• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Creating a Wireless Home Network for Video

Posted on: Saturday, 20 August 2005, 03:00 CDT

What are the technical requirements for deploying reliable wireless networks to stream video around the home? Alex Ashley & Ray Taylor assess the suitability of current technologies and suggest a way they can be utilised

Over the next decade we will see a huge increase in the use of in- home networks for sharing printers, Internet access, music collections and television programmes. Some of these home networks will have been purchased as components from retail outlets and some home networks will have been deployed as part of a subscription service offering.

In both business models the consumer will expect their home network to be easy to set up and work reliably, but in the case of the subscription business model, the consumer's expectations are likely to be more demanding because they will be less tolerant to failures.

Installation costs and ongoing support costs are an issue to service providers because they erode profits by increasing the cost of acquiring subscribers and increase the risk of churn. The home network must be easy to install, easy to manage, highly reliable and it must be simple to diagnose and correct faults.

The focus of this paper is the concept of a home network that is able to stream video content with levels of reliability that consumers currently expect from broadcast television services (ie working at least 99.8% of the time).

Equipment costs for the storage devices ("server") and playback devices ("client") are an important design factor. In neither case are networking features likely to be the main driver for purchase, and the incremental cost of the networking will need to be small.

Client devices will probably be attached to secondary TVs in bedrooms, the kitchen etc. The client device is merely an enabler that allows the consumer to watch TV programmes remotely and as such we expect it have little value in the mind of the consumer. The client device needs to be very cheap; and we would expect that it must cost no more than a DVD player or a free-to-air digital TV receiver. As a reference, it is sobering to recognise that the digital AV home network has to compete with a 30 analogue video sender

Installation complexity can be a barrier to home network adoption because of aesthetics, effort or cost. It is an irony that the more affluent proportion of the population, who are more likely to be interested in the extra features of a home network (and less put off by cost), are the least likely to accept anything they do not consider aesthetically pleasing, such as extra wires or large aerials.

Installation effort and aesthetics rapidly lead us to consider that the only sensible approach is to choose networking technologies that require no new wires to be installed. Cost of installation is a more difficult area to quantify because it has large regional variation. For example the cost of installing coaxial cable in a brick built home (common UK practice) will be significantly higher than a wooden build home (common US practice).

REQUIREMENTS

It is important to estimate how many AV streams are likely to be simultaneously active in a household and just how many adjacent households are likely to be simultaneously using their home network.

A sensible basis is to consider how many television sets are used in each household, because this will be a reasonable upper bound on the number of active video steams. In Western Europe this figure is around 2.5 television sets, whilst in North America the figure is higher at just under 3 television sets.

The next factor that affects the number of video streams is how many storage devices we expect in a home and whether these devices are directly connected to a television. The worst case figure for number of video streams occurs when the storage device is not connected to a television, as this requires all video playback to be streamed over the network.

We think that the home network must support at least one video stream per household, and ideally it should support at least one video stream per television.

An estimate of the number of households who are likely to install a home network is rather difficult to produce. We can assume that a sensible upper limit is the number of households who currently have a television. The US census in 2003 (D showed that 94% of the US homes had at least one television. Figures for Western Europe vary between 94% (France, Switzerland) and 100% (Norway, Spain). (2)

It is unlikely that home networking will reach the same levels of penetration as television (at least not for quite a long time), but there are some factors to consider that could mean that penetration figures may be rather large. The increasing popularity of the Internet (and especially broadband Internet access) is likely to drive home networking adoption amongst computer users. 66% of US households own a computer. European figures for computer ownership (3) vary between 42% (Ireland) and 58% (Denmark).

Table 1 - Availability of Existing Wiring

Another factor that is likely to drive up home network adoption is the switch off of analogue television. The US, UK, France, Germany, Japan and China all have plans to turn off analogue television. Multiple television homes will either have to throw out all of their analogue sets (highly improbable), add digital receivers to each TV or route signals from a digital receiver to each "legacy" television.

The increasing adoption by consumers of personal video recorders certain to increase home network adoption. As the consumer becomes familiar with the benefits of a personal video recorder, they will expect to be able to have the same functionality and content available on other sets in the home.

We think that it is entirely reasonable to assume that if a suitable set of technologies can be found, penetration of home networking will exceed 50%, possibly even as high as 75%.

If the home network is offered as part of a subscription television service, it will be important that all features available on the primary set top box are available on home network enabled devices. For example premium channels, t-commerce and pay-per-view events should not be restricted to just the primary television set.

What are the currently available wired and wireless solutions that could be used in a home network?

RE-USE OF EXISTING WIRING

When pursuing a "no new wires" approach there are two options to consider - existing wires or wireless. To analyse the existing wiring approach we need to consider the wires we are likely to find in a typical home.

TELEPHONE WIRES

Most homes have a telephone line to the property. It is quite common for this telephone line to have been extended inside the home to multiple sockets to enable multiple handsets to be placed around the home and to facilitate Internet access for a PC.

The HomePNA standard provides a way of using existing telephone connections for in-home networks. The current standard (HPA v3) provides up to 100 Mbits/sec. It uses a frame structure and CSMA/ CD system that is very similar to IEEE 802.3. A difference between the 802.3 MAC is that HomePNA v2 introduced a QOS system with 8 traffic priority levels.

This priority system is similar to 802.11e EDCF QOS in which stations with higher priority traffic wait for less time after the medium becomes clear before starting to transmit. Two or more packets with the same priority level would still have to contend with each other using their exponential back-off algorithms. HomePNAvS adds another QOS option where a master node controls access to the medium. The master node transmits a Medium Allocation Plan (MAP) for each cycle. HomePNA v3 nodes have to obey the plan and transmit only at time slots allocated to them.

A problem with using telephone wiring for video streaming is that telephone sockets are rarely close to television sets. Telephone sockets are normally placed at locations where it is convenient to have a telephone or near to a PC, neither of which may be close to a television. The requirement to install a telephone cable to a set top box has been a problem for pay-TV operators for many years.

MAINS WIRES

The HomePlug standard enables networking using the mains power cables in a home. The current standard provides up to 14Mbit/sec using mains power cables and the next version claims to provide 200 Mbits/sec.

Using the mains wiring seems like an ideal solution because devices need to be plugged in to get power, no new wires are needed and it does not assume any other type of wiring has been installed in the home. There are, however, some significant issues with the use of mains wiring.

One issue is that mains wiring is an unshielded nontwisted cable, which makes it very susceptible to picking up and also transmitting RF noise. For example if current HomePlug standards were deployed in the UK it is expected that long wave radio reception would be degraded.

Another issue with mains wiring is that multiple homes share the same power source from a local transformer, with a difference in the practice from country to country and region to region. The HomePlug standard uses DES to encrypt data on the network so that multiple homes on the same transformer cannot access each other's cont\ent. Other HomePlug users are essentially considered as extra noise on the line, reducing the signal to noise ratio. In regions where many households share a transformer output, HomePlug may have problems achieving a reliable link with sufficient bandwidth.

COAXIAL CABLE

In both Western Europe and North America many have coaxial cable installed to provide aerial feeds to multiple televisions in the home. Such installations remove the ugly set-top aerial and increase signal quality.

In a small number of households this coaxial cabling has been used to form a simple analogue home network to allow the UHF output of a VCR or set top box under the primary television to be viewed on other televisions. Some installations are also able to use the coaxial cable to send remote control signals from remote televisions back to the originating VCR or set top box.

The Multimedia over Coax Association (MoCA) has created a specification that allows this coaxial cable to be used for home networking. The MoCA specification enables transmission speeds of up to 270 Mbits / sec.

Using coax has a big advantage over using telephone cabling because it is more likely to be found next to a television. The disadvantage of using coax is the low number of homes so equipped and the practicalities and aesthetics of making such installations. Another disadvantage of using coax is that signals do not propagate backwards though an active amplifier.

STRUCTURED CABLE

For completeness we mention structured cabling that uses twisted pair cables (such as cat5) to provide telephony and computer networking. Structured cabling is used extensively in business premises, but is very rare in the home environment. This has improved slightly in recent years, with roughly 10% of new builds in North America having structured cabling installed. Compared to the total number of homes in North America, this installed base is still very small.

When used for networking using the 802.3 standard twisted pair can support 10, 100 or 1000 MBits/sec, normally in a full duplex mode that allows these bitrates to be achieved in both directions at the same time.

INSTALLATION OF NEW WIRING

Any of the cabling candidates in the "Re-use of existing wiring" sub-section could be used for a new cable installation.

Table 2 - Cost of New Wires

The following table illustrates that the cost of the materials is irrelevant compared to the cost of installation (in the region of 400 to 1000 for a house).

As the cost of the cable is small compared to the installation costs, one option would be to install almost all types of cable. In a new build this is a reasonable option, but in an existing property the disturbance this would cause probably makes an "every cable" approach unacceptable.

The decision as to which cabling system to install when only one or two systems can be selected will need to be based on the usage pattern of the consumer. A household that has an extensive collection of analogue devices (televisions, VCRs, set top boxes with UHF outputs, etc) might choose to use UHF coaxial cable.

For most other situations it makes sense to use Cat6 (twisted pair Ethernet) cable because this can be used for both networking and telephony. The star topology has the added advantage that network traffic can be selectively routed down individual cables. This selectivity may be very useful in a home with multiple servers and multiple clients that are streaming high definition video streams.

WIRELESS TECHNOLOGIES

For wireless we need to consider the network bandwidth provided, its ability to reach all the required parts of the home and its ability to scale to very widespread deployment (i.e. when in use in most households).

433 / 868 MHZ

Wireless RS-232 adaptors are available that operate in the 433MHz or 868MHz bands, and provide data rates varying between 10kbit/ sec and 40kbit/ sec. The transmission range depends on the transmission power and the gain of the receive antenna, but typical range figures are between 20 and 100 metres.

These data rates are not suitable for video transmission, but it could be used as a return channel in combination with another wireless technology.

ZIGBEE

ZigBee was designed as a low power radio system for remote monitoring and control. It uses the physical and MAC layer of the 802.15.4 standard on which it builds its own application layers. Typical transmission ranges are between 70 and 30Om.

It provides a bitrate of 20kbps using the 868MHz band or a 250kbps using the 2.4GHz band. The network can operate in a star or mesh mode containing up to 64000 nodes, with one device designated as the network controller.

The data rates supported by ZigBee are not suitable for video transmission, but it could be used as a return channel in combination with another wireless technology.

DECT

There are 10 carrier frequencies available in the core DECT frequency band. Each frequency is time-sliced into a number of data and control timeslots.

At all times a base station is transmitting on at least 1 or 2 carriers. In the absence of any real data control data is still transmitted, allowing handsets to measure received signal quality and determine if they have access rights to (i.e. are they "paired" with) the base station.

When a base station is transmitting only control information on an otherwise idle carrier, it monitors for least interfered channels and if appropriate, switches its transmissions to the new least interfered channel.

At call set-up, the handset selects the strongest base station and the least interfered channel and attempts to contact the base station requesting access. If the base station responds appropriately, the communications path is established, otherwise another combination of channel and base station is tried.

Wi-Fi is one option for home networking

Whilst a call is in progress, the handset is continuously scanning the other carriers and adjusting its lists of least interfered channels and strongest base stations. When another combination is determined to be better (with manufacturerdefined hysteresis) a call is setup on that channel/base station combination and then the existing call is handed-over in a "make-before-break" manner.

The major problem with DECT for home networking use is that the data rates it can provide are not suitable for video transmission. It could however be used as a return channel in combination with another wireless technology.

DVB-T

The DVB-T standard (ETSI EN 300 744) was designed to allow the transmission of MPEG-2 transport streams over long distances to many homes, using the VHF and UHF frequency bands. It is designed to use 8MHz, TMHz or 6MHz wide bands to transmit at up to 31.7MBits/sec.

Due to the long distances of transmissions, the DVB-T system has been designed to cope with significant multi-path distortion and interference from transmissions on nearby channels such as existing analogue TV transmissions. DVB-T uses 2048 or 8192 carrier OFDM modulation and a Reed-Solomon FEC.

The robustness of the encoding of DVB-T makes it an interesting candidate for in-home networks. It has two major disadvantages however, the first of which is that it is a one-way system (transmitter to receiver) and that the transmitter constantly outputs a signal on its UHF/VHF channel.

The lack of return path from receiver to transmitter could be solved by using one of the other wireless transmission systems such as ZigBee or DECT.

The fact that the transmitter is always outputting a signal is a problem for scalability because frequencies cannot be shared and one channel per transmitter is. If a household has multiple transmitting devices (e.g. multiple PVR home, home with PVR + IPTV) it would require multiple channels.

Table 3 - Spectrum Efficiency of Existing RF Modulation Schemes

Finally, the ability for consumer electronics grade components to produce a sufficiently stable DVB-T signal is questionable. As a broadcast standard with narrow guard bands DVB-T requires a very accurate oscillator to avoid drift into adjacent channels.

802.11 "WIFI"

The 802.11 group of standards provide wireless local area networking using two different licence exempt bands.

The 802.11b standard uses the 2.4GHz ISM band to transmit at up to UMBits/sec. The 802.1Ig standard builds on top of the 802.11b standard by adding OFDM (64 carriers) and QAM transmission options from the 802.11a standard to increase the maximum physical data rate to 54MBits/sec. There are 14 channels(7) available in 802.11b/g, but because they overlap at 5MHz intervals, there are only three nonoverlapping channels.

The 802.Ua standard uses 2OMHz channels in the 56 GHz range, to produce a physical data rate capable of up to 54Mbits/sec per channel. 802.Ua is normally used in 10OMHz bands at 5.15 GHz "Band A", 5.47 GHz "Band B" and 5.725 GHz "Band C". In the UK band A and band B are available for use by 802.Ua equipment. Band A provides 8 non-overlapping channels and band B provides 11 nonoverlapping channels. In the United States there are three bands (5.15GHz, 5.25GHz and 5.75GHz) that each provide 4 non-overlapping channels.

The 802.11 standards have advantages over DVB-T for home networking use because 802.11 is good at dealing with multiple transmitting devices and because its popularity in the computer industry has made these devices relatively inexpensive(8)

The 802.11a/g/b standards do not provide a QOS system, but the 802.1Ie standard provides an extension that can be used on top of these standards to provide QOS. It defines two different techniques that can be used to prioritise traffic flows, called EDCF and HCF.

EDCF modifies the time stations wait to transmit so that prioritised traffic is more likely to start transmitting before best effort traffic. The disadvantage of this system is that traffic is prioritised based on class (e.g. video, voice), which means that two or more streams of the same class will have to compete with each other usingthe normal exponential back-off medium access control.

HCF uses a polling system where by the access point allocates contention free periods to stations that have QOS flows. This system is more complex but it should be more suited to multiple streams of the same class. There is a problem with HCF, in that it assumes a channel is not used by another access point. However, to date no products are available that have implemented HCF.

802.16"WIMAX"

The 802.16-2004 standard was created to enable wireless broadband networks using fixed point-to-point or pointto-multipoint links. Its primary use is for line of sight links using frequencies in the range 1OGHz to 66GHz by modulating a single carrier. It can however also be used in the 2-11GHz frequency range using OFDM (using 256 or 2048 carriers), for example in the licence exempt bands at 5GHz. It is designed to be able to support many terminals all communicating with the same base station. It uses a request-grant system to allocate bandwidth between active terminals.

Table 4 - Summary of RF Candidates

The system can be configured to use one channel using time division multiplexing, or multiple channels using frequency division multiplexing. Multiple coding schemes may be used on the same channel so that robustness and bitrate rate can be negotiated on a stationby-station basis. Transmissions are ordered so that the most robustly coded packets are sent first, which enables stations to easily estimate their maximum bitrate by noting the error rates of the currently active coding schemes.

The 802.16 standard has good QOS support, mostly due to its base station controlled request-grant architecture. It is not designed to cope with multiple base stations on the same frequency, probably because its main application was in line of sight links. Normally traffic only flows between a station and the base station, however 802,16 has a mesh networking mode that allows station-to-station transmissions.

Figure 1 - RF Bandwidth vs Number of Supportable Services

The suitability of 802.16 for home networking use is currently unknown. Its QOS architecture and its ability to form collaborative mesh networks make it an attractive candidate, but its ability to scale when many homes are all running 802.16 base stations is unclear.

Table 5 - Required Guard Bands

WIRELESS SUMMARY

There are many modulation schemes that could be used as the physical layer for a wireless home network. Of the schemes that are most realistically useable for a home network (802.11 and DVB-T) it can be seen that they are very similar in their spectrum efficiency.

Table 4 provides a summary of the technologies described in the previous sub-sections

RESULTS FROM AN RF STUDY

We commissioned a paper study(11) into the feasibility of creating a wireless home network capable of streaming multiple video streams. The study examined RF propagation at various frequencies between 47OMHz and 5GHz.

Table 6 - Spectrum Efficiency at Various Channel Sizes

Part of the results of this study was a mathematical model of RF propagation that could be used to estimate coverage and transmit power level requirements for a range of typical UK homes.

The (abridged) conclusions of the report were:

It is likely to be possible to transmit RF signals in the frequency range 470 - 5000 MHz through most of the building fabric of typical domestic houses, allowing a simple DVB domestic wireless LAN using COFDM to be designed. It should be noted that the low microwave frequencies considered of 2.4GHz and 5GHz undergo much higher building attenuation than frequencies within the UHF band. White goods such as kitchen appliances and room radiators will significantly attenuate the RF signals and careful sighting of the RF equipment away from these devices is recommended. Certain materials such as metal re-enforced concrete and foil-backed plasterboard are likely to act as perfect RF screens and this will prevent RF transmission through these types of materials. Materials having high water content, such as undamp-proofed stone cottage walls may also significantly attenuate RF signals at the frequencies considered.

The most challenging aspect of this project is the frequency planning as this application may require a large number of frequencies to produce a practical plan in the long term. However, it may be possible to start the service with a small number of frequencies first and then increase this number as the service develops over time. Moreover, different regions of the country may also have different numbers of channels available at any one time, enabling a different local frequency planning treatment to be employed regionally. It is highly recommended that a paper study of the frequency planning of this service be started.

Using the mathematical model we created a model of a typical UK house (2 floors, less than 88m^sup 2^ total floor space, situated in a 600m^sup 2^ plot)(12) for which we estimated that a transmit power of -58 dBW was sufficient at the lower range of tested frequencies. We estimate that a signal in this frequency range at the given power level will be above the noise floor in 9 other homes.

This estimate implies that we need 10 different channels for a system that can scale to every home using a home network. This figure can be reduced if several homes are able to cooperatively use a channel, as each channel has enough capacity to support multiple services at the same time. The most likely method for channel sharing would be some sort of time division multiplexing, however if a suitably stable oscillator can be found this could also be achieved by reducing the RF bandwidth of each channel (e.g. using 5MHz channels).

Figure 1 shows a plot of total number of supportable services against the total amount of bandwidth allocated to the wireless network. We have plotted currently available modulation schemes (DVB- T, 802.11a/g) along with a current state-of-the-art modulations scheme (OFDM with turbo codes that achieve 0.5dB below Shannon limit).

It can be seen from the table that approximately 4OMHz of RF bandwidth should be sufficient for supporting 30 simultaneous SD services. Using the -58dBW figures from before, this would be sufficient for an average of 3 SD services per household even when every household has installed a home network. Note that the figures above are based on physical rate transfer speeds. A real system will produce lower numbers of simultaneous services because of the overhead of other layers of the network protocol (e.g. MAC layer overheads).

The RF bandwidth is not required in one continuous block, and could be split in to multiple smaller frequency allocations (e.g. 2 x 2OMHz, 4 x 10MHz). The limitation on the minimum size of a frequency block is based on resistance to fading and the size of guard band required for avoiding interference with neighbouring channels.

As the size of frequency allocation decreases, its susceptibility to fading increases because fading tends to be frequency selective. A larger frequency allocation is less susceptible to frequency selective fading because the OFDM carriers that are lost due to the fade can be recovered by the error correcting code.

The guard band that is required between channels is based on the expected drift of the oscillator in the transmitter and the receiver. A stability of 50ppm (parts per million) is a reasonable assumption for a crystal oscillator at a reasonable price. If the transmitter or receiver uses a super-heterodyne design, it might include two crystals. This gives a total expected drift of four times the ppm of the crystal. As this drift could be either to increase or decrease the frequency, the guard band between channels will need to be twice the maximum drift (i.e. eight times the crystal stability). The rolloff of the front end filter also increases the guard band. If the front end filter is based on passive components, a reasonable estimate for this roll-off is 1MHz.

As the size of each frequency allocation decreases, the percentage of wasted spectrum due to guard bands increases. At lower frequencies this is less of an issue because the guard band can be smaller.

REQUIREMENTS THAT ARE NOT MET BY CURRENT CANDIDATES

DYNAMIC QOS

The term "quality of service" is used for many different things when people talk about networking and in particular home networking. In this paper we make the assumption that a home network can be said to have sufficient QOS for the consumer if once a video stream has been started, it is able to deliver television pictures regardless of what else is occurring on the network.

We want a home networking solution that provides the consumer with a reliability of service at least equal to the level of reliability they receive from their existing television service provision.

If a home network contains a device with a tuner but no storage (or a full storage device) it may be necessary to stream video from the tuner device to a storage device. In this situation, it would be desirable to be able to configure the QOS system to prioritise a recording stream over a playback stream.

Wireless networks are subject to rapid variation in achievable link quality, especially as the frequency increases. The QOS system will need to be able to adapt to the current link quality in order to maintain the most critical QOS flows for as long as possible.

BIT ERROR RATE

When we consider QOS, we also need to consider the bit error rate of the network. A QOS system might provide a guarantee of transmit time on the network and packet timing jitter, but this will not be acceptable to the consumer if the network has a high error rate, because they still cannot use the network to reliably watch a television programme.

Figure 2 - Example of "Brown Field" Frequency Re-use

The home network must be able to provide guarantees that a video stream can be delivered with a low bit error rate (BER). The bit error rate will n\eed to be at least equal to the DVB requirement of a quasi error free rate of 1x10-11 BER. Error correcting codes and/ or retransmissions could be used to achieve this requirement, subject to still being able to meet other QOS requirements.

The 802.11 standards use retransmission to achieve near error free delivery to layer 3 of the network software stack. These retransmissions require bandwidth and increase latency. The 802.11a standard defines a packet error rate of 10% (i.e. 1x10-1) at the limit of receiver sensitivity. To achieve 10-11 BER requires a worst case of 11 retransmissions because a transmission is only successful when the entire packet has been received error free. If the round- trip time for packet Tx and NAK is 10ms, this implies a latency of at least 110ms.

At first it might seem that this system would require a lot of bandwidth to achieve 11 retransmissions, but it should be noted that only 10% of packets need a retransmission, and reaching re- transmission has a 90% chance of success. The total extra bandwidth requirements for retransmissions at a PER of 10% is roughly 11%.

Block code error correction has been used successfully in IP networks to decrease bit error rates below 10-11. These schemes use UDP datagrams, and continuity counts to detect lost packets. These lost packets are then corrected by the block code (Erasure correction). 802.11 MAC does not at present have a true datagram mode. The MAC layer retransmits packets until they are received without error. This limits the use of such block code correction schemes.

The 802.11 MAC does not provide many options for error correction. It is not possible to prevent retransmission from occurring. This forces the 11% overhead and associated variable latency onto the client device. It would be far more beneficial for video broadcasting to turn this feature off, and allow higher level error correction codes to reduce the bit error rate. For instance adding a simple Reed-Solomon forward error correction, with suitable interleaving, would provide the required BER, but would require only 8.5% overhead, and provides a fixed latency. Modern block codes can provide the same correction capability with even less overhead.

However, correcting lost packets is not always the best way to use block error codes. Higher correction capabilities are possible if the data in error is flagged and passed to the error correcting code. At the moment it is not possible to pass incorrect data to the higher layers because there is no way to flag in the transmitted packet that the packet level CRC should be ignored. In the IEEE 802.3 standard it is possible to use a zero CRC to indicate that CRC checking should be skipped, but this was not adopted for the 802.11 MAC.

FREQUENCY AGILITY

Table 7 - UK Frequency Allocations

Table 7 - UK Frequency Allocations

It is clear that more than one channel would be required to meet co-channel interference requirements within a neighbourhood. As such the home network must be frequency agile and perform some kind of clear channel selection.

Currently 802.11 devices perform passive clear channel assessment. They listen on the channel they wish to use and determine if the noise level is low enough to use. The quietest channel will be selected. This sensible approach would be suitable for the home network we envisage.

It is not however without its limitations. The receiver sensitivity limits the measurement of the background noise level. secondly only the access point performs the channel assessment, but the client stations also have to transmit, and transmissions to the access point can interfere with adjacent networks.

We would like to allow adjacent networks to share the same channel. If the two access points can communicate reliably and can support all the QoS streams on a single channel, it makes sense for them to collaborate and use the same channel. A third station would be more likely to register the channel as "in use" because there would be at least two stations transmitting from different locations. Additionally sharing a channel provides more channels open for use by other home networks.

A feature of the audio visual home network is that it operates as an edge device to a broadcast/broadband service. As such, it is relatively easy to send large amounts of information to the device. If the device knew its geographic location, the device could positively determine channels which were not available for use in its location. It is hard to imagine a solution where this could be used solely as the basis for channel allocation, but in conjunction with clear channel assessment it provides the potential for home networks to use a "Brown Field" frequency.

A "Brown Field" frequency would be a frequency that is already in use, but has geographic regions where the frequency is not used. This is usually a result of frequency reuse and planning considerations, limiting the proximity of two transmitters using the same frequency. However the very small footprints of home networks, could potentially allow them to use frequencies which could not be used by larger transmitters.

As the home network is connected to an access network (e.g. ADSL, DVB-S, cable) it is possible to update the frequency maps. This means that allowing "Brown Field" sites does not necessarily fix the frequency planning. Primary users of the spectrum can maintain the right to re-plan the frequency map at any point in the future. New regional channel availabilities could be quickly disseminated to all home networks.

MANAGED FREQUENCY

The licence exempt ISM and SRD bands are always going to be popular with equipment manufacturers due to their more relaxed regulatory framework. The 2.4GHz band (and the 90OMHz band in the US) is now highly congested with many devices competing for use of the spectrum.

Despite the congestion in the 2.4GHz band, many companies are successfully selling analogue video senders, and IP devices with 802.11b/g or Bluetooth connectivity, etc. The critical difference between current products and the type of home network we wish to enable is the sheer bandwidth and reliability required for streaming digital video, needing more bandwidth than an 802.11b Internet connection or a Bluetooth headset.

At the moment the 5GHz bands are relatively unused, mostly due to the higher cost of equipment at this frequency band and because the licences were only granted relatively recently. (13) However, a problem is looming for the 5GHz bands as more people start to use them. For example the rollout of WiMAX at 5.8GHz for Internet hot spots is already reducing the available channels for users of 802.Ua equipment who are unlucky enough to be close to a hotspot (e.g. living near a railway station).

Another issue with ISM bands is that they are technology agnostic - any technology is allowed to use the band as long as it stays within the transmission power limits. This is an issue because most of these technologies are not aware of each other and make no attempt to gracefully share time on the wireless medium. This becomes an issue as the number of users increases because this puts more pressure on the need to re-use channels.

It would have been extremely beneficial if the requirements for use of the licence exempt bands had included some sort of requirement to be a "polite" protocol, but this is probably very difficult to achieve in a manner that does not limit technological advances.

It would be ideal if we were able to identify a frequency band that can provide sufficient bandwidth for a widely deployed AV home network that has some sort of regulatory framework. This framework would allow cooperative protocols to make efficient use of the spectrum without the risk of being jammed by another uncooperative protocol.

It is probably not desirable to mandate one particular technology because this would curtail innovative use of new coding and modulation techniques, but it would be sensible to mandate use of a band for use for home audio-video transmission to avoid it becoming saturated with Internet traffic. However we accept that technologies such as IPTV and MPE blur the line between broadcast TV and the Internet.

Based on the results of our RF study, we estimate that about 4OMHz of spectrum is required for a wide scale deployment of a home network that can support standard definition TV. If the home network needs to support high definition TV to the same level of penetration as SD, the bandwidth estimate would increase to 80 MHz.

In the following section we provide a summary of some analysis of the entire radio spectrum as managed by Ofcom in the UK, looking at possible frequency bands and the pros and cons of their use.

ANALYSIS OF THE UK RADIO SPECTRUM

Table 7 summarises the radio frequency spectrum and provides pros and cons for a frequency band's use for home networking.

MODIFYING 802.11 TO MEET REQUIREMENTS

Although current IEEE 802.11 technologies do not meet all of our requirements, we believe that with modifications the specifications could be made to be suitable.

Most fundamentally, modifications to the MAC layer are required to allow for improved channel usage and QoS, and to the PHY layer to enable its use in new frequency bands.

We would like a modification to the MAC layer that allows multiple access points to gain knowledge of each other, so they can co-operate on a single channel. This would allow two or more access points to use HCF on the same channel, and allow two or more homes to share a single channel, whilst maintaining QoS. This maximises the use of any available spectrum in a dense environment, such as a block of flats.

To allow higher level error correcting algorithms, we would like to modify the MAC layer to allow packets that fail the CRC check to be flagged and passed up the network stack, as opposed to triggering a retransmission. This modification would need to be supported at both the transmitting and rec\eiving devices.

The transmitting device will need to add the "do not retransmit" flag to its packets and not re-transmit a packet if it fails to receive an ACK. The transmitter will also need to add an appropriate error correcting code to the data it is transmitting. This would allow the receiver to correct packets with errors or missing packets without having to use re-transmissions.

CONCLUSIONS

We see wireless home networks as an extremely attractive proposition for consumers, platform operators, service providers, broadcasters and CE manufacturers. Our desire is to be able to create a home networking solution that can be widely deployed, and provide platform operators with products that are highly reliable and able to scale with the growth of their business.

We believe that at the moment there is no single wireless technology that is able to meet our reliability and scalability requirements. We think that the IEEE 802.11 appears to be the most likely set of technologies that will be deployed for home networking. However the current 802.11 standards cannot provide the necessary QOS and scalability requirements without modification.

A fundamental question is whether there is any possibility of spectrum being made available to implement the wireless home AV networks we all need. We have seen that licence exempt bands rapidly become congested with many competing technologies that do not cooperate with each other. We would like to see a frequency band chosen that provides reasonable RF propagation within typical housing stock and which is controlled by a managed process.

if current HomePlug standards were deployed in the UK it is expected that long wave radio reception would be degraded.

the ability for consumer electronics grade components to produce a sufficiently stable DVB-T signal is questionable

We want a home networking solution that provides the consumer with a reliability of service at least equal to the level of reliability from their existing television service provision.

a key question is whether spectrum will be made made available to implement the wireless home AV networks we all need

REFERENCES

1. Section 24 Information and Communications, Statistical Abstract of the United States, US Census Bureau (http:// www.census.gov/prod/2004pubs/ 04statab/infocomm.pdf)

2. WorldScreen.com (http://www. worldscreen. com/europe.php)

3. Information Technology, ICT Statistics, International Telecommunications Union (http://www. itu. int/ITUD/ict/statistics/ at_glance/Internet03.pdf)

4. Main telephone lines, subscribers per 100 people, ICT Statistics, International Telecommunications Union (htip://www. itu. int/ITU-D/ict/ statistics/at_glance/main03. pdf)

5. Multimedia over Coax Alliance (http://www.mocalliance. org/en/ aboutus/ourmiss/on.asp)

6. Some homes may not have mains wiring, but such homes are highly unlikely to want a home network!

7. Not all of these 14 channels are available in any one country

8. Whilst 802.11b/g equipment is relatively inexpensive, 802.11a equipment currently still carries a price premium

9. Seated to 8MHz channel

10 Scaled to 8MHz channel

11. "An RF Study on Home Audio/Visual Distribution using COFDM", Professor Arthur Mason OBE

12. This represents 80% of UK homes, as reported by the 2001 government review of housing stock

13. Use of bands A and B in the UK was granted in October 2002

ABOUT THE AUTHOR

Alex Ashley & Ray Taylor work in research and development for NDS, the leading provider of conditional access systems.

Copyright International Institute of Communications Aug 2005

Source: Intermedia

More News in this Category