Network Rail claim that their existing radio systems are obsolete and must be replaced as soon as possible. They are getting old, it's true, but they're not obsolete yet by any means - when necessary, Network Rail always seem to find that their useful lives can be prolonged.

As far as we know, Network Rail's first attempts to replace them with digital technology was a short-lived radio system called DART. This is the first reference we could find to it:

…we are developing a new system called DART (Digital Advanced Radio for Trains). Last year a £13 million contract was awarded to Siemens UK to develop the system hardware.

Memorandum by Railtrack (RS 01), 9 December 1998

More information is given in the 1998 Network Management Statement:

We have...launched a project to develop a digital radio system, initially to replace NRN but with the capability also to replace CSR where the additional performance of a 'secure' system is required.
The International Union of Railways (UIC) has led development of a new international train radio based on digital cellular telephony standards. We have been participating in this project and we have decided to proceed with a digital system which we have called DART, using the public cellphone network as the radio bearer.
DART will involve working with the selected network operator to provide the required coverage levels along the lines of route where it is implemented (including tunnels and cuttings where the 'secure' mode is required). This will have the secondary benefit of giving much improved performance to other users of the selected network travelling by train. Certain 'special' features also need to be provided to ensure emergency calls get appropriate priority and that train drivers cannot be called by unauthorised persons.
It is planned to have a pilot DART radio system operational during 2000. Once proven, the system is planned to be used to replace the existing Cab Secure Radio equipment in Scotland and currently used to support the Strathclyde Manning Agreement. The benefits of providing a good quality train radio have been outlined in a number of accident inquiry reports. DART has the potential to provide such a radio system in a cost-effective manner.

Network Management Statement 1998, page 60-61 (3.1mb PDF)

How eminently sensible: sharing existing mast capacity with one or more mobile phone operators in order to minimise wasteful overlap, and so provide a modern in-cab radio system "in a cost-effective manner", while extending mobile phone coverage to rail passengers.

It was too good to last, of course: just a few months later, DART had been unceremoniously dropped. This was alluded to in the Cullen Report (paragraph 12.29): "Counsel for First Great Western rightly submitted that it was most unfortunate that there was still no national system of radio communication between trains and signallers. Counsel pointed out that First Great Western had pressed for a modern radio system. In April 1999 Railtrack had cancelled the DART project (to introduce such a system nationwide) without explanation. It was common ground that such a system is desirable and,...there appeared to be no insurmountable technical difficulty."

The nearest we've been able to find to an explanation for the scrapping of DART is this:

During the course of 1998/99, we established that the National Radio Network (NRN) can be maintained as an effective operational radio system until about 2008, which is about 4 years longer than previously envisaged. In addition the tendering process has shown that use of the public cell phone network for our operational railway purposes is not cost effective in the context of renewal options. We have therefore decided not to proceed at present with Digital Advanced Radio for Trains (DART). As such, we are re-evaluating options for the long-term renewal of NRN and Cab Secure Radio (CSR). We intend to discuss the options for developing a modern digital based approach to meeting the radio requirements for the railway with the industry during the course of 1999/2000.

The Annual Reconciliation Statement, Report of Progress Against Plans Set Out in the 1998 Network Management Statement, Network Rail, July 1999 (416kb PDF)

The second sentence, “the tendering process has shown that use of the public cell phone network for our operational railway purposes is not cost effective in the context of renewal options”, is almost impenetrable, but the implication is that Network Rail did not enjoy the experience of cooperating with external organizations such as mobile phone operators.

Network Rail is, apparently, notorious within the mobile phone industry for its stand-offishness and not invented here syndrome. Given the Government’s efforts to force mobile phone operators to cooperate by sharing masts, it is outrageous that Network Rail should be allowed to withdraw from cooperation without explaining its reasons in detail.

In the Cullen report, it was stated that “Railtrack was developing a national radio project in a form of CSR which was European compatible. This was for all passenger lines and that was being looked at as a matter of urgency.” This new radio system was almost certainly GSM-R.

The GSM-R network started as a straightforward digital replacement for the existing analogue in-cab radio systems, as suggested in Lord Cullen's recommendation. In a document published in January 2001 (1.3mb PDF), Railtrack estimated that they would require 1425 GSM-R masts at 12-kilometre intervals - roughly the same number as they have already. The document states that “Where possible existing NRN and CSR sites will be re-used.”

So far, so sensible. However, within a year the GSM-R radio project had been swallowed up in the grandiose ERTMS Level 2 project:

The Railtrack GSM-R network is currently being developed as a replacement for existing radio systems and is currently authorised (and funded) to support the voice application only. Support for ETCS [European Train Control System, part of ERTMS ] data would represent an enhancement to the GSM-R network and would require separate, additional funding to be made available.

ERTMS Program Team [EPT] Final Report April 2002 (2.1 megabyte PDF).

Modifying the GSM-R network to support ERTMS not only costs more, it also requires more masts, as the same document makes clear:

The radio signal strength for ERTMS data is higher than for voice and would require 10-20% more base stations along the line. The National GSM-R Network is being planned to have the signal strength for data confined to the high speed TENs Lines, although agreement on funding has yet to be reached. The rest of the network is currently only planned to support voice, and for ERTMS to function satisfactorily, the signal strength for data should be provided wherever ERTMS Level 2 is required.

Because of Network Rail's subsequent decision to extend ERTMS to all UK lines, not just the three high-speed lines originally envisaged (the west and east coast main lines and the Paddington-Bristol/Cardiff line), the number of extra masts required went up by 40% rather than the 10-20% estimated here. That, of course, is not how Network Rail or the Government explained it - here's an excerpt from Hansard (31 March 2003):

Mr Jamieson: I understand that Network Rail is upgrading its analogue radio systems network of around 1,400 telecommunications masts to a national digital network of approximately 2,000 masts. The increased number arises from the extended coverage of the new system across the entire rail network, including all tunnels and cuttings.

The truth is that virtually all the masts in new locations are dictated by the demands of ERTMS, not the GSM-R in-cab radio system (which, according to Network Rail, could be supported by the same number of masts as the existing CSR and NRN network).

The additional height of the masts is also dictated by the demands of ERTMS, since a voice radio system based on GSM should require both fewer and shorter masts.

The explanation is, apparently, that "the Network Rail system has been designed to facilitate contact not just to the adjoining cells, but to two cells in each direction. This provides a 'failsafe' safety factor so that the system will continue to function even if one transmitter is out of action, but the consequence of this is that taller masts are required – the proposed masts are typically 30 metres or more in height, and are to be installed at appropriate points along all main lines throughout the country." Such a degree of redundancy can only be warranted for safety-critical data communications required for automatic train control - that is, ERTMS. For driver-to-signaller voice communication, some redundancy may be required in very congested sections of track with multiple points and signals - but certainly not throughout the country, and certainly not to that degree.

Only 188 GSM-R masts had been erected by March 2004, most of them on the West Coast Main Line. None of them is being used for GSM-R voice communication. According to Alistair Darling, Secretary of State for Transport, in a letter to Patsy Calton MP, GSM-R hasn't even been tested in the UK yet - trials are scheduled for the end of 2004. It is hoped that the radio system will be operational on some lines by 2006. Some of the masts are being used now for yet another Network Rail radio system called Interim Voice Radio System (IVRS). A Marconi press release explains that IVRS is "a GSM-R solution, supplied by Nortel Networks, to support Railtrack's existing radio system" - presumably NRN.

So the masts won't even be used for GSM-R radio communication until 2006. And they won't be used for ERTMS (the reason for almost all the new ones) until 2015 at the earliest. By then, GSM-R will be at least as obsolete as CSR and NRN are now.