the question another way, say I find that a 20°C temperature rise is too limiting for my design. If I decide instead to allow a 40°C temperature rise, am I likely to run into either short-term or long-term reliability problems?
Bonus points to anybody who can cite standards that give a reasoning for the numbers, or who have historical evidence for why those numbers were picked.