The customary one-dimensional thermal resistance model used for single chip modules is not well suited for use with multichip modules. In this paper, it is proposed to model a multichip module (MCM) using a two matrix approach. An external matrix of influence coefficients is used in conjunction with cap temperatures measured above each chip site to determine individual chip powers within the MCM. These chip powers are then used with a matrix of internal influence coefficients to determine chip temperature rise relative to the cap temperature. The proposed matrices are determined using a “thermal test module” or “computational model” where chip power, chip temperature, and cap temperature can be explicitly measured. This dual matrix model is then available for use with production modules to determine chip powers and temperatures from a measurement of cap temperatures above each chip-site. The proposed model is demonstrated for a four chip MCM using data from numerical simulations. A sample of test runs using the model predicts the chip powers and temperatures to within 3.5 and 5.5 percent, respectively, of the values from the numerical model. Module physical properties are also parametrically varied to determine their effect on matrix elements. The matrix model is also proposed as a design tool to augment more numerically intensive methods of thermal analysis.

1.
Bar-Cohen
A.
,
Elperin
T.
, and
Eliasi
R.
,
1989
, “
Θjc Characterization of Chip Packages—Limitations, and Future
,”
IEEE Trans. Comp., Hybrids, Manuf. Technol.
, Vol.
12
, No.
4
, pp.
724
731
.
2.
Biskeborn, R. G., Horvath, J. L., and Hultmark, E. B., 1984, “Integral Cap Heat Sink Assembly For IBM 4381 Processor,” Proceedings, International Electronics Packaging Society Conference, pp. 468–474.
3.
Kraus, A. D., and Bar-Cohen, A., 1983, Thermal Analysis and Control of Electronic Equipment, Hemisphere Publishing Corporation, Bristol, PA, pp. 558–563.
This content is only available via PDF.
You do not currently have access to this content.