| Created: 10/24/95 |
Updated: 8/3/98 |
Integrated Thermal Analysis of Air Cooled Systems: Modeling and Measurements
Project Number : C96-14
Point of Contact:
|
Dr. Yogi Joshi
CALCE EPRC
email: joshi@calce.umd.edu
Phone: (301) 405-5528
Fax: (301) 314-9269 |
Objectives
To further develop combined system and board level thermal analysis capabilities
for air cooled electronic systems, through modeling and measurements.
Background
Traditionally, thermal analysis of electronic equipment has followed one
of the two approaches. Board level analyses involve the solution of a heat
conduction problem, with specified convection coefficients and reference
temperatures at the component surfaces. The convection coefficients and
reference temperatures to be used in such analyses are often unavailable
and resulting thermal predictions are prone to large errors. The second
approach is relatively recent and includes the use of computational fluid
dynamics/computational heat transfer (CFD/CHT) for the entire system. Due
to the large number of resulting grid points and the complexity of such
approach, excessive computing time is required for adequately detailed
simulations. The development of an integrated system and board level modeling
capability was initiated at CALCE EPRC during the previous year for natural
convection cooled and indirect air cooled systems to address these limitations.
This effort extends the integrated modeling methodology to mixed and forced
convection cooled systems and provides experimental validation of the methodology.
Work Accomplished
An air cooled enclosure was selected to demonstrate the approach.
The computational model considered the combined effects of natural convection
as well as forced flow. Using a coarse grid, a CFD/CHT type solution was
carried out. From this, the local thermal information near the printed
wiring board was extracted. This information was interpolated on a fine
grid for use in board and component level thermal analyses. Validation
was performed by taking temperature measurements in a direct air cooled
enclosure equipped with horizontally oriented boards, one of which is populated
with thermal test packages. The experiments were performed in a low speed
wind tunnel facility. The test board consisted of a 4 by 4 array of ceramic
modules on an epoxy-fiberglass substrate.
The thermal packages were powered at different levels and the resulting
junction temperatures were measured using temperature sensing silicon diodes.
In addition, thermocouples were mounted on the top center of each package
to measure component surface temperature. Calibration of the wind tunnel
and diodes was done. The experimental runs at two power levels and two
air velocities were performed, covering both forced and mixed convection
regions. The resulting component surface and junction temperatures were
recorded for model validation.
The electronic board with its enclosure as used in the experiment was
first modeled with the CFD system level solver using a coarse grid. A multi-layer
package model with effective thermal properties for each layer was used
here. Realistic velocity and temperature boundary conditions were applied
for the system level modeling. The power levels and air velocities were
chosen according to the experimental runs. After the convergence of system
level solution, thermal data were extracted and interpolated for component
level analysis. The results were presented in forms of component junction
temperature, component temperature distributions, and surface convection
coefficients.
