London Underground: The Forgotten Digital Railway


The railway’s attitude to signalling has changed a lot in the past few decades. In times gone by it was often regarded as something one unfortunately needed to run trains safely. On a modern metro railway it is now more positively seen as an enabler that helps maximise capacity.

With the publication of an item on legacy signalling for the Programmes and Investment Committee and various other recent publications, it seems a good time to look at the London Underground Signalling strategy, and the digital railway revolution already quietly underway underground.

The other digital railway

We hear a lot these days about the ‘digital railway’ – which is often taken to mean automated trains. Network Rail and the Secretary of State for Transport miss few opportunities to tell the public that things will be better in future because we will have a digital railway. On Network Rail there is very little to show currently except for a few miles of Thameslink track – although to be fair, what they have got is quite impressive.

Much less talked about is the fact that London Underground have had a digital railway on a number of lines for some time, are busy implementing it on four more lines and are planning to convert the remainder. We might be almost twenty years from a fully digital Underground railway, but there is a detailed plan to do so and much has already been implemented.

The name of the game (changer)

The game-changer is undoubtedly Automatic Train Operation (ATO), where the driving of the train is automatic and, nowadays, driven by the signalling system which with an electronic ‘central mind’ that can oversee a route or part of it. Note that this is a considerable advancement on the original Victoria line system, which did little more than automate the existing concept of track speed limits and a restriction on getting too close to the train ahead.

No gain without pain

The technology associated with ATO has gone through various phases. In the late 20th century it was painful to do and sometimes technology was pushed beyond the limits of the day. The failed promise of 36tph on the Jubilee line by the end of 1999 showed for probably the first (and certainly not the last) time what happened when people relied too much on promises of technology that didn’t yet exist.

The first decade of the 21st century saw what was achievable and the DLR and the Central line started to show how beneficial it could be once one succeeded in getting it working. In other words, the gains from success meant it was worth enduring a lot of pain, if need be, to get there. But it is only really now that there is a ‘can do’ attitude that isn’t over-optimistic. Sure, there will be problems – such as those Crossrail are battling in the Heathrow tunnels – but there are also clear successes. ATO using ETCS on Thameslink is a forgotten ‘good news’ story. It was installed on Thameslink and it worked with no discernible inconvenience to the travelling public.

Time, distance, movement authorities

This revolution can be seen almost as a third generation of railway operating philosophy. Initially trains were separated by time. A train departed and the next one was not allowed to leave until 15 minutes had elapsed. Needless to say, this often didn’t end up well. The next generation saw trains restricted by space (‘block sections’) so that only one train was allowed to occupy a block at a time. Initially there was nothing physically to stop a driver entering a block section he shouldn’t. The only power the signaller had was to restrict a train to a particular route by means of the points. In an emergency, signallers had no capability of ordering a train to a stop within a block section and on lightly used, long distance routes these block sections could be many miles long.

It is only now that the block working signalling concept is being overturned with either the elimination of block sections or, with some implementations, block sections so short that a train can occupy three or more of them at the same time – so crudely mimicking the effect of not being restrained by block sections at all.

With traditional ‘block working’ a train driver can move his train unless explicitly prevented from doing so (typically by a red signal). Modern train working replaces this with a system where the train will only move when the ‘driver’ has approved it being in automatic mode and the train receives positive confirmation from the signalling system that it has permission to move. With a modern system, if that authority to move is rescinded, or not received for any reason, the train will immediately brake and then come to a stop. What allows this is the removal of the need for communications to be rigidly geographical. In the previous generation, authority for a driver to move could really only be rescinded by putting the next signal at red or, far less likely, by use of detonators. In either case, one had to wait until the train reached a location where the necessary message that the train needed to stop could be given.

Train based, not track based?

A recent paper presented to TfL’s Programmes and Investment Committee provided much of the details of what is planned and, by using this and other sources, it is now possible to get a fairly comprehensive overview of the future signalling that will be implemented on London Underground.

The paper presented to the committee describes the change in signalling as

Advances in technology have enabled traditional track based systems to be replaced with train based signalling systems.

Now signalling engineers might have a lot of issues with that statement, but it is certainly true that a lot of signalling system equipment is now train based. There is still track-based equipment, indeed more so than before, but it is not so obvious. For example, there are no visible signals. In addition to trains and track, there is also a signalling centre, or multiple signalling centres, that do much more than just ensuring that trains don’t collide.

Automatic Warning System (AWS) and beyond

A few years ago, the most you had on a train was a bit of equipment to detect an magnet in the middle of the track. If a permanent magnet was detected the system would be ‘armed’. If a second magnet of opposite polarity didn’t ‘disarm’ the system shortly later the system issued a warning to the driver and if he didn’t cancel it the brakes would be applied. The second magnet (the ‘disarming’ one) would be energised if the train was about to pass a signal that was green (‘all clear’). If the driver acknowledged the warning signal yet carried on regardless, there would be nothing physically prevent the train from continuing to move.

Nowadays, with ATO, there is more on the train related to signalling. Initially there was uncertainty as to whether it was better for the train provider to provide the on-board equipment or whether it was down to the signal contractor. The current practice is for the train to be specified so it is ready to receive the signalling equipment (eg space for equipment, power supply, interface with motors and brakes).

A simple strategy

The aforementioned paper is concise in describing signalling policy on LU:

LU’s strategy is to implement train based signalling systems across its network

So, all the qualities of a good objective. Simple to describe, worthwhile, achievable. ‘Entire network’ would seem to leave no doubt that it really was intended to apply this to all Underground lines but, as it is, it does leave a little wriggle room to exclude Bakerloo services north of Queen’s Park if necessary.

We can now break this down to a strategy for various lines in different time frames. Here again, the paper is helpful because it separates the Underground lines into various groups and describes the policy more specifically for them. We go a little further and look at individual lines where appropriate.

Signalling strategies on London Underground. DTUP is the Deep Tube Upgrade Programme

Make do and Mend

There is one group of lines that has a policy of minimal work to be done. These are the lines that are due to have their signalling replaced within the next four years. These will be resignalled and converted to ATO as part of the Four Lines Modernisation program (4LM) being overseen by Thales, who are implementing their own system. The policy for these lines is quite clear:

During the upgrade, the existing systems are being maintained but not enhanced unless required to interface with new equipment.

Signal failure by line as portion of total failures

This is understandable, but inevitably signalling systems that are oldest (and therefore generally the least reliable) are the ones being resignalled, so there is the double whammy of an old system and upgrades only done when essential. Not surprisingly, this is reflected in the frequency of signal failures on the lines involved. So, as reported via IanVisits, more than 25% of all signal failures, or more accurately, signalling system failures, occur on the District line. In fact, over 49% of failures occur on lines involved in the Four Lines Modernisation, despite it only accounting for 40% of route mileage and a couple of other lines also being quite problematic.

Failures per track kilometre for individual lines

Even if you take into account the lengths of the individual lines, things look no better for the District line other than being only the second worst line, but then the Waterloo & City does have a disproportionate number of signals and points for its length. Of course, a portion of this may be down to Network Rail, so London Underground may be unduly castigated. The fairly poor performance of the Bakerloo line is almost certainly due to failures north of Queen’s Park rather than anything under London Underground’s control.

Of some concern must be the outer reaches of the Metropolitan line which were last comprehensively resignalled in the 1960s. This was as part of a scheme to electrify the line and provide quadrupled track as far as Moor Park. This is the final area that is due to be resignalled under the Four Lines Modernisation programme and it is not due to be completed until the end of 2021. This means that the signal engineers need to keep the existing signalling working for another three and a half years, yet there are reports that it is in a terrible state. This does not compromise safety, but it does impact reliability.

Four Lines Modernisation

The London Underground Railway Society’s (LURS) April 2018 magazine edition gave a good comprehensive update of this resignalling programme for the Circle, District, Hammersmith & City and Metropolitan lines. Since then the new combined control centre at Hammersmith has gone live, although it does not yet actually control any trains. The first section (Hammersmith – Latimer Road) is due to go live on 27th May 2018. In contrast to the Northern line, is appears that actual ‘switch on’ will take place without any closures being involved – though a lot of closures have taken place to run a complete skeleton service to ensure reliability of the system.

Sequence of resignalling the Circle line

This first short section will have the benefit of giving drivers an early experience of the system. It will be followed a month later on 24th June when ATO will be extended as far as Paddington (Circle and Hammersmith & City) station to complete phase 1. This combined migration area, for technical reasons, turned out to be the one that logically made sense to be brought into operation first. It was closest to the control centre at Hammersmith and it has the advantage of being a plain track section of railway with no junctions in normal use. There is the added complication of a terminus, but to offset this if it all goes horribly wrong at any time you have only disrupted a short section of the railway.

Much more dramatic is phase 2, which extends ATO from Paddington to Euston Square. It will also include Finchley Road to Baker St on the Metropolitan line. So, this will control Praed St Junction, Edgware Road station (four through platforms two of which are used for terminating trains), two terminating platforms at Baker St and Baker St Junction. Phase 2 also requires every single S8 Metropolitan line train to be equipped for ATO, since virtually all Metropolitan line services either terminate at, or pass through, Baker St station.

Such is the confidence in the signalling for Four Lines Modernisation that the date for phase 2 has already been decided. It is due to go live on 2nd September 2018. We will cover Four Lines Modernisation in more detail closer to this time.

Indeed, the LURS article, goes as far as to suggest that the most challenging sections to implement involve those that have to interface to other systems (Chiltern Railway running on the Metropolitan and District line trains running over Network Rail tracks). A few years ago it would be the junctions around the Circle line which were generally regarded as the most challenging part of the implementation.

Obsolescence Management

The Victoria, Jubilee and Northern lines are in a category of their own. The title of obsolescence management highlights a very real concern of modern signalling systems. In the past there have been far too many cases of engineers desperately searching eBay to source replacement electronic components that are no longer manufactured. In general, the pace of progress is advancing so fast that equipment can be obsolescent in a few years after introduction and the equipment is hard to source. In most cases there is a more modern alternative available that is more reliable, but sometimes the existing system must be modified to accept the new component.

When a signal failure can not only lead to thousands of hours of passenger delays, but also mean substantial lost revenue, it can often make sense to replace relatively modern equipment with the latest available. This will this reduce the likelihood of failure in future and, if a failure does occur, it may well simplify the procedure to replace the faulty component.

A simple example of obsolescence management is the replacement of filament lamps with LED equivalents in signals. Not only is the replacement product better, but it is much more reliable and, because it is in demand for new schemes, it is unlikely to be difficult to source.

A more relevant example is that on the Victoria line some of the non-safety-critical hardware consists of standard personal computers initially running Windows 7. The implementation team had their concerns about this choice but the argument in favour was that one would always be able to source a Windows-compatible PC and, hopefully, after testing, upgrade the bespoke software from Windows 7 to a later version.

Specific Upgrades to the Victoria line

The Victoria line signalling is unique on London Underground. Supplied by Siemens, not Thales, it is generally well thought of. Night Tube has meant is has been upgraded since installation and there have been various tweaks for 36tph.

More significantly, the Victoria line is probably about to undergo another significant signalling upgrade – this time to the depot at Northumberland Park.

To quote:

5.3.2 The proposal is to replace the existing signalling system at the depot with a modern programmable system with a visual display unit replacing the existing shunter’s panel and modern train detection (replacing existing track circuits). The project will also co-locate control of the depot signalling with the Victoria line’s signalling and service control centre.

What is strange is that, at one time, this was regarded as an essential requisite to enable the peak period working of 36tph on the Victoria line to be extended from a claimed 90 minutes (it is actually about 80 minutes) to three hours. The problem is simply one of getting all the trains into service by around 0700.

According to a paper from last June

No further asset changes are necessary to support introduction of the full three hour peak service from June 2018.

It is therefore surprising that no mention has been made of this in the Update on Major Projects – especially as a planned similar extension to peak working on the Jubilee line (due to start on 20th May) is mentioned.

Jubilee line

There are no known significant signalling upgrades due on the Jubilee line. As the order for new trains was cancelled (or ‘paused’ in TfL parlance) there would appear to be no necessity to do so. It would appear that the necessary capacity and flexibility is already available for any changes that may take place. Indeed the new May timetable sees an improved service to Stratford (24tph all day, up from 22.5tph peak and 18tph off-peak) which is largely achieved by introducing stepping-back at Stratford and terminating some trains at the recently relaid and improved turnback at West Hampstead. No signalling changes were involved.

Northern line

The Sub-surface lines, being resignalled under Four Lines Modernisation, may appear to be where all the action is at. A lot is happening on the signalling for the Northern line, however. This is because of signalling that will be installed on the Northern line extension to Battersea for which the contract was recently signed.

The Northern line is currently signalled with the TBTC (Transmission Based Train Control) system made by Thales. It therefore comes as no surprise that Thales will signal the line from Kennington to Battersea Power Station.

Although the extension is not particularly long and, Kennington excluded, fairly straightforward, for various reasons it was decided that it merited a new Vehicle Control Centre. A Vehicle Control Centre is the ‘brains’ behind the system, but it is impractical to have just one for the whole line. Currently the Northern line has eight VCCs and each one controls a specific section of the line.

What is highly relevant to the signalling strategy is that the existing TBTC system is already obsolescent. In simple terms it is generation 2 (2G) whereas the line to Battersea Power Station will be 3G. This does mean there will be the issue of interfacing 2G VCCs with the 3G VCC. But it does also mean that signal engineers will become familiar with 3G on the Northern line and, when the time comes to upgrade one of the existing VCC areas, then there should not be the problem of staff not being familiar with the latest equipment.

With a lot of the cost of TBTC in the software and in the track loops (and associated equipment) throughout the length of the line, replacement of the VCCs over time will probably be cost effective. A planned replacement programme will also probably avoid a situation where the line becomes a hostage to unavailable spare parts.

As part of the signalling work to be done, the Kennington loop will be made bi-directional. This was already being considered but the ‘pausing’ of the additional trains probably made this even more desirable, as the latest plan appears to be to stable just two trains at Battersea overnight. Trains would arrive at platform 1 at Kennington from Battersea then reverse around the loop to platform 2 before reversing again to return to Battersea.

How to go the wrong way round the Kennington Loop. Based on Carto Metro and used with permission

Making the Kennington loop bi-directional makes it possible to run an early morning or late night shuttle on the branch should that be considered desirable. It would also mean that, if the Charing Cross branch were to be closed for any reason, a service could be maintained to Nine Elms and Battersea Power Station.

A further outstanding signalling upgrade would enable 32tph to run between Morden and Kennington. This was originally due to be implemented in December 2014 but was postponed. It was excluded from the work that was recently ‘paused’ and assurances were made that this would still happen. In a sense it is the most important upgrade on the Northern line as it would provide more trains to the most crowded section of the line. It was originally delayed by the need to give priority to signalling modifications to cater for Night Tube. It is not now due until September 2019 (until recently it was April 2019) and one wonders whether signalling of the Battersea Power Station branch and a consequent lack of signalling resources is the cause of the latest delay.

Life extension

Finally we come to the lines that are due to be resignalled under the Deep Tube Upgrade Programme (DTUP). The upgrades for these lines are still a long way in the future. For this we need to look at each line individually. In its simplest form, life extension involves replacing components before they fail. As mentioned, ideally this is not like-for-like but with modern equivalents. You avoid the need for redesign and the upgrade can be piecemeal. What is more, upgrades to legacy signalling systems do not involve alterations to the trains or pushing the limits of technology.

The principle of Life Extension have been applied very successfully by Network Rail. When they discovered that 40% of all delays they were responsible for on Southeastern were down to signalling and track failures between London Bridge and Lewisham, a policy of replacing life-expired equipment along that section reduced delays dramatically. Basically, you change components when the risk of it failing is no longer worth taking even though it probably will work reliably for many years to come.

Piccadilly line

The recent order for new Piccadilly line trains does not include a signalling upgrade and it is made quite clear that the initial tranche of trains is expected to be able to upgrade the frequency from 24tph to 27tph under existing signalling. This is, presumably, achieved by better acceleration rates and reduced dwell time. This means that the existing signalling has to last at least until 2026, and probably a few years beyond.

On the Piccadilly line it is not the signalling itself which is the main problem. The signalling is old, but is of proven design. Also, presumably, once the Four Lines Modernisation is complete, there will be a glut of spares. In many cases, these can be refurbished and effectively be as good as new.

What is a problem on the Piccadilly line is the control system. Anyone who has waited for a train to Rayners Lane or Uxbridge only to see multiple trains to Heathrow will have witnessed this problem first hand. Without a decent overview of the entire line and the necessary tools to enable good decision making, a small problem can rapidly escalation into a substantial deterioration in the service provided. To quote from the paper:

The existing Piccadilly line signalling and control system is vulnerable to failure. The frequency and impact of failure is predicted to increase rapidly as asset degradation and obsolescence are exacerbated by diminishing spares and technical expertise. Assets and building conditions at the service control centre, LU’s oldest serving, date from the 1950s and offer sub-standard working conditions.

Full Programme and Project Authority is already in place for the design and implementation of a new control system. This project, the Piccadilly Line Interim Control Upgrade (PICU) project is now fairly advanced, with fit out of a new service control building taking place and the control system nearing go-live for the first section. The project appraisal delivered a benefits cost ratio of 11.5:1.

The project is well underway. It is an excellent illustration of how London Underground have encountered the worst of all worlds. The aim was for an earlier line upgrade, which should have happened by now. When that didn’t happen the decision was eventually taken to replace the service control side of things, despite the fact that now a lot of the work may see only a few years use. A side-effect may be that it will be more difficult to justify a further replacement in the near future as the benefits will not be as great.

Critical dates for PICU

Bakerloo line

The Bakerloo line only gets a brief mention in the Programme and Investment Committee paper:

Replacement of obsolete and high risk equipment including transferring software to an up-to-date programming language and replacement of depot control panel.

The reference is surprising because it would appear to refer to the ancient control system which, as a post on District Dave pointed out, has been put out to tender for maintenance. It could be that the tender was just London Underground examining all possibilities or, maybe, the lack of positive response means that an alternative is being considered.

This lack of planned signalling updates until future replacement is understandable. North of Queen’s Park, Network Rail is currently responsible for the assets except at the depot. South of Queen’s Park there is not much complexity involved. There is a small depot at London Road near Elephant & Castle for which there is a crossover north of Lambeth north and a single line junction on the northbound running line. There is also a trailing crossover at Piccadilly Circus which is not in normal use. Apart from that it is a simple stretch of plain line between Queen’s Park and Elephant & Castle in a warm and dry tunnel. Nor is the service level of 22tph especially challenging to run with the current signalling.

Central line

The Central line has its own ATO system. While it appears to work well, no-one would design such a system today. For instance, the Central line is the last line to have a train describer system installed to keep track of the trains. Whilst it does interface to the signalling and is itself digital and does appear to be reliable, it is a completely unnecessary layer of complexity. Nowadays, the train describer functionality is built into the command functionality of the signalling system.

Rather like the trains on the Central line, the signalling seems to have been installed at a time when technology was changing rapidly. As the paper states:

The existing Central line signalling and signalling control system was installed in the late 1990s and many sub-assets and components are now obsolete. This is mainly due to the age of the electronic components. Repairs to existing components are becoming increasingly difficult due to the availability of compatible spares and the assets present an increasing risk of major service affecting failures.

The plan is to extend the life of the current equipment. Tellingly, the report continues:

Siemens Rail Automation (the Original Equipment Manufacturer) will carry out the works, on the basis that they are the only supplier who can modify and assure the system the[y] installed.

A completion date of 2025 is indicated.

Waterloo & City line

Last and least is the Waterloo & City line. In the Deep Tube Upgrade Programme it is currently in fourth and last place, having at various time been considered for positions one, two and three. The line has always had the problem that it is harder to make a case for the Deep Tube Upgrade. The passenger flow is almost entirely one way in peak hours and very light outside of them, and to do the job properly would involve a lot of costly work to the underground (small ‘u’) depot.

In the paper on signalling, the Waterloo & City gets just one mention and that is only that the signalling will be replaced as part of DTUP. This is understandable. The signalling was replaced (on a like-for-like basis) when the line was upgraded a few years ago. Nevertheless, it is slightly concerning that by the end of 2018 the two-station Waterloo & City line is expected to have had almost as many signal failures from the 1st January 2015 (73) as the sixteen-station high-frequency Victoria line (97).

Signal failures by line and year

The challenge

It is clear that LU sees the massive benefit in modern signalling and what it can achieve. It is also clear that they have been quietly working towards their own ‘digital railway’ on the ground for some time. Yet, equally clearly, they are aware it is expensive to do so. For TfL it will be challenging to fund future schemes with finances the way they are. The big issues for the organisation are how to get value for money on the lines that are not automated, but will be one day, and how to continue to get money and ensure long-term reliability on the lines that already are automated.

Thanks to Jonathan Roberts and ngh for their assistance.

Written by Pedantic of Purley