Finding Case Temp

[BusaJames]

Sorry if this is a dumb question. I'm new and an Intern. I have some thermocouples mounted into the base of some heat sinks to get temp readings while running a stress test in my chamber. How can I find out what the case temp is of the chips? Do I just need to worry about the thermal conductivity of the thermal greese since the thermal couples are mounted in the base of the heat sink? Thanks ahead of time for any help.

 

[logbook]

Case temperature is the thing you measure with thermocouples. Die temperature, the temperature of the chip, is calculated from the case temperature, the power dissipated in the chip and the thermal resistance from chip (junction) to case.

 

[BusaJames]

Wouldn't there be some difference because I'm not actually in direct contact with the case though?

 

[itsmoked]

Yes certainly.. And it won't be easy to quantify.

If I were trying to do that I would attempt to use the absolute smallest T/C you can possibly get your hands on and then make sure it is touching the actual IC case.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[sreid]

Putting a thermal couple in a hole in the base of a heatsink is a common way to measure "case temperature." We all know that it is not exactly true but it is practical. Filling the hole with thermal grease is a good idea, it tends to rise to the same temperature as the aluminum around it.

 

[melone]

Or just calibrate your system.....

1) Use thermocouple to measure the chip in a couple of known steady state power consumption scenarios.

2) Use the thermocouple to measure the heat sink under the same thermal loads.

Now you know the difference between case and heatsink.

 

[itsmoked]

Nice idea melone.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[IRstuff]

Another, approach is to use the either the input protection diodes or the output driver drain or collector diffusion to determine the die temperature. Since the forward current is exponentially related to the bias voltage and temperature, one can characterize the diode over temperature and then measure the current for a given voltage to determine the die temperature. The accuracy is probably no more than about 5ºC, but that might still be useful data.

TTFN

FAQ731-376

 

[ScottyUK]

If you know roughly what the case temperature is you can use tiny temperature indicating labels which adhere directly to the case. I have some which are small enough for a DIP-8 package. They cover a range of about 8C so you need to know the approximate temperature you are working with. The type I have are one-shot: they do not return to original state after they cool.

Other option is a thermographic camera. These are expensive but can normally resolve 0.1C or better. Depends on just how accurate you need to be - if you are running a device so close to the limit that a degree or two is the difference between survival and failure then there is a problem. That's why the heatsink temperature is normally a good enough measure.


----------------------------------

Sometimes I wake up Grumpy.
Other times I just let her sleep!

 

[zeitghost]

I've seen the diode forward bias voltage used to determine die temperature in PowerMos die attatch testing.

Sort of works.

 

[melone]

If you are running that close to the limit, then the case temp may not be good enough. In a BGA package, most of the heat goes into the PCB (through the balls). Therefore, it is often better to try to use ThetaJ-B (Junction-to-Board) versus ThetaJ-C (Junction-to-Case). Note: If you do use ThetaJ-B, you must get your thermocouple right next to the package (virtually touching) to get an accurate measurement.

Technically, there is another factor, PhiJ-B, that may be more appropriate to your situation. It depends on your system, i.e. system exposed to open enviornment with air movement, system in enclosure with minimal air movement / no appreciable cooling of the flowing air, circuit connected to a large ground plane (a.k.a. huge heat sink / spreader :-D), etc..

The best method for estimating die temp is to use the method outlined by IRSTUFF. I have used that method successfully in the past, and found it be VERY accurate (ran a correlation exercise with the component manufacturer and the correlation factor was like 98%). Furthermore, it makes the most sense. You are measuring a property on the actual die that is linear over temperature. Plus you don't have to worry about the thermal lagging / smoothing effect of the case and/or PCB.

 

[IRstuff]

The only caveat on the diode approach is that if there are hotspots within the die, your particular diode might be too far away to get the correct temperature, so you might need to sample more than 1 diode.

TTFN

FAQ731-376

 

[ScottyUK]

Hang on guys, are we trying to measure case temp or die temp? The diode-based die temperature methods are good but they don't address the OP's question which was how to measure case temperature. Of course, it would be useful to understand why the OP wants to measure case temp in the first place Some more information about the application would be useful.


----------------------------------

Sometimes I wake up Grumpy.
Other times I just let her sleep!

 

[BusaJames]

I'm evaluating a product made by another company for our use. We expect this product is thermally challanged and we need to measure temps and compare to the datasheets to see how close to the limit we are. As I said I'm and intern with almost no experience with this kind of testing. So far I have found temps of many different componants but now I am trying to find out what to do with them I guess. I would like to extrapolate them to compare with data from the datasheets. Thanks to all that have replied

 

[IRstuff]

Ultimately, what's important is the die temperature, unless we're talking about a safety issue with the heat sinks. Whatever the case temperature might be or might be measured to be, the die temperature is what eventually kills the part.

TTFN

FAQ731-376

 

[musiklab]

I guess it´s too obvious to mention, but to pinpoint hotspots and case temps in power devices, I use a small (keyfob size) IR thermometer, cost me 30$ at the time.
it can make remote measurements on larger surfaces like heatsinks , and closeup case temps on single power devices without making physical contact.
to decide if something is thermally challenged, check the thermal derating of those hot devices along with the actual thermal dissipation taking place in a worst case situation.
hope this helps.

 

[MetalworkerMike]

Maybe this is too obvious, or maybe it wouldn't work, but in the plastic industry we use what we call brass-shim thermocouples. You could have them made so that they were just large enough to cover the whole surface of the chip case. They can be made with brass or stainless shim stock, or other materials. Any reasonable size. The thermocouple would be sandwiched between the chip and the heat-sink (with some grease, of course). This should, as far as I can see, give you the true temperature of the case while not interfering unduly with heat transfer. You could also try cementing one to the underside of the chip (with suitable insulation to guard the pins and whatnot).
Just a thought.

Mike