Mobile TV is the next in a very long string of next big things. The competition to deliver it just heated up.

Mobile TV services are currently carrying very little actual content, although they do bear the hopes of major network operators keen to finally extract big returns from their 3G investments. Especially in the UK, large-scale trials of a swathe of different technical approaches are being bandied about as proof that television is a viable application for mobile telephones. But TV is also perhaps the most challenging mobile data application technologically.

Not only is streaming video the most bandwidth-intensive application there is, but the very idea of "mobile TV" means that the user experience has to be similar to traditional TV-as good as instant connection, excellent picture and sound quality, full motion and no drop-outs. All this means more network capacity and more complexity. One unique selling point for mobile TV technologies - immediacy - creates even more difficulties for systems that save capacity by working only on-demand.

In the majority of mobile markets, most early TV implementations are based on the following lines - essentially packing TV content in media files and either streaming or downloading them over the WCDMA data channel. With a top speed of 384Kbits/s if you're lucky, this is always going to have its limits. Surely, though, the HSDPA network upgrade will deliver? After all, speeds heading towards 10Mbits/s ought to be enough, and the uplink is of limited importance in a TV application.

Streaming, of course, is an especially inefficient way of delivering anything over a bandwidth-constrained network, and subject to unreliability in a mobile application (go under a bridge, and the stream drops out). But intelligent use of existing web tools can offer an almost-immediate downloaded TV service - for example, common RSS technology can be used to push out a URL and description of new content to the users, who can then choose to download it and watch it at their leisure.

It's unlikely, though, that this approach (known as unicasting) will rule the earth. The bandwidth consumption is multiplied by the total number of users, an especially serious constraint if streaming is used. For example, if 30 per cent of the subscriber base use seven minutes of streaming TV a day, a WCDMA network will reach capacity with no bandwidth for any other data applications. HSDPA only extends this to 11 minutes of TV per user per day.

Quite clearly, what might be described as "centralised" TV is not going to be technically feasible over 3G networks. (Note, though, that video download from the internet is likely to be a workable proposition for small content providers.)

The alternative to sending individual streams of video to users, with its cripplingly high bandwidth requirement, is to broadcast the stuff in the same way as TV has always been broadcast. Hence the rush of technologies intended to produce broadcast data for mobile devices. DVB-H, perhaps the best known and the one championed by Nokia, is a variant of the Digital Video Broadcasting system intended to replace analogue TV stations. DAB-IP (also known as DMB) is a repurposing of the well-tried Digital Audio Broadcast standard, but (as the name suggests) in this instance it is internet protocol packets that are broadcast. Qualcomm is touting a CDMA-based, high-power broadcast system called MediaFLO, and IPWireless last month launched its own TDTV system, an adapted form of UMTS-TDD.

DVB-H seems to have the upper hand, but suffers from two key problems. The first is that in most of the potential markets, the spectrum that it uses is taken up by analogue TV transmissions. The UK has voiced plans to shut down analogue TV by 2012, which is by no means soon, and the switch-off remains far from certain. The other key problem is shared by DAB and MediaFLO, and is fundamental. A new network is needed.

Operators are unlikely to be keen on trying to make their 3G licences and networks pay by buying more licences and building another network. Worse, there is nothing to guarantee that network operators will be the only DVB, DAB or MediaFLO network operators, so the possibility exists that either the operators will have to pay a third party to broadcast the TV signal, face competition from other TV operators, or both.

This also applies to DAB. DAB has the advantage that networks already exist, although under the control of other firms. BT's successful Movio test-run around London involved the use of existing DAB infrastructure, but the technology is limited in the number of channels available, with a maximum of around 15. Another serious difficulty with the D-networks is that the spectrum situation is very different across national borders and no institutional arrangements for roaming exist - hence mobile TV using DAB/DVB-H is unlikely to be capable of roaming or interworking.

Qualcomm's system is intended to fit alongside an existing CDMA2000 network, but still requires a new deployment, and will be harder to roll-out where CDMA2000 is not present - in other words, outside the United States. It is also an all-proprietary system.

Long-time UMTS-TDD specialist IPWireless is, unsurprisingly, attempting to get around some of these problems by using the often overlooked UMTS-TDD standard. Unlike WCDMA, which creates two radio channels for each user, one for the uplink and one for the downlink, UMTS-TDD creates timeslots for the user and the base station (Node B) to send and receive alternately. Most UMTS operators hold, in addition to their traditional UMTS 1900MHz and 2100MHz bands, 5MHz of spectrum just above the 1900MHz uplink band which is specifically intended for unpaired TDD working.

Apart from the distinct approach to duplexing, TDD is very similar to the more familiar UMTS-FDD (WCDMA) system. There should be no significant differences in radio-planning or in the core network, and the radios at each end are similar. TDTV is intended to take advantage of this by broadcasting TV from the same cell sites as the WCDMA signal, possibly even with the same antenna fit. In the core network, the only change is to introduce a dedicated Radio Network Controller (RNC) for the broadcast implementation and sufficient backhaul capacity to feed the TV broadcast signal to the cell sites.

TDD typically might create eight downlink and four uplink timeslots in a TCP/IP broadband access application, plus three used for control purposes. Television requires only a minimal link back from the user to the broadcaster for interactivity, which can be handled by the WCDMA data bearer, so TDTV uses 14 out of the 15 timeslots to send. IPWireless claims that the system can deliver 53 channels at quality suitable for small cellphone screens from the 5MHz of available spectrum, that is to say 1b/s/Hz. As this is a broadcast application, the backhaul requirement could be met by a single 5Mbits/s satellite downlink transponder.

This approach seems to have much to recommend it at first sight. It permits the use of existing infrastructure, existing spectrum, and the Mobile Multicast-Broadcast Subsystem defined in UMTS Release 6 to produce an economic and capable mobile TV system. The demonstration seen by MCI was a success, with high-quality TV streaming to smartphones and, indeed, full-screen TV streaming to a laptop from a base station installed nearby.

However, as ever in the industry the reality is a little behind the hype. IPWireless boasts that the additional cost of enabling a handset for TDTV reception will be as little as $10. But this is not expected to be a reality until sometime next year. For the moment, the dual-mode FDD/TDD chipset is much more expensive and is still in the form of either an SD card or an external sleeve for smartphones and PDAs. Or at least it would be if the chips were ready.

At IPWireless's demonstration in London, the highly impressive TV streams to the laptops were being received by the firm's UMTS-TDD data cards, as used in its major TDD wireless-broadband deployment in the Czech Republic with T-Mobile. In fact, they were exactly the same data cards, in shocking T-Mobile pink. And the bevy of HTC smartphones? The HP iPaqs? The Oqo?

Well, the demos weren't using the SD card version. That was for sure. In fact, the TV was being streamed from a Netgear Wi-Fi router in the corner to these devices. To be fair, though, the router was itself getting its data from a UMTS-TDD card plugged into its expansion slot.