The performance of free-space optical communication systems is impaired by atmospheric fog where the droplet size is on the order of the carrier wavelength and multiple-scattering effects are significant. Our research considers methods of estimating the atmospheric channel transfer function and impulse response, to design receiver equalizers which correct signal distortion. This dissertation research continues work done by Akira Ishimru, Yasuo Kuga, Richard Cheung, Sermsak Jaruwatanadilok, and Uruchada Ketprom, and others at the University of Washington Department of Electrical Engineering.

We approach the inverse problem of determining the channel given the transmitted and received signals by using a parametric physical model based on the modified pulse-vector frequency-domain radiative transfer equation. Details of the radiative transfer method formulation and the main parameters of the model are discussed, including: the aerosol size-distribution function, the particle refractive indices, and the optical depth. Several different fog distribution models are evaluated.

We present a dual-wavelength method for estimating the channel impulse response function with M-ary pulse amplitude modulation. We refine the method to use a single wavelength for the on-off keyed non-return-to zero unipolar signal, and we develop performance optimizations based on numerical curve fitting and optimization of fit parameters. We consider the radiative transfer model results for the channel transfer function and examine the behavior of solutions in both the time- and frequency-domain.

We compare our radiative transfer model against field test data which we gathered over 2009-2010 at Point Loma, San Diego, where heavy atmospheric fog conditions are favorable. We present an overview of the field data acquired using a custom free-space optical communications system which we developed. We then describe the statistical methods used for estimation of the atmospheric channel transfer function and related analysis quantities from the data. We compare the radiative transfer model results and the channel transfer function estimates from the experimental data, and evaluate their agreement and limitations.