System Level Cumulative EM Radiation Prediction From Stacked Machines in Data Centers

This work proposes a methodology for predicting the cumulative electromagnetic radiation from multiple stacked sources, starting from the knowledge of the radiation from the elementary sources. Nowadays, the capability of predicting the total radiated emission level coming from stacked trays within a data center rack system is becoming more and more appealing for system designers and this is one of the possible application scenarios in which the proposed methodology perfectly fits. The developed strategy is based on the knowledge of the radiation characteristics of each single source. Each tray, considered as a single EM source, can be individually measured inside an anechoic chamber. This is a simpler process compared to the measurements of a more complex, larger, and completely assembled system. The radiated E-field from each source is used as the objective function of a genetic algorithm (GA) developed for this specific purpose to find the spherical wave expansion (SWE) coefficients whose associated radiated field best matches with the reference measured data. The SWE coefficients of all single sources are summed to obtain the spectral representation of the entire system. The developed method is validated by full-wave simulations based on simplified tray sources, although the radiation pattern of the full-wave model is very similar to the one of a real tray source. An experimental validation is also carried out based on three different tray machines. Both the full-wave and the experimental validations demonstrate the feasibility of the proposed methodology.