Using Light to Translate Information for a Computer

Henry Chin
Electrical Engineering
Stanford University
May 2001

I am working with other students to demonstrate a system that would enable computers to process certain signals, such as radar or cellular phone transmissions, at very high speeds. Along the way, we also hope to gain a better understanding of the physical principles behind devices that interact with both optics and electronics.

Electronic processors, like the Pentium chip in your home computer, communicate using digital signals. Information in the real world, however, exists in a completely different form, one consisting of analog signals. Systems that translate between the two, called analog-to-digital or digital-to-analog converters, are necessary whenever a computer must process information from the real world or send out information into the real world. Fast converters allow digital processors, which are relatively easy to re-program, to replace electronic circuits that can process analog signals but are very expensive to design and build. My research is part of a larger effort to make an analog-to-digital converter that is much faster than anything currently available. We are attempting to do this by incorporating lasers and other optical components into a system that has traditionally been purely electrical.

Analog signals have two basic characteristics: they can change at any instant in time, and they can take on any value. The pitch of a human voice, for example, is an analog signal. During a song, a singer’s voice can change pitch at any time. Moreover, if the singer is off-key, the pitch is not limited to the twelve notes (including sharps and flats) of a musical scale.

Contrast this with a pianist who is playing in strict time with a metronome. Only when the metronome ticks can he or she play a note. Moreover, the notes must correspond to one of the eighty-eight keys on the keyboard. These two features, a steady rhythm and a limited range of possible values, characterize digital signals. In computers, digital signals are often in the form of electronic “ones” and “zeroes.”

How do analog-to-digital converters translate real-world information into signals that a computer can understand? Let's say we want to take a song and arrange it for the piano; and, as before, we restrict the pianist to playing notes at a constant beat with a metronome. The process requires two steps:

The converter simultaneously monitors the vocalist's song and the pianist's metronome. Every time the metronome ticks, the converter records the exact pitch of the song at that instant in time. The pitch that is recorded is not a note in a scale, but is instead an exact numeric frequency. A note that is sharp of B-flat, for example, might be recorded as 490 hertz. This process is called "sampling."
The converter examines this list of frequencies and assigns to each entry the musical scale's note that is closest to the recorded frequency. Even though 490 hertz does not match any note, the converter rounds it off to the closest one, in this case "B." When the converter has performed this approximation for every recorded frequency, the result is a list of notes. This entire list constitutes a digital signal, one that a computer processor would be able to analyze.
The speed of current analog-to-digital converters is primarily limited by the regularity of the "metronome" used in the conversion. This metronome, called a “clock” by engineers, must tick at a constant rate; any variation causes the final digital signal to be inaccurate.

This is where the advantage of optics over electronics comes into play. Special "mode-locked" lasers emit very short bursts of light with remarkable regularity, one unmatched by traditional electronic clocks. By using these laser bursts to trigger the "sampling" of the electronic analog signal, we can operate at the requisite high speeds without sacrificing too much accuracy.

We are designing, building, and testing two devices that will be used in this sampling process. Both these devices would reside on a single semiconductor chip. The first device samples the analog electronic signal when a pulse of laser light hits it. The second device takes the sampled signal, which is still in electronic form, and transforms it into an optical signal. This allows us to relay the signal to a second chip, which performs the rest of the conversion process.

On this latter chip, a more conventional electronic circuit can perform the second step in our two-step process: conversion into discrete values. However, as stated earlier, we are attempting to convert signals at a much faster rate than can currently be achieved with electronics. Our design thus coordinates the operation of a hundred of these electronic converters in a sequential fashion, so that the overall speed of the entire system is a hundred times that of the individual electronic circuits. The ultimate aim for our design is to convert 100 billion values per second. This is dozens of times faster than what is available through current technology.

An analog-to-digital converter operating at such high speeds can have important applications in wireless communication and radar. With such a converter, a wider range of radio and microwave signals can be directly converted into digital data, which can then be processed by computers. As a result, costly, custom-designed circuits could be replaced with computers running easily modified software.

In addition to these aims, though, we are studying the physics of basic electronic devices that interact with light. Our research, then, does not only involve building a fast analog-to-digital converter, but is also more broadly applicable. The ideas that we are studying here could be used to design other devices that might be used in such fields as optical networking or sensitive optical and electronic detection.