Cable network was originally designed to broadcast analog TV. The signal was expected to come from headend only. Other end points were all receivers. Therefore there was very little problem. Even some negligence such as unterminated splits can have little effect or limited effect to small area only. This fact allows amateur installer to make a lot of minor mistakes that later can cripple the system when converting to two-way.

The network can carry signal from 5 MHz up to 800MHz depending on the amplifier capability. The spectrum is divided into 6MHz channels for NTSC system and 8MHz channels for PAL system. Any add on services have to be located properly on the spectrum so that they will not interfere with the existing channels.

Cable network with two-way capability is built with amplifiers that work both ways. We do not have to concern about the media since they can naturally deliver signal both ways. However, there is a problem about how to distinguish signals from each side of the network. Normally this is done by using one portion of the spectrum in one way and the other portion in another way.

Main problem with two-way network arises from the network topology itself, especially in the distribution network. The network has a tree structure which contains many branches and leave nodes holding together with passive devices like splitters and taps. Signal that travels down the tree is attenuated at each splitter and tap. The noise is also attenuated at these points. Therefore, the signal to noise ratio is not getting much worse down the stream. However, in the upstream direction, splitter's outputs become inputs and inputs become outputs. The signals travelling up the tree are combined and so are the noises. The result is that headend receives signal from one leave node but noise from all nodes and joints. This problem is called "noise funneling" or "ingress" which significantly decreases the signal to noise ratio in the upstream signal. The channel capacity directly depends on this signal to noise ratio.

Most noise comes from external sources such as electrical appliances, radio signal, electrical motors, etc. The noise leaks through splitters, taps, or even amplifiers. An amateur installer can also contribute to this part by ignoring loose end cable or making bad splits and joints. Fortunately, we can ease this problem by reducing the size of the distribution networks. For example, splitting a 5,000 node network into ten 500 node networks may increase the signal to noise ratio by 10 dB.

The portion of the spectrum that we use for upstream also has effect. High frequency signal has more attenuation than lower frequency over the same distance. To get a better signal to noise ratio, we should use the lower portion of the spectrum for upstream communication. However, since low frequencies are mostly occupied by analog TV signal, there is not much room left for new applications.

Each NTSC channel occupies 6MHz bandwidth. We usually allocate bandwidth from 42 MHz up for downstream signals and 5-42 MHz for upstreams. A single downstream channel has been tested that it has reliable capacity of up to 43 Mb/s. For a hundred-channel system, aggregate bandwidth can be up to 4 Gb/s. Upstream capacity for a single channel is a much lower due to noise problem. It is approximately 20 Mb/s for a single 6 MHz channel. Four 6 MHz channels in the range 5 to 42 MHz can accomodate just 80 Mb/s (in a real system, an upstream channel occupy less than 6 MHz).

Reallocation of spectrum usage has been considered seriously since digital television was proposed. If the digital television technology is successful, we will be able to transmit more channels into lesser bandwidth using digital compression technology. That means we will have more bandwidth for upstream communication and other applications.

Analog to digital target dates:

12/1998 : Major networks will offer digital broadcasts to the top 10 television markets.

2001 : All commercial networks will broadcast in digital.

2002 : All public broadcasting networks will broadcast in digital.

2006 : Analog spectrum will be returned to the FCC for auction.

source: Gecko Research and Publishing

The regulation for cable network is not as strict as the one imposed on telephone network. The system therefore has less reliability as a result. Most of the problems are power failures along the line which can be improved by inserting backup power sources, and by using amplifiers that can automatically bypass signal when they fail. Subscriber end devices can also cause trouble to the network when they fail. We need more sophisticated end device which automatically detach itself from the network when failing.

Cable network has very little physical security. It is unlike a corperate lan which we can restrict physical access or a dedicated service line that no one can access other lines unless intentionally tapping into. Cable network is a hugh shared network. Any signal transmitted over the cable network can be intercepted or forged by anyone who share the same distribution network. This is a major weak point which has been exploited by other competitive technologies in the same market. To overcome this problem encryption schemes were introduced into the system.

The cable medium is shared by all subscribers hook up to the same distribution network. The maximum delay on the cable can be estimated by the longest path between a subscriber and a headend which can be more than ten miles long. This maximum delay can be measured in the range of tens of milliseconds. Conventional shared medium scheme cannot put to work here. We have to find a new effective way for media access controlling. Several groups attempted to propose specification on how to access the cable network. There are three major groups that pushed their ways through.

The IEEE 802.14 working group was organized during the early 90's

to develop a specification for data over cable network. It was hoping

that the specification will later become an international standard. However, its attempt were not satisfied by a group of major vendors and cable operators in the US which later formed a consortium called Multimedia Cable Network System (MCNS). MCNS started working on its own

specification for North America and later the specification became an international standard . However, IEEE802.14 attempts was not all wasted. Its specification for advanced physical layer has been widely adopted even by MCNS. The IEEE802.14 working group still continue working with other groups such as ATM forum, European vendors and cable operators, on other specification.

This specification was developed by the groups of IEEE 802.14, ATM Forum, European companies, and DAVIC(Digital Audio Visual Counsil) based on ATM technology. The specification has been recognized as a preferred technology by European vendors and cable operators despite the ITU recognition of DOCSIS as an international standard.

A group of North American major cable operators has established

a private organization call CableLabs since 1988 to do research and

development for their industry. This group later join North American vendors to form MCNS to develop a specification later known as DOCSIS (Data Over Cable Service Interface Specifications). It was submitted to ITU in March 1998 and has been accepted as an international standard. All materials in this report are based upon DOCSIS standard.

OPENCABLE is a CableLabs project for developing a new generation of interoperable set-top boxes. This new standard will make the system requires less amount of hardware (for conversion between different operators) and reduce the cost of set-top boxes.

The OpenCable specifications will include the following:

Downstream: 64/256 QAM

Upstream: QPSK/16 QAM

Digital video: MPEG-2

Audio: Dolby Audio AC-3 Compression

Signal security: GI's DES Encryption system

High-speed connections to the Internet: DOCSIS

The dataflow in cable network is asymmetric, with abundant of bandwidth in downstream direction and relatively very small bandwidth in the upstream direction. Therefore we divide the MAC protocol into two separate sections. The upstream MAC protocol is a centralized protocol with headend acting as the center hub of the network. The downstream MAC protocol is more like the one used in a conventional shared medium.

Upstream MAC scheduling

An upstream channel is divided into basic scheduling time unit called "minislots." A minislot can be as short as 32 bits or 12.5 m s at 2.56 Mbps or 200 m s at 160 Kbps. The headend will inform downstream modems of minislot structure in the next scheduling period. These minislots are categorized into several types according to their usages. Some are used in contention-based which are shared between modems that make requests. Some are used in unicast for specific modems to send data bursts. Some are used for maintenance and controlling purposes.

Headend has an admission control algorithm to determine whether or not a flow can be accepted. Once a flow is acccepted, it issues the corresponding modem a new service identifier (SID) and control the flow using a token bucket. The modem can send data through this SID only if it complies with SID classification. An SID may be created, deleted, modified by headend at anytime. This SID can be compared to an ATM virtual circuit that it provides a logical channel with certain QoS.

There are 4 modes for upstream flow each corresponding to different classes of QoS:

1 Unsolicited Grants

When a flow is accepted in this mode, the CMTS schedules fixed-sized grants periodically to the modem. The modem do not have to contend for the channel.

2 Real-Time Polling

In this mode, the CMTS reserves some time slots and periodically unicasts request polls to the corresponding modem. If the modem answers the polls, the time slots are assigned to it. If no answer is made, the time slots will be reassigned to other flows.

3 Committed Information Rate

In this mode, the CMTS forces modems to use contention-based request. Each time a modem needs to send a datagram, it has to make a request.

4 Tiered Best Effort

This mode is just best-effort service combining with Layer 2 priority mechanism. There are eight priority levels defined at this time.

Upstream Frame Format

(Variable length)

Downstream MAC scheduling

The downstream MAC protocol is very much simpler because there is only one transmitter and all the rests are receivers. Besides, there is a lot of bandwidth available. Separate channels are assigned to different class-of-service.

Downstream Frame Format

We do not concern much about quality of service in the downstream direction because of its hugh bandwidth. We can assign a whole channel (or even several channels) to each class of quality of service. However, we are more concern about the upstream direction. Four modes of flow are defined so that certain class of quality of service can be met within each mode.

Mode 1 Unsolicited Grants

This mode was designed to serve as close to constant bit rate service as possible. Any delay or jitter can be controlled by the timing of the grants. Voice and video flow can be comfortably supported by this service.

Mode 2 Real-Time Polling

This mode is comparable to variable bit rate service. However, the unicast polling method requires some extra latency. Variable bit rate voice and video signal can be supported in this mode.

Mode 3 Committed Information Rate

This mode is for non-real-time flow. Using a fair contention and queuing algorithms, Minimum upstream bandwidth can be guaranteed to a flow.

Mode 4 Tiered Best Effort

This mode is for a flow that does not have any specific delay, jitter, or minimum bandwidth requirement. Regular Internet traffic such as WWW, FTP, email, can be assigned to this mode.

Layer 3 IP level

The QoS controlling mechanism in this level is defined by the IETF. There are identifiers which can be used: RSVP classifier and IP Type-of-Service byte. IETF is working on how to merge the RSVP with the Layer 2 technologies. In the cable network, it is safer to assume that on the other end is not connected to a cable network and layer 2 alone cannot guaranteed the required QoS. Layer 3 mechanisms should be employed in all cases. We can allow usage of layer 2 mechanisms in some specific cases. However, this will make system more complicated due to the fact that it has to identify whether to use layer 2 or layer 3 or both.

Most cable operators are still concern about the amount of data passing through their network. Therefore, they are quite reluctant to offer new kind of services until they are sure that certain amount of profit can be generated. Internet access for home users is a still major target. In the near future, more services will become available.

Most of the applications we can do over the cable network now are those of the IP applications. They are

- Internet access

- IP telephony and fax services

- Video conferencing

- Webcasting

- Virtual Private Network

Some cable operators offer a VPN service which allow employees to work from home connecting through cable modem while the company is connected through a dedicated T1 or T3 line. This type of service is expected to generate even more revenue than the Internet access service.

- Educational network/Distance learning

- Home monitoring/Security

FFTx is a series of networking technologies that connect users and the central office via optical fiber. Although the cost of optical fiber is not much different from coaxial cable, optical equipment are very expensive. However, some Japanese companies claim that the cost for building a FFT network is not significantly different from building a HFC network in Japan. Therefore, this technology is expected to be developed within Japan while other parts of the world prefer other cheaper technologies.

The cost for ISDN BRI has been significantly reduced over the years. However, its bandwidth is still limited to 128 Kbps which is incomparable to multimegabit per second of other technologies with about the same (or even lower) price. The only advantage is that it is more widely available and claimed to be more reliable. The speed of ISDN PRI is comparable to that of cablemodem but cost is too high.

The concept of using satellite to provide high speed data channel is the same as in cable network except that it is nearly impossible to build a two-way system. All of the services offered need a telephone line for upstream channel. The cost is also still too high to be competitive in the public market. However, it has one advantage that other technologies do not have. We can access its service from anywhere that the satellite signal covers. In some systems, this can be anywhere around the world. This technology is more suitable for field communication.

xDSL is a series of Digital Subscriber Line technologies which use existing telephone line to deliver high speed data. Optical fiber will carry signal from central office to a neighborhood node which convert signal and put into a telephone line to user. In some cases, a direct copper line can be used if the distance from central office to user is not too great. DSL has a similar problem that upstream bandwidth is much smaller than downstream (except in symmetric system which is much more expensive). The cost for building this network is still higher than that of HFC. However, in the long run, it would be comparable to the cable network. If cable network cannot solve reliability and security problem, this technology may win in the competition. The following tables compare DSL with cable modem.

source: 3com, March 1998

source: Wired Magazine, May 1999

The cable network has a lot of potential. Although it has some problems, these problems have already been identified and they are not too difficult to solve. We just need a little more time. Cable network may not be suitable for all kinds of networking needs. However, it is obvious that it can serve very well in some part and will become one of dominant technologies for a long time.

[1] Robyn Aber xDSL Local Loop Access Technology Delivering Broadband over Copper Wires, Feb 23, 1999

[2] David Gingold Integrated Digital Services for Cable Networks, Sep 1996

[3] Jessica L. Kemp Cable Modem Impact on Service, Dec 10, 1996

[4] Walter S. Ciciora Cable Television in the United States - An Overview -, May 25, 1995

[5] Cisco Systems Inc. Quality of Service in Cable Data Networks, Dec 1998

[6] CableLabs Inc. The Cable Connection: The Role of Cable Television in the National Information Infrastructure, 1995

[7] CableLabs Inc. DOCSIS Radio Frequency Interface Specification SP-RFI-104-980724, 1997

[8] http://www.catv.org/modem