In a previous life, I worked as a research technician for a microwave radio manufacturer that specialized in providing data connections for hard-to-reach areas. One of the things it bought home to me was how much of a difference even a single phone line to a village could make. And this was pre-dot-bomb, so imagine how much of a difference a reliable, fast Internet connection could make now.
Unfortunately, a single 2Mbps microwave channel doesn't really cut it anymore, especially when that connection is shared between individual householders, schools, and local government. In the end, the answer is that someone is going to have to lay some fiber. The problem is that it's not cheap. One of the most expensive parts of short-haul (~60-mile) networks is the set of diode lasers that generate the different wavelengths used to carry the traffic. A recent Optics Express paper takes a new approach that replaces many diode lasers with a single amplified spontaneous emission source. The result? A relatively inexpensive 1.25Gbps link.
Let's step back a bit and talk about what makes these lasers so expensive. A standard optical network consists of an optical fiber carrying many different wavelengths of light. Each of these light streams might be modulated at 10Gbps, but high capacity is obtained by carrying many of these lower bit-rate streams. To do this requires that each laser diode emits light at a fairly precisely defined wavelength.
When left to their own devices, diode lasers will emit over a fairly broad range of wavelengths. Manufacturers of telecom diodes build structures into the diodes that provide a huge amount of feedback for a predefined wavelength and very little for others. This results in a laser with a very precise wavelength and a very high price tag (relatively speaking).
The odd thing is that in long-haul networks, this isn't good enough, because the signal must be periodically amplified. The amplifiers don't just amplify the signal, they amplify the noise as well. Left alone, light in between—and in—each channel builds up and eventually you end up with a mess. To reduce this, filters are put before and after each amplification stage, reducing the amount of light sitting in between channels and slowing the build-up of noise. The important thing is that these filters are nearly as precisely defined as the original laser diodes, but are much much cheaper.
A group of researchers from Taiwan took note of this and thought that filters combined with an amplified spontaneous emission source might replace several expensive laser diodes. Here's how an amplified spontaneous emission source works. Imagine you have a laser crystal which you have just excited, so there is a whole bunch of ions just sitting there waiting to emit. One of them does emit, and, as that photon passes through the crystal, it stimulates other photons to emit. This is the first part of what makes a laser: amplification. But, in a laser you also have feedback—mirrors reflect the amplified spontaneously emitted light back to stimulate more emission. Without the mirrors, the crystal just glows weakly and everyone goes off to buy a fluorescent light.
If we replace the crystal with an optical fiber, a certain fraction of the spontaneously emitted light is guided by the fiber. As this light travels along the fiber, it becomes amplified by stimulating other excited ions to emit. In the end, you can end up with hundreds of watts of power being emitted from an optical fiber, but it is not laser light. (That's because measuring the properties of the light in one instant of time tells you nothing about what the light will be like at some later point in time—there is no feedback to allow light emitted earlier to stimulate the emission of light later.)
On the end of the active fiber, the researchers placed a filter. They ended up with a comb of wavelengths that happen to correspond to exactly what is expected on telecom networks. One problem: laser diodes occupy a tiny fraction of the filter bandwidth and more of the bandwidth is occupied by adding information to the signal. In this case, however, the source completely fills the bandwidth. Even so, the researchers demonstrated links operating at 1.25Gbps.
However, because of the nature of the source, only every second channel can be used, so in terms of link capacity, this sits at half of that of a traditional fiber network. But balanced against what we gain—the same source can be used for every village, and higher-capacity networks cost nearly the same as low-capacity networks—I think they are onto a winner, even in the developed world.
Compared to the microwave solutions I worked on, and traditional fiber networks, this is much cheaper—depending on the terrain. But for a 60-mile ring network over relatively easy ground, it would work out to be cheaper still. This is perfect for those situations where power is available, but the phone lines absolutely stink, like parts of rural India, southeast Asia, and Africa.
Optics Express, 2009, DOI: 10.1364/OE.17.022246
reader comments
19