| In this document, we consider the performance required of an optical-pulse source for use in ultrafast optically time-division-multiplexed communications systems. For such a source, the jitter is a critical parameter for achieving a desired bit-error rate.; We describe a pulse source, consisting of a gain-switched laser diode, a linear pulse compressor, an adiabatic soliton pulse compressor, and a pedestal suppressor. This source meets most of the established criteria. We also examine a means of reducing the jitter of a gain-switched laser diode, the phenomenon of temporal-multiplexing-induced amplitude jitter, and two techniques for characterizing any pulse train's jitter.; We experimentally compare gain-switching of a laser diode by an electrical pulse train (created by driving an impulse generator with a sinusoid) or directly by the sine wave. When the power of the waveform driving the laser is equal in both cases, pulse-train drive yields lower pulse width, timing jitter, and amplitude jitter. In contrast, when the power of the sinusoid is held constant, no advantages are found to result from pulse-train gain-switching.; We derive an expression for the power fluctuations of a temporally multiplexed pulse train in terms of its pedestal level; this expression quantifies how temporal multiplexing couples pedestal level to slow, correlated amplitude jitter. Experimental results (obtained with two pedestal-suppression devices) confirming the theory are presented.; Finally, we derive two methods of simultaneously extracting the uncorrelated and correlated contributions of the amplitude and timing jitter (and the correlation times of the correlated components) of a pulse train from radio-frequency spectra. The first technique extracts these contributions directly from the detected train's spectrum. The second extracts the jitter from two spectra obtained from mixing the detected train with sinusoids that are in phase and in quadrature phase. Simulations detailing the conditions under which these methods are accurate are provided. Since the methods' errors are systematic, we can compensate for them, enlarging the high-accuracy domain. Experiments confirming the accuracy of the first scheme are also included. |
