Abstracts Computer Science

Noise characterization of devices for optical computing

by Samuel Mark Storrs

Optical computing has proven useful in several specialized applications. However, the general development of optical computing is limited by the computational inaccuracies typically found in optical information processing. One source of these inaccuracies is the noise generation and transformation properties of optical devices. The signal and noise models of optical devices are inadequate for information processing system development and optimization. This is specifically true for spatial light modulators. Progress in optical computing requires more accurate models of these optical devices and their signal and noise characteristics. System modeling of optical devices is complicated by the multiple-port and nonlinear nature of those devices. In this dissertation, the Volterra series model is shown to adequately represent a multiple-port, mildly nonlinear device and is adapted to represent a spatial light modulator. Furthermore, a method is described to experimentally measure the Volterra series model characteristics directly. We develop a three-input, single-output, second-order Volterra series model for a spatial light modulator and describe an experimental method to measure the nonlinear transfer functions using sinusoidal perturbation and synchronous detection with a lock-in amplifier. We review the spectral noise transformation properties of Volterra systems and adapt these results to the multiple-input case. The spectral noise transformation characteristics of optical devices are shown to be calculated directly from the Volterra system model using the nonlinear transfer functions. These techniques are applied to experimentally measure the Volterra series model characteristics of a Hughes 4050 liquid crystal light valve and estimate closed-form expressions for the nonlinear transfer functions. The validity of the model characteristics and spectral noise transformation is demonstrated by comparing predicted spectral noise characteristics with measured device performance.