Manufacturers' electrical component Boundary Condition Independent (BCI) thermal resistances 2

[Tunalover]

Anyone out there know what component manufacturers provide the BCI resistance networks for their components rather than just the rough JEDEC Compact Thermal Model (CTM) two-resistance model consisting of only Θ[sub]JCtop[/sub] and Θ[sub]JB[/sub]? Also, I'm not really clear on how the BCIs are defined and determined so would love to find sources that do that.
TIA

H. Bruce Jackson
ElectroMechanical Product Development
UMD 1984
UCF 1993

 

[Tunalover]

I need to correct my post. I used the term CTM for a two-resistor model. The CTM as devised by DELPHI is actually a nine-resistor model.

H. Bruce Jackson
ElectroMechanical Product Development
UMD 1984
UCF 1993

 

[IRstuff]

I've never heard the term in reference to any components I've run across, and a Google search hit academic papers, suggesting that this approach is not widely used.

JEDEC has standards on how to generate the models, FWIW:

https://www.jedec.org/system/files/docs/JESD15-4.pdf

https://www.jedec.org/system/files/docs/JESD15-3.pdf

Note that they talk about a "high degree of boundary condition independence." The development of the DELPHI model seems awfully tedious, since it mandates a "validated" thermal model to begin with, and you're essentially looking to abstract that model down to a 9-component, or 2-component, model. While the 9-component model looks interesting, its utility seems to be limited, to me, simply because a board containing hundreds of components would require thousands of nodes using 9-component models. At this point, one might reasonably ask whether it's worth the bother, given that the granularity would seem to demand that the board itself have thousands of nodes, and the 9-component models all assume no proximity effects from adjacent devices, and is that really a plausible approach? And, at that point, you're essentially more than halfway to a full-blown FEA thermal model.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[Tunalover]

Thanks, IRStuff for your thoughtful insights.

Well the JEDEC standards have language implying that suppliers are/will provide the nine-resistor compact thermal model (CTM), the figure of merit W and objective function F, together which give the analyst all he needs for a given component. I will call JEDEC Tuesday, January 2nd to ask if this CTM is used and, if so, what suppliers provide that information. As for your implication that someone would implement CTM on "hundreds of components" in any system thermal model is pretty far-fetched. No one in their right mind would do that since it's most likely that only a dozen or two of those "hundreds of components" would have enough power dissipation and/or temperature sensitivity to worry about.


H. Bruce Jackson
ElectroMechanical Product Development
UMD 1984
UCF 1993

 

[tkordyban]

Back to the original question: what component suppliers provide Delphi-type resistor models for their component packages? The answer is, in my experience, not very many. It is a chicken/egg problem. It takes a relatively large effort to derive the multi-resistor models. At this time, nobody uses them in thermal analysis. Why? Because nobody supplies them and they are too hard for the typical engineer to derive. To make your own you need to have construction details of the inside of the package, and the ability to create a Finite Element Model of it. These construction details are very hard to obtain, since they are proprietary to the manufacturer. I am not sure how to prime the pump. If the models were freely available, more thermal engineers would use them, including me.

In the mean time, I limp along with 2-resistor model information, which is slightly more available. Most often there is only the useless Theta J-A.

The 2-resistor models are better than nothing. But they are definitely NOT independent of boundary conditions.

It is NOT impossible to solve models with tens of thousands of nodes and resistors. It might be tedious, and unnecessary, but not impossible. Aren't you using a computer to do this solving for you? You are already solving many thousands of equations. Resistors are pretty easy for the computer. That being said, the typical thermal engineer will only include a few important components in detail on a typical circuit board model. Most of the remaining parts dissipate no heat, and so there is no point in finding their junction temperatures. So the multi-resistor component models, when they are available, can be very useful. I have put them on my wish list for 2018 (again).

 

[Tunalover]

Thanks, Tony.

Actually, if you read my response to IRStuff, I don't advocate solving more than a dozen or so components using CTM models on even the busiest of CCAs, be they two-resistor or nine-resistor CTMs. Also, I am always one to "limp along" all the time because I am just a regular electronics packaging engineer who only puts on the thermal engineer's hat now and then to break the tedium. In fact, I would love to be doing even electronic packaging now because I've been out of work since August.

But, as I always do when I've been between jobs, I dig into things to increase my knowledge. This times its thermal stuff (last time it was shock and vibration stuff). I always wanted to specialize in thermal, but although I did five graduate courses in an MSME ThermoFluids program, I never managed to get that opportunity.

I had some LinkedIn messaging with Clemens Lasance recently and he piqued my interest in the DELPHI CTM. As for thermal analysis tools on my personal computer, I don't have any. But I know how to do network resistance modeling and CFD for electronics and normally don't waste my time trying to get an FEA to do the job.

I am calling JEDEC on the 2nd to ask if they have a list of suppliers who participate and collaborate with JEDEC and provide customers with the DELPHI CTM. If I find anything I'll be sure to "share the wealth".

BTW have you heard of the Mentor Graphics T3Ster test station for component thermal characterization (for use with FlowTherm and FloEFD)? If so, what do you think of that system?

H. Bruce Jackson
ElectroMechanical Product Development
UMD 1984
UCF 1993

 

[Tunalover]

Oh, I forgot to ask. Have you ever seen a need to "trust but verify" (ala Ronald Reagan) suppliers with their thermal resistance values? It seems there may be times when Marketing, Sales, or even Engineering Management at a supplier may help make the numbers look a little rosier than reality. Also, as you know there are many times when the supplier starts selling the components with data sheets and catalogs that conspicuously LACK that information. Do you ever send components out for determination of the two resistor values or even test them yourself in-house?

H. Bruce Jackson
ElectroMechanical Product Development
UMD 1984
UCF 1993