Anyone who has ever tried to make a call from a mobile phone only to be denied access because the network was too busy, will understand the need for third generation (3G) telephony. As mobile phone use has expanded, the existing networks have begun to creak under the strain.

The promise of new radio spectrum is a major reason why the world's telecoms operators have paid such high prices for 3G licences. Existing cellular systems operate either on the very crowded 800MHz (megahertz) and 900MHz wavebands or at the less crowded 1800MHz and 1900MHz wavebands.

In contrast, many 3G systems operate at 2GHz (gigahertz), a waveband that is being cleared of existing traffic-thus, 3G licences give operators room to expand their services, especially the new value-added data variety.

"In urban areas, 800MHz is the most efficiently and highly used part of the radio spectrum," says Mike Short, chairman of the Mobile Data Association. "In most European countries where 3G licences have been awarded, we have seen a 70 to 80 per cent increase in capacity based on the number of megahertz available.

" Steve Walker, marketing director for UMTS at Ericsson, the Swedish mobile phone company, compares existing second-generation networks with a motorway crowded with cars. General Packet Radio Service (GPRS), the generation 2.5 upgrade, helps users to share cars, but there are still only three lanes. Enhanced data rate for GSM enhancement (Edge), an available upgrade for networks that do not get a 3G licence, is like putting double-decker cars on the same motorway. In contrast, 3G networks will use an entirely new method of communicating between the handset and the base station, known as the "air interface". This will be more will be like building a new six-lane motorway alongside.

Many existing networks use a system called time division multiple access (TDMA). It divides the frequency into six or eight repeating time slots. Two time slots are assigned to each caller, one for talking and one for listening. Voice is compressed and encrypted and sent in tiny bursts. According to Mr Walker, GSM transmits voice for half a millisecond every four milliseconds.

The base station then handles the calls by knowing which handset is using which time slots. However, this is inefficient because each conversation has to have a dedicated slot and even silences are compressed and transmitted. Also, adjoining base stations have to use separate frequencies in order to avoid interference.

TDMA networks will change to wideband code division multiple access (W-CDMA) when they upgrade to 3G systems. CDMA is a narrowband technology that is already in use on voice networks in the US where it is now referred to as "cdmaOne". Networks using cdmaOne will upgrade to CDMA2000, a similar standard to W-CDMA. According to Joe Barrett, head of 3G market positioning at Nokia Networks, the equipment supplier, W-CDMA has twice the capacity as cdmaOne.

Under these protocols, each conversation is split into separate packets that are transmitted together, like a computer network. Each conversation is given a different code, which is sent with each packet, enabling the network to reassemble them into continuous speech.

The main problem with CDMA is that, as more users communicate concurrently with the base station, they compete for finite power at the transmitter. This reduces the range of the cell, a phenomenon known as "cell breathing".

"The reduction in cell size of an isolated cell carrying high data rates or heavy traffic can easily be 50 to 80 per cent," says Christopher Davis, director of Quotient, a telecoms consultancy that publishes the Quantum 3G mobile network design tool.

W-CDMA and CDMA2000 have been designed to manage this process. They also make more efficient use of the available spectrum because all base stations can use the same frequency. The coding identifies the conversation and eliminates interference from adjoining base stations.

However, 3G networks will require more cells than 2G networks. The range of radio waves decreases as frequency increases, so theoretically a 2G network operating at 1800MHz or 1900MHz requires twice as many cells as one operating at 800MHz or 900MHz, but in practice this is not always the case.

Good planning would allow a 3G network to be installed on existing base stations. Although this would give coverage, it would not give the ability to handle higher data rates or handle the volume of traffic anticipated.

"The higher 1800MHz to 2GHz frequencies travel in straighter lines and don't go round corners," explains Robin Potter, director of UMTS business development at MSI, Marconi's mobile network consultancy. "In urban areas it doesn't make a lot of difference, but it impacts heavily in rolling rural countryside. In practice, if you make some compromises on rural and suburban coverage, operators will need about twice the number of cells."

Base station equipment will need to be upgraded to handle the new air interface. Antennae will need to be replaced by dual-mode versions that can work with both 2G and 3G handsets. The equipment in the base station cabinets can be upgraded to provide limited 3G capacity. However, this is only a temporary measure and new equipment will eventually be required.

The transmission network from the base station does not require any technical changes. However, many of these are being upgraded to provide additional capacity to handle the expected extra traffic.

"Third-generation networks bring 70 to 80 per cent more spectrum and new technology that is more efficient in carrying traffic," concludes Mr Short. It is important to understand that 3G networks represent an enormous change for network operators, whereas GPRS is a relatively minor upgrade. However, users will notice a significant change when they move to the "always-on" GPRS that will bring them a wide range of new data services and devices. For them, the move to 3G networks will be relatively minor, as most of these new services will migrate with them.