Evaluation of Airflow Prediction Methods in Compact Electronic Enclosures

Rebecca Biswas, Raghu B. Agarwal
San Jose State University
San Jose, CA 95192
Ph: 408-295-9237

Avijit Goswami, Vivek Mansingh
Applied Thermal Technologies Inc.
3255 Kifer Road
Santa Clara, CA 95051
Ph: 408-522-8730
 

Abstract
During the design of forced convection cooled electronic enclosures, one of the most important parameter needed is the airflow through the enclosure. The airflow through the enclosure mainly depends upon the pressure drop in the enclosure and fan characteristics. Fan curves are often used in conjunction with the system pressure drop (impedance) characteristics to determine the airflow. The accuracy of the computational fluid dynamics tools depends mainly on accurately modeling pressure loss in the system (grilles, filters, etc.) and the accuracy of the fan curve data. These fan curves, which show the air delivery capacity of the fan at various pressure drops, are usually generated with no obstructions close to the fan. However, most modern electronic systems contain densely packaged components including airflow obstructions such as inlet and outlet grilles in close proximity to the fan. It is therefore possible that methods that use fan curves can often be inaccurate for predicting the airflow. Inaccuracies can also occur by using pressure loss data of grilles from engineering handbooks. The main objective of this study is to understand the accuracy of different methods of airflow prediction that rely on pressure loss and fan curve data
when compared to experimental results obtained in a wind tunnel. The system used in this study is chosen to be representative of typical electronic systems, which include the major components such as fans, inlet and outlet grilles and an array of stacked printed circuit boards (PCB). Additional components such as capacitors, inductors, transformers and heat sinks are also included to increase the total pressure drop in the system. Variations to the base configuration are made by changing grille open area, fan size and using fans in series and parallel configurations. It is found that differences of up to 20% can occur depending on the method used to calculate the flow relative to the experimental method.
Key words: Fan curves, IcePak, wind tunnel, pressure loss coefficient, system pressure drop (Impedance) characteristics, CFD, series, parallel.

Introduction
With microelectronics technology steadily packing more chip power into small packages, cooling issues are becoming increasingly critical in the design process. Most modern electronic systems containing densely packaged components uses fans or blowers for forced air-cooling. Designing forced convection electronic systems requires one or more reliable ways to predict the airflow and pressure drop characteristics of the system. Although experiments continue to be important, especially when the flow is very complex, the trend is clearly towards greater reliance on computer based predictions in design because of cost effectiveness.
Most airflow prediction methods use one or both of the following: fan curves obtained from the manufacturer and pressure loss data of grilles and filters obtained from engineering handbooks. Before these methods can be used confidently some kind of validation needs to be done to verify whether the fan curves and the pressure loss coefficients of the grilles are estimated accurately. Biswas, Agarwal, Goswami with a simp lified enclosure, which included only the
major components as grilles, fan and stack of pcbs.
The study showed that even in a simple model, differences of up to 20% can occur relative to the experimental results.
Mansingh and Misegades (1990) [2] used iterative technique with another CFD package called FIDAP to determine the operating flow and pressure drop through a computer system-processing unit.
Numerically stimulated particle traces were recorded using video equipment. The particle traces showed some extremely interesting flow characteristics.
Although they reported good agreement between measured and calculated pressures, the process required twelve iterations to reach a flow value within 10% of actual. This required ten hours of run time on a Cray Y-MP supercomputer. Additionally,1.5 man-months were invested in creating the computer model. The authors indicated that qualitatively accurate flow paths were calculated A problem of similar complexity could be run on modern high-end workstations within a couple of hours.

.....

Conclusions:
In compact electronic enclosures the pressure loss due to the presence of the inlet and outlet grilles accounts for a substantial part of the total system pressure loss. For higher accuracy, the pressure loss coefficient should preferably be measured experimentally instead of using the values from data handbooks. The fan curve obtained from the manufacturer should be used with caution when predicting the airflow in a compact, densely-packed system, especially if the airflow is closely ducted around the fan. This study also confirms that CFD is a very useful and accurate tool in the cooling design of electronic systems.

熱設計論文下載：Evaluation-of-Airflow-Prediction-Methods.pdf

標簽： 點擊： 評論: