For high power component, I think it is also good to system designers if the component manufacturers can provide more case temperature data with a particular layout (size, number of vias, copper thickness etc.) and related air speed just over the surface where temperature is taken.
There is no risk of ambiguity in the meaning of ambient temperature with respect to a practical implementation it is always uncertain!
My biggest concerns are not the hottest parts but parts that do not dissipate much power. If they are rated for an ambient of say 85°C, does this mean their maximum junction temperature is 90°C? If a low power part is placed next to a high power part, both rated for 85°C ambient, then either the low power part is operating above its maximum junction temperature or the system ambient has to be significantly reduced.
Data sheets often do not say what the maximum junction temperature is. Even if all data sheets did say, what the maximum junction temperature was this is as un-measurable as the ambient in a modern small system.
The most certain temperatures that can be measured in practical system are the case temperatures. Case temperature measurement does have its own problems; how and where to attach the thermocouples, thermocouples acting as a heat-sink, thermocouple wires restricting air flow, surface finishes affecting thermal camera images etc. But I would still strongly encourage all semiconductor manufactures to quote case temperatures for all devices. It would mean OEMs with less than a full understanding of the thermal issues would be encouraged to measure the case temperature and consequently would avoid unknowingly running the devices above the maximum junction temperature leading to failures and poor reputation for the semiconductor manufacture. Similarly, the OEMs with an understanding would be able to raise their operating ambient, reduce their cooling cost or offer more power hungry functions and still be confident their devices are not operating above the maximum junction temperatures.
Romig - Good point. Ultimately, it is that junction temperature that counts. I suppose the best definition of "ambient" depends on where you are in the design through usage chain and from what perspective you are looking.
If you're the design engineer, you may even have two "ambients" to worry about. You have to design based on right at or in the chip. But you also have to design your cooling system based on room operating ambient to keep the device internal ambient from raising to the point where the junction temperature is exceeded.
And I think that may be the take-away from your article. It's an ambiguous, confusing mess. If you have a lot of space and a flexible budget, you can usually just put in a bigger fan, but when space and budget are constrained, then you need someone skilled in the black arts of thermal design.
Keep those app notes like the one you linked to coming. We need them.
Duane- yes and no, our goal is always to make our components as easy for you to use by using system reference points. But the junction temperature is the real physical driver for component function and reliability.
So the reality is anytime you see a "spec" of ambient temperature (from any component supplier- we're all guilty and it goes back decades), there are a lot of buried assumptions in there. The best true "spec" is really a max junction temperature (in our datasheets we generally call it max operating temperature).
A happy medium that I like is to say that "XYZ component is designed to operate at (say) 70C ambient temperature, provided that care is taken to ensure that the maximum operating conditions are not exceeded". This gives you guidance as to what a reasonable ambient temperature should be, but still leaves room for you to design and confirm in your unique system (some of which are pretty tough!).
You can see some reference points and terms in TI Apps Note SPRA953 on www.ti.com. And be on the lookout as I just put together a few videos walking through a few of these points, which should be published shortly (I'll post a note here when they are). And we'd always appreciate suggestions on related topics that we can write about!
hm- Yes, as a component supplier I fully agree that junction temperature is always the aim, to make sure the device functions reliably over it's lifetime. And everything around the device- PCB, other components, airflow (or lack thereof)- will affect the junction temperature. Especially in aerospace, as you mention, where the convection cooling is greatly reduced. It definitely has to be accounted for.
Very informative article. From the outside looking in (an end user's perspective), Number one, or a slightly modified version: "the air temperature of the area in which the device is being operated" seems to be the most useful. Ultimately, isn't the purpose of the ambient temperature spec to determine if the device can still operate within it's temperature envelope based on the maximum thermal rise added to the ambient temperature? Of course, there are still plenty of variables such as wind and solar effects for an outside design.
Eventual aim for thermal consideration is to find out junction temperature of the device. Junction temperature must be within limit specified in datasheet so as not to damage part and it should be below certain margin to provide enhanced reliability and calculated prolonged life. For this, ambient temperature is one parameter for thermal consideration. Other parameters are air flow or effective heat transfer around the individual part and system as whole. Should we also consider altitude of operation and operation in vacuum in general? Device employed in aerospace application does encounter this high altitude and effective heat transfer will be very low.
My Mom the Radio Star Max MaxfieldPost a comment I've said it before and I'll say it again -- it's a funny old world when you come to think about it. Last Friday lunchtime, for example, I received an email from Tim Levell, the editor for ...
A Book For All Reasons Bernard Cole3 comments Robert Oshana's recent book "Software Engineering for Embedded Systems (Newnes/Elsevier)," written and edited with Mark Kraeling, is a 'book for all reasons.' At almost 1,200 pages, it ...