|Issue:||Asia-Pacific I 2013|
|Topic:||Economic gigabit broadband access using DSL and WiFi|
|Author:||Dr John Cioffi|
Dr John Cioffi, is the CEO and Chairman of the Board of ASSIA, Inc. - Adaptive Spectrum and Signal Alignment, Inc. Prior to founding ASSIA, he served as founder, CTO and vice president of engineering at Amati until its acquisition by Texas Instruments. Dr Cioffi held a tenured endowed professorship at Stanford University in the Department of Electrical Engineering from 1985-2009 and is now an active Professor Emeritus. He has published over 400 papers and holds over 100 patents. Dr Cioffi is a member of the United States National Academy of Engineering and an International Fellow of the United Kingdom Royal Society of Engineering. He is a winner of the The Economist Computing and Telecommunications Award (2010), IEEE Alexander Graham Bell Medal (2010), the IEEE Kobayashi Award (2001), and the Marconi Society's Marconi Prize (2006). He currently serves on the boards of Teranetics, ClariPhy, Vector Silicon, and Alto Beam.
Dr John Cioffi holds a BSEE from the University of Illinois and a PhD in electrical engineering from Stanford University.
One of the principle limitations to the growth of WiFi as a broadband medium is the lack of sufficient, affordable, backhaul. Networks that combine WiFi with DSL provide a promising solution. Since DSL uses existing copper wiring and the speed it provides is the sum of the speeds of each DSL circuit used, it offers more economical WiFi backhaul than fibre. A combination of managed WiFi and DSL is the most affordable way to obtain gigabit broadband access for all.
The state of broadband
Broadband connection-speed increase continues to be a great Internet-growth enabler. Figure 1 shows Nielson’s law of exponential connection-speed growth over the past three decades, which recently is fueled by residential use of mobile devices and future machine-to-machine enabled appliances, connected within the home by WiFi and to the Internet by fixed-line broadband connections. Increasing video-based application and content use by home communication devices particularly increases required connection speed.
Figure 1 – Nielson’s law of high-end internet-user data rate with time. Source: J. Nielson useit.com
The exploding demand for applications generates enormous commercial opportunity for content and application providers, while increased connection speed itself may derive at best only incremental additional revenue for connection-service providers. Thus, the demand for higher speed must be satisfied in a profitable way for the connection-service provider. This article provides a very economical means for such satisfaction, different from other means currently considered.
Mobile device data use appears in Figure 2 both in total and in the portion offloaded to WiFi. While emerging licensed-spectrum wireless technologies like LTE increase spectrum efficiency relative to previous wide-area wireless technologies, these technologies cannot accommodate the total growth shown in Figure 2. Offloading to ‘small-cell’ unlicensed-band WiFi is increasingly ubiquitous and necessary to ease the licensed spectra’s burden, allowing tremendous increases in data delivery in any geography.
Figure 2 – Mobile data, and corresponding WiFi offload, growth
There are roughly 700 million WiFi access points worldwide today, estimated to grow to at least two billion by 2015. WiFi is a tremendous asset for all Internet users, but it requires an equally ubiquitous fixed-line technology to ‘backhaul’ the WiFi data to the Internet.
The indicated WiFi backhaul technology today, and for the foreseeable future almost everywhere, is the rapidly expanding 400 million DSL connections worldwide. Figure 3 shows DSL’s steady large growth over the past decade, which experts expect to continue for several years. There are roughly 1.3 billion (and growing) fixed-line copper connections worldwide into which this growth can continue.
Figure 3 – Fixed-line broadband growth (source: Point Topic with VDSL plotted as DSL)
DSL dominates fixed-line broadband connections because of its economical use of existing copper phone lines to create connections. Fibre-based passive optical networks (PON) have high bandwidth, but are exorbitantly expensive. Recent publications - show that a fibre connection to a customer’s residence costs US$2,500 to US$10,000 per customer, worldwide. This investment is too high with respect to the US$20 to US$50 per month subscription fees reasonably expected for residential broadband service. Much of fibre’s high cost is the labor, right-of-way construction licenses, and time to install. No known consumer-based return-on-investment strategy justifies such expense. The cost of converting the over one billion worldwide copper connections to fibre would thus exceed US$3 trillion.
DSL is sometimes viewed as a limited broadband technology because of its connection speed. However, advances in both DSL and WiFi solve that problem, as Figure 4 illustrates. DSL speeds will rise to 100 megabits per second (Mbps) at one kilometer in the near future, while a single 802.11n 20-MHz WiFi channel can provide up to 150 Mbps over a few tens of meters inside a residence, and to surrounding residences.
Figure 4 – DSL/WiFi shared bandwidth of R1+R2+R3
A good WiFi access point can use any of over 20 available distinct WiFi channels. The coverage of many WiFi access points often overlaps, which allows us to select the one we’d like to use. It is possible to use different WiFi channels from two or more WiFi access points simultaneously. The available Internet data rate is the sum of the connection speeds when each access point has its own DSL backhaul; there is no such summing effect on a PON or cable since those media are already shared by all users – the effect is unique to DSL. The result is that the sum of the shared DSL/WiFi connections is available for all WiFi applications and devices. Such sharing already occurs on PONs, cable systems, and even on wireless 2G/3G/4G and does not fundamentally present any new security or other issues beyond what other systems already see. In Figure 4, if each DSL ran at 100 Mbps, then 300 Mbps would be shared by all the users. This shared data rate for just three systems is well above the data rate that PON and cable systems can share among the same number of users. The use of 20 access points could have shared speeds exceeding the speeds shared by more than 20 PON users, making the multiple copper/WiFi sharing much faster than a PON. Further, such a shared DSL/WiFi system is much cheaper to provide because no new fibre or cable infrastructure needs to be deployed to the home.
DSL state of the art
Currently, 100 Mbps download/30 Mbps upload DSL services are already sold in various Asian cities including Hong Kong, Tokyo, and Seoul. These services use very-high-speed DSL (VDSL) technology (ITU standard G.993.2 released in 2006), with a maximum speed of 150 Mbps. The copper lengths in these 100 Mbps VDSL services today are typically less than 300 meters, and require a fibre to the basement or to a point close to the residence. Such a fibre is sometimes economical and sometimes not. It is economical when the fibre’s cost is shared by DSL customers and the fibre is extended neither to nor into the residence. ITU standard G.993.5 (released in 2010) vectored VDSL technology extends the 100 Mbps range to one kilometer. Deployment of fibre to within one kilometer is almost always economical because the number of customers over which its cost can be divided is sufficiently large. 100 Mbps DSL speed is a good match to 108 Mbps WiFi speed, since neither communication medium slows the other.
The use of G.993.5 vectored VDSL at 100 Mbps and one kilometer, or any DSL at sufficiently high speed to strain the connection, requires dynamic spectrum management (DSM) technologies detailed in national standards bodies’ reports and standards, like the U.S.  and the U.K. . DSM methods require a high-speed server and software that routinely scan all DSLs for developing issues and make rapid adjustments of parameters to keep DSL connections stable and fast. DSM technology details are beyond the scope of this article, but below are tutorial references  - . Such management systems can implement or assist the sharing described above.
A copper line length of one kilometer is sufficient almost anywhere to cover the subscribers served by a distribution point or junction. Fibre can connect from the network to this point, a small vectored DSL access multiplexer (DSLAM) that handles a few hundred subscribers, at modest cost per customer, typically less than US$500, much less than the several thousand for fibre to a customer’s residence. Existing copper at no cost completes the individual connections to customers. Further, since the 100 Mbps DSL is about the same speed as the maximum of WiFi on a single 20 MHz channel, it provides a seamless 100 Mbps path from the consumer device through the access-point/DSL modem to the DSLAM and beyond to the Internet - presuming the fibre is fast enough to handle a few hundred such 100 Mbps users. Sharing, then, simply multiplies the 100 Mbps by the number of cooperating DSL/WiFi connections.
WiFi connections, like DSL, often perform below speed and need management. Even at the lowest DSL speeds of a few Mbps, tests show that 25 percent or more of DSL/WiFi connections are limited by unstable, unmanaged, WiFi connectivity. On those few small PON broadband networks attempted in various world cities, customer call rates remain high in part because of WiFi connectivity issues, defeating the fibre’s speed. There are many complex WiFi effects that involve multiple access points colliding on the same channel, home noises, and placement of WiFi devices. Most of these issues cannot be addressed within the access point devices and require, as DSL requires DSM, a high-speed server that statistically characterizes problems, frequency of occurrence, and most likely successful dynamic remedies within a neighborhood of access points. With remote dynamic WiFi management, it is possible to manage most connections safely and stably to 100 Mbps between users and the access points within and between neighborhoods.
It is feasible to magnify/aggregate WiFi connections. WiFi client devices increasingly have low-cost embedded WiFi transceivers, often partly in software on a processor used for other device functions. It is feasible to include more than one low-cost WiFi transceiver in devices, and a number of devices entering the market already have two WiFi transceivers. IEEE WiFi standards 802.11n and 802.11ac already mandate multiple transceiver interfaces, albeit they presumed connecting with the same WiFi access point. At an architectural level, the bonding, or separation/addition of downlink/uplink signals from/to several access points is feasible as long as a corresponding device further upstream in the network (shown as the ASSIA server in Figure 4, but there are many locations possible for this device) provides the corresponding inverse function to make the client device appear as a single internet address. The speed then becomes the sum of the link speeds.
The combination of the logical IP connection and the DSL/WiFi connections then becomes a shared medium, similar to that of a cable or PON system - even a wireless base station uses such sharing. With up to 22 WiFi channels, and enough participating DSLs, the shared bandwidth is enormous, magnifying both the WiFi and the DSL bandwidth available to any broadband customer within the system’s reach, at a very affordable cost and without the long waits for construction of fibre systems. Combining DSL and WiFi in this manner, it is possible to provide economical Gbps broadband access for all.
 Farzad, Robert, “Questions Arise about Verizon’s Big FiOS Bet,” Bloomberg Businessweek, March 21, 2011.
 Report on Google’s Kansas City Fibre Project, Telecoms.com, February 2012.
 Outside Plant Magazine, June 2011
 FCC Broadband Report, February 2010.
 “Dynamic Spectrum Management,” Alliance for Telecommunications Industries Solutions NIPP-NAI Technical Report ATIS-PP-0600007, 2006.
 Report on Dynamic Spectrum Management Methods in the UK Access Network, UK NICC std. ND 1513, January 2010.
 J. Cioffi and M. Mohseni, “Dynamic Spectrum Management,” Proceedings ISSLS 2004, Glasgow, Scotland, April 2004.
 G. Ginis, M. Mohseni, and J. Cioffi, “Vectored VDSL to the Rescue,” Outside Plant Magazine, April 2010.
 J.M. Cioffi, H. Zou, and A. Chowdhery, “Greening the Copper Access Network with Dynamic Spectrum Management,” International Journal of Autonomous and Adaptive C