Devil's Advocate: In search of wireless nirvana

To realise the dream of a wireless world, we need to find a new way to deliver power to computers and other electronics, says Martin Brampton. If only this were an easy task.

A major attraction of Bluetooth is the elimination of wires. Most computers are a mess of connections and a rat's nest of cables. Quite apart from wanting to be tidy, the messiness and inflexibility of cable limits flexibility in offices.

Although wireless might eliminate cabled data connections, we are left with the problem that so many devices need a power supply. Offering this through the ethernet link has attractions but only if the network link is a cable. Unfortunately wireless power supply looks to have insuperable problems at the levels currently required.

While the IT world struggles to establish and extend standards for power over ethernet, the old fashioned telephone engineer is probably laughing into his beer. The telephone system has been providing power along with the signal for decades, usually with deep backup that is one of the factors making the system so resilient.

On the computer side, the steadily increasing addiction to power is not helping matters. While thin screens are drastically reducing consumption, processors are going in the opposite direction. It is now common for a desktop processor to be consuming as much as 100 watts. That is the reason for all those hefty processor fans that make it difficult to achieve a quiet life.

Power over ethernet is currently offering a maximum of 13 watts, plenty for a low power device but nothing like the requirement of a typical computer. Negotiations are going on to extend the specification to support much higher wattages. They are confronted with practical problems.

Twisted pair was never designed to be power supply cable. Moreover all the schemes implemented so far have relied on direct current transmission at quite low voltages. That makes for significant power losses over long cable lengths. Changing to alternating current and higher voltages would present all kinds of interference and safety problems. The power losses naturally turn up as waste heat, creating problems of temperature rises at high concentrations.

An appealing application for power over ethernet is support for a centrally powered, distributed network that channels data packets on fewer and fewer cables as they get closer to central points that house servers. Even that faces the problem that high data rates tend to demand high processing power and thus high power, though.

So we seem to be a very long way from providing all the power needed at the average desk along the network cable. That pushes the argument towards trying to eliminate the data cable altogether by replacing it with wireless. Taking both power and data to desks, which in modern flexible offices frequently have to be moved, more than doubles the complexity and work involved.

With even voice over IP now deliverable wirelessly, there is a strong temptation to push in that direction. Unfortunately data rates still do not compare with fixed wiring - nor does overall reliability, especially in areas of very high density such as office buildings.

How well are market forces serving us in dealing with these questions? There seems to be a contrast at present between wide area communications, local communications and computing. We are seeing liberalisation of telecoms leading to a catching up in what can be achieved with fast, low-cost wide area communications.

In the local network, existing technologies all have some merit. But there is no clear direction towards solutions that mesh easily with the typical demands of modern organisations. The limitations impose choices and historically people have rarely made clear decisions on IT matters. The tendency is to hanker after the best of all worlds and to finish up with the best of none.

Finally, on the computer itself, it is hard to see that we are getting value from the huge increases in processing power that chip makers have delivered. While there are notable software innovations, much processing power seems to be absorbed in marginal improvements and dubious new features. How can we do better?