GSM-R: a migration strategy to its successor?

The thorny question of what should replace GSM-R as the future radio bearer for track to train communication has recently been the subject of conferences, discussions and, yes, even articles in Rail Engineer. Considerations of 4G, 5G and Wi-Fi options are being assessed, all having their pros and cons, and a few commercial interests have emerged as well.

Whatever and whenever the decision is made for the replacement technology, a major consideration will be how to migrate from GSM-R to its successor. Very little thought seems to have been given to this but Kapsch, as one of the leading suppliers of GSM-R infrastructure, presented a paper in London recently setting out the challenges involved and how they might be overcome. The logistics are somewhat frightening, but they nonetheless will have to be tackled sooner or later.

The present UK position

The GSM-R network in England, Wales and Scotland has been fully operational since 2015 and now comprises two MSCs (Master Switching Stations), 3,000+ masts and base stations, 9,500+ cab radios, 4,500 connected trains and 900 signaller terminals, all covering some 15,000km of railway. It’s quite an impressive portfolio, but having to change it all over to something else will need a very carefully thought out migration plan.

The GSM-R industry group has confirmed that it will support GSM-R until 2030, which sounds a long time off but, in reality, is only just over 10 years. In project terms, that deadline will come all too soon.

The Future Railway Mobile Communication System (FRMCS) study, being led by the UIC (Union Internationale des Chemins de fer – international union of railways), is making progress with developing a functional specification that the new radio network will need to provide, which is good news, but it offers little advice as to how this might be delivered.

It now seems likely that 5G will be the chosen way forward under the 3GPP R15/R16 (third generation partnership project) terrestrial coverage specification, but this has still to be confirmed. Some requirements are, however, sacrosanct – there must be bearer independence for railway applications, the system must have high robustness and high availability. Those requirements are easily said, but they are not so easy to quantify within a specification.

Functionality

The emerging vision for the new radio system is looking at three areas:

Critical communications – the bearer for ETCS and ATO operation, secure voice communication between driver and signaller, provision for emergency and group calls, real-time video imagery for any occurring incidents;

Performance communications – on-train telemetry, maintenance of non-critical infrastructure, non-critical real-time video, wireless communication for on-train-staff, on-train voice announcements;

Business communications – applications including information to the public and passenger communication connections.

These are listed in order of priority and it may not be possible to deliver all of them at the same time in any migration plan.

However, the FRMCS thrust is to have a single network under the railway infrastructure manager’s control for all safety critical application, but also to allow greater flexibility between dedicated rail and public network operators for performance and business applications. Indeed, it is seen as very likely that business applications will be entirely given over to public operators.

This mix of networks might entail having a ‘Mobile Communications Gateway’ as part of the on board system.

Radio spectrum

GSM-R enjoys a European-led dedicated allocation in the 900MHz band, comprising both a 4MHz uplink and downlink bandwidth, which is now recognised internationally. Recent channel allocations adjacent to these have created some interference problems but improved filtering has largely overcome the problem.

Knowing the future challenges that a migration will cause and the wider use of radio for other purposes, the EU has made a declaration to allocate an additional 2 x 1.6MHz of spectrum adjacent to the present GSM-R uplinks and downlinks. This will be critical to any migration plan as it should permit in-band co-existence of present and future services.

The present GSM-R infrastructure is built around 900MHz operation, with radio masts and towers sited for the coverage patterns of that band. The possibility of having both multiband and wideband aerials on the same sites could cause problems with mast loading, so engineering calculations will be needed. Appropriate separation of aerials on the tower or mast will be sensitive to avoid unwanted interference.

Train aerials, which can be a challenge to fit on the different classes of rolling stock, should ideally be capable of being used for both old and new systems if the spectrum allocation is the same, but there are technical challenges to this which may mean more than one aerial. No change to the permitted radiated power limits is expected.

A special case for deploying 5G technology in the 900MHz band will be needed, as this is not currently permitted for public 5G usage. It may be necessary to provide some additional radio sites to minimise interference from neighbouring users and to create additional traffic-handling capability by having smaller cells.

There is also the potential for having FRMCS services moved to either the 1900-1920MHz or the 2.3GHz bands. These might end up as complementary bands to provide additional spectrum to the 900MHz allocation. There may be complications in using these bands as different countries (including the UK) have parts of them allocated for other users and purposes.

There is lots of competition for spectrum allocation and it may require the railways to co-exist with other users. However, this would almost certainly mean that new infrastructure would have to be to be planned and it would create fitment problems on rolling stock, so, for now, it is not a preferred option.

Planning for migration

It is clear that any migration of this magnitude cannot be achieved overnight. The development of a 5G railway network is unlikely to produce new systems before 2023 at the earliest. Assuming all goes well, there will then need to be several years of co-existence of both old and new networks.

These sorts of challenges have happened before when planning changeovers involving both track-based and train-borne equipment. It is usual to either duplicate the infrastructure or duplicate the train kit, which is fine for a self-contained line such as a metro but much more difficult for a main line railway with many interconnecting lines and different types of rolling stock.

In view of these problems, Kapsch seems to be recommending a mix and match of duplication. Any plan must firstly concentrate on the critical communication need, primarily the bearer requirements for ETCS and voice communication to and from the driver. The former, although at first sight appearing to be the most important, will only have that status if ETCS with no lineside signals is the only control system for the line. If the ETCS is an overlay, then a more relaxed way forward might be adopted whereby, for a short period of time, trains are driven traditionally to whatever aspects are showing on the lineside signals.

By 2023, it is difficult to assess just how many rail routes in the UK will be operating with full ETCS Level 2 and no signals. Certainly, the Cambrian line, as that is already equipped in that fashion, and probably the southern part of the East Coast main line. Beyond that, it is anyone’s guess and much will depend on the roll out programme of the Digital Railway team.

It must be said that predictions for ETCS adoption over the past five years have been wildly optimistic, so any dates into the future should be treated with a degree of caution.

The Kapsch recommendation is that radio infrastructure would need to be duplicated in as short a time as possible, but concentrating on the ETCS-equipped routes as a priority. This will mean the provision of new core network elements, new base stations and, possibly, new aerials, with the necessary redundancy to meet the high availability requirements of the railways.

The next generation of network design will be much ‘flatter’ than at present, meaning that MSCs and base station controllers may disappear. Whatever the outcome, the new network must operate alongside the existing GSM-R network. Assuming the 900MHz band is continued, then the existing masts and towers would be suitable, subject to loading limits for new antennae.

Where coverage is provided in tunnels by radiating cable, then a means of mixing the old and new radio signals will need to be devised, but this should not be a major problem.

The new radio infrastructure will require a separate land-based network that, logically, will be based on IP communication, including all the backhaul links to and between the base stations. Clearly this network will need full resilience, synchronisation for both frequency and phase (PTP – Phase and Time Protocol), path asymmetry, accuracy of time stamp and packet delay variation, plus, of course, minimising the threat of cyber attacks. With regard the latter, the security offered by 5G should be markedly improved over GSM-R.

Even when decisions have been made on duplication, a carefully planned migration strategy will be required to manage the actual changeovers. Kapsch has come up with what is termed a ‘whitespace’ plan, whereby a wideband carrier is superimposed on to the GSM-R channel to enable the continuance of GSM-R services plus up to 5.6MHz of 5G operation. This would allow the gradual introduction of FRMCS services on to the existing GSM-R spectrum but assumes at least an additional 1MHz of bandwidth is allocated.

Included in this plan will be the smooth decommissioning of GSM-R frequencies, which can then be immediately used for FRMCS services, giving much more flexibility to the migration programme.

Rolling stock considerations

The retro fitting of rolling stock is both complex and expensive, regardless of what equipment is involved. The cab radios do not take up much space, so having to have two radios in the cab may not be as onerous as some retrofits.

As well as the cab radio, there will need to be provision of the train aerial(s), additional power supply requirements and a second train data radio associated with the ETCS kit. None of this will be straightforward and significant expense will be incurred.

The Kapsch view is that, once the decision on the new radio standard is finalised, then fitting the equipment to the trains should become a franchise requirement. It would also be advisable for any new rolling stock procured following the radio decision to be made ready for (or even provided with) two train radio configurations, including multiple aerials along the train.

Changeover en-route

With a mix of both infrastructure and train-borne kit, there will then be the challenge of which radio is used on the train depending on where it is on the journey. For the voice radio in the driver’s cab, this should not be too much of a problem as both radios will be powered all the time and it will be obvious which one is active to transmit or receive messages.

For the train radio provided for the ETCS equipment, however, the problem will require more thought since the ETCS movement authority (MA) will require a near-constant transmission of confirmation data from the control centre. In the event of non-receipt of the data within a few seconds, the MA will be lost and the train will automatically brake to a standstill until the MA is restored. Thus, the continuance of the radio signal must be guaranteed during any changeover from GSM-R to the FRCMS and vice versa.

The use of eurobalises to achieve this may be one solution, although it is a clumsy one as the changeover point will be altered from time to time as the radio infrastructure is updated. The reprogramming of balises to reflect the new condition will be an unwanted burden and therefore it is more likely that some on-air signal will need to be devised.

Capacity and priorities

The question is posed as to whether the existing 900MHz allocation, without any additional bandwidth being provided, could support both the GSM-R and FRCMS systems. This seems unlikely, hence the negotiations with the CEPT to obtain the additional 1.6MHz of spectrum.

Although the ETCS operation requires near constant reception, the bandwidth required for individual trains is low – about 4kbps. Even with many trains operating within one radio cell, the now proven and adopted use of packet switched data has considerably eased the problem for both old and new to be contained.

However, should any voice communication take place at the same time, and possibly from multiple trains, it is quite likely that insufficient spectrum will be available. In very busy areas, such as city centres where radio signals will spill over from one line to another, this is already recognised as a problem with GSM-R. One solution would be to re-engineer the cell structure so as to create smaller cells but, even with this, it is a complex engineering and coverage challenge and the chance of spill-over remains.

To provide all the services envisaged in the FRCMS specification would be unrealisable, so priority will have to be given to the critical communications services, with other usage being restricted. The business and passenger communication requirements might have to wait for another chunk of spectrum in another band to become available and/or agreement with other network operators to provide these.

As such, the case for additional spectrum is reinforced. Much will depend on how usage works out in practice and some traffic modelling will be necessary.

Some final thoughts

That someone is giving consideration to how GSM-R will be migrated to a new radio standard is commendable. It is not going to happen tomorrow and the first migration will probably not be until 2023. However, that falls within Network Rail’s CP6, which leaves no great time for the planning process.

There is a significant risk that this advice will be ignored. Many signal engineers have a real problem in understanding the importance of the radio link within the overall ERTMS provision. It is as if they think ‘it will always be there’, a dangerous assumption. Even radio planning authorities have been known to publish papers that state “the railways have GSM-R so all is sorted”.

The harsh reality is that, however much ETCS equipment is installed on both trains and control centres, if the radio link is unavailable and/or unreliable, then it is all a waste of time. Which should provide network planners with food for thought.

Thanks to the IRSE London & South East section for organising the presentation and to Pierre Tane from Kapsch for delivering the paper.