By John Cioffi
Hitachi America Professor
Engineering Department of Electrical Engineering
Stanford University

Today, digital subscriber lines are recognized as a successful delivery mechanism for services like Internet access, television, voice-over-Internet Protocol and various video and file-sharing applications. But this was not always the case. DSL had to transcend many difficult and lean times. It owes its success to the dedicated efforts, optimism and enthusiasm of several individuals.

1979: A naive suggestion
During my first year at Bell Laboratories as a 22-year-old, 4,800-bit/second full-duplex voiceband modem designer, I was invited along with my boss and a co-worker to attend a meeting about converting the last copper-line segments to digital transmission of voice. In my opinion, the birth of DSL occurred in that Whippany, N.J., meeting.

Everyone else in this meeting was at least a decade older, and I was pretty intimidated, but happy to be there. I computed the maximum speed for a 4-mile twisted-pair based on a text-book formula. I calculated "about 1.5 Mbits/s-pretty fast, maybe even fast enough for compressed video," I said, when someone finally let me speak at the meeting.

My boss turned red and some other senior engineers laughed and half-politely rebuked that young kid. "We're shooting for 144- to 192-kbit/s two digital voice channels here. Your calculations are off; go read your book again and stop dreaming."

I kept my mouth shut but thought, "Someday I'm going to show these guys we can go a lot faster." ISDN (the first DSL) started many years later at 160 kbits/s.

1986: A chance meeting
The 1984 divestiture of AT&T sent me permanently westward and eventually to a faculty position at Stanford University, where two events allowed me to return to the DSL dream. The first was Reagan's new Presidential Investigator program for young faculty, which allowed one enough money and academic freedom to invest in folly such as copper.

The second event was a chance meeting with Dr. Joe Lechleider at a conference. Joe was near retirement age, had tremendous respect in the old Bell system, was Mr. ISDN and also had calculated the same 1.5-Mbit/s number, especially if one made DSLs asymmetric-a match to most video applications anyway-with a higher speed toward the customer than back up from the customer.

Joe provided a little extra funding to my Stanford group and a lot of encouragement. He retired from Bellcore and DSL a few years later as the phone companies were making widespread announcements about their intentions to lay fiber to all customers within five years. We should forget DSL; no need for this lousy copper anymore, they said. Nonetheless, my Stanford group had just enough dollars to research and detail DSL systems that could do 1.5-Mbit/s speeds.

1989-91: The wayward VC
As our DSL designs progressed, a brochure from a very unusual investment group arrived in my Stanford mail. "Got ideas to transfer to industry? We'll fund your group for two years," it said.

The name of the venture capital firm was University Technology Transfer. I sent in a proposal. It was the last proposal the group funded before going bankrupt-and then only half the promised funds ever came. But it started a cycle of commercialization that eventually led me in July of 1991 to found Amati Communications on my sabbatical and subsequent extended leave from Stanford.

Amati had no money for seven months, but I worked for free on this interesting DSL. Finally, a few companies, most notably Northern Telecom, funded development of the world's very first asymmetric DSL modem prototype, which actually uses a design virtually identical in all its basic elements to the hundreds of millions now in use.

Amati hired a handful of engineers to help us design this modem-all of whom were my friends or former students. Throughout it all, very few outside our little group believed in DSL, but a few companies-most notably my old group at Bell Labs-were then attempting to get 1.5 Mbits/s on twisted pairs. No one mentioned the 1979 meeting, nor the laughter, if they remembered it.

The majority of the telecom world still did not believe in DSL. A respected executive at what became the largest DSL service provider in the United States told me at the time, "I promise you we will never, ever deploy a single line of ADSL. You are wasting your time."

No VC would touch Amati with that kind of endorsement, and it never had any venture investment other than from the original bankrupt group. Nonetheless, Amati was off and running for ADSL.

1993: The Bellcore Olympics
I disagreed with my old group and friends at Bell Labs on the transmission method for DSL. They quickly changed from old friends to aggressive competitors. AT&T broadly promoted a technology known as quadrature amplitude modulation (QAM) or CAP1 (carrierless amplitude phase modulation v1) for DSL. Amati-which then amounted to a few matriculated graduate students, me and a few dear friends-supported a newer, highly adaptive technology called Discrete Multitone (DMT). Even Northern Telecom had a second QAM effort and refused to acknowledge publicly that its nontechnical marketing department had helped fund the Amati DMT effort.

In January 1993, Amati entered a competition organized by communications research lab Bellcore, which is now Telcordia. Amati was given less than a year, and had far less than one-tenth the money of any of the others to deliver a high-speed DSL product. I was worried Bellcore would not even let us enter the competition.

Somehow, working long hours around the clock, we got our modem to work. It was actually smaller than any of the others in the competition, despite being constructed from discrete components. We could not afford VLSI. It was not fully optimized.

At the competition, Amati had to demonstrate its system first. Then three QAM systems developed by Bellcore, GTE and British Telecom were tested on a variety of phone lines.

The results of the tests shocked the world: The Amati modem ran four times faster (6 Mbits/s) on the same lengths of telephone line, or between one-half mile and one mile longer in fair tests at the same speeds. One of the inherent features of DMT technology was that it could be easily programmed to run at the highest rate while the QAM technology was fixed at one speed only. That said, we did not mention the 6-Mbit/s capability until just before the Olympics-it completely surprised the competition.

There were a lot of embarrassed people-some of whom were still a bit angry over a decade later-but those test results could not be refuted. It was like Palo Alto High School beating the Patriots in the Super Bowl-unheard of.

On March 10, 1993, the 117th anniversary of Alexander Bell's invention of the phone, DMT was selected for the U.S. standard for ADSL. It was an emotional experience. I cried for the first time in 20 years on the plane ride back, but no one saw me. A few days later, I passed out and fell cold as the hard work caught up with me. A doctor said it was overexertion.

Over the next five years there were several attempts to reverse the DMT ADSL standard to QAM, but all failed. Most QAM supporters eventually adopted DMT or moved to develop chips for other industries. DMT became the international standard for ADSL transmission a few years later.

1997: Web surfing and sale
Unnoticed at the time, we had written into DMT the ability to go upstream at one-tenth the downstream rate (up to 600 kbits/s). The QAM systems would not support more than 16 kbits/s-a 1:100 upstream/downstream ratio. The 1:10 ratio of DMT was intended to match the Internet Protocol and the emerging needs of Internet traffic.

Back in 1993, many ADSL trials focused on delivering movies on demand. But we saw Internet access rising as a major application. GTE engineers deserve a lot of credit on this observation and encouraged Amati to use the 1:10 ratio when no one else would.

People quickly realized what a difference DSL made for what was then called the World Wide Wait. The telecom industry got excited, and ADSL went on to become internationally standardized as phone companies braced to offer fast Internet services with ADSL.

Amati could never get VC funding, so we prematurely took Amati public at the time the DSL/Internet wave was rising. The share price went from 50 cents to $36 in about a year. We could finally raise money, but the company was never really profitable.

Texas Instruments Inc. was the world's largest supplier of voiceband modem chips and saw DSL eventually replacing that industry. At the time, TI had nothing in DSL, so it did not take much for me to talk them into purchasing Amati, which they did in 1998. TI went on to become the world's No. 1 supplier of ADSL components as the market grew, and TI still holds that position today.

The Koreans
Samsung saw that the United States was headed for fast Internet access and started a Korean program in 1998 to get there. Somehow, I wound up being dragged into South Korea's parliament house as the government considered supporting DSL.

Today, South Korea leads the world in DSL deployment density, but I think it only thought it was trying to keep pace with the rest of the world in the late 90s, perhaps not realizing how slow the rest of us would actually be. It goes to show what a motivated government can do.

Curiously, Japan went in the opposite direction, investing instead in an all-fiber program. No one wanted me anywhere near the Japanese parliament house. The fiber never made it all the way to the home, but the Japanese DSL lines got shorter. Today, 15 million DSL customers in Japan enjoy the highest DSL speeds in the world. Again, a little government investment, even if only for fiber, went a long way.

2003: An Olympic rerun
Later, remnants of the old QAM/CAP suppliers tried again to force a shift away from DMT for the world's highest-speed DSLs, known as very high-speed DSLs. VDSLs have rates of up to 100 Mbits/s on short phone lines.

QAM supporters claimed their technology was cheaper and worked the same, now that the 1993 bugs were fixed. The 2003 Bellcore Olympic results were nearly identical to the 1993 Olympics, with DMT displaying much higher speeds and longer ranges in a variety of shorter-line situations. DMT was again selected for American and then international standardization, although this second time it was supported by most of the industry instead of just a tiny startup. Today there are no active QAM/CAP supporters.

2004: The future of DSM
Back in 1992, it was first noted that the DMT technology accidentally provided a complete line and binder diagnosis. Taking advantage of this long-dormant diagnostic capability, Dynamic Spectrum Management (DSM) is the latest exploitation of the adaptive spectral characteristics of the DMT modem. Tremendous gains are possible because today, DSL's worst enemies are other DSLs that occupy the same telephone cable. DSLs radiate into one another just like wireless systems do. When DSLs are polite, they use Adaptive Spectra. Better yet, if they cooperate, they use Signal Alignment, and their speeds go way up. Watch this area as it evolves, particularly the company concatenating those acronyms-Assia, also my wife's name! I'm optimistic we'll all have at least 100-Mbit/s DSL connections before long.