Analysis of flow distribution in a power supply using flow network modeling (FNM)

This paper presents an analysis of flow distribution in a new-generation power supply used in telecommunication applications. The power supply involves several boards, each containing several magnetic components, EMI screens, and heat sinks, arranged in a rack. The flow in each passage is driven by four suitably placed fans. The study utilizes the technique of Flow Network Modeling (FNM) for efficient prediction of the flow distribution in the power supply system. The novelty of the study lies in the use of a two-level flow network approach to achieve modularity and gain further efficiencies in the analysis of flow distribution. A flow network model of a passage formed by two successive boards is first constructed to determine the characteristics that describe the behavior of the flow in and out of an individual power supply. Two types of flow characteristics, namely resistive and pumping, are outlined and used for constructing these equivalent compact characteristics of the power supply. Network model for the overall system is then constructed by arranging the compact models for each passage in a manifold arrangement corresponding to the physical arrangement in the rack. Analysis is carried out for three placements of cabinets, namely a isolated cabinet, two cabinets side by side, and three or more cabinets side by side. Results of analysis show that for a single cabinet, the flow is uniformly distributed among the passages because of the presence of side vents. However, with multiple cabinets plated side-by-side, the forward flow streams from individual passages are forced to merge in the header passage at the back of the cabinet. The resulting combining manifold gives rise to a maldistribution of flow among the passages with lower flow rate in the bottom passages. The two-level network modeling approach outlined in this study is very efficient in terms of the time required for model definition, analysis, and examination of re-suits. The modularity of the approach is well suited for a top-down design and packaging of large scale electronics systems for achieving the desired thermal performance.