Loading summary
Transcript1 lines
- [00:02]
A
Other than a few minor size changes, AT&T has been using the same type of copper wire technology since 1887. Millions of miles of copper wire that for decades carried just analog voice calls. Then came along a remarkable technology that takes that decades old copper wire and somehow makes it able to carry millions of digital bits per second. ADSL in the 2000s, ADSL brought fast always on Internet to millions of people around the world and did it using their existing telephone lines. What is perhaps Most funny about ADSL's rise is that it was never intended for Internet broadband. Many of us might have forgotten it today, now that we have fiber 5G and Starlink. But ADSL truly helped make the modern Internet in today's video. The Accidental Copper Miracle the 1990s brought forth a new formidable force in media cable TV. Cable dates back to 1948. With a simple communication service called Community Antenna Television. CATV received and retransmitted TV signals to households with bad reception due to mountains or isolation. Such video signals are first received via antennas and sent through coaxial cables, which then split into tributary cables that eventually reach people's homes. For the first few decades, the US regulator FCC kept the tight grip on what content that CAT TV can transmit through their cables. But in the mid-1970s, these rules were loosened, birthing premium content channels like HBO and Ted Turner's SuperStation. Events like the momentous Thrilla in Manila boxing match in 1975 and other pay per view events like it helped supercharge cable's adoption in the United States. Then in 1984, the Cable Communications Policy act deregulated cable TV rates. Cable operators lifted their rates 60%, which allowed them to fund huge buildouts across the nation. From 1985 to 1990, the number of cable systems in the United States surged 40%. The rapid rise of cable in the second half of the 1980s greatly bothered the telephone companies like the Baby Bells, which had spun off after AT and T's breakup. Key part of that threat was how the cable companies had this direct relationship with their customers, the coax cable delivering video signals. It did not seem a stretch to imagine them eventually delivering other services like phone calls through that cable, and they would later the Baby Bell Telecoms eventually felt that they had to deliver video to their customers too. But antitrust and telecom law prohibited them from buying a cable operator operating in their own service areas. They had to do it themselves. In mid-1992, the US FCC allowed the Baby Bells to deliver video content through their wires, thus birthing a service called VIDEO DIAL TONE Just pick up a telephone and have a movie stream to your home. The catch was that the telecoms cannot own the video content themselves like the cable operators do. They can only transmit it on behalf of third parties. But the ruling allowed them to start getting into the streaming business. The problem was copper. Thanks to their big fat coaxial cables, the cable companies can push a lot of rich video to people's homes. The telecoms did not have access to that coaxial cable. What they did have were the legacy copper cables used for telephones. These are called twisted pairs, a name that describes how these insulated pairs of copper wires twist around each other. This is done to reduce interference. The twisted pair has been used by AT&T since the very first days of the telephone in the 1880s. So the install base by then was absolutely massive, which can be a good and bad thing. Another good and bad thing is the fact that 99% of the whole system's wires, as measured by length, are last mile wires. May be one to two miles long or shorter that go right into a household. These are the least economical to change, not only because the replacement would be optical fiber and that then cost a lot of money, but also because it meant going into the house, and that was a huge hassle. Some last mile twisted pairs haven't been touched in decades. So technicians had to make video dial tone work entirely over existing copper lines, originally meant for plain old telephone service. Now, how are we going to do that? Joe Lechleiter was a researcher at Bell Labs studying signal processing back in the 1960s, and AT&T began converting their network from analog to digital for cost and quality purposes. By the 1970s, the network had digital transmission and switching, but the network still had to convert data from digital to analog for the last mile copper lines going into people's homes. Yet telecom engineers still dreamt of a system that was digital end to end without needing to replace every last mile wire. And in 1980, Bell Labs started upon this. A few years later they had a set of international communication standards called the Integrated Services Digital Network, or isdn. ISDN establishes within the copper line two coexisting voice and data channels, each with bidirectional capacity of about 64kbps and one overhead channel with 16kbps. So from end to end, we have digital signals. I'm not sure if we can call this the first digital subscriber line or dsl. It is a contested term. Leck Leader uses the term to describe technologies that help deliver digital content through the twisted pair copper wires. Others have a more expansive definition. If we go by that phrase, the honor should probably go to the business oriented T1 lines and created by AT&T in the 1960s. But ISDN is still a pioneer, gaining widespread adoption in Europe and Japan, Though in The United States, AT&T did not widely adopt the standard, partly for technical reasons and partly due to the chaos of its breakup, which went into effect in 1984. This breakup also moved Lech Lieder to a research lab spun off from AT&T called Bell Bell Corp. A research consortium owned by all the Baby Bells that served as kind of like their own Bell Labs. One of the technical reasons why ISDN did not catch on in the United States was that its speeds were slow. So slow that people in the industry mocked its name as meaning innovation. Subscribers don't need can we make that faster? But there existed an impediment to just cranking up the bandwidth speeds, a technical problem known as crosstalk or near end crosstalk. Next to be specific, near end crosstalk happens due to the signals of one twisted pair coupling with a neighbor electromagnetically. I know it sounds naughty. Folks behave. Imagine a household connected to a carrier station a few miles away via a last mile twisted pair. That house's pair leaves the house and joins with other last mile pairs inside a cable bundle before arriving at the carrier station. At the near end electromagnetic coupling coming from one twisted pair transmitting data ends up interfering with the data in the other receiving twisted pair. This is next to deal with it, we split the frequencies at which we transmit and receive data inside the copper. The intuitive split is to do it half for transmit and half for receiving. But is there a split that gives you more overall bandwidth? There was. Leckleder was a man who liked a radical idea or two, born with wide ranging interests. His obituary noted an admiration for how Einstein challenged accepted wisdom. Leckleder saw that the carrier's receiving end is way louder than the household receiving end because that is where all the twisted pairs come together. We do not have this jumble at the household where we typically just have one or two twisted pairs. Lek leader realized that this quieter household lets it receive way more data than what the carrier side can. So how about we allocate a larger portion of frequencies to the data flow going downstream from the carrier to the receiver? Of course, in doing this, the slice of frequencies dedicated going upstream from the receiver to the carrier must decline. But since crosstalk worsens at higher frequencies we can alleviate it by giving it a lower, more ideal frequency. The result is an asymmetric downstream and upstream setup that greatly raises the overall bits per second sent without as much crosstalk. When Leck Leader first came up with this asymmetric concept, nobody considered it very practical until the telecoms decided to get into the movie streaming business with video dial tone. Engineers like Charles Dudis quickly recognized that video streaming needed more downstream bandwidth than upstream and pushed asymmetric as a potential solution. The next step is to agree on a pattern scheme to take the digital bits and map them to analog signals for sending through wire. Such modulation schemes are called line codes and they play a critical role in determining how much data we can send through copper wire. The choice of line code was heavily debated in the early 1990s, and to determine the line code standard for their asymmetric DSL or ADSL service, BelCore in 1993 held a competition colloquially referred to as the BelCor ADSL Olympics or the BelCor Shootout. There were three line code competitors. The first was from AT&T itself. Their line code entrant was called CAP, which stands for Carrierless Amplitude Phase. The second was QAM, which stands for quadrature amplitude modulation. It was brought forth by Broadcom Corporation, not Avago, the chip giant, but rather the young wireless and broadband company before its growth and subsequent acquisition. CAP and QAM are quite similar, with subtle differences only in the implementation. They employ well known battle tested algorithms to encode digital data into an analog signal and then send it off using a single big frequency band. But the third entrant was a bit of an outsider. Discrete multitone or dmt. It was championed by a young startup called Amati, led by a Stanford professor named John Chaffee. Chaffee actually recalled being inspired by Leckleder himself to take a leave of absence to start Amati. DMT works kind of like another technology that we talked about earlier while I was going through my wireless kick, Orthogonal frequency division multiplexing, or ofdm. OFDM splits data into many streams and sends them in parallel. So DMT splits the 1.1 MHz of available copper frequency band into slices. Early DMT had 256 slices, each about 4kHz wide. It then splits up the data, encodes it using something like QAM and transmits it through each frequency slice in parallel. So multi carrier rather than single carrier. In the end, DMT's multi carrier approach won the bake off. It was 50% more efficient, allowing us to push significantly more data through the same copper wire. The resulting speed was said to be shocking and up to 4 times faster than other candidates at 6 Mbps. Another reason why DMT1 was its adaptability. Copper attenuation, as I mentioned, worsens at higher frequencies. So if the higher frequency slices DMT can lower the bit per Hertz rate to as low as 4 bits per Hertz for better signal to noise ratios. For lower frequency slices can do the opposite. There are millions of copper loops out in the world with huge differences between them all. DMT had the flexibility to probe any particular copper wire and optimize per its individual tendencies to send the most data near the telephone line's theoretical limit. Actually, DMT's downsides were its complexity, that engineers were less familiar with it, and its computational intensity. That last one in particular, with 256 channels working in parallel, you're going to need some hardcore silicon. What Amati had brought to the competition was just the prototype. Now someone in industry had to make actual DMT standards compliant silicon and gear that can be sold to telecoms to implement insteps. Alcatel in the early 1990s, Alcatel NV was the world's second largest telecommunications firm. A sprawling company formed after a merger between France's CGE and the European assets of the US firm ITT. Alcatel competed with other giants like Ericsson and AT&T to sell big ticket telecom items like digital switches for telephone networks. A few years before the Belcore Olympics took place, a small team at the Alcatel Research center in Antwerp came together for a special project. The Antwerp location worked on broadband access, focusing on coax cable, but mostly optical fiber. Fiber performed the best, but its high unit costs made it impractical for wide use. Until those costs came down, copper had to fill the gap. Yet at the same time, it seemed like copper's maximum data rate capacity was what was offered by ISDN. So about 144kbps total. Then, during a coffee break, Martin de Praeger and Willem Verbeeste, two employees within Alcatel's research division, got to talking about trips they had both made separately to Belcorn in 1990. There they had both seen a prototype of a symmetrical broadband DSL system capable of 1.5 million bits per second through twisted pair lines. The system was impractical. It was about the size of a small fridge. But it evoked the possibility of fast broadband Internet access over ordinary copper. So de Praecker and Verbeest could got permission from management to set up a small 3 person internal team to work autonomously on DSL technologies. The members had no background on twisted pair transmission, but they were smart and that's all that mattered. To start, Willem gave the three team members IEEE journals on the topic as reading homework over the holidays in 1991. Then in January 1992, the team got started. When the team began, they focused on single carrier transmission line codes like QAM and CAP, which we discussed earlier. Then in March 1993, news came down the pike that DMT had won the Belcore Olympics. The results had been controversial and the battle of the line codes would go on for several years. Some providers like AT&T stick with CAP, going on to produce chipsets with partner paradigm. Nortel, which was then pretty committed to optical fiber, also rejects DMT based Copper adsl. They later developed the proprietary hybrid system called Etherloop as their high speed Internet offering. The Alcatel team had also been surprised. Verbese later recalled in a case study, I had a meeting in the United States concerning CAP technology and standardization. We were developing CAP in cooperation with Belcor and we were certain that the CAP technology would become the standard. After the meeting, I left to travel back to Belgium, and when I arrived in Belgium, I heard that DMT was chosen as the standard instead of cap. This meant we had developed the wrong technology. Alcatel Decides to Pivot despite knowing little about dmt, they decided to license and adopt it right away. Recognizing an opportunity to have a say in crafting the ADSL standard, the team quickly expanded to 100 people and management took a leap of faith. In 1995, Alcatel introduces the first DMT based ADSL chipset consisting of an analog asic, a digital signal processor and and a channel processor. Designed by Alcatel and fabbed on a 0.7 micrometer node, the chipset came out less than two years after the DMT pivot and a few months before the final ADSL standard was finalized. In October 1995, Alcatel demonstrated an end to end prototype at the Telekom Geneva Fair. By mid-1997, Alcatel had a video streaming product ready for customers. But one small those customers didn't need it anymore. As late as 1996, telecom industry experts asserted that video dial tone was going to be a hit. One analyst wrote then, video on demand is expected to be the next big wave. Rather than walking to the local video shop, customers will get films transmitted electronically to their homes and by just dialing in their requests. Not wrong. The video rental market was then valued at $12 billion and today Netflix is a thing. But Video Dial Tone, the product failed to work for a variety of reasons. Interestingly, the product did not seem to be terrible. Video Dial Tone was reminiscent of cable TV with many hundreds of channels. When you landed on a channel, a menu showed up offering other various services like faxes and video mail. In supporting this, ADSL did its job. In a 1995 interview with Wired magazine, Bell Atlantic CEO Ray Smith talked it up, saying ADSL is not an interim technology, at least not in the sense that it's second best or doesn't work well. It has excellent quality. You can do the virtual VCR over it, you can fast forward and back and you can have a whole batch of channels. It's server based, it's digital. Testers of the Video Dial Tone beta product appear to echo the sentiment. It wasn't terrible. The channel changing lag was less than a second and the visual and audio quality comparable with broadcast satellite. Moreover, the economics of the install offered some advantages compared to cable. Raysmith continued, the cost of installing ADSL is higher, yes, for each house. But remember when you cover a whole batch of houses for cable, not every one of them takes cable television. When you do adsl, each house costs more, but you do it only after the sale is made. End quote. So if the product and the install economics weren't terrible, why did VDT fail? One proximate cause of death seems to have been a lack of content. Video dial tone was built around the assumption that telecoms cannot own the content, so they had to entice content owners like the movie studios to lease channels on video dial tone. Many ultimately did not bother. But what was probably the real final nail in the coffin for video dial tone was regulation. Prior to 1996, the FCC struggled to put together a set of coherent rules for video dial tone, causing several carriers to drop plans to support it. Then in 1996, US Congress passed the 1996 Telecommunications Act. The telecoms made VDT because they couldn't own cable systems in their areas. The 1996 act lifted that ownership restriction, obviating the video dial tone's whole reason to exist. And with that, ADSL seemed doomed to the trash heap. But you know what the fortune cookie says? When one door closes, another opens. And by that I mean the Internet. The early 1990s saw the release of the World Wide Web project by Sir Tim Berners Lee at cern, plus the release of the Mosaic browser. These together made the Internet a user friendly experience that helped drive adoption and and by the mid-1990s, you had about 20 million households nationwide going online. The Internet experience got better the more people were on it. The problem was access. The way most people got online was using the so called voice band modems, which turn digital computer data into analog signals sent like an ordinary voice call. To get online, these modems seized control of the phone line. Automatically dial the number of the ISP and establish a voice circuit. This process creates your iconic mixtape of static screeches and whines. Unfortunately, the voice band grants just 3.3 kilohertz of available frequency to send digital content. So they slow. As AOL and others boomed across the United States, dial up modem speeds topped at 28.8kbps. We mock 56k nowadays. But those modems didn't even arrive until late 1997. And of course, voice band modems used up pricey voice minutes and blocked the telephone line. Few things got my dad more riled up than me trying to get online while he was on a phone call, or if he missed the call because I was surfing the Web. Dad was not alone. A study by Pacific Telesis Telesis in one region about 24 miles in diameter cited by US Today in 1996, found that 16% of local calls did not connect because of high Internet usage. Historically, just 1% of calls failed to connect. ADSL did not block the telephone. Recognizing that Internet demand was real and they had something useful, Alcatel quickly pivoted ADSL from video to broadband. Fortunately, there was a lot of technical overlap between the two. Small uploads to request content, big downloads to get it. However, the telecoms, perhaps burnt by video dial tone's failure as well as the scale of the work ahead, were initially hesitant to follow Alcatel's pivot to Internet. Was the technology ready? How can they support millions of customers across the nation? So they balked. But eventually rising competition forced their hand. Many of the cable firms by the early 1990s had spent billions to install new cable infrastructure. Facing now their own challenge from satellite tv, they realized that they had to provide something different. That was Internet. The 1996 telecom deregulation also created this new category called the competitive local exchange carriers. The Bells were obligated to lease their copper lines and equipment to these startups, who can then resell them to Internet service providers or ISPs. This all eventually forced the Baby Bell telecoms to recognize that they needed their own Internet play too. But to offer broadband Internet, the telecoms needed to buy and install ADSL equipment which included, among other things, first, customer premises equipment installed inside the household, a splitter and ADSL modem. The splitter splits the analog voice signals from the data signals and the modem is the device through which the computer communicates with the Internet. Second, they need telecom equipment like the Digital Subscriber Line Access Multiplexer or dslam, a big server like thing with many cards installed inside. The DSLAM sits inside the telecom's central office. It receives signals from connected households and bundles them to send to the right ISP core to bringing down costs and making the service as affordable as possible would be standardization and group purchasing. This was critical in helping the cable providers bring down the cost of their own cable modems. In August 1996, four Baby Bells, Ameritech, Bell, South Pacific Bell and SBC Communications, serving 65% of the U.S. population, banded together to form a buying group called the Joint Procurement Consortium, or jpc. The JPC then put out a request for proposals. Alcatel aggressively went after the request and was shortlisted alongside a small US provider called Westdal Ericsson. In October 1996, Alcatel was announced the winner. On the basis of price and having a full end to end solution from modems to D slams, it was a huge win for their DSL equipment division. Early commercial deployments in the United States started sometime in 1999, backed by aggressive advertising and sales by the bells. By 2000, the US had about 2 million DSL subscribers. That nearly doubled the following year, eventually reaching nearly 12 million in 2004 and staying on track with cable's 15 million ADSL became the primary Internet access technology backed by the Baby Bell telecoms. Though I do not want to discount the many other DSL technologies that floated around during this wild HDL, SDSL, IDSL, RAD, SL. The confusion led some people to create a new term called N xDSL. But ultimately xDSL and all others became just a DSL. The JPC's buying power brought initial scale in the United States, but Alcatel as a global firm, quickly sought to sell to telecoms everywhere. In this they succeeded wildly. By 1999 they had shipped a million lines already. It was overseas where DSL made especially deep inroads. One early adopter was Singapore's Singtel, which held significant trials in the late 1990s and early 2000s, though they quickly then moved to fiber, probably because Singapore is so small. Korea, Japan and Taiwan quickly followed. Their dense apartments and legacy copper loops were well suited for ADSL's strengths. Korea in particular was a rapid adopter, counting over 10 million users by 2002. In the early 2000s, DSL became the leading broadband technology in the OECD countries, especially in Europe, where cable Internet never really took off and regulators forced incumbents to open up their lines to DSL only startups. DSL's membership peaked in the late 2000s and early 2010s with roughly 360 million lines. Alcatel maintained 40 to 50% market share throughout this era and and continued extending DSL transmission speeds. The performance they managed to get out of twisted pairs was remarkable. VDSL2, sorry for all the acronyms, is capable of reaching up to 200Mbps. In the end, however, DSL gave way to faster speed fiber or cable. Today it is a legacy technology said to be on the decline, but things like this decline very slowly. Aol, after all, only just finally phased out its dial up service in September 2025. So even now there remains maybe 100 million plus DSL subscribers still out there getting their Internet through copper wires. With its speed still good and international penetration deep, I reckon that DSL is going to be around for a very long time. Still, I want to thank a number of IEEE luminaries and former Alcatel people for reaching out to me to suggest this topic, as well as walking me through the history of adsl. DSL occupies a funny place in our memories. It was not how many of us first experienced the Internet, and it is not how many of us access the Internet today. But DSL's near miraculous repurposing of existing copper infrastructure smoothly brought hundreds of millions of people online. It, along with cable modems, helped make the Internet the always on ever present experience it is today. Its speed allowed people to first experience the richness of content that would go on to build the world's big Internet businesses. Not bad for a technology that was never made for web browsing in the first place. A fortunate accident that helped accelerate the one of the most powerful tools in history all right everyone, that's it for tonight. Thanks for watching. Subscribe to the Channel. Sign up for the Patreon and I'll see you guys next time.