4 December, 2017
The development and use of compact component models or compact thermal models (CTMs) has been a key element in producing accurate thermal analysis and simulation of electronic components and assemblies, not only in the initial design of a system but also in the test and validation phase.
CFD (Computational Fluid Dynamics) based thermal analysis tools for electronics cooling have been around now for the best part of a quarter of a century, and the importance of thermal models became quickly apparent to users and the industry in general. The standard junction-to-ambient thermal resistance often quoted on component datasheets is not nearly good enough and does not really represent how heat is dispersed from a silicon chip’s package. On the other hand, a fully detailed thermal model (DTM) that replicates in detail the physical geometry of a chip package was - and still is - IP that semiconductor vendors do not wish to disclose. In addition to which, this is a model that would be impractical to solve by a CFD tool, especially for a board-level simulation involving hundreds of components if each package part was defined in detail.
This need for a better representation of component models led to the inception of the European Delphi project (DEvelopment of Libraries and PHysical models for an Integrated design environment) in the mid 1990’s. The Delphi mission, in conjunction with its successor SEED and PROFIT projects, was to create CTMs that comprised fewer elements than a DTM and did not disclose package IP, but offered an abstracted representation with a high degree of accuracy. In addition it should require significantly less computational time in thermal simulation. Importantly it also needed to be boundary condition independent: in fact, Delphi set a number of boundary conditions (initially 48 but revised down to 38 cases) using uniform heat transfer coefficients to predict temperature at a number of discrete points on a package such as at the junction, regardless of the environment and any forced-convection cooling system.
In 2008, the methodology for relatively simple two-resistor (2R) thermal models (junction-to-board and junction-to-case) and more complex resistance/capacitance-network compact models was absorbed into JEDEC standards (JESD15-3 and JESD15-4, respectively). These compact models have been shown to be highly accurate, often within 4 per cent of maximum temperature. In addition, the Delphi project also defines measurement calibration to enable engineers to validate compact thermal models.
However, as one can imagine, this does not mean all thermal analysis problems are now solved. For example, the widespread use of packaging technologies such multi-chip modules and also the development of Dynamic Compact Thermal Model (DCTM) for the simulation of transient thermal phenomena are current major issues for engineers working in thermal simulation.
A further issue, and a very important one, is the creation and distribution of compact models. Sometimes models of specific components are available in a numerical tabular form or occasionally in a proprietary file format for a particular software tool. What is required to drive our industry forward is a vendor neutral format to report and share compact models – this is an issue we strongly support. In fact, this initiative is currently going through a JEDEC standards committee, but results may take some time to come through.
6SigmaET supports a wide range of component modelling levels including: detailed geometry, two-resistor thermal models, compact resistance/capacitance network models (DCTM) and simple blocks. This allows users to select the appropriate modelling level for the type of analysis they are performing.
Unfortunately, few people are using compact models today and the industry needs increased commitment from component manufacturers to provide them, preferably in an industry-standard format which would accelerate and proliferate usage. This would bring significant benefits to all, as the increased adoption of compact component models would undoubtedly enhance the accuracy of thermal simulations across the electronics industry.
Tom Gregory, Product Manager