Original URL: http://www.theregister.co.uk/2006/11/24/spectrum_analysis/

Pico GSM cells, what's the fuss all about?

We won a spectrum, now what do we do with it?

By Steve Kennedy

Posted in Mobile, 24th November 2006 12:34 GMT

Analysis What if you built a network, and nobody came?

In May this year, the frequencies 1781.7-1785MHz paired with 1876.7-1880MHz (known as the GSM Guard bands) were made available to 12 licensees under the Wireless Telegraphy Act (WTA).

They are known as the Guard bands as they separate parts of GSM spectrum from DECT spectrum. In the past radio equipment was known to "bleed" at the edges, so the guard bands were there to stop the bleeding spreading into spectrum that it shouldn't.

In the 20 or so years since GSM networks have been around, radio equipment is much more efficient and doesn't suffer from these effects, so Ofcom decided to make the spectrum available. Two sets of 3.3MHz is a valuable commodity in terms of spectrum.

Ofcom awarded the spectrum by holding an auction with a reserve price of £50,000 per license. Up to 12 licenses would be awarded. Ofcom allowed anyone to bid on any number of license slots (i.e. from one to 12 licenses) and the award was made purely on a financial basis.

Ofcom published the complete matrix of bids as the award was for between seven and 12 licenses. It was a close thing at eight licenses as a few bidders put in high entries for low numbers of licenses and dropped the amount as the license numbers increased.

Ofcom arranged the auction in a sealed bid process in a "what you bid is what you pay" arrangement, which led to the lowest price paid at £50,110 by Spring Mobil and the highest £1,513,218 by COLT.

Some have argued that the highest bidders paid over the odds, but Ofcom is putting a good spin on it saying it's in line with its mobile strategy. The total amount of the licence fees paid was £3.8m - not bad for Ofcom's first spectrum auction.

Of course, compared to the license fees paid for 3G spectrum (around £6bn per license) it's peanuts.

In the end it was close, but 12 licenses were awarded. They are national UK licenses, though the operators of the licenses have to cooperate so they don't interfere with each other. As part of the license condition all licence holders had a obligation to Ofcom to agree on an engineering co-ordination plan for the joint use of spectrum. The industry group is called Mobile200 and 11 licensees joined - O2 was awarded observer status for the discussions as it refused on principle to pay to negotiate the code of practice.

On 2 November the 11 members of Mobile200 handed the agreement into Ofcom, still without O2 having joined.

Much of the agreement concerns siting cells (like the Sitefinder database for GSM and 3G networks) and this database will be owned by Mobile200.

Though the licenses are only low power (sub 200mW compared to tens of Watts for traditional GSM systems), they are suitable for services such as in-building GSM, local area GSM (such as in a theme-park) or other constrained areas. There are 15 GSM channels available, each one being able to carry eight voice calls.

Having a reasonable number of channels will allow multiple operators to co-exist in an area and also allow single operators to cover larger areas (in such a way that multiple GSM basestations won't interfere with each other). If the cell is mounted on an external mast, it can't be more than 10 metres high, however, in-building use can be to any height (so build a plastic greenhouse on a terrace on Canary Wharf and you can get an "internal" cell quite high).

Winning Licensees

The 12 companies winning licenses and the prices they paid were:

British Telecommunications PLC - £275,112
Cable & Wireless UK (England) - £51,002
COLT Mobile Telecommunications Ltd - £1,513,218
Cyberpress Ltd - £151,999
FMS Solutions Ltd - £113,000
Mapesbury Communications Ltd - £76,660
O2 (UK) Ltd - £209,888
Opal Telecom Ltd - £155,555
PLDT (UK) Ltd - £88,889
Shyam Telecom UK Ltd - £101,011
Spring Mobil AB - £50,110)
Teleware PLC - £1,001,880)

A license, but what to do with it?

Having a license is all very well, but now licencees must be wondering what they've got themselves into. Just because they can run a GSM service doesn't mean anyone will use it, in fact it may well be difficult to get people onto your network. It's extremely unlikely the existing mobile operators are going to want to have anything to do with these new upstarts, they've invested millions (err, billions) to get to where they are today. The last thing they want is new entrants poaching customers or moving users off their networks when they move into, say, an office environment. They especially don't want their customer doing it with equipment (i.e. handsets) that they've heavily subsidised.

Unfortunately, what this means is that the new players are going to have to issue new SIMs (subscriber identity modules) and they won't work on existing GSM networks, or users will manually have to select the new network when they're in range. This makes it all very difficult, and users won't bother if it's hard.

New entrants could enter into roaming agreements with the current operators, but unless Ofcom mandates this (which is unlikely) there's likely to be strong opposition. Since some of the license winners already have GSM networks, they can offer localised services knowing there's no interference problems with existing infrastructure.

One way ahead is for a licensee to make an agreement with a foreign operator and the localised network just becomes an extension of their network, but then when users roam on to the network they'll be subect to roaming charges which, as both Ofcom and the EU Government know too well, can mean very high charges for the end-user. If roaming charges do decline, this may well be a way forward.

There's also a big potential opportunity for the Channel Islands GSM networks here, as they abide by UK numbering plans, so though they are considered "foreign" their numbers look like UK numbers, including mobile ranges. They could offer roaming agreements and even offer SIMs which would still look like UK numbers, unfortunately as they are foreign operators, high roaming charges still apply.

The clear opportunity ahead is for O2, Opal and Cable and Wireless (C&W). O2 has an existing GSM network, C&W owns Guernsey Telecom (and it has a mobile network) and Opal is owned by Carphone Warehouse which has an MVNO (mobile virtual network operator) agreement with O2 (O2, of course, has to let it utilise its back-end network).

Building a GSM network isn't cheap

Just because the licensees have got a license, that doesn't give them the facilities to run a GSM network, it just allows them too. There's much more that's actually required to put a GSM network in place.

A cell site is actually known as a BTS (basestation), BTS' are controlled by a BSC (basestation controller). A BSC can control around 100 BTS's.

BSCs connect to an MSC (mobile switch centre) which is like a telephone exchange.

The network has to have an HLR (home location register) which stores the state of mobile phone (i.e. which cell they're in, are they on or offline), really just a big database. If the network needs to support roaming customers, a VLR (virtual location register) is needed too, though that can be part of the HLR.

If SMS' (or texts) are going to be supported, then an SMSC (short message service centre) is also required.

The mobile operators spend millions of pounds on this kit, and a GSM guard band operator needs to have at least that as a minimum. Though the big operators have to contend with millions of users, even a small network is looking at at least £500,000 to get started.

Once all the network is in place connections are required to other networks so calls can be made and received. This requires telco interconnects (which also don't come cheap) and everything uses a telco protocol known as SS7.

There are companies which make pico cells (used on ocean liners and such like) but these cost around £2,000 each.

O2 ahead of the game

If a big mobile operator was to offer a pico cell solution, they generally have a major problem as they don't want the pico cell to interfere with their existing network.

Operators use what's known as a seven cell repeat patern. Each cell has six sectors (the six aerials that can be seen on a cell site). Cells and sectors can't use adjacent frequencies or they'll interfere. The seven cell repeat pattern ensures that frequencies can be re-used without interfering, but in the next seven cell pattern.

O2 now has a very efficient way of dropping new cells into places, using new frequencies which are guaranteed not to interfere with O2's existing network - or any other major operator's. It does have to mind other GSM guard band operators, but as the power output is small, even this can be easily mitigated.

It has a choice of 15 frequencies, so even if another guard band operator is nearby, they can both choose different (non-adjacent) ones.

Where O2 really gains is in the in-building/home use. One of the major costs of providing a GSM network is getting the calls back from the cell. By utilising a pico cell and a broadband connection, O2 has reduced this to practically zero, so it can offer cheaper calls just on that reduction.

O2 does need to convert the calls to VoIP (voice over IP) and that increases the bandwidth required (GSM encoding uses 13Kb/s, by the time that is packetised and put in an IP packet it may well be 26Kb/s, but that easily fits into a broadband connection).

O2 happened to have bought a broadband provider (Bethere), so it's not just a marriage of convenience, it actually gives it a very strong proposition in the marketplace.

The big question is how cheaply it can make a BTS. O2 is saying it can produce a combined cell/ADSL router for under $100 (around £60) which is around what a normal Wi-Fi ADSL router costs.

If it can pull it off, it has the broadband service and can offer local GSM connectivity using normal handsets without the problems that other services like BT Fusion suffer (specialised handsets with poor features, battery life etc).

It, of course, could also offer data services like GPRS (general packet radio service) or even EDGE (Enhanced Data Rates for GSM Evolution - which came first the acronym or discription?) and use them for localised data hotspots much like there are for Wi-Fi, but using nothing more than a GSM phone connected to a laptop.

Though O2 has a headstart, it will be interesting to see what innovative services the other guard band operators will bring to the game and who will survive in the long term. If the licenses aren't used for real services, they're just expensive bits of paper. ®