How can ETCS (or other forms of automatic train protection) help to prevent overspeed incidents?

[Belperpete]

Moderator note: Split from

Overspeeding at Cambridge Junction (Hitchin) -- 11/12/25

Saw this earlier: https://www.gov.uk/government/news/overspeeding-at-cambridge-junction At 10:39 on 11 December 2025, train reporting number 1T20, the 10:12 London Kings Cross to Ely service, operated by Great Northern (a trading name of Govia Thameslink Railway Ltd), was involved in an...

www.railforums.co.uk

Having the train pass the signal to single yellow does create a hazard that the driver is accustomed to seeing it and then finding the next signal showing a proceed aspect.

A root cause of this and similar incidents is the vastly better acceleration of modern trains compared with those that existed when this signalling was designed.

At one time, the controls for approach released from red signals would have been designed so that the junction signal would be held at yellow until the train had passed the AWS magnet.

However, that was felt to be too restrictive, as in many cases it slowed trains much more than necessary. And, as you say, it also led to drivers anticipating the signal clearing up as soon as the train passed the AWS magnet.

The basis of modern junction signalling is to give information about what route the train is taking, and that it is up to the driver to then control the train to the appropriate speed for that route.

Trying to get a conventional signalling system to control train speed with a wide variety of train characteristics is unrealistic, it is either going to be over-restrictive for some trains or not restrictive enough. It needs something more sophisticated, like ETCS.

 

[MarkyT]

At one time, the controls for approach released from red signals would have been designed so that the junction signal would be held at yellow until the train had passed the AWS magnet. However, that was felt to be too restrictive, as in many cases it slowed trains much more than necessary. And as you say, it also led to drivers anticipating the signal clearing up as soon as the train passed the AWS magnet.

The basis of modern junction signalling is to give information about what route the train is taking, and that it is up to the driver to then control the train to the appropriate speed for that route. Trying to get a conventional signalling system to control train speed with a wide variety of train characteristics is unrealistic, it is either going to be over-restrictive for some trains or not restrictive enough. It needs something more sophisticated, like ETCS.

Traditional speed-based aspect systems, as used in mainland Europe and the USA, often restrict speed more severely than the UK's route-based paradigm.

In many systems, if you get a 'proceed 25' aspect you must comply from the signal all the way to and through the turnout(s), even if half a km away, whereas the route-based paradigm can release at the full readability point of the route indication given, and the train can follow a more natural deceleration envelope down to the restriction starting exactly at the switch.

Note junction indicators are preferred in many locations over alphanumeric route indicators, because they can be clearly read easily over a greater distance, up to ~half a mile away, assuming no obstacles are in the way. Hence, the signal can often clear earlier, and the train can maintain a higher speed for further before reaching the divergence.

On the other hand, some existing UK junction arrangements can also impose overly restrictive approach release conditions, sometimes due to poor approach visibility, or the ideal route indicator can't be provided, if there were no further JI positions available for turning right for example.

These cases can suffer from the long slow crawl to the distant switch phenomenon, but with no explicit limit until the pointwork, there is also the risk of the train speeding up again significantly before a slow turnout is reached, as occurred at Peterborough.

From the 1970s, early ATP systems, such as Ebicab, helped some pioneering European railways overcome this limitation by allowing a deferred speed restriction for a particular turnout route to be transmitted to a train at a preceding signal beacon, from which a precise smooth braking curve could be generated to the turnout.

ETCS provides similar functionality that should assist greatly in optimising junction capacity and performance on complex, busy infrastructure, while fully supervising movement speed, and replacing complex aspect sequences and sometimes confusing route indications with clear, standardised cab indications.

 

[zwk500]

This location could be quite unique in that the same protecting signal offers diverging routes both left and right, but both go to the same destination; in this case, the Cambridge Branch.

The real issue could be driver confusion of turnout speed, dependent on if diverging left or right. Which may have been the issue with this overspeed, so flashing aspects to diverge to the left might be the solution...

At this specific location, the best fix is already in progress - ETCS. The question is whether there are any other locations that could benefit from the issues raised here.

 

[ChilliSauce]

At this specific location, the best fix is already in progress - ETCS. The question is whether there are any other locations that could benefit from the issues raised here.

As I understand the ETCS level 2 project on ECML for the initial roll out is an overlay on the existing Colour light signalling ...The lineside signals will be withdrawn later, at which point all locomotives must be Level 2 equipped. Until then non Level 2 equipped trains will still be able to operate on the route.

So at day 1 not quite the same as the Northern City Line to Moorgate where the Colour Light Signals and the Train Stop Protection has all been removed.

The interesting question has to be when a Train on Level 2 has the route set what aspect will the still in place signals display? If the same model is followed as per LZB in Germany when a train running on LZB and is in section the traditional colour light signal goes dark so as not to give a misleading indication to the driver. Once the train has passed the LZB block the signal then displays Red. Thus a none LZB fitted train can run in-between LZB fitted

 

[Belperpete]

The interesting question has to be when a Train on Level 2 has the route set what aspect will the still in place signals display? If the same model is followed as per LZB in Germany when a train running on LZB and is in section the traditional colour light signal goes dark so as not to give a misleading indication to the driver. Once the train has passed the LZB block the signal then displays Red. Thus a none LZB fitted train can run in-between LZB fitted

My understanding is that in the UK, just extinguishing a signal is unlikely to be acceptable if the same driver is also required to run over conventional signalling where an extinguished signal has to be treated as stop. Having a driver routinely ignoring extinguished signals could lead to him ignoring one when he shouldn't.

It is practice to override approach release controls and the like for ETCS trains, as the ETCS will regulate the train speed.

 

[zwk500]

As I understand the ETCS level 2 project on ECML for the initial roll out is an overlay on the existing Colour light signalling ...The lineside signals will be withdrawn later, at which point all locomotives must be Level 2 equipped. Until then non Level 2 equipped trains will still be able to operate on the route.

AIUI yes that is the plan, with a transition period while driver training takes place. However if a train is fitted and the driver is trained it will be using ETCS. But the point is that the solution is in the process of being fitted, even if it will take time to go into full operation.

So at day 1 not quite the same as the Northern City Line to Moorgate where the Colour Light Signals and the Train Stop Protection has all been removed.

the NCL had the benefit of being able to do overnight testing and training, but even then IIRC it did have a period where it was using ETCS overlaid on the conventional signalling for service trains, before the signals were completely removed.

The interesting question has to be when a Train on Level 2 has the route set what aspect will the still in place signals display? If the same model is followed as per LZB in Germany when a train running on LZB and is in section the traditional colour light signal goes dark so as not to give a misleading indication to the driver. Once the train has passed the LZB block the signal then displays Red. Thus a none LZB fitted train can run in-between LZB fitted

AIUI from a previous post about the GWML overlay, if the interlocking detects an ETCS MA is issued it can step up the aspects immediately. And on Thameslink I think it's also been reported that the signal will clear to Yellow if the ETCS MA is issued, even to only the intermediate block marker.

As another point of comparison, I think London Underground display a blue aspect for cab-signalled trains operating on sections shared with non-cab signalled trains.

 

[MarkyT]

AIUI from a previous post about the GWML overlay, if the interlocking detects an ETCS MA is issued it can step up the aspects immediately. And on Thameslink I think it's also been reported that the signal will clear to Yellow if the ETCS MA is issued, even to only the intermediate block marker.

On TL, I recall the project considered using a distinct new aspect, but settled on yellow as the 'minimum proceed'. On the ECML and GWML overlays, will the proceed aspects also be limited to yellow as well as stepping up junction approach release immediately, when the interlocking is satisfied an appropriate MA has been issued for the set route?

As another point of comparison, I think London Underground display a blue aspect for cab-signalled trains operating on sections shared with non-cab signalled trains.

Even fully cab-signalled CBTC lines often have limited fixed short range signals around critical positions such as platforms with screen doors and junctions. These can be useful for failback if radio MA systems are disabled for example, as they locally check points are in the correct position and platform doors have closed. Such signals don't need to be placed for optimum long range visibility, important in traditional installations, as approaching them is safely controlled on board and indicated on cab screens.

 

[ChilliSauce]

It is practice to override approach release controls and the like for ETCS trains, as the ETCS will regulate the train speed.

I think you would expect in ETCS level 2 mode for AWS and TPWS to get suppressed ... Similar to LZB in that when in LZB mode Indusi is suppressed. Remember DB-AG have been at this in-cab signalling since the mid 1960s .. ETCS level 2 is basically the same except rather than exchange Data through an inductive loop laid between the track it's Data messages sent over GSM-R

 

[Belperpete]

On TL, I recall the project considered using a distinct new aspect, but settled on yellow as the 'minimum proceed'. On the ECML and GWML overlays, will the proceed aspects also be limited to yellow as well as stepping up junction approach release immediately, when the interlocking is satisfied an appropriate MA has been issued for the set route?

The GWML overlay is basically just overlaying ETCS on the existing signalling system. Unlike Thameslink which I think was designed as bimode signalling from the start. So as I understand it, on GWML the conventional signalling will display it's usual aspects, and ETCS trains will only get authorities from signal to signal (no mid section authorities). Approach release controls will be bypassed for ETCS trains. I think there are also differences for the approach locking for ETCS trains, but my memory is a bit fuzzy about what.

Can't help about the ECML overlay.

== Doublepost prevention - post automatically merged: 18 Jan 2026 ==

I think you would expect in ETCS level 2 mode for AWS and TPWS to get suppressed ...

They are, but that is done on board the train, not by the interlocking or lineside equipment. I think there may be a time during the transition from one signalling system to the other when both systems might be active.

== Doublepost prevention - post automatically merged: 18 Jan 2026 ==

My understanding is that in the UK, just extinguishing a signal is unlikely to be acceptable if the same driver is also required to run over conventional signalling where an extinguished signal has to be treated as stop. Having a driver routinely ignoring extinguished signals could lead to him ignoring one when he shouldn't.

The other issue raised with extinguishing signals is the impact on multi track lines. If you turn off one signal on a gantry, for example, that could lead to drivers on other lines misreading the other signals on the gantry. Even on a double track line that is bidirectionally signalled, if you turn off one of each pair of signals, leaving just the red aspect on the other line, then you will effectively be conditioning ETCS drivers to ignore red aspects.

As an aside, there is an interesting article on approach lit signals in a recent SRS journal, which was a method of avoiding drivers seeing a red aspect ahead on a parallel line, where there is no corresponding signal on the line he is on.

I recall that when ETCS was being considered for the WCML, we got a heads up from HMRI that they were minded against signals with extinguished aspects.

 

[MarkyT]

The GWML overlay is basically just overlaying ETCS on the existing signalling system. Unlike Thameslink which I think was designed as bimode signalling from the start. So as I understand it, on GWML the conventional signalling will display it's usual aspects, and ETCS trains will only get authorities from signal to signal (no mid section authorities). Approach release controls will be bypassed for ETCS trains.

That all makes sense. Anything that relies on long-range aspect and indicator visibility can be relaxed as much better information will be available for the route ahead on the cab display.

I think there are also differences for the approach locking for ETCS trains, but my memory is a bit fuzzy about what.

Perhaps an alternative release if the train can explicitly tell the interlocking it's at a stand and hasn't been given a MA, rather than inferring that where possible from sequences and times of track circuit occupancy?

Can't help about the ECML overlay.

== Doublepost prevention - post automatically merged: 18 Jan 2026 ==


They are, but that is done on board the train, not by the interlocking or lineside equipment. I think there may be a time during the transition from one signalling system to the other when both systems might be active.

On the Siemens TL and GN trains, I understand the AWS/TPWS is fully emulated by the onboard ETCS computers and DMI cab screens. I guess the system ignores the legacy sensor inputs when in full ETCS mode. You definitely wouldn't want to have to suppress all the permanent AWS inductors. In ETCS, a train shouldn't receive any conflicting TPWS indications, as the ATP should prevent any overspeed or SPAD.

Edit: I know there are release speeds defined in ETCS based on length of overlap available, so a SPAD is still possible, but stopping within the locked overlap is guaranteed, and the ETCS would be intervening with an emergency brake application, so there's no conflict with the TPWS trainstop.

 

[Belperpete]

Perhaps an alternative release if the train can explicitly tell the interlocking it's at a stand and hasn't been given a MA, rather than inferring that where possible from sequences and times of track circuit occupancy?

They may be doing that on the ECML, which I seem to recall has more complex interfaces than the GWML. I think on GWML they just ensure that any approach locking timeout allows for any potential ETCS delays in revoking a movement authority.

The issue I was thinking about was where you release aspects early for ETCS trains. Routes with approach released aspects may have limited approach locking controls. This can cause a problem if you remove the approach release. What you don't want to happen is, if the signaller pulls the button, for the approach locking to say "I see no approaching train" and allow the route to be cancelled in the face of an approaching train that has an ETCS movement authority to pass the signal. Likewise an approach locking time based on a route being approach released may not be sufficient if you remove that approach release.

 

[zwk500]

They may be doing that on the ECML, which I seem to recall has more complex interfaces than the GWML. I think on GWML they just ensure that any approach locking timeout allows for any potential ETCS delays in revoking a movement authority.

The issue I was thinking about was where you release aspects early for ETCS trains. Routes with approach released aspects may have limited approach locking controls. This can cause a problem if you remove the approach release. What you don't want to happen is, if the signaller pulls the button, for the approach locking to say "I see no approaching train" and allow the route to be cancelled in the face of an approaching train that has an ETCS movement authority to pass the signal. Likewise an approach locking time based on a route being approach released may not be sufficient if you remove that approach release.

AIUI, If an ETCS movement authority has been issued, the train's onboard EVC will only allow it to be shortened and the route released if the braking curve can be recalculated for a safe stop without activating the emergency brake.

(Caveat that I'm not a signalling engineer so some of this may not be 100% accurate).

 

[Annetts key]

It should also be noted that at some junctions (or diverging routes over crossovers), the time delay has been reassessed since the signalling was originally installed. As a result, the time delay has been removed or reduced making the release less restrictive.

 

[D365]

On the Siemens TL and GN trains, I understand the AWS/TPWS is fully emulated by the onboard ETCS computers and DMI cab screens. I guess the system ignores the legacy sensor inputs when in full ETCS mode. You definitely wouldn't want to have to suppress all the permanent AWS inductors. In ETCS, a train shouldn't receive any conflicting TPWS indications, as the ATP should prevent any overspeed or SPAD.

This is not quite correct - TPWS and AWS functions are not emulated by the EVC.

As per the later Class 350 subfleets (/3, /4), each Class 700 driving vehicle uses a dedicated Unipart "TPWSfour Control Unit", although on Class 700 this control unit interfaces with the ETCS displays to provide a virtual DMI. Instead of the dedicated TPWS DMI on Class 350.

I'm not sure whether the same applies to Class 7xx, but the TPWSfour implementations that I am familiar with receive 'discrete' digital inputs from the EVC in order to suppress AWS/TPWS when the train is protected by ETCS.

 

[ChilliSauce]

Traditional speed-based aspect systems, as used in mainland Europe and the USA, often restrict speed more severely than the UK's route-based paradigm.

In many systems, if you get a 'proceed 25' aspect you must comply from the signal all the way to and through the turnout(s), even if half a km away, whereas the route-based paradigm can release at the full readability point of the route indication given, and the train can follow a more natural deceleration envelope down to the restriction starting exactly at the switch.

Note junction indicators are preferred in many locations over alphanumeric route indicators, because they can be clearly read easily over a greater distance, up to ~half a mile away, assuming no obstacles are in the way. Hence, the signal can often clear earlier, and the train can maintain a higher speed for further before reaching the divergence.

On the other hand, some existing UK junction arrangements can also impose overly restrictive approach release conditions, sometimes due to poor approach visibility, or the ideal route indicator can't be provided, if there were no further JI positions available for turning right for example.

These cases can suffer from the long slow crawl to the distant switch phenomenon, but with no explicit limit until the pointwork, there is also the risk of the train speeding up again significantly before a slow turnout is reached, as occurred at Peterborough.

From the 1970s, early ATP systems, such as Ebicab, helped some pioneering European railways overcome this limitation by allowing a deferred speed restriction for a particular turnout route to be transmitted to a train at a preceding signal beacon, from which a precise smooth braking curve could be generated to the turnout.

ETCS provides similar functionality that should assist greatly in optimising junction capacity and performance on complex, busy infrastructure, while fully supervising movement speed, and replacing complex aspect sequences and sometimes confusing route indications with clear, standardised cab indications.

Just to remind a number of countries in Europe use Speed signalling unlike the UK that has Route signalling .... ETCS 2 and it's predecessor the German in-cab LZB apply speed signalling principles.

A good example of Colour Light Speed Signalling could be that deployed by the SBB with the now legacy "L" signals. In that, in complex areas the signal has both a "Home Signal"...displaying the speed to pass this signal as in 3 Greens = 90km/Hr, 2 Greens = 60, Green & Yellow = 40, 2 x Yellow = 40 and expect short stopping distance. And Single Green = Line Speed

And the "Distant Signal" displaying the speed to pass "The next signal"

So none of this approach release type stuff is needed.

As example if the junction turn out is 60km/Hr and the route is set the signalling protecting that junction would be 2 x Green (pass at 60) and the Distance signal at the same point could very well be displaying pass next signal at line speed (also 2 x Green) .. So Indicating when the complete train length is clear of the junction "Speed up" to Line Speed ..whatever that is.

The signal prior to the signal protecting the junction would in the Distant aspects be displaying 2 Greens and a Yellow with the Yellow in the top LH corner. Indicating pass next signal at 60

Clearly much more complicated that Route Signalling but can deliver so much more .

 

[godfreycomplex]

AIUI, If an ETCS movement authority has been issued, the train's onboard EVC will only allow it to be shortened and the route released if the braking curve can be recalculated for a safe stop without activating the emergency brake.

(Caveat that I'm not a signalling engineer so some of this may not be 100% accurate).

If the distance to the End of Authority is too short for the braking curve to be adjusted with a full movement authority, the ETCS will put the train into trip mode, and the emergency brakes will be applied.

 

[zwk500]

If the distance to the End of Authority is too short for the braking curve to be adjusted with a full movement authority, the ETCS will put the train into trip mode, and the emergency brakes will be applied.

Does trip mode hold the route locked?

 

[godfreycomplex]

Does trip mode hold the route locked?

Not as such, the conventional interlocking holds it locked until time out if the train is close enough to the EoA that’s been cancelled. Route time out being connected with modes is level 3 only

 

[bahnause]

Traditional speed-based aspect systems, as used in mainland Europe and the USA, often restrict speed more severely than the UK's route-based paradigm.

In many systems, if you get a 'proceed 25' aspect you must comply from the signal all the way to and through the turnout(s), even if half a km away, whereas the route-based paradigm can release at the full readability point of the route indication given, and the train can follow a more natural deceleration envelope down to the restriction starting exactly at the switch.

This is not necessarily the case, as not every restriction applies until the next signal. Additionally, the system can be optimised for maximum speed or capacity during the planning stage by selecting the locations of the signals accordingly.

ETCS provides similar functionality that should assist greatly in optimising junction capacity and performance on complex, busy infrastructure, while fully supervising movement speed, and replacing complex aspect sequences and sometimes confusing route indications with clear, standardised cab indications.

However, the switch from highly efficient external signalling to ETCS can result in significant capacity losses, regardless of the level of ETCS selected. The SBB network currently demonstrates the need for significant optimisation by removing some safety margins from the ETCS system to allow for a similar speed profile and similar route capacity that is achieved today. Of course, this optimisation process takes years and consumes planning resources. And we're not even talking about the compatibility issues that are now cropping up, even though the train equipment and track equipment comply with the ‘standard’.

 

[MarkyT]

This is not necessarily the case, as not every restriction applies until the next signal. Additionally, the system can be optimised for maximum speed or capacity during the planning stage by selecting the locations of the signals accordingly.

I expect that often results in extra signals in complex layouts so the distance from signal to slow divergence is reduced. UK practice has difficulty doing this. Its possible to go from 3 to 4 aspect, but it's frowned on to switch back and forth between 3 asp and 4asp frequently. Hence, the old trick on secondary lines of introducing an extra signal for closing up at the station with the outermost distant signal being able to show a first caution of single yellow for the first stop signal at red or a double yellow for the next stop signal is no longer permitted. Complex approach released sequences and repeated yellows are also considered to have anticipation hazards so must be risk assessed. With speed-based aspects, an underbraked approach can be controlled by lower speed aspects to rear.

However, the switch from highly efficient external signalling to ETCS can result in significant capacity losses, regardless of the level of ETCS selected. The SBB network currently demonstrates the need for significant optimisation by removing some safety margins from the ETCS system to allow for a similar speed profile and similar route capacity that is achieved today. Of course, this optimisation process takes years and consumes planning resources. And we're not even talking about the compatibility issues that are now cropping up, even though the train equipment and track equipment comply with the ‘standard’.home

You've written about your experience with the switch from Euro Signum/Zub to L1 LS. SBB seems to have suffered somewhat from being the earliest adopter of this. In retrospect, I think UK did the right thing in waiting until ETCS was more mature before rolling it out widely.

 

[zwk500]

I expect that often results in extra signals in complex layouts so the distance from signal to slow divergence is reduced. UK practice has difficulty doing this. Its possible to go from 3 to 4 aspect, but it's frowned on to switch back and forth between 3 asp and 4asp frequently. Hence, the old trick on secondary lines of introducing an extra signal for closing up at the station with the outermost distant signal being able to show a first caution of single yellow for the first stop signal at red or a double yellow for the next stop signal is no longer permitted. Complex approach released sequences and repeated yellows are also considered to have anticipation hazards so must be risk assessed. With speed-based aspects, an underbraked approach can be controlled by lower speed aspects to rear.

Of course, with ETCS L2 there is no aspect sequence to need to worry about.

 

[bahnause]

I expect that often results in extra signals in complex layouts so the distance from signal to slow divergence is reduced. UK practice has difficulty doing this. Its possible to go from 3 to 4 aspect, but it's frowned on to switch back and forth between 3 asp and 4asp frequently. Hence, the old trick on secondary lines of introducing an extra signal for closing up at the station with the outermost distant signal being able to show a first caution of single yellow for the first stop signal at red or a double yellow for the next stop signal is no longer permitted. Complex approach released sequences and repeated yellows are also considered to have anticipation hazards so must be risk assessed. With speed-based aspects, an underbraked approach can be controlled by lower speed aspects to rear.

Not only to the rear. Example: Optimisation of train speeds when approaching a conflict point. Something that is now mostly transferred to the train driver's iPad. The system optimises the speed of trains leaving Zürich at almost the same time by using the available speed limits of the lineside signals to slow the trains below the max line speed. This keeps the second train from approaching the conflict point where the tracks converge at full speed and then having to brake or stop.

You've written about your experience with the switch from Euro Signum/Zub to L1 LS. SBB seems to have suffered somewhat from being the earliest adopter of this. In retrospect, I think UK did the right thing in waiting until ETCS was more mature before rolling it out widely.

This does not only affect Level 1LS; the capacity losses and journey time increases are also significant when switching to L2. The additional complexity of the overall system is also becoming increasingly worrying. There are increasing undesirable side effects when specifications are adjusted. And since these are valid throughout Europe and are also safety-relevant, adjustments usually take years and are extremely time-consuming and expensive. Apart from interoperability, ETCS is now pretty much the opposite of the desired cost saver.

In addition, there are numerous problems in the area of the user interface due to the different interpretations of the specifications and designs by OBU (On Board Unit) manufacturers. It has become apparent that human factors have been given far too little consideration.

It might work better if the number of manufacturers and vehicles were manageable. However, in such an open system with countless stakeholders in transalpine transport, the number of variables is very high. If stickers indicating the manufacturer and software version of the ETCS are already required on the vehicle, then it is not as standardised as one would like.

 

[HSTEd]

My, admittedly limited, understanding is that train protection systems using intermittent communication with the train, for example balises, reduce capacity because trains cannot be updated when signals become less restrictive until they pass a balise.

If we were to use ETCS Level 1LS to prevent overspeeds on diverging routes, surely the speed limit would only become less restrictive on approach if the train arrives with the wrong route set?
Is that a common problem?

(EDIT: I think a move to ETCS would likely require the replacement of TPWS etc since I doubt it likes sharing control)

Beyond that, from reading the spec I think you could implement de-facto TPWS using Level 1LS, as TPWS is just a speed check. You could have an arbitrarily short speed limit commencing effectively immediately at the balise, if the train is moving too fast the emergency brake will trip. That would also reduce the impact of the signal becoming less restrictive because the train would only be in the restrictive zone for its own length, as you'd select the speed limit to be over the shortest possible length.

So, would replacing TPWS (and AWS if you can make it actuate some sort of warning?) with ETCS L1LS equipment, and then implementing diverging route speed protection, actually reduce capacity? Does the capacity only become an issue if you try to implement 'proper' train protection?

 

[Annetts key]

My, admittedly limited, understanding is that train protection systems using intermittent communication with the train, for example balises, reduce capacity because trains cannot be updated when signals become less restrictive until they pass a balise.

If we were to use ETCS Level 1LS to prevent overspeeds on diverging routes, surely the speed limit would only become less restrictive on approach if the train arrives with the wrong route set?
Is that a common problem?

Beyond that, from reading the spec I think you could implement de-facto TPWS using Level 1LS, as TPWS is just a speed check. You could have an arbitrarily short speed limit commencing effectively immediately at the balise, if the train is moving too fast the emergency brake will trip. That would also reduce the impact of the signal becoming less restrictive because the train would only be in the restrictive zone for its own length, as you'd select the speed limit to be over the shortest possible length.

So, would replacing TPWS (and AWS if you can make it actuate some sort of warning?) with ETCS L1LS equipment, and then implementing diverging route speed protection, actually reduce capacity? Does the capacity only become an issue if you try to implement 'proper' train protection?

Yes, systems using intermittent communication with the train limit the capacity - the train has to get to the next balise (or equivalent) before the on-board computer on the train gets updated information. However, where this occurs regularly, this can be improved by having an "in-fill" balise (or equivalent) so that the on-board computer gets updated sooner.

I'm not an expert on ETCS. But I do know about GWML ATP. This takes signal aspect information from the existing signalling system. But it also has fixed data. This includes details of all the speed restriction information for the line ahead for the next section. Including diverging junction speed limits and permanent speed restrictions. The speed restriction data includes the distance until the start of the restriction and the length of the restriction. The train knows where it is, as it knows the speed of the train and counts the wheel rotations from the last signal beacon. Hence can work out distance.

So with this system, you don't really need approach controlled/released junction signals from a safety point of view. You only need approach control/release so that the train is going at a slow enough speed that the driver can see the junction or route indicator. But on modern signalling schemes, there are additional route indicators on approach to the junction signal anyway (although only normally provided for some routes/junctions).

GW ATP, when combined with in-fill beacons (it's equivalent to balises), or with in-fill loops (cables that can be up to 800 metres long), the on-board computer can be updated as soon as the train passes over the in-fill beacon (which are carefully positioned). Or within seconds if the train is over an in-fill loop. Hence when the signal ahead changes aspect, the driver does not have to wait until the train gets to the signal before they can apply power.

If you combine all these systems and features (GW ATP with in-fill beacons or in-fill loops and additional route indicators on approach), there would actually be no loss of capacity and it may even slightly improve capacity compared to a traditional British colour light system for junctions.

Trains with working GWML ATP don't really need TPWS. But still have it because they work over non-ATP fitted lines.