The cable modem is the latest craze to hit the consumer side of the US computing market. Whilst much has been said and written about the benefits of the cable modem, the subject is technically somewhat more complex than may be immediately apparent. Moreover the primary vendors of access to the underlying infrastructure are techno-culturally firmly embedded in the analogue world, and thus it is reasonable to expect some teething troubles until the technology matures. The promise of the cable modem lies in the ability to deliver Ethernet like speeds to the household for cable television like prices.

The cable modem is clearly a threat to the established telecommunications providers, particularly to supposedly "fast" services like ISDN. As such it is also a potential threat to the voiceband modem industry, who are constrained in throughput performance by the slow medium which they are tied to. Indeed the sudden burst of activity in the Telco camp, aimed at producing standards for high speed transmission over telephone twisted pair (ADSL/SDSL/HDSL/VDSL - to be covered in a following feature), is directly attributable to this situation. Properly implemented, the cable modem industry has the potential to blow a fatal hole in the telco's long established and comfortable monopoly on the delivery of domestic, and possibly commercial digital communications.

For us humble consumers, long suffering over-inflated pricing for such services, the competition presented by the cable modem industry has the potential to drive prices per bandwidth down quite significantly. Viewed in the broader context of the provision of Internet Protocol services to households and businesses at the lowest possible cost, this development has the potential to be a critical enabling technology for the Internet and multimedia industries. To fully develop the potential of the cable modem, the computer industry will need to throw its weight behind the cable modem industry and make the necessary investments to allow seamless integration at the earliest possible point in time. Whether this occurs remains to be seen. Our industry has suffered technological strategic myopia on many occasions in the past, and the potential for a significantly bigger collective pie to feed from has often taken second place to very short term objectives in carving up and plundering existing markets.

Cable TV Technology and Networking

The cable TV technology base has been around for many decades, and is a well developed and mature service in the US. The central idea in this model is to distribute TV carriers to household sockets through a coaxial cable. A single cable can be threaded through multiple sites, with a high impedance tap delivering the signal to the user's wall socket. Cable TV is inherently a broadcast service, analogous to a file server providing one way traffic to clients. This presents some obstacles for the provision of digital data services, but not insurmountable obstacles.

The classical CATV topology is a tree like structure, with branches fanning out from linear segments. Because a coaxial cable can provide a bandwidth of about a GigaHertz, cca 150 television channels can be carried on the single cable. With about 6 MHz bandwidth allocated to a channel, and typically only a fraction of the channels used, it was only a matter of time before somebody would think of exploiting the unused bandwidth and extensive cable infrastructure market penetration for household data transmission.

Interestingly, the connection between Ethernet and cable TV technology is historically strong. When Metcalf and Boggs originally conceived Ethernet, they merged two then quite well established ideas - packet radio networking and CATV cabling. Ethernet uses the packet radio collision detection model, where multiple hosts using radio modems contend for a single carrier, and share the same radio frequency channel. All hosts continuously listen on the shared carrier frequency.

To translate the model into a cabled, baseband environment, they used the most readily available technology base at the time - the CATV scheme. The original Ethernet trials were conducted using cable TV cables, terminators, taps and tapping tools. It was only when the technology was translated into a defacto standard, that the 75 Ohm TV cable was supplanted by a lower impedance 50 Ohm cable. Nevertheless, the 10 Base 5 backbone Ethernet uses a very similar linear segmented scheme as CATV does, with an actively buffered twisted pair drop cable, rather than a passive transformer coupled coaxial drop cable typically used in CATV.

The simplicity of the CATV scheme was possible because there was no upstream component of the signal, ie all traffic is downstream from the central distribution site to the users. Because Ethernet had to provide a symmetrical bidirectional transmission scheme, the transceiver ("vampire") box had to be able to transmit and receive which required buffer hardware and a power feed from the host.

While Ethernet rapidly overwhelmed the market, crushing numerous proprietary schemes as well as the token ring family of protocols, variations on the Ethernet scheme appeared. These where designed to exploit existing CATV wiring by asserting Ethernet packets as modulation on a radio frequency carrier, rather than the baseband Manchester code used by Ethernet 1, 2 and 802.3 . These schemes are the historical precursors to the cable modem, and can be regarded technically as a resurrection of the packet radio model, using a cable rather than open air transmission.

The idea that the cable modem is something new is quite wrong. The cable modem is merely an evolved variant which combines attributes of a number of schemes dating back to the seventies.

The fundamental limitation of schemes which attempted to modulate RF carriers with Ethernet-like packets was bandwidth utilisation. If you wished to squeeze the signal into a 6 MHz bandwidth, it was at the expense of channel throughput. If you wished to get full Ethernet bandwidth, you would end up devouring at least two channels. In either instance, you would require the radio frequency equivalent of an Ethernet transceiver, with a baseband drop cable to the host.

To translate this idea into something which could be piggybacked on to the existing CATV infrastructure required some further technological evolution.

Cable Modems and Market Players

Three central technical issues constrain the cable modem. The first is that the medium and topology is not built for a two way symmetrical service. Whether passive or active taps are used, problems arise with sending traffic upstream from the subscriber. In a wholly passive tap arrangement, the typical problem is in getting enough signal pumped into the backbone, which needless to say concentrates noise and interference from the large number of user sites at the upstream distribution point. In an active tap arrangement, the one way repeaters simply block upstream traffic.

A further issue in this context is the increasing use of hybrid fibre coax schemes (HFC), in which optical fibre links are used to distribute the signal to suburban repeaters, which use either existing coax or new coax cables to distribute the signal to the users.

The second problem lies in bandwidth efficiency. An Ethernet-like signal must be squeezed into a 6 MHz bandwidth. This calls for some clever thinking, particularly since you may have several thousand subscribers contending for a few dozen carriers on the cable.

The third problem lies in providing a suitable interface at the host end. Does one use a proprietary interface, an Ethernet interface, or does one simply devise a new standard ?

A cable service provider seeking to implement a cable modem network will have additional requirements, such as network management, IP address management and importantly pricing.

The US cable modem industry is at this time in a similar position to the networking industry in the mid eighties. The IEEE 802.14 Cable Modem standard is still being defined, while a plethora of vendors is contending for the existing customer base with a range of generally similar but essentially proprietary and mutually incompatible schemes. What we can expect to see in the short term, is specific CATV providers fitting their networks for particular vendor's proprietary cable modem schemes. All subscribers on that particular network will be locked in to a particular vendor's cable modem protocol and modulation, and more than likely a subscriber will rent the modem and service from the CATV provider.

Once the 802.14 standard is defined providers will progressively across to the generic standard. Users will then be able to supply their own cable modems, and rent the service connection alone from their provider. This is not unlike what we have seen with voiceband modems and the now ubiquitous Ethernet interface.

While the 802.14 spec was at the time of writing still being defined, it is expected to use a QAM 64 scheme delivering 27 Mbps downstream, and QPSK delivering between 1.5 to 2.0 Mbps upstream. One of the difficulties with the 802.14 spec is that it is intended to support the error rate requirements of a range of different types of service (eg video, data, multimedia). Two dozen different MAC level protocols for modem to modem signalling have been evaluated by the IEEE committee members.

At the time of writing no less than nineteen international vendors were contending for the market. These include ADC, AT&T, Com21, DEC, First Pacific Network, General Instrument, HP, Hybrid Networks, IBM, Intel, LANcity, Motorola, Netgame, Nortel, Terayon Corporation, Toshiba, West End Systems and Zenith Electronics. The performance offered for upstream channels varies between 96 kbps and 15 Mbps, and for downstream channels between 500 kbps and 30 Mbps. Features sets also vary widely, as do available modem interfaces. Some modems are implemented as plug-in cards, which rely on a device driver executed by the host to deliver the smarts.

Closer to home we are unlikely to see rapid movement on the cable modem modem front as out CATV providers are still climbing the learning curve on the basic analogue technology. It is not unreasonable to expect to see alliances develop between the larger ISPs and the CATV industry, over the next 2 to 3 years. This would be the safest strategy for the CATV vendors to pursue. Some interesting possibilities could arise from a widely deployed cable modem and CATV base. One is the provision of full bandwidth between sites in metropolitan areas, using private links between routers rather than the expensive and skinny Telecom national backbone. Another is the substitution of fixed ISDN point to point services with virtual circuits over the cable modem network. Unless Telstra is protected by legislation, it stands to lose a substantial proportion of its lucrative commercial monopoly to other players.

It will be most interesting to see future developments in this area.