Possibly the most common and familiar thermal parameter known to electronics packaging engineers and system designers is theta-JA (JA). Theta-JA is typically provided on a device data sheet, and using a single number, it tells the story of how hot the junction of a device will run in a given ambient, if you know the power dissipation in the device. Sounds great, doesn't it?
The problem, as any thermal specialist will tell you, is that >JA isn't really a device characteristic. It depends, rather, on many factors external to the device. Although the device certainly makes a contribution to the overall theta value, it is often one of the smallest contributors.
Other significant factors are reasonably obvious. For instance, there's the nature of the airflow around the device, whether a fan is providing forced-air movement and so forth. Then you have the amount of circuit-board metal (signal traces, power and ground planes, heat spreader areas) in the vicinity of the device. Depending on the package style, there is also the possibility of an external heatsink being attached directly to the device.
Perhaps the most under-appreciated external factor is the presence of other nearby power-dissipating devices. You might think this was because they preheat the ambient air, and, therefore, if you just had an accurate measurement of the air temperature immediately surrounding the device, you could use the published JA value. Don't count on it!
Just as moving the air around cools the device primarily by blowing over the board around the device (unless the device has its own dedicated, external heatsink), neighboring power-dissipating devices interact with the heat flow within the board itself. In this way, they significantly affect, albeit indirectly, the JA of the device you thought you knew.
So, how does this have anything to do with your mother? At some point, your mother taught you how to boil an egg or to make a tuna sandwich. Before that, she did it for you, but eventually she just started pointing to the refrigerator and saying, “Make it yourself.”
Sure, part of this had to do simply with her wanting you to become more independent. But she also knew that with independence, your needs would expand beyond her ability to provide. To push the image a bit further, maybe you've become a marathon runner, and your dietary and nutritional requirements have become too specialized and complex for her traditional steak-and-potatoes perspective.
Just so, there might have been a time when thermal numbers on a data sheet could get you somewhere, but now you're grown up and you really have to do your own thermal analysis. Only you know your application, and thermal reality has gotten too complicated to collapse into a single number.
Here's an electrical analogy: Can you boil the electrical function of a device down into a single number? It's unlikely. Depending on the device and the application, there's frequency response, parasitic inductance, gate capacitance, maximum current, maximum breakdown voltage, and on and on.
Do all these come into play in every design? Of course not! A skilled electrical engineer can look at the data sheet and say, “We don't care about frequency response in this application, or the max breakdown voltage; only leakage current matters here. While we're at it, maybe we should double check that we don't exceed single-pulse drain-to-source avalanche energy.”
The truth is that the traditional single number, JA, might be fine for comparing part A to part B (if you carefully read and compare all the footnotes). Unfortunately, it's not going to tell you how hot the device will run in your application; you have to know how to dig in and do your own thermal analysis.
That means you ignore JA, and focus on thermal characterization parameters like psi-JL (ΨJL). Then you figure out how your circuit board ties all your heat sources together and what the capacity of your cooling system is, and how it varies with fan speed or board orientation and ambient (external) air temperature.
Remember, no data sheet is ever going to be able to digest everybody's thermal analysis down into a one-size-fits-all number. The data sheet is not your mother!
Roger Stout has authored more than 40 technical papers and presented thermal seminars at SemiTherm, ITherm and APEC. Stout joined Motorola in 1979, spending most of his time with the company as a thermal-mechanical engineering consultant. He is now a senior research scientist at ON Semiconductor, heading the SCG Thermal/Mechanical Characterization Lab.