The baseline temperature in an electronic system is a very important parameter for thermal design and must be carefully considered. But the term ambient temperature has developed a lot of ambiguity in its usage as well as its value. System and component designers can greatly improve the accuracy of their thermal considerations by clearly communicating what the baseline temperature truly is for their implementation, and finding a way to estimate the temperatures at or close to the components of interest.
What is Ambient Temperature?
Thermal analysis is an essential part of nearly all electronic system designs these days. We have to make sure that the components do not heat up so much as to affect their functionality or make the system practically unusable. And for most of us, it’s a delicate dance of environmental factors, system design, component power dissipation, and a range of usage conditions. We must establish some baseline temperature conditions, then figure out through modeling, calculations, rules of thumb, and ultimately measurements – how much will things heat up above that temperature.
So what if that baseline temperature and the assumptions around it could vary so wildly, that all the other calculations are rendered as noise, and the results of such a calculation could be off by 100 percent or more? In other words, what is ambient temperature?
The interesting thing is that while it is a very common term, there seem to be different assumptions on what “ambient temperature” really means. Here are some interpretations used by some of us day in and day out, as also illustrated in Figure 1:
- The temperature of the system, unaffected by the temperature rise of the system itself (either far enough away or before the system is turned on). The equivalent of a true thermal ground.
- The temperature of the air inside the system when it is running. Often measured far away from the component of interest so as not to be too much affected by it.
- The temperature of the air moving over a component.
- The temperature of the board or chassis of the system, which is considered to never rise above a certain maximum temperature.
- The temperature of the air (usually driven by fans) at the entry or exit of the system.
- The temperature of the room when a component or system is tested on the bench, on an evaluation module (EVM) or a real system board (with or without an enclosure).
- The temperature setting of the oven when a component or system is tested at high temperature (with or without an enclosure).
- The temperature to which a system is preheated for an operational test (of unspecified duration).
Figure 1: Eight potentially different ambient measurement points.
(Click on image to enlarge)
For example, an end system designer will probably use 2, 4, or 5, because this is the temperature they have experience with from previous systems and will ultimately be able to measure on the system in the field. While a system test engineer will probably use 6, 7, or 8, since they need to focus on the temperature they can control during testing. Alternatively, a component supplier will probably use 1 or 3, since these are conducive to component modeling and the component suppliers are generally not privy to the level of detail needed to consider the entire system effects.
There are two common misconceptions when using ambient temperature alone for calculations of component temperatures.
The first is that each component can be viewed in isolation. You simply go to the datasheet for the component of interest, get the theta-JA value (or maybe even a fancy table or curve or derating factor), drop in your ambient temperature and power dissipation, and there is your component temperature.
But this calculation assumes only one component on the board, and that it therefore has the entire board (or at least a 3x3 inch square of it, if using a JEDEC board) dedicated to cooling it. I’ve not seen many systems where this is the case. In reality the other components around that one will be heating it up and skewing that calculation.
The second is that the ambient in one system is equivalent to the ambient in another. Specifically, that the ambient in a JEDEC system, a one cubic foot box with only one component inside, is equivalent to the ambient in a real system such as a cell phone, base station, or even a car.
Looking at the comparison in Figure 2, it’s pretty clear that this is not what real systems look like, and yet that is the implicit assumption when a JEDEC theta-JA (ΘJA) value is used to calculate component temperature rise above ambient. In fact, the very definition of theta-JA (JEDEC standard JESD51-2) states that the parameter is intended for comparison only!
Figure 2: System comparison.
(Click on image to enlarge)
If ambient temperature means different things to different people, and the basic calculation of component temperature rise above ambient temperature has some flaws in it, what is a system or component designer to do? Here are three suggestions.
First: agree on terms when working through the design stages. Don’t accept a simple specification of (for example) 65°C ambient temperature or industrial temperature rating, but dig into such details as where is the baseline temperature measured, under what operation conditions, and for what period of time?
Second: find a way to get closer to the components than the ambient. This will make the estimates much more accurate, because of the implications of each system’s uniqueness as well as the surrounding components. One good method is to use the board or case temperature as the reference specification, rather than ambient.
Third: take case temperature measurements on the actual system to confirm the analysis. A good way to bring these data earlier into the design cycle is to use measured data on similar systems (often a previous generation) to get an estimate and calibrate, and then simply consider the changes rather than doing thermal analysis of a new system from scratch.
So ultimately, what is ambient temperature? It is a commonly used term. It carries the risk of ambiguity because it is not clearly defined. It carries the risk of inaccuracy because it suggests that junction temperature calculations can be oversimplified. When carefully defined, it can help to start the conversation on thermal design. But, ultimately, it is probably not the best place to end the conversation because there are better options available.
Download a datasheet: “IC Package Thermal Metrics”, Darvin Edwards, Texas Instruments, SPRA953A–June 2007: http://focus.ti.com/lit/an/spra953a/spra953a.pdf.
About the Author
Matthew Romig currently serves as the Packaging Technology Productization Manager in the Analog organization at Texas Instruments, where he is responsible for packaging technology development and implementation for TI’s broad range of Analog product lines and packaging technologies. He has developed specialties in thermal analysis, flip chip packaging, and power management packaging. He is also a member of TI’s Group Technical Staff. Matt received his BSME from Iowa State University, Ames, Iowa. He can be reached at firstname.lastname@example.org.