Experimental Validation Methods for Thermal Models

W Temmerman, WNelemans, T. Goossens, E. Lauwers
Alcatel Bell Telphone
F. Wellesplein 1,
2018 Antwerpen Belgium
C Lacaze
Alcatel Espace
26 Av. J.F. Champollion
31037 Toulouse, France

Abstract
The present paper evaluates the most important experimental thermal characterisation techniques for electronic components. Weak points are indicated in the .standardised methods for the still air, cold plate and fluM bath environment. Improvements have been developed: Suitable measures determine well-controlled surface temperature profiles over the whole component surface in the double cold plate. Two impinging jets impose reproducible heat transfer coefficients in the fluicl bath method. Hence, the  reproducibility, the scope and the accuracy of the measurements are enhanced.
1 Introduction
1.1 Thermal Modelling
Due to the impact of ever progressing integration and speed in electronic systems, which is not completely offset by the reduction in power dissipation per function, the heat load per unit area or volume continues to grow, reaching critical values in many application areas. This has brought thermal problems from an afterthought to an early system design issue.
Consequently, recent years have seen a sharp rise in the use of enclosure-level and PCB-level thermal analysis software. Enhanced computer power allows calculation of local air temperatures and heat transfer coefficients. If the analysis packages are supplied with good thermal models of the components then it becomes possible to calculate the junction temperature with sufficient accuracy to serve as an input for later reliability analyses. The PCB-level thermal analysis packages all contain thermal:
models of components, of varying degrees of sophistication, ranging from a singie thermal resistor using the manufacturers value of R~c through to quite complex thermal resistor networks. However, the accurate prediction of the temperature of critical electronic
parts is still seriously hampered by the lack of standardized, reliable, input data:
two different networks for the same part cannot be expected to produce the same results. The conclusion is that both enclosure- and PCB-level analysis require correct thermal models from a component database, according to an internationally-agreed format.
This paper discusses model validation as part of a structured approach to develop libraries of models for thermal simulation. It is based on work done in DELPHI, a CEC-sponsored project under the ESPRIT III Programme. For further details, we refer l,o [1]. Empirical results, specific simulations and some general project results are fi~'ed to compare and enhance existing measurement methods. An actual systematic comparison between the simulations and the experiments is presented in [2].

1.2 DELPHI
1.3 Thermal resistance concept
1.4 Device under test (DUT)
1.5 Calibration

2 Model validation requirements
3 Measurement methodologies
3.1 Fluid bath (FB)
3.2 Natural convection
3.3 Cold plate
3.4 Double cold plate (DCP)

4 Conclusions
The single cold plate is not an accurate calibration environment. All others are, but the oven typically requires more than 6 times longer settling times than the rest of them. The settling times for FB (SDJI) and DCP are determined by the cooling aggregate : high capacity systems need more time.
The DCP with both plates at the same temperature promises the best overall performance. A range of well-defined boundary conditions for verification of the corresponding conduction-only simulations is available. SDJI avoids the interface resistances which may trouble cold plate measurements, but it is not trivial to realize the low HTCs in this test. The following table summarizes the performance of the investigated measurement methods with respect to the aforementioned requirements.
5 Acknowledgements
Th.e authors acknowledge the support for this work by the Commission of the European Communities under the ESPRIT contract DELPHI (9197). They also wish .to thank the other members of the project: P. Zemlianoy from Alcatel Espace, "H.I.
Rosten and J. Parry from Flomerics, C. Cahill from NMRC, C. Lasance from Philips CFT and T. Gautier and Y. Assouad from Thomson CSF. At IMEC a lot of ~ork was done on fluid bath tests, especially by F. Christiaens and E. Beyne. .'FLOTHERM is a registered trademark of Flomerics Limited.
6 References

熱設計論文下載：Experimental-Validation-Methods-for-Thermal-Models.pdf

標簽： 點擊： 評論: