for the last few months, we have been talking quite a bit about jedec characterization and compact modeling. let me digress a little now and talk about building quality detailed models.

recall that a detailed model is a model that attempts to represent or reconstruct the physical geometry of a package to the extent feasible. thus the detailed model will physically always look similar to the actual package geometry.

a properly constructed detailed model, once validated (and there are debates about what this exactly means), can be expected to reliably predict the junction, case, lead or any other temperature within the package body to an acceptable accuracy, say 10%.

before you build your detailed model

the key is the phrase "properly constructed". because we are working with a diverse range of package families designed for a broad range of applications, the construction of a "good" detailed model still has an element of art. simply trying to mimic the geometry to every excruciating detail will not always do. one has to ask questions such as:

• do i need to build a detailed model?

sometimes you are quite certain that a given component is going to satisfy its thermal constraint. this could be due to past experience with a very similar design, or relevant test data from the supplier. in such a case building a detailed model may not be worth the effort. a cuboid block with some lumped conductivity value may sufficiently represent the air blockage and lumped heat generation.

another reason why a detailed model may not be necessary is if there exists a validated compact model. a boundary condition independent (bci) model, such as one created using the methodology from the delphi project is ideal, but a simpler, two-resistor model can still give a good enough accuracy for an early design stage.

two-resistor models can now also be generated in minutes using commercial web-based software tools, and users should take full advantage of these methods. due to computational constraints, using a compact model is the preferred method in a system or even a board-level analysis (see below).

however, there is one situation in which building a detailed component model is absolutely essential. this is the case when you are a component supplier, or are otherwise interested in doing package characterization or design.

• what do i want out of my model?

if air temperature is all you need, then often a lumped cuboid block-type model can do (though, not always). this is especially true if the internal thermal resistance of the package is significantly less than the external thermal resistance of the environment around it.

in that case, the lumped cuboid model will predict the heat flux partitioning between the top and bottom paths reasonably well, which is what is required to get the air temperature right.

however, if you are looking for an accurate value of the junction or case temperature, and you do not have a validated compact model at hand, it is likely that you will need to build a detailed model.

• do i have enough computational power to run my detailed model?

this could be the biggest roadblock in using a detailed model in your simulation, especially if you are using a computational fluid dynamics (cfd) tool. cfd tools are computer intensive, and overnight runs are common for full system designs.

using a detailed model in a large problem can really stretch your computational resources. usually using more than a couple of detailed models in a system-level analysis is almost impossible.

one way to get around this is to perform what is known as a "zoom-in" type analysis, in which the full system level model is solved first. at this level, the components are modeled as smeared heat sources or lumped cuboids. then a second analysis is done of the region around the component of interest. at this level, the component can be modeled in detail without great difficulty.

• is my model compatible with my analysis tool?

as we saw above, it is often difficult to use detailed models in system-level analysis using cfd. however, if you are using a conduction/radiation-only type analysis tool, then modeling the component in detail can be quite manageable.

after you have decided to build your detailed model

well, let us suppose that you have asked yourself all of the above questions, and have decided that you do indeed want to build a detailed model of your part. what next? there are many things to consider. modeling is as much an art as a science, and two models built using different approaches, could be equally acceptable. therefore there are no perfect recipes for building the ideal detailed model.

that having been said, there are indeed broad do's and don'ts we should pay attention to. we will talk about this more next month.

about sarang shidore:

sarang shidore obtained engineering degrees from iit madras (india), texas a & m university (college station), and university of texas (austin). he worked at flomerics inc. in various roles in engineering and product management with a special focus on package-level thermal modeling and analysis, a field in which he has authored several papers and articles.

in addition, he worked for mentor graphics as product marketing manager and for several years as a consultant for various organizations. he is currently a visiting scholar at the lbj school of public affairs at the university of texas focused on energy and climate policy and future strategies.