Despite what you may have read elsewhere, delays in the availability of digital video and other broadband services to homes have little to do with too-costly set-top boxes and everything to do with the delivery networks. Now, potential service providers have defined several suitable wired- and wireless-network architectures. Moreover, signal-modulation schemes and the ICs necessary to deliver broadband data services to homes have finally arrived. An understanding of the applications, networks, and data-transmission schemes can help you avoid the potholes and bumps that still litter the so-called Information Superhighway.

Capacious data pipes into homes form the foundation of many emerging technologies, such as the $500 Internet appliance, multiplayer games, movies on demand, and video teleconferencing. The pipes also enable societal trends, such as telecommuting. Even the future direction of the PC industry depends on how service providers implement broadband services.

A look through the engineering classifieds in any major newspaper verifies the magnitude of digital-video and broadband-data efforts. You can also expect the next wave of high-tech entrepreneurial success stories to relate to broadband networks.

For three years, potential service providers have been bustling about planning broadband networks. Leading the interested parties are regional and national telephone-service providers and cable-TV (CATV) service providers, or "multiple-systems operators" (MSOs). The interested parties also include independent entrepreneurs and even power companies. All of the potential service providers see broadband networks as a way to boost revenue by expanding their customer bases and menus of services.

Digital TV leads the charge

Industry observers originally saw digital-TV services as the first compelling application to pay the tab for constructing a broadband network to homes. Telephone companies viewed any type of video-delivery service as a new revenue generator. MSOs, meanwhile, reasoned that the hundreds of TV channels and video-on-demand services that digital video promises could boost revenues among existing customers and compel new customers to subscribe.

MPEG forms the basis of the digital-video potential. Lossy MPEG video compression combines with robust modulation schemes that provide lossless compression. This combination allows service providers to transmit four to six video programs in the 6-MHz-wide frequency band that accommodates NTSC (National TV Systems Committee) or PAL (phase-alternation-line) analog channels.

This approach also allows a 100-channel analog cable system to suddenly provide users with approximately 500 channels, making possible near-video-on-demand services. Near-video-on-demand implies a roster of top movies that start every 10 to 20 minutes. New network architectures promise true-video-on-demand services and even the ability to use interactive, VCR-like controls, including rewind and fast forward.

While service providers were developing network architectures, they also held worldwide service trials. These trials convinced providers that video-on-demand services alone cannot justify the cost of new networks with interactive capabilities. Consumers will use the service only if it matches the price of local video-rental businesses that have spread from specialty stores into supermarkets and convenience stores.

Service providers aren't ready to abandon enhanced-video services, as some recent news publications would have you believe, however. Rather, the service providers now recognize video service as one of several applications that must contribute to the cost of big data pipes to the home. Moreover, the providers acknowledge that a combination of analog and video services will coexist into the next century.

Despite a bearish turn on the outlook for video on demand, the MSOs and telephone companies are still sprinting to install broadband networks to the home. The reason is twofold.

The new "killer application"

First, the service providers intimate that they have found the killer application that can pay for the broadband networks: high-speed Internet access. Several of the network architectures under consideration can deliver substantially higher two-way data rates than even integrated-services digital-network (ISDN) phone lines. Moreover, users can employ the data services for far more applications than Internet access, including telecommuting, access to commercial on-line services, video teleconferencing, and multiplayer games.

Realistically, however, a second reason behind the aggressive network-deployment plans of MSOs and telephone companies proves to be the true driving force: Deregulation has opened the door to direct competition in providing telephony, video, and data services.

Both the MSOs and telephone companies believe that they must offer low-cost broadband services to continue to prosper. The competition will benefit consumers by providing better services at lower prices, and the huge investment in R&D, products, and services can only be good for the electronics industry.

Potential service providers are pursuing several network architectures in an attempt to quickly offer broadband services to as many customers as possible. The most aggressive organizations are laying new fiber- and coaxial-wiring plants based on robust network architectures, called hybrid-fiber coaxial (HFC) and switched digital video (SDV). Each of these architectures will support telephony, video on demand, and two-way data services. Before examining HFC and SDV, however, note that some providers are implementing several other network options. Decisions about these potential implementations involve geography, potential subscriber base, existing infrastructure, and time to market.

The most widely installed digital-video service today is the Direct Satellite System (DSS), which countrywide consumer-electronics stores are selling for $500 to $800 (Reference 1). Approaching 2 million units sold, the systems have exceeded even optimistic market projections. Many of the sales, however, have been to customers in rural areas who previously lacked any CATV service.

The DSS technology offers near-video-on-demand service and more than 100 channels of digital video with picture quality exceeding that of analog TV. Still, the technology has limitations. For example, a shared broadcast media, such as an array of satellites, can't offer sufficient dedicated bandwidth to any one customer to support true-video-on-demand service or Internet access. Moreover, even if you solve the dedicated-downstream-bandwidth issue, the satellite architecture lacks an upstream data path for accessing any two-way data services.

MMDS and LMDS services

The market success of DSS, however, has attracted the interest of some service providers, particularly the telephone companies. The companies would like to bring DSS-like video service on-line quickly to areas in which the phone companies lack a video-capable wired infrastructure. The companies can't offer two-way data services but see wireless-digital-video services as a way to compete immediately with established MSOs.

Several of the Regional Bell Operating Companies (RBOCs) are aggressively moving forward with a terrestrial wireless technology called Multichannel Multipoint Distribution Service (MMDS). The microwave line-of-sight technology can deliver more than 100 digital video channels over approximately 30 miles (see box, "Wireless MMDS goes digital").

Pacific Bell (San Francisco), for example, plans to deploy MMDS services in the Los Angeles basin. The Los Angeles area is particularly appropriate for MMDS due to geography. From two hilltop transmission antennas, Pacific Bell can offer line-of-sight service to 5 million customers.

An even more robust cellular-like wireless service is on the way. Called Local Multipoint Distribution Service (LMDS), the technology can carry several hundred digital video channels, telephony traffic, and two-way data services. Real deployment is at least a year away, but public trials should begin this year (see box, " Cellular-like LMDS wireless loop").

ADSL supplements analog video

Other RBOCs are looking at ADSL asymmetrical digital-subscriber-line (ADSL) services as a way to offer video on demand and two-way, high-speed data services over existing phone circuits (see box, "Existing phone lines host ADSL"). RBOCs can offer ADSL today by adding an ADSL interface in a "central office" (in phone-company terminology) and a home—a card on each end of the local loop.

ADSL can deliver as much as 8 Mbps of data downstream and 640 kbps of data upstream. The downstream bandwidth is plenty for a high-quality MPEG-2 video program. Moreover, an ADSL link to the Internet could offer better performance than that office users who connect via LANs to the Internet enjoy.

Analog Devices now offers a modem IC, the AD6444, that complies with the ANSI ADSL standard. Aware Inc developed the discrete-multitone (DMT) modulation scheme the chip uses. Analog Devices is currently offering samples of the chip and expects to offer a volume price of approximately $170. Aware has also introduced a box-level modem using the IC. Aware calls the $2500 modem the Internet Access Transceiver and targets it at evaluations and trials of the ADSL technology.

Most RBOCs that are pursuing ADSL see it as a transitional technology, especially because it can carry only one video program. Customers will likely use it, therefore, to supplement analog cable. The RBOCs plan to offer ADSL and move to more robust networks as new wiring plants emerge. The RBOCs can, in turn, redeploy ADSL equipment to other customer areas.

HFC architecture

The RBOCs, MSOs and other would-be service providers see either HFC or SDV as the future network of choice, and most have taken one side or the other with a near-religious fervor. Figure 1 depicts the HFC architecture. HFC uses fiber optics to deliver a broadband signal from a central office or "head end'" (in cable company terminology) to nodes that typically serve fewer than 500 homes. A coaxial bus connects the node to the homes and carries a mixture of analog video, digital video, telephony, and data.

HFC networks will use a segmented frequency spectrum to handle signals. Pacific Bell's HFC network, which the company is deploying in San Diego and the Greater San Francisco Bay area, provides an example of how to mix these different signals on one cable. Lucent Technologies (formerly, AT&T Network Systems) was the primary developer of Pacific Bell's HFC implementation. The Pacific Bell system assumes a total bandwidth of 750 MHz—the total bandwidth available on the most advanced analog-cable systems.

You can use analog NTSC video channels in their normal frequency range of 54 to 550 MHz to carry analog or digital video that is broadcast to all subscribers. The 550- to 750-MHz frequency range is dedicated to narrow-cast downstream services, including telephony, digital video (true video-on-demand), and data.

The 5- to 40-MHz passband, meanwhile, carries upstream telephony and data. Portions of the upstream and narrow-cast downstream spectrum is dedicated to telephony traffic, ensuring service to all telephony customers. The remaining upstream and downstream capacity is available on demand for video or data uses to customers.

Pacific Bell's network connects 480 homes per node and uses the 64-QAM (quadrature-amplitude-modulation) approach to transmit data downstream and quadrature phase-shift keying (QPSK) to transmit data upstream. Standards organization the Digital Audio-Video Council (DAVIC) (Geneva), have specified these modulation techniques for HFC networks, and independent CATV research organization Cable Laboratories (Louisville, CO), funded by the leading MSOs, has tested these techniques.

Pacific Bell claims that its network can provide each customer an average of 4-Mbps downstream and 140-kbps upstream transmission in narrow-cast capacity for data or video-on-demand services. The company promises to double the capacity with signal-coding enhancements over time. Moreover, the architecture also provides for reducing the number of homes (to 240 or 120) that share narrow-cast services in areas with many consumers of those services. Such a change would boost available bandwidth with a more granular network topology.

Switched digital video

The SDV (Figure 2) architecture, also called "fiber to the curb" (FTTC), pushes fiber optics deeper into neighborhoods than does HFC. SDV is essentially very-high-speed ADSL (VDSL) over shorter loops. SDV implementations typically require an optical network unit (ONU) within 1000 ft of a home. Typical central offices in phone networks, on the other hand, can be 12,000 ft or more from a home.

The ONU in an SDV network can serve a maximum of 64 homes; 12 to 30 homes per ONU are typical. A 1-Gbps fiber link connects an ONU with the host digital terminal (HDT) in a central office. A dedicated twisted-pair or coax cable connects each home to the ONU. As the name implies, SDV networks deliver dedicated switched bandwidth in the form of asynchronous-transfer-mode (ATM) packets to each home.

DAVIC has specified 16-CAP (carrierless amplitude/phase modulation) downstream signal modulation for SDV networks, yielding a 51.84-Mbps ATM stream. (The number 16 refers to the size of the signal constellation; see box, "Digital-video glossary. ") QPSK yields a 1.6-Mbps return channel. The architecture can simultaneously deliver six digital video programs to the home. Moreover, the subscriber can dedicate as much of the bandwidth as needed to two-way data services.

Figure 3 depicts the frequency spectrum of an SDV implementation. Standard telephony services and ISDN can travel over SDV networks, just as in normal phone loops, in the band below 1 MHz. The downstream 16-CAP and upstream QPSK signals also fit conveniently below the standard FM audio and analog-video frequencies.

SDV implementations that use twisted-pair wires between the ONU and home can't carry analog video traffic. Instead, the SDV and analog video signals combine at the side of the house onto a coaxial cable. Once in the house, a passive device splits video and telephony signals.

The same or different service companies could provide SDV and analog services. When a single company supplies both types of service, a coax drop from the ONU to the home can carry both SDV and analog video.

Analog support required

After you examine the SDV and HFC architectures, you may wonder why analog video support is important. The reasons are more logistical than technical. For one, all-digital video requires the consumer to pay for a set-top box at every TV. All-digital TV is unlikely to become a reality until all TVs directly integrate MPEG decoders and digital networks are as ubiquitous as NTSC; these events could happen by 2010.

HFC and SDV have other surprising similarities. Both types of networks require an identical telephony backbone and an identical switched-ATM network behind the central office or head end. The ATM network provides video and data services, and the telephony and data networks will eventually merge into one network of ATM packets running over the synchronous optical network (SONET).

Both networks also require power distribution near the home. In SDV implementations, power must come to the ONU, both to provide a telephony ring voltage on the local loop and to power active ATM electronics in the ONU. HFC networks require power all the way to the side of the house, so that a ring voltage can be present for telephony services.

Traditionally, telephone companies insist on supplying power from centralized locations with backup power present. Centralized power distribution allows the phone network to operate throughout power outages.

Companies planning to deploy SDV networks are planning to use a coax link (side by side with the fiber optics) to the ONU for power distribution. Such an implementation simplifies offering analog video service over the same coaxial cable.

HFC networks, meanwhile, will likely send power down the coaxial cable all the way to the home. Companies planning to deploy these networks view local power and batteries as too risky for telephony services.

Deploying HFC and SDV

Discussing the theory behind HFC and SDV is simple compared with making the networks operate in real deployments. The challenge with SDV is investment in infrastructure, because no service provider has anything approaching a suitable cable plant in place, except for trial networks that cover small areas. Moreover, SDV requires that a service provider invest in the full interactive video and data capability up front rather than slowly phase in interactive services.

Service providers can now buy SDV equipment from the partnership of Lucent Technologies and BroadBand Technologies. Both Bell Atlantic and Southwestern Bell will roll out SDV networks this year.

With an infrastructure in place, designers should have little trouble making the point-to-point SDV transmission scheme work. Lucent Technologies Microelectronics Business (formerly, AT&T Microelectronics) offers the Multipoint Broadband Access family of ICs that addresses the requirements of both the ONU and the set-top box. The three-chip set for the set-top box costs as little as $45 in volume. The chip-set price means that an SDV data-modem card for a PC could cost as little as $100.

HFC networks offer service providers the advantage of phasing in enhanced services as demand warrants. HFC's architecture is similar to that of the installed base of cable plants. The installed plants, however, often serve as many as 2000 homes from a single coax tap. HFC will require more fiber nodes to reduce the number to 500 or fewer. Moreover, only a few percent of the installed cable systems currently offer any upstream communications.

MSOs have been busily upgrading their wiring plants to support HFC, and other providers, such as Pacific Bell, are installing new plants. At first, the companies will continue to offer analog services as standard fare and digital video, telephony, and data products as premium services.

Figure 4

From a technology perspective, making video-on-demand work over an HFC plant is relatively simple. The upstream requirements simply entail low-speed data conveying menu choices or VCR-like controls. Sending digital video downstream requires merely modulating a digital bit stream in an analog frequency slot.

Telephony and data services, however, substantially complicate the picture because of the need to send higher data rates upstream. HFC networks' point-to-multipoint topology makes the upstream data transfers problematical for several reasons.

First, noise, which the CATV industry calls "ingress noise," enters the coax from vacuum cleaners, microwave ovens, and other ac appliances in all houses using the network. The noise makes it difficult for the HFC nodes to accurately decode upstream data.

Second, no standards exist for sharing the narrow upstream channel. Moreover, an application such as Internet access has far different media access requirements than phone service. Pacific Bell, for example, will use time slots to schedule upstream QPSK telephony and data traffic. Such a media-allocation scheme, however, fails to provide optimal usage of overall bandwidth. Even Pacific Bell admits that new schemes will evolve.

The ICs to handle upstream and downstream modulation are available from several vendors. Broadcom has perhaps the most highly integrated HFC modem chip, BCM3115 QAMLink dual-channel receiver. The $70 (1000) IC integrates a QAM receiver with forward error correction and a QPSK transmitter. The QAM receiver supports 16-, 64-, and 256-QAM formats and can also operate in SDV applications to receive 16-CAP signals.

Other companies supporting HFC networks with ICs include LSI Logic, Philips Semiconductor, and VLSI Technology. All three of these companies offer QAM and QPSK chips as standard parts, and LSI and VLSI also support the functions in ASIC libraries.

The available chips operate in video, data, and telephony applications. Data access—the so-called cable modem—currently has garnered the spotlight, because the service providers believe that users will readily pay for high-speed Internet access. Even MSOs that are far from ready to roll out digital video or telephony services want to support cable modems for data applications.

Today, no fewer than a dozen companies, including such industry stalwarts as Intel and Motorola, have demonstrated cable modems. The modems take a fairly standard approach to getting data downstream. One or more of the 6-MHz NTSC channels is dedicated to downstream data, yielding from 4 to 40 Mbps of bandwidth, which all users connecting to the same HFC node must share.

At worst, the downstream data flow should support performance similar to what a user of a busy LAN experiences. The upstream implementations range from using the telephone and a dial-up modem to various modulation and media-access techniques.

At least three organizations are now working on modem standards for HFC. DAVIC and Cable Laboratories will likely establish recommended standards by April or May. The leading MSOs will almost surely follow Cable Laboratories' lead, because the MSOs fund the organization. DAVIC, on the other hand, primarily comprises vendors to the service providers and could develop a different standard. The IEEE 802.14 committee is also exploring cable modems but is a year away from a spec. The market will almost certainly have set a standard by then.

Chaos dominates the cable-modem arena, and, similarly, the QAM, QPSK, and CAP modulation schemes for use in HFC and SDV networks are making some companies unhappy. Zenith, for example, has always championed its vestigial-sideband (VSB) modulation as more efficient and requiring less signal-processing power than QAM. The company is also using VSB in its cable modems, and the industry has slated the scheme for use in high-definition-TV transmission.

Analog Devices and Aware state that DMT and discrete-wavelet-multitone (DWMT) modulation perform better than the established schemes. According to the companies, an SDV DWMT implementation could match 16-CAP performance at 1000-ft cable runs and longer runs at reduced rates (26 Mbps at 3000 ft and 13 Mbps at 5000 ft). Analog Devices also plans to offer a DWMT IC for use in HFC telephony that could support triple the traffic that Pacific Bell's implementation supports.

Newcomer Orckit Communications is touting another alternative for SDV systems. The company has developed its own version of QAM that it claims could operate at 51.84-Mbps at 500m distances.

Despite the progress in implementing broadband networks to homes, the rest of the year should reveal even more traffic bumps. Big players, including Motorola (Phoenix) and National Semiconductor (Santa Clara, CA), are poised to add chip sets and, potentially, their own modulation schemes to the fray. By year's end, however, systems will be operating, and most standards will be in place.

You can reach Technical Editor Maury Wright
at (619) 748-6785; fax (619) 679-1861.

1. Schweber, Bill, "Direct satellite broadcast brings downlink directly to your set-top box," EDN, Dec 21, 1995, pg 53.