New 3G Partnership in OEM’ Building 3G Systems

An article by Richard Deasington ( inset ) of WFI Consulting

A recent study for a client building a new 3G network in Europe identified that power consumption of Node Bs varies significantly between vendors and can in theory be significantly greater than for GSM BTSs. In practice however we concluded that the differences are not really significant because the average power consumption is way below the theoretical maximum. The extra costs could in any case be more than offset by effective re-negotiation of the electricity tariff with the power company.

Lets start with the fundamental ‘laws of physics’ type of argument. The data rates used in 3G are much greater than for 2G (GSM) and therefore more sophisticated modulation schemes are employed to achieve this greater throughput. The more complex waveforms used by the 3G modulation schemes are less tolerant of distortion. The waveforms produced by the WCDMA Node B Transceiver should be almost pure – the primary source of distortion are the Power Amplifier (PA) modules used to increase the milliwatt level output from the Transceiver to the usual 20 – 40 watts to be fed to the antenna.

We can measure the ‘efficiency’ of the PA modules in terms of the amount of power they consume in watts (i.e. the DC input power feeding the PA) compared with the watts of radio power output they produce.

In order to achieve the low distortion levels needed for 3G transmission to work, the PAs are operated in a mode that is much less efficient than can be achieved with 2G transmission: typically less than 10% at maximum output power compared with 40% for GSM. Additionally the efficiency falls as the output power falls, generally to less than 2% at 2 watts. What does this add-up to across a network?

If we take a medium sized country we will need to provide say 10,000 cell sites, each with three sectors and one carrier per sector. Of course there will be ‘hot-spots’ with more than one carrier, but equally there will be areas of low-density coverage using ‘omni’ or ‘OTSR’ single sector sites. Thus there will be about 30,000 PAs across the whole network. If we compare the power consumed just by the PA elements of the 3G Node Bs or GSM BTSs and assume that we are using 20 watt PAs in both cases the power consumption (in kWhr / year) for the three cases would be:

20 watts GSM = 0.020 kW ? (100%/40%) ? (24 ? 365) ? 30,000 = 13,140,000 kWhr / year
20 watts WCDMA = 0.020 kW ? (100%/10%) ? (24 x 365) ? 30,000 = 52,560,000 kWhr / year

Additionally we could consider the power consumed by the air conditioning units to extract the waste heat, usually adding about another 50% on top of the basic consumption, giving in round numbers a network-wide annual consumption of 20 million kWhr for GSM PAs versus 80 million kWhr for WCDMA PAs. The difference of 60 million kWhr at a typical energy cost of €0.1 per kWhr gives a difference of €6,000,000 – interesting, but not make or break for an operator.

At the typical overall average utilisation one finds in a mobile network the PAs will only consume half of their maximum rating, making the difference only €3 million – even less interesting. The same arguments mean that claims by some vendors to be able to improve PA efficiency by several percent in future versions looks even more uninteresting in the context of the overall business.

Looking at the overall electricity consumption of an operator – added up across all the cell sites, switching centres, call centres and offices – the annual consumption could easily exceed €20m, and at this level the introduction of competition in the energy sector in most countries can lead to substantial savings. Our advice would be to ignore the PA power efficiency debate and focus on doing a better deal with the electricity supply company!