Empirical modelling of electromagnetic acoustic transducer data

We characterize the observed phase and amplitude of acoustic birefringence data collected with an electromagnetic acoustic transducer. Our characteriz ation models are extended versions of an idealized model for acoustic biref ringence. In the extended models, angle-dependent terms account for observe d variability in phase and amplitude which is not predicted by the idealize d theory. Possible sources of this extra variability include material inhom ogeneity and angle-dependent sensor gain. The adjustable parameters in the model are determined by minimizing the sum of the squared phase residuals p lus the sum of the squared amplitude residuals. To facilitate convergence, we determine the model parameters by fitting the extended models sequential ly according to model complexity. Ten experimental data sets were collected from the same sample. We estimate the mean value of each model parameter a nd its associated standard error. For each extended model, we estimate the mean phase delay between the fast and slow modes. We also estimate the mean rotation angle of the pure-mode polarization directions (relative to a ref erence coordinate system in the specimen). From run to run, we observed pha se data drift. However, the difference between the phases of the slow and f ast modes did not follow this drift.