Practical Case Study: Train Satellite

For more rural routes, to fill not-spots and to offer diversity of connection it is likely that Satellite communications may be chosen.  This will need one or more antennas to be installed on each train in a location that maximises visibility of open sky – most likely a carriage roof.

The UK is regarded as the birthplace of railways and was a very early adopter of rail as an alternative to horse-drawn wagons and water freight.  Bringing a new transport mode, and in a race with other transport pioneers at a time of great industrial expansion, railway builders worked with limited budgets and in a hurry.  This first-mover penalty leaves UK rail with tunnels and bridge arches of relatively small sectional (gauge) profile, meaning UK trains continue to be constrained in terms of the available space between train roof and trackside structures. 

Until recently, for the relatively large and flat antennas needed for satellite communications, a curved train roof profile represents a significant technical challenge.  A new generation of antennas has made the task somewhat easier, with less requirement to significantly modify a train roof.  However, there are still challenges around being able to install in optimal locations on the train roof due to a secondary complication.
Due to the constrained UK railway profile, train builders have utilised structural (monocoque) body-shells to maximise the amount of usable space within passenger trains.

Their design challenge has been further compounded by increasing train complexity meaning a lot of extra equipment needing to be installed, both in the void between the train roof and the ceiling in passenger areas, and on the train roof itself.

This complexity and congestion, on and under the train roof, presents significant challenge for those looking to retrofit satellite equipment.  In practice, this results in increased costs for retrofitting satellite panel antennas and routing the associated cabling.  This cost penalty is further compounded by complex train systems often needing to be temporarily removed and reinstated.

A supplementary challenge will be the approval of antennas for electrical, EMC, fire and robustness standards compliance.  This involves a time-consuming and costly process of testing and certification which needs to be undertaken for each product.

Finally, whilst UK Government has shown great commitment to explore the role of Non Terrestrial Networks (including satellite) in addressing poor train passenger connectivity, this does not appear to be joined-up with a parallel consideration of how UK trains will migrate operational connectivity to a new FRMCS (Future Rail Mobile Communications System) approach.

UKTIN are extremely grateful for considerable support received from both Clarus Networks and CGI, not-only in helping to clearly describe the problems faced by UK rail, but also for suggesting how engineering, logistical and commercial challenges might be addressed.
 

Image

The train roof is already likely to include multiple antennas to allow the MGC to access MNO networks. The satellite upgrade would involve adding one or more flat panel phased array antennas/terminals

From a commercial implementation perspective, the solution may be treated as an upgrade to the existing train WiFi system by providing a supplementary data bearer for the MCG to utilise.  However, due to the various generations of Wi-Fi/Comms solutions deployed from a number of vendors (including Icomera, Nomad and McLaren), replacement of existing on-train networks may be required to best exploit the available bandwidth that satellite communications could provide.  However, if an upgrade is viable, then the installation and integration of the Satcom solution would be relatively straightforward (away from the mechanical and electrical challenges previously described), as the service would be presented to the MCG via an Ethernet cable.

Operators of viable satellite data services comprise the long-standing GEO (geostationary) operators e.g. Inmarsat, Eutelsat, Iridium etc and the new LEO (low earth orbit) operators e.g. Eutelsat (OneWeb), Starlink and Kuiper.  

Use of GEO services continues to be limited by the large dimensions of the antennas required.  Also, latency issues have not been, and probably won’t be, overcome to enable effective aggregation/bonding techniques to be used.  The less than advantageous orbit of GEO constellations adds a further burden, perhaps explaining low take-up for train communications in more Northern latitudes. Work undertaken by CGI for Project SODOR concluded that line of sight to a GEO satellite is available <75% of the time on many UK routes.

Until recently, the form-factors of LEO antennas have also been problematical for the majority of UK rolling stock.  However, at the time of writing, a physically viable Starlink antenna is understood to have undergone compliance testing and suitable OneWeb train antennas are likely to follow.

For LEO services, Starlink currently offer upto 250Mbps downlink with the aim to deliver 500Mbps with the currently generation constellations by the end of 2025 (which may be as high as 1Gbps).  At the time of writing, the Starlink website states that no new consumer services are being sold in the Southeast of the UK, indicating that the service could be congested and actual bandwidth to trains could be limited.  However, in densely populated areas, it is likely that satellite services would be aggregated with MNO and some private trackside networks by the MCG (also known as the FRMCS TOBA box) which could provide an acceptable level of service for short to medium-term requirements in many scenarios.

Security arrangements are unlikely to differ from existing solutions, where data transmission is secured (encrypted) to and from the vehicle.  Multiple techniques are already effectively applied, and these are seen as sufficient, particularly as most aggregation/bonding techniques and secure by their nature (splitting packets etc), in addition to using secure data bearers.  Onboard the train, the security of the MCG (and associated infrastructure – WiFi access points etc) is likely to be assured by intrusion and configuration inspection software.

In considering the choice of data connectivity to provide or upgrade train services, consideration of the relative merits of the different network options is likely to include:

MNO(Cellular) – Likely to represent the lowest cost of provision due to utilisation of existing networks which are nationally ubiquitous.  Also, potentially the lowest cost of upgrade through using existing on-train antennas and RF feeds.  But has multiple disadvantages: Best endeavours approach to trackside coverage (not spots), extensive service contention (periodic low data throughput) and high data usage costs.

Private Trackside Mobile Network – A dedicated network is likely to offer the highest bandwidth and a consistent data connection, where data usage costs could be considered “unlimited” (ie paid-for in roll-out investment cost).  But again, with challenges:  High costs for deployment (cap-ex), support (inc future technology upgrades) and maintenance (op-ex) of the trackside equipment.  The potential proliferation of multiple non-standard solutions may also result in high cost of train deployment for “go everywhere” fleet (eg CrossCountry Trains) needing to deploy a greater number of technologies and antennas.

Satellite – High Bandwidth and High Availability (LEO) with no new ground infrastructure needing to be deployed.  Per-Gb data usage costs have fallen to a level comparable to public MNO rates.  Disadvantages include the higher antenna hardware cost (vs MNO or private) and challenges of train installation work to maintain physical clearances for tunnels and bridges on some routes.

 The realisation of a marketplace of technology solutions would be predicated on industry being able to realise a viable financial model for both train fitment and affordably priced satellite connectivity service models.  Any future development of a multi-network (eg OneWeb, Telesat and Kuiper) antenna may also help the financial model by driving agile service competition between the satellite operators.  Service interoperability would also allow rail to benefit from resilience to technology, commercial and political risk factors though a diversity of supply.  The following section (Commercial Model) advocates government-funded research which would stimulate stronger and wider market engagement in offering attractive solutions.

Image

The UK rail network currently includes over 3,000 passenger train units (locomotive and coaches plus self-propelled vehicles) which are predominantly WiFi-fitted but would require supplementary installation of satellite terminal equipment and antennas.  Whilst each class and sub-class of train is likely to require a separate design and approvals process, this still represents a significant market with economies of scale.

The UK rail industry is self-sufficient in the capabilities to undertake rolling-stock modifications, including design and approval.  The industry operates a geographically diverse and significant numbers of train maintenance facilities where modification work can be undertaken.  The specialist staff manage an ongoing programme of access to rolling-stock for periodic examination, maintenance and modification activities.  No specific up-skilling would be required for installation as the predominant specialist skillset required will be rolling-stock modifications.

The typical cost for first-in-class train modification and testing for satellite comms upgrade is expected to be in the order £35k to £45k plus associated Engineering Change (design review and acceptance) costs from the OEM if the vehicle is still under warranty, or if the OEM maintains an interest in the vehicle (acting as Design Authority).  This is based on integrating the satellite communication upgrade solution with an existing Wi-Fi solution.  For the remaining trains in the class and fleet a per-unit modification cost for the of around £20k per train – including hardware, metalwork and labour can be assumed.

If full replacement of a life-expired on-train WiFi system is required, then a 25% uplift on first-in-class design and installation may be assumed.  Hardware costs are likely to double for a new system fitment on a typical 4-5 car train. 

Upgrades and replacements are likely to be treated as a cap-ex deployment for an enhancement to a long-lifetime asset which may be amortised as an increased annual rental for the train or through an up-front payment. Ongoing maintenance and fault rectification costs are not foreseen as being significant and data usage charges are likely to be comparable to those charged for fixed terminals.  The proposed research work highlighted above will also be required to determine the necessary data throughput per passenger/journey and to potentially determine usage policies which may differentiate funded and chargeable usage of train WiFi.

Don’t underestimate the process to achieve railway specific compliance, it is costly, time consuming and many test labs don’t fully understand the requirements for satellite hardware.

Don’t expect Network Rail or other rail industry bodies to be experts or expect them to support or engage as they have their own specific objectives which may not be aligned.

Don’t look at FRMCS (the future 5G replacement for Network Rail’s GSM-R network) and passenger comms separately – they are both about connecting the train to the trackside and have very similar challenges and objectives.

Don’t assume that finding the capital funding to install satellite connectivity will solve the problem of connecting train passengers. The revenue model for free WiFi on trains is dysfunctional with the majority of “Free WiFi” services offering a poor experience.

Do understand where parts of the railway are different, and possibly more relevant for satellite.  Whilst GSM-R (a dedicated trackside mobile network) and ETCS (an in-cab signalling system) are well suited to the busy mainlines, the most rural of lines (notably in Scotland) are never likely to benefit from improved terrestrial train data connectivity.

Do engage hearts and minds across the whole rail industry and government to avoid the blocker of “not invented here” from an often-conservative decision-maker community.  Network Rail’s ETCS test facility at Hitchin delivered little technology enhancement, but convinced the industry that in-cab signalling could offer a reliable upgrade.

Do demonstrate that it’s valid technology through a pilot on service trains that currently offer poor passenger data connectivity. Delivery by the train operator and their maintainer will further reinforce industry confidence that they can be fast-followers and implement their own satellite upgrades.

Do prove on-train operation of a TOBA box (FRMCS on-train unit) to close-down the doubters and open a dialogue on how FRMCS and passenger connectivity can both be addressed in a single train fitment.

Do specify ongoing support of on-train satellite upgrades such that necessary maintenance activities will align with existing vehicle maintenance schedules, thus avoiding extended and expensive maintenance access periods.

Do try and limit proprietary hardware to avoid vendor lock-in.  As an example, MCGs utilise industry standard hardware and should be able to operate using any MCG vendor’s software.