Perhaps the most pervasive trend in communications over the last decade has been the drive to exploit the established infrastructure to carry digital data transmissions. The emergence of the high successful cable modem technology is a prime example, in which a globally well established wideband medium was adapted to provide a high speed data service to the household user.

The cable modem is not alone. The driving force behind the ADSL protocol was the idea of exploiting existing baseband telephone wiring to subscribers. In practice the performance achievable will depend strongly upon the quality of the wiring in place, and many subscribers will require new wiring to support the standard.

From a strategic perspective, the ultimate solution to the bandwidth bottleneck into consumer households is the provision of a universally available single mode optical fibre connection. This would provide for the forseeable future bandwidth limited only by the capabilities and cost of the subscriber terminals and exchange interfaces.

Will this solution be seen at any time in the forseeable future? The answer is most likely NO, since the explosion in demand for bandwidth has coincided with a period during which the telecommunications industry has been deregulated. Short term commercial profit imperatives and the slow rate of return upon a household high speed service will act as impenetrable brick walls to growth in available low cost high speed services. With the enormous cost overheads of supporting massive marketing machines and greatly redundant low speed transmission infrastructures, the industry is unlikely to pursue any initiatives in this direction.

Given the enormous capital costs of universal fibre connectivity, and the slow payback period, the expectation in the nearer term is that the trend to squeeze every kilobit/s out of established wiring infrastructures will continue.

In this context the unglamourous but very useful technology of Power Line Transmission (PLT) deserves some careful exploration. While it might not offer the dazzling performance of cable modems, 802.11a/g wireless Ethernet, or indeed fibre, its universal availability via the ubiquitous electrical power distribution network makes it a useful tool for a great many applications.

Data Through Power Lines

Traditionally the POTS telephone cable has had the distinction of being labelled with the least flattering language used to describe a data transmission medium. Alas it is in the process of being knocked off its perch as the least well behaved medium used for this purpose. The usurper of this dubious title is the power line transmission cable.

Electrical power cables have never been designed for data transmission. The primary criteria in the design of such cables have been low cost, low current losses, high insulation strength, durability and low flammability. Typical power cables are untwisted and where twisted, the absence of characteristic impedance specifications results in arbitrary cable performance. Cited specifications of commonly used cable types span 74 Ohms up to 143 Ohms, or a 2:1 ratio in cabling alone.

From the communications engineering perspective, power cables are about as bad as it gets. The widely varying cable impedance and imhomogenous cable mixes installed in households, commercial buildings and industrial plants result in a transmission environment which exhibits enormous variations in impedance characteristics.

This variability in cable impedance is further exacerbated by the widely differing high frequency behaviour of distribution panels, connectors, wall sockets, switches and other items of distribution hardware. Unlike hardware designed for RF or high speed applications, power hardware is optimised for robustness, low cost and low flammability. As a result, the impedance behaviour of such hardware is generally undefined. As wiring connections into such hardware do not impose impedance constraints, arbitrary birdsnests of connections are possible with the inevitable impact on impedance behaviour.

Is this the only impediment to uniformity in power line impedance? The answer is no, since a wide range of devices will be connected to the power wiring. Of key importance will be highly capacitive devices such as EMI control capacitors in the power supplies of consumer electronics, but also resistive heater elements of the ilk used in ovens and toasters. At hundreds of kiloHertz such hardware has impedance values typically one half to one quarter that of the wiring itself.

The traditional view of household or building wiring is of a capacitively loaded cable, since under unloaded conditions the impedance declines as the frequency increases. Factoring the multiple low impedance loads attached in a typical environment, the wiring typically behaves in a strongly inductive manner.

For a carrier frequency in the hundreds of kiloHertz, typical for PLT hardware operation, impedance increases with frequency and attenuation or loss can be extremely high.

Echelon Corp (www.echelon.com), one of the very visible players in the market, recently published the results of a large survey performed across a population of hundreds of houses, in which attenuation was measured between pairs of sockets at 130 kHz. Measure loss varied between 6 dB and 84 dB, or across many orders of magnitude. The statistical summary for the experiment indicated that 96% of the population measured had less than or equal to 54 dB of loss at 130 kHz.

In practical terms, the result is a transmission environment in which the variation of carrier loss, carrier phase and cable impedance with frequency changes continuously in time, and varies widely within a building and across buildings.

From a data transmission engineering perspective, this is about as bad as it gets. But this is hardly the end of the story, either.

Data transmissions on power wiring have to coexist with a wide range of power consuming devices, many of which inject very substantial levels of interference into the wiring.

Some notable examples are summarised thus:

Suffice to say this is a hostile interference environment at the best of times. If one adds into this brew older electric motors with brushes and largely dried out electrolytic EMI filtering capacitors, together with spikes and interference coupled into the household wiring from the externals mains grid, the result is by any measure ugly.

Combined with the woeful transmission characteristics of the cabling itself, the problem of piping ones and zeroes reliably through power lines verges on the boundaries of difficult to intractable.

It should therefore come as no surprise that typical devices designed for this environment achieve dazzling performance figures of the order of kilobits/s, or comparable to older technology POTS modems.

The immediate conclusion a typical user of digital networking hardware will reach is the obvious one - why bother?

The answer is however a little more complex than that. One of the areas which is expected to see major growth in coming years is household automation via networking. The aim of such automation is to provide a hosuehold in which all appliances can be digitally controlled from a central point. Heaters, air conditioning, refrigerators, lighting, indeed any appliance in a household would be networked to a central household managment computer which can then monitor the status of the device and if needed control it as required.

In practical terms, the house could be programmed to activate lighting and heating / air conditioning before the occupant arrives home from work, yet maintain a low energy burn state when the house is unoccupied. Indeed, the unoccupied house could be made to emulate the behaviour of an occupied residence by programmed activation of appliances, using activity patterns stored from periods during which the house is occupied. By monitoring the movements of occupants within the house, lighting could be activated or turned off automatically.

That common annoyance to house owners, the lawn irrigation system, is another candidate for this technology.

The precondition for such an automated household is the availability of a digital channel into each controlled appliance. How many kilobits/s are needed to control a fridge, turn off a heater, switch a television channel or interrogate the state of a stereo system? In practical terms, very few.

The biggest constraint to growth in the household automation market is the cost of providing digital connectivity in a highly standardised format to consumer appliances and electronic devices. While the bandwidth demands might be trivial, the cost of rewiring a building to provide data cables into appliances could easily exceed the incremental cost of the digital interfaces on the household hardware.

There are no technological issues in putting a TCP/IP literate interface into a television, stereo package, microwave oven, fridge or other appliance. With public domain Linux/BSD readily available for a wide range of processor chips, the incremental cost of adding the back end of such interfaces could be as low as tens of dollars per product item, in a mass production environment. Devices which already have microprocessor controllers might incur even a lower cost overhead.

So the technological bottleneck preventing the wider use of this model is the unavailability of a cheap low bit rate digital channel which interconnects these appliances.

One solution which has been argued to be the panacea is the use of ISM band wireless networking for this purpose. A low power spread spectrum signal is used to provide a digital channel. The drawback of ISM wireless networking is its vulnerability to narrowband interference and often a very poor ability to penetrate building walls and floors. Solid brick or reinforced concrete can indeed become the proverbial brick wall barrier to transmission.

While in-building powerline transmission technology may verge on irrelevant for the lucrative internet access and entertainment markets, it has very large long term potential in the provision of much more basic digital connectivity for automated appliances. That is a niche where it will be difficult to compete against, since most household appliances are connected to the mains power grid. Addition of a standardised PLT digital interface into an appliance in production could incur very low production cost overheads yet result in a large gain in appliance functionality. As noted earlier, devices with existing embedded microprocessor controllers would be the cheapest and quickest to adapt.

Modulation Techniques

The pathological transmission environment encountered in the PLT game is the principal reason why this technology has yet to become ubiquitious. The combination of time varying propagation losses and signal shape distortion typical for the cabling environment, when combined with the severe and highly time variant interference environment, will defeat most common modulation techniques.

The options available to an implementor of a PLT device fall into three basic categories, reflecting modulation techniques widely used in other data transmission media:

Various vendors have implemented and indeed marketed products using a range of these techniques. The performance and integrity of the products can vary quite widely.

The broadest division between these modulations is by bandwidth. In practice wideband schemes such as spread spectrum and OFDM do not cope well, since the transmission environment can vary widely in behaviour across the bandwidth used by the signal. The resulting distortion can severely impair the performance of the modulation. Moreover, OFDM schemes frequently do not cope well with interference, while spread spectrum schemes may require unusually high process gains (ie spreading ratios) to cope, the latter at the expense of useful data transmission rates.

A key issue in the PLT environment is the rejection of impulse noise, which is a difficult challenge especially if analogue filtering techniques are used.

An example of the use of adaptive techniques is the approach followed by US vendor Echelon Corp. Their PLT-22 series products employ an dual carrier scheme using a DSP for active impulse noise suppression and distortion correction. The carrier frequencies are adaptively selected to bypass in band interference - if the primary 132 (86 in the EU) kHz carrier is impaired, the device switches to 115 (75 in the EU) kHz carrier. While Echelon have published detailed comparisons of their respective adaptive technique based PLT-22 and spread spectrum based products, they have not disclosed (for obvious commercial reasons) the specific distortion handling and impluse noise suppression techniques used in the dual carrier product.

The Regulatory Framework

The industrialised world runs on several packages of industrial standards, and the PLT community is constrained in this manner no differently than other industries. With the US operating in a 115 VAC / 60 Hz environment and the EU in a 220-250 VAC / 50 Hz environment, it should come as no surprise that the standards constraining PLT devices also differ considerably.

The principal issue in the regulatory context is the interference produced by PLT systems, especially where high density of devices exists in an urban environment. The European CENELEC EN 50065-1 sets a number of constraints on PLT system behaviour. Nevertheless, serious concerns remain in the amatuer radio community who do not wish to see harmonics of the PLT signals obliterating their HF (shortwave) band radio reception. As is the case with ADSL, the central issue is whether the combination of a pervasive transmission environment and low cost transceiver equipment can be engineered in a manner where the harmonics of the link transmissions do not further add to the noisy urban RF background environment. If your neighbour two houses away is using PLT technology, is it acceptable for his installation to impair your radio reception? Cable modems and other technologies which constrain the RF signal by cable design can be controlled much more easily than schemes which use untwisted cable pairs.

PLT is a technology which despite its performance limitations offers some very useful capabilities to the consumer base. The principal issues for PLT will remain in engineering design standards which allow robust performance, yet do not exacerbate existing RF environmental problems. While cheaper computing power will facilitate better modulation techniques, the challenge for the PLT community will remain in finding schemes with good RF compatibility behaviour and unit production costs compatible with the consumer marketplace. As always there are no free lunches.