Semiconductor Optical Amplifier

With the development in the late 1970s of single-mode fibers, whose chromatic dispersion was considerably lower than the modal dispersion of the multimode fibers used until then, it was attenuation that had become the main limitation to the range of terrestrial or submarine systems: simple amplification, without the need for signal reshaping, could considerably simplify repeaters. The concept of direct optical amplification therefore seemed a viable alternative to complex optoelectronic repeaters, and the first experimental work on solid-state optical amplifiers (SOA) was published in the early 1980s. But it wasn't until the late 1980s that research into direct optical amplifiers really took off. Until then, optical transmission systems had only used a single carrier. However, to meet the ever-increasing demand for long-distance transmission capacity, it seemed wise to juxtapose several carriers modulated in the fiber, using spectral multiplexing to exploit the phenomenal bandwidth of optical fibers (several tens of terahertz). However, the need to periodically amplify this multiplex of optical channels with optoelectronic repeaters poses a huge cost problem. Since these repeaters can only handle one channel at a time, it was essential to install as many repeaters in each regeneration station as there were channels in the multiplex. By contrast, a single direct optical amplifier operating on the principle of stimulated emission, with a fluorescence linewidth of a few terahertz, would enable the entire multiplex of optical channels to be amplified simultaneously.

It was this prospect that accelerated research into optical amplifiers. Two different direct optical amplifier techniques were studied and developed in the late 1980s: the SOA, the subject of this article, and the erbium-doped fiber amplifier (EDFA). The primary objective was to develop an in-line amplifier capable of amplifying several channels simultaneously. Quite quickly, it was the EDFA that showed the best performance on key points: total independence from the signal's polarization state, due to the fiber's constituent geometry, a low noise figure, negligible in-line insertion losses, and total absence of signal distortion for modulation frequencies above MHz.

As for the SOA, despite numerous technological efforts to improve its sensitivity to polarization, this is still currently of the order of 1 dB, too high a value to consider cascading numerous components in a very long-distance system. What's more, the SOA distorts the signal as soon as its power reaches the saturation threshold, which is relatively low for the application in question. However, these non-linearities can be exploited to perform signal processing functions (mentioned later) that EDFA does not allow. In addition, by modifying the...