Electro-thermal Analysis Of Multilayered Boards And Substrate Integrated Waveguide Filters

With the widespread use of the system on package technology, thethermal environment of IC chps are getting more complex than ever. For mostapplications, the majority of heat generated by IC chips is conducted to printedcircuit boards through various packages. So the packages and boards areimportant factors that affect the thermal performance of IC chips. It is alwaysdesired to generate thermal model to characterize the heat removal capabilityof packages and boards. Further, substrates in system in package applicationsare more than interconnect layers and they always integrate high performancecompact passive components. Microwave passives are heat sourcesthemselves, and on the other hand, they also serve as heat dissipation paths forother components in the system. So the thermal analysis is getting morecomplex.On this purpose, we prososed a fast and efficient method in this paper toperform steady-state thermal analysis of multilayer boards and electro-thermalanalysis of substrate integrated waveguide (SIW) filters. The multilayerfinite-difference method for frequency-domain electrical analysis is migratedto obtain and slove thermal conductance network. With this method, thetemperature rise distribution of multilayer boards can be achieved efficiently.Further, the fast electro-thermal analysis of multilayer SIW filter is carried out,with its distributed heat source, i.e. electromagnetic losses, derived from thecoupled resonator circuit model. Finally, in order to validate the applicabilityof our method to high-efficiency thermal optimization, a compositeconfiguration is tested, which consists of a SIW filter with a self-heating chipattached on it. Both the optimal temperature rise distribution and the locateon-dependent maximum temperature rise are obtained. Our idea hasbeen demonstrated by the good agreement between the results simulated withthe proposed method and the commercial finite element method (FEM)software. Especially, the worst-case error of our method is5.3%while the runtime is significantly reduced by95%, when compared with the FEMsimulators. The thermal equivalent model derived here is a semi-compactmodel, which is both computational efficient and flexible in differentstructures. To further extend the method in solving thermal transient problems,we incorporated the method with Krylov subspace transform, which can helpreduce the order of the model, and, thanks to the semi-compact modelgenerated, the accuracy is also guaranteed.