And then there was light.

AND THEN THERE WAS LIGHT

The communications industry turned on the lights--lightwave, that is-- fewer than 10 years ago. Also known as fiber optics, the technology (even in today's most advanced applications) is still fairly primitive compared to what we can expect from it in the future. In fact, we're still probing the possibilities.

Several years ago, researchers at AT&T Bell Laboratories rigged up 10 lasers, each producing a signal of 2 billion bits (2 gigabits) per second, resulting in a multiplexed signal of 20 gigabits per second carried without errors over 42 miles of single-mode optical fiber. That 20 Gb/s multiplexed signal is capable of carrying enough information to supply each of 10,000 customers with 10 ISDN signals, or the equivalent of 20 private lines to each customer.

That is still less than 1% of the theoretical potential capacity of fiber optics technology.

Limitless Horizon

Astonishing as such exhibitions can be, they reflect the applications of the past. Researchers were intent on proving that "fiber can do anything copper can do, better." The ability to transmit thousands of channels on a finger-thick cable has a lot of appeal, especially in communications systems that share ducts already crowded with othe cables.

Fiber's immunity to stray electrical fields is another great advantage, since it eliminates electrical "noise" as a disrupter of signal clarity. Another benefit is that lightwave signals are not easily intercepted by unauthorized people.

It is not clear that sometime soon, photonic systems will be a lot more than just glass pipelines for huge flows of data. Photonic beams in free space (even if the distance is measured in micrometers) can pass through each other without interference. This enables switching systems and digital access and cross-connect systems (DACS) to transfer signals from channel to channel without having to convert them from light pulses to electrons and back again.

Electronic switching systems--and computers--might see their dominance start to fade within this decade.

These developments have not been unexpected. Like the vacuum tubes that preceded the transistor, electronic systems have theoretical performance limits, and they're getting closer at the time in these days of increasing demands from network users.

Exciting Research

During 1990, two important Bell Labs innovations inv olving optical communications were demonstrated for the press. In January, Alan Huang unveiled the world's first optical digital processor> and in April, Scott Hinton showed the hardware required for photonic switching fabrics based on free-space digital optics.

Alan Huang likes to compare electronic computing to optical computing by drawing an analogy to Manhattan.

When cars and people try to cross the waters, they have only a few crossing points choke up quickly at rush hours, slowing traffic to a crawl. If you think of the people and cars as data, the same thing happens on a chip that is connected by relatively few wires to its surrounding circuitry.

In optical computing, this traffic congestion is eliminated by creating high-capacity light paths that no longer need a phsical carrier (such as fiber) to move from one switching or storage medium to another. It's as if traffic on any street in Manhattan could simply flow across the water to the mainland, without having to use a bridge or a tunnel.

Today's electronic computers, including supercomputers, usually are able to handle the workloads imposed on them, but most network administrators are aware that the computers capacities are sometimes strained. Some terminals show the load imposed on a computer by a task, sometimes displaying a bar so long that you cannot almost hear the computer grunting and groaning.

These peak load demands reduce the processing power that can be allocated to maintaining a network. Optical computers may be able to process more than a thousand times as much information as their electronic counterparts. But not right now.

Keep in mind that the Wright brothers' first airplane (to which the Huang optical processor can be compared) carried only one passenger. As Huang sees it, this radically new technology will make its first impact in some of the components used in electronic computers.

Racing-Car Analogy

Like a racig car that contributes improvements to the family car's carburetor, suspension, steering, and engine, the prototype optical processor will generate a gradual upgrading of today's all-electronic computers.

Inside the optical processor are Symmetric Self-Electro-optic Effect Devices (S-SEEDs). These are optical switches with a potential speed of 1 billion operations per second and a switching energy of about 1 picojoule. (The S-SEED was invented in 1987 by Anthony Lentine, Scott Hinton, and David A.B. Miller, all of AT&T Bell Labs> Miller invented the SEED concept in 1984.)

There are 32 S-SEEDs on each of four arrays within the processor> each S-SEED can drive two inputs. Don't look for them with the naked eye--each device measures only 5 micrometers square and contains two microscopic mirrors with controllable reflectivity to infrared light.

Modulated near-infrared (850 nanometers) light from two laser diodes, each using 10 milliwatts, is divided into many separate beams to provide communications between the arrays.

Tiny masks and lenses separate the four arrays, either blocking or passing the beams and so determining connectivity within the machine. Input/output functions can be handled by optical fibers or laser beams transmitted in free space.

Free-Space Switching

Another advance in optical communications technology is Scott Hinton's free-space photonic digital switching fabric, demonstrated in April at Supercomm '90. He is convinced that this new technology will not only change switching fabrics and systems of the futer, but will also affect fundamental architectures of computing structures.

A switching "fabric" is the composite meshing of components used in switching voice, data, and/or video from one place to another. Again, the S-SEEDs are used as either logic devices, or memory cells, or crosspoints switches.

Each S-SEED can switch on or off in less than a billionth of a second when illuminated by a low-power beam of light.

Of course, a switching fabric has an aggregate capacity.

For a fabric supported by the first digital switch in America (the AT&T 4ESS toll switch), that fabric capacity was 8 gigabits per second. The 5ESS digital central office switch supports 6 gigabits per second of aggregate capacity.

Various new optically-oriented systems are designed for fabric capacities in the range of 40 to 47 gigabits per second. They will all be left in the dust by the free-space photonic digital switching fabric, which is expected to start in the terabit range. (A terabit is 1 trillion bits.)

AT&T Service Net-2000 switching system is expected to incorporate this new technology.

A switching fabric with 1-terabit-per-second capacity can support:

* 198 million fax terminals,

* or 15 million 64 kilobits-per-second DS0 channels,

* or 669,000 vide teleconferences,

* or 100,000 high-speed LANs,

* or 19,800 studio-quality TV channels,

* or 6600 high-definition TV (HDTV) channels.

Although the power of 170 digital CO switches would be able to fit on a desktop, this doesn't met we have to be afraid of concentrating too much traffic in a very small device. What it means is that we can implement far more backup protection and monitoring capabilities to ensure more reliable switching and transmission control.

Watch The Basket

"Put all your eggs in one basket and watch the basket," said Mark Twain, and that's what these advanced optical systems will let us do.

The higher computing speeds and vastly greater capacities, plus freedom from much of the electronic "noise" that plagues electrical systems, will allow system administrators to design their own protective systems, without regard for the switching limitations that currently handicap them.

In addition, various photonic integrated circuits and optoelectronic devices will gradually be used in advanced DACS systems and other network components.

These will greatly increase the reserve power of a network, affecting its reliability and its flexibility.

This is all due to the capabilities and advantages of light as it is used in photonic systems. Judging from past experience, our hopes and expectations for the future will probably be surpassed by an enormous margin.

So in 2000, you can look me up and ask why I was so shortsighted in 1990.