Thesis for the degree of Doctor of Philosophy Signaling for Optical Intensity Channels

With the growing popularity of social media services, e-commerce, and many other internet-based services, we are witnessing a rapid growth in the deployment of data centers and cloud computing platforms. As a result, the telecommunications industry has to continue providing additional network capacity to meet the increasing demand for bandwidth. The use of fiber-optic communications plays a key role in meeting this demand. Coherent optical transceivers improve spectral efficiency by allowing the use of multilevel in-phase and quadrature (I/Q) modulation formats, which encode information onto the optical carrier’s amplitude and phase. However, for short-haul optical links, using noncoherent optical transceivers, also known as intensity-modulated direct-detection (IM/DD) systems, is a more attractive low-cost approach. Since only the intensity of light can carry information, designing powerand spectrally-efficient modulation formats becomes challenging. Subcarrier modulation, a concept studied in wireless infrared communications, allows the use of I/Q modulation formats with IM/DD systems at the expense of power and spectral efficiency. This thesis addresses the problem of optimizing single-subcarrier modulation formats for noncoherent fiber and wireless optical communication systems in order to achieve a good trade-off between spectral efficiency, power efficiency, and cost/complexity. For the single-subcarrier three-dimensional signal space, denoted as raised-QAM in the literature, we propose a set of 4-, 8-, and 16-level modulation formats which are numerically optimized for average electrical, average optical, and peak power. In the absence of error-correcting codes, the optimized formats offer gains ranging from 0.6 to 3 dB compared to the best known formats. However, when error-correcting codes with performance near capacity are present, the obtained modulation formats offer gains ranging from 0.3 to 1 dB compared to previously known formats. In addition, laboratory experiments using the obtained 4and 8-ary modulation formats were carried out. The performance improvement over the previously known formats conforms with the theoretical results. To address transceiver complexity, a two-dimensional signal space for optical IM/DD systems is proposed. The resulting modulation formats have simpler modulator and demodulator structures than the three-dimensional formats. Their spectra have in general narrower main lobes but slower roll-off, which make them a good choice for single-wavelength optical systems. The three-dimensional formats are more suitable for wavelength-division multiplexing systems, where crosstalk between adjacent channels is important.