Development of a New Electro-thermal Simulation Tool for RF circuits

28 2.4. Compact Thermal Model
the creation of a 3D thermal model, based on the detailed physical structure is a
cumbersome task.
Ideal CTM should be (1) accurate, (2) able to manage multiple heat sources and cooling
surfaces, (3) taking nonlinear material properties into account, and (4) Boundary
Condition Independent (BCI). Nevertheless, CTM structure has to remain simple, and
easy to extract. Existing methodologies usually focus on a few of abovementioned
qualities; otherwise, the resulting CTM would become too complex. Several methods
have been already developed to provide accurate CTMs. One of them is the Delphi
method [43, 44], which is able to give accurate static BCI CTMs for single-chip
electronic devices and systems. An important drawback of Delphi, is the amount of
3D simulations [41]. In addition, dynamic models are difficult to extract. Delphi CTMs
with more than two heat sources result in a very dense and excessively complex resistor
network, where the number of 3D simulations or measurements becomes exorbitant.
On the other hand, [45, 46] describe "traditional" CTM generation methods, with the
way to make the thermal coupling between heat sources in static and dynamic modes.
However in these methodologies, the CTMs are BCI only for systems with one cooling
surface.
Beside that, the device manufacturers may not wish to publish the confidential data
about the dimensions, materials and technology. However from the customer point
of view, the knowledge of a thermal behaviour might be necessary. This problem
has been noticed by the semiconductor industry [39, 40, 41, 42]. As the result, the
standard equivalent thermal networks libraries have been developed, which are in use
with standard CAD software. However, the drawback of this approach is the amount of
3D numerical simulations to be performed, in order to produce the CTM parameters.
A new CTM methodology will be demonstrated, which offers reduced amount of
3D simulations. The model has been under investigation during the PhD stage in
LAAS-CNRS laboratory, Toulouse, France. Based on existing methods, the approach
gives innovative solutions, in order to improve known CTM generation procedures.
Firstly, a BCI CTM extraction method for systems with multiple cooling surfaces will
be demonstrated. The CTM is an evolution of the star thermal network. Secondly,
a new method conceived especially for multi-chip devices, i.e., multiple coupled heat
sources [47, 48] will be presented. The thermal coupling is based on a definition of an
Optimal Thermal Coupling Point (OTCP) between heat sources [49].
2.4.2. Methodology for Multi-Cooling Surface Structures
The star thermal network is a simple representation, which allows dealing easily as
well with static as dynamic multiple cooling surface problems. In typical configuration
the model (Fig. 2.17) has three cooling surfaces. Each one is represented with thermal
resistance: Rth_top, Rth_bottom and Rth_side. The main drawback of the star model is with
boundary conditions (Rh_top, Rh_bottom and Rh_side) change, the model becomes not
valid. . The presented methodology is an extension of the star model representation,
modified in order to deal with changing boundary conditions (BCI). The detailed