Consumer Installable ADSL: An In-Depth Look at G.Lite Technology
Executive Summary
Full ADSL, up to 8.2 Mbps downstream and 768 Kbps upstream, originally intended for video applications is now targeted for bringing high-speed Internet access to residential and small business users. For this new market, access speeds at 10% of full ADSL capabilities are quite adequate for surfing today's Internet. Key to this is the inherent constraint of the Internet backbone, which cannot support throughputs above 300 to 400 Kbps. Although backbones are built on super high-speed fiber optic networks, throughput is based on many other elements such as: Internet traffic conditions, routers, servers, and PCs. Having a sports car that can drive at 150 mph does not mean you can drive from New York to Boston averaging 150 mph. In reality, our sports car will do little better than an average family car.
In mid-1997 many ADSL vendors and carriers recognized that a reduction in speed would also result in a reduction in complexity. The notion of reducing speed to reduce hardware complexity was very attractive to vendors that wanted to implement DSL modems on the same digital processing chips that are used in today's dial up modems. Telephone companies also saw a reduction in speed as a way to simplify installation of ADSL service. Telephone companies looked for technical solutions that had the potential of eliminating the need to send a technician to install ADSL service, enabling users to install ADSL modems just like they do with typical dial up modems. One major difference between ADSL and dial up modems is the need for a telephone splitter. This device keeps the telephone and ADSL signals separated, giving it the capability to provide simultaneous Internet access and telephone service on the same line. Splitters also eliminate the destructive interference conditions caused by telephone sets. The telephone splitter had to be installed on the line at the point of entry to the residence and would still require a technician to install. An idea developed which would sacrifice the full speed in favor of operating without a splitter. A reduced rate ADSL modem looked like it could eliminate some obstacles, improving widespread deployment of ADSL services.
Thus, a new version of ADSL that would be consumer installable was born. This lighter version would enable ADSL to compete against other high bandwidth access options such as cable modems. The Universal ADSL Working Group (UAWG) was formed in January 1998 to address these issues, form a collaboration amongst vendors, service providers and others, and develop a proposal for submission to the International Telecommunications Union (ITU) for adoption as a standard called G.Lite, which was adopted in October 1998. There has been much written in the press, both fact and fiction on the work being done on the G.Lite standard, and how it may effect the industry. This paper attempts to examine the driving forces for developing G.Lite, the technical issues that need to be overcome, and the market for G.Lite.
The Growth of the Internet
The Internet is a story about a network that demands more bandwidth as it evolves. Today's Internet began with the ARPAnet, a network positioned to allow people with dumb terminals to access computers with speeds of 300 to 1200 bps. The arrival of the PC was the initial catalyst in transforming the network to an e-mail and information exchange network for Personal Computers using access speeds of 2400 bps to 9600 bps.
A second major paradigm change started when picture (GIF) files and sound files started to appear, which could be accessed via FTP sites and later viewed off line. Modem manufacturers began offering speeds up to 19,200 bps. In 1993 a program called Mosaic was the next step in revolutionizing the Internet by enabling users to view text, images and sound simultaneously. Thus began the frustration over slow dial-up access speeds.
Today's dial-up modems allow access of about 28.8 - 56 Kbps over the PSTN network. Approximately 8 to 10 times this rate, or 300 to 500 Kbps, is needed to take full advantage of the Internet's current capabilities including sophisticated audio and animation applications. While corporations do provide access speeds approaching these rates via frame relay or dedicated T1 circuits, most residential users must funnel through slow dial-up access lines limited to no more then 56 Kbps.
As applications become more sophisticated, needing more and more bandwidth, demand for Internet services and the number of Internet connections is growing rapidly. Today there are 70 million adults, one-third of the U.S. population over age 16, on the Internet. International Data Corp. (IDC) says the number of users will surpass 163 million in the year 2000. The number of businesses that can afford full-time connection and publishing on the Internet is expected to quadruple between 1997 and 2000. Users are not going to want to sit around while a 28.8 Kbps modem paints their screens with data. They will demand faster transmission of data from Web pages to their PCs.
Basic rate ISDN was a fast data service offered by telcos. It is faster than a typical analog modem, with top speeds around 128 Kbps; but deployment in the US was slow, expensive and not fast enough for today's multimedia Internet. A full T1 connection, at 1.544 Mbps, provides more than adequate speed but the cost is prohibitive for a residential or SOHO (small office/home office) user.
ADSL: Using the Existing Copper Network
The answer is ADSL (asymmetric digital subscriber line) service in one of its several forms. ADSL, like other DSL technologies, was designed to increase the information carrying capacity of plain old telephone (POTS) wires. ADSL operates on the existing installed copper wire infrastructure and does not require new transmission media such as fiber optic cable. ADSL enables high-speed multimedia services, Internet access, distance learning and video conferencing for everyone with a standard, copper phone line.
ADSL offers fast data rates in the direction toward the user. While the downstream rate is up to 8 Mbps (full ADSL), ADSL provides slower upstream data rates to the network -- ranging from 32 Kbps to 768 Mbps. That unbalanced flow is part of the attraction of ADSL. With ADSL, interactive multimedia services can be provided to almost anyone with a telephone line. How fast is ADSL? This is one of the most misunderstood concepts of ADSL. For example, if there were no constraints of the Internet backbone or if fast servers were located in every telephone central office, an ADSL modem could download the entire Encyclopedia Britannica to a user's laptop in 16.6 minutes, compared to 6.4 days using a typical modem speed of 14,400 bps.
But beyond the high speed DSL connection are other constraints. Orckit has done many tests with ADSL access speeds from as low as 64 Kbps to a full 8 Mbps connected to the worldwide Internet. What was found is the typical backbone throughput averaged around 250 Kbps and rarely exceeded 400 Kbps. If the ADSL modem access speeds were set below this value then the throughput would be limited by the ADSL modem. If the modems were set above this value then the throughput was limited by the Internet backbone. Doing this same test with local servers (servers located in the same central office with the ADSL equipment) resulted in much higher throughputs. In this case, server constraints and protocol settings dominated the throughput. Speeds of over 2 Mbps were observed. We also observed that the Internet throughput limitation varies widely depending on what backbone providers are used, location and time of day. Also we have observed a steady improvement in the Internet even over a short period of one year.
The question How fast is ADSL? depends on the service ADSL is providing. More important can ADSL support improvements expected in the Internet backbone and new and exciting services? The answer is yes in all areas. ADSL is a technology that not only is well suited for bringing high-speed access for today's Internet but also is very well suited for tomorrow's services.
This concept (ability to mature with the technology changes) is a key ingredient in the relationship between full rate ADSL and G.Lite. Both are based on a single foundation. G.Lite cleanly addresses the need to reduce complexity in modem design for today's Internet access speeds. This will clearly benefit consumers who use ADSL services now. Today, ADSL already has the wherewithal to address increased traffic demand and new enhanced applications. The G.Lite version of ADSL is still 8 to 10 times faster than the ISDN services offered for Internet access, capable of providing 1.5 Mbps downstream and 386 Kbps upstream.
G.Lite
The G.Lite version of ADSL is a subset or form of ADSL service. It has also been referred to as "splitterless" DSL or as Universal DSL. Until the conclusion of the recent standard, the Universal ADSL Work Group (UAWG) termed G.Lite technology as Universal ADSL. Its key goal was to develop a technology that could be easily installed by end users. This refers to G.Lite's promise to allow telcos to provision DSL, and customers to set up G.Lite DSL connections on their own -- without the delay and expense of a telco service visit.
In October of 1998, G.992.2 was adopted as the standard that began as the Lite standard. Formal ratification of the new G.992.2 standard will take place in June 1999.
User-Installed DSL
One of the initial driving factors of creating a G.Lite standard was that the customerend technology had to be user installable, eliminating a telephone company installation or telco truck roll. Thus expediting and lowering the cost of deployment. The goal is to do away with a telephone technician install of the POTS splitter as a necessary piece of customer premises equipment. With this version of service, the G.Lite modem and the POTS operate together on the same internal home wiring system, allowing the customer to plug both a telephone and a computer modem into a standard wall telephone outlet. This contrasts with the original concept where ADSL service is provided over separate home wiring fed from a centralized POTS splitter, located on the telephone line where it enters the home, usually in a NID (network interface device). This would also require a new wire from the NID to the ADSL modem.
For G.Lite, the telephone and ADSL service are carried on a common in-house wire to the computer.
ADSL vs. G.Lite Technical Challenges & Solutions
A G.Lite installation must cope with a more difficult operating environment than the original ADSL splitter concept. G.Lite must work on the house wiring with a telephone without the benefits of a POTS splitter.
It is the sharing of the telephone and ADSL signals on the same in-house wiring that creates the greatest potential for difficulty with G.Lite. Figure 3 shows that noise generated from a telephone in the same frequency range as the ADSL signal can be disruptive to the ADSL signal. In addition the impedance of a telephone when off-hook may be so low that it essentially shunts the strength of the ADSL signal.
When a POTS splitter is installed at the entry point where the line comes into the home, it will filter the telephone signals before combining the ADSL and telephone signals transmitted and received. The issues of noise and impedance are eliminated with a single POTS splitter installation.
Testing ADSL and telephone sets on the same wiring showed that the disturbance was drastically different depending on the model telephone used. Some handsets are non-linear and convert voice signals to a modulated high-band signal that results in interference on the modem. The opposite can happen as well, and ADSL signal can be heard as annoying sounds in the telephone.
For these cases, a simple solution is to install an in-line microfilter (low-pass filter) between the wall jack and the telephone, thus eliminating the disturbance problem.
Such a filter is inexpensive and can be easily installed by the customer. In effect, this accomplishes the same thing for G.Lite as did the centralized splitter concept. Instead of having the filter at a single point, the filtering takes the form of a distributed splitter (Figure 6). This in-line microfilter eliminates disruptive noise and raises the impedance at ADSL frequencies to prevent shunting out the ADSL signal. Since they cost only a few dollars, and eliminate the need to send a technician to the customer premises, they may be the best answer to the truck roll challenge.
For some users who are unable to solve the noise problem with a single in-line filter, a single centralized consumer-installable splitter is available. There are those who maintain that a single centralized splitter is no more difficult to install than G.Lite. They point to one telco test where full-rate ADSL was offered with a self-installable POTS splitter option. About 50 percent of customers ordered it that way and 90 percent installed it without assistance. The job is not that complex.
If microfilters are installed correctly on all telephone jacks, then a house may also get full rate ADSL. This concept is one of the biggest misconceptions about G.Lite. Orckit's testing of G.Lite in field trials demonstrated that without microfilters there was often a reduction in service rates due to interference from telephone handsets. With proper installation of microfilters basically full rates can be realized. This makes the definition of G.Lite with no splitters, G.Lite with microfilters, and centralized splitters more vague. What appears to be developing is the concept that G.Lite will be a technology that may be less expensive due to reduced rates and installable without any filters (if telephone sets are non interfering), with microfilters (if telephones interfere) and with centralized ADSL splitters (to eliminate installing many microfilters).
Perhaps not so simple to resolve will be the problem of G.Lite installation at the PC (personal computer). The G.Lite standard does not address the complexity of software and the drivers that are needed for PCs. It takes little imagination to foresee the possibility of irate consumers complaining to the telco that their G.Lite service does not work when the problem is with a PC software driver. Nothing in G.Lite resolves this problem, although Windows 98 and the new computers that are G.Lite compliant should ease the trouble of installing new devices. Customers who purchase a PC with the modem and drivers pre-installed should have little of this sort of difficulty.
Another consideration is the fact that home wiring has no standards. Often the in-home telephone connection is run on irregular cable, not well suited to ADSL transmission. In addition, there may be questions on who is responsible for isolating faults. The customer may have no telco maintenance plan for home wiring, in which case the whole wiring issue also falls into the user's lap.
G.Lite Market
The hope for G.Lite ADSL is to accelerate widespread deployment of the service by making installation of a DSL modem as easy as installation of the familiar analog computer modem. Since companies like Compaq, Microsoft and Intel pushed G.Lite even before it hit the market, the outlook for ADSL and G.Lite demand is good. Keep in mind that most SOHO, small business and even home computer sales today are either upgrades of earlier systems or represent the purchase of additional computers. The buyers are knowledgeable, if not sophisticated users and are willing to spend dollars for Internet access at speeds faster than yesterday's modems.
The product for consumers is simple, very fast and online all the time, enabling home Internet users to get the same speed and easy access at home as they get in their work place. The point is to provide a product that phone companies can market at less than $50 per month -- within the acceptable range for such customers. The major market for G.Lite will be in the SOHO and advanced home sector. Home users want speed, but are only willing to pay $37 a month, according to a study by Dataquest, San Jose, CA.
G.Lite is attractive to two groups for two separate reasons, telephone companies and computer manufacturers. The telcos who will provide DSL services to their customers are attracted by the ostensible simplicity of installation of G.Lite as opposed to full ADSL. Any time a phone company can avoid rolling a truck to install customer equipment, it saves money. The RBOCs and other telcos appreciate the "do it yourself" technology which allows the customer to plug in the DSL modem and go to work (presuming the telephone company has pre-installed DSL at the Central Office). The telcos' customers for DSL service are expected to see G.Lite as a service they can buy from a local computer or electronics store, plug in on their own, and begin using.
Computer manufacturers will offer bundled G.Lite technology in new computers shipped. This comes in the form of a V.90/ADSL NIC (network interface card) in their PCs. The computer firms see G.Lite bringing them an expanded market for new software applications and expect the market for faster processors will surge if the current bandwidth bottleneck is opened. Compaq Computer and several other manufacturers have announced they will offer a V.90/ADSL NIC in their PCs.
Conclusion
By reducing the complexity of on-site installation and minimizing the need for new wiring at the user's home, the G.Lite version of ADSL will make it possible to increase bandwidth for the customer on a more cost-effective basis. Small business and SOHO operators are expected to rush to get its powerful access to the Internet and corporate LANs. Over the next 18 months, the 1.5 Mbps speed provided by G.Lite should be more than sufficient for most residential and SOHO Internet access applications. Just as the market for larger PC hard drives continues to accelerate as demand for more speed and storage on computers grows, so too will the demand for access speed grow, increasing the demand for G.Lite.