Optical signal processing

I decided to see a real physics talk - a colloquium about optics and experiments.

Kelvin Wagner from University of Colorado at Boulder was speaking about optical signal processing. While Eli Yablonovitch complained about the end of the semiconductor roadmap and the disaster that it will bring to the society, Kelvin Wagner was celebrating it.

In fact, the digital processing and electronics is a competitor of optical signal processing. The people in electronics are getting much more money, and it is only because they are cheating, Wagner was explaining: for example, they always use a fuzzy picture of the Los Angeles area obtained by optical signal processing if they want to show that their digital methods are superior.

What is optical signal processing all about? Wagner chose some old-fashioned optical computers as his first example. They were able to factorize the Mersenne number 2^{79}-1 decades ago, by a sophisticated combination of gears, holes, and light rays - this was about 1 million times faster than any other method available at that time. Well, today we have to look (digitally) for larger Mersenne numbers in order to find greater primes.

Wagner also showed the Fourier transform of various letters and argued that the methods of optical signal processing are useful for pattern recognition - and for some tasks, they are superior over the digital techniques.

This field of physics uses various two-level atomic systems. If the splitting is fine (e.g. hyperfine structure), the system is suited for long-term memory purposes. For example, Wagner argued that a system may be used to store 10^{14} bits in a piece of crystal (the memory decays after one day or so). On the other hand, other atomic systems with bigger energy splittings are appropriate for computing, fast response, and the application to wideband optical processing is the natural one. One is expected to use these methods in the religion, for example, to localize the signals from the ET aliens ("SETI").

One of the key methods he discussed was the so-called spectral holography. This concept is analogous to the usual, spatial holography (which itself is analogous to holography in quantum gravity). Instead of the interference pattern as a function of the position on the screen, which occurs in spatial holography, spectral holography creates "interference patterns" in the space of frequencies - in other words, it is about modulation. Such modulation may be obtained by superposing two signals that are shifted in time.

They are working to increase the bandwidth, the resolution - in order to solve massively parallel processing tasks - and he expects that unless electronics will cost 100% of the GDP, the optical signal processing over may take over around 2020, at least in some applications. I have not understood most technical details of the talk well enough to illuminate them.