Optimizing Low-Frequency Mode Stirring Performance Using Principal Component Analysis

We formulate and perform principal component analysis (PCA) of mechanically stirred fields, based on tuner sweep data collected across a frequency band at a single location of a sensor inside a reverberation chamber. Both covariance- and correlation-based PCA in undermoded and overmoded regime are performed and intercompared. The nonstationarity of the stir performance as a function of the angular position of the stirrer is demonstrated. It is shown that this nonuniformity can be quantified and exploited to select a set of optimal stir angles. The rotated principal components are found to be interpretable as energy stirred by specific angular sectors of the stirrer and are related to the correlation structure of the data. The analysis leads to the concept of eigen-stirrings (stir modes), which form an orthonormal set of empirical basis functions for expanding stir data.