News:

Simutrans Chat Room
Where cool people of Simutrans can meet up.

Discussion of new signalling system

Started by Octavius, October 26, 2015, 09:37:24 PM

Previous topic - Next topic

0 Members and 2 Guests are viewing this topic.

Octavius

Mod note: This post has been split from this topic as the discussion has strayed into substantive issues of signalling rather than the videos themselves.

Quote from: jamespetts on August 10, 2015, 07:59:20 PM
Whenever a train stops at a station or a reversing waypoint, it reverts automatically to drive by sight until it passes a signal: this is why there usually need to be signals at the end of a platform.
I agree with the reversing, but I think a train should not revert to drive by sight automatically at a station. This would still necessitate the special long block signals to protect a single track section with intermediate halts. And on modern double track lines (at least over here) there are only automatic block signals at regular intervals, ignoring the positions of the halts (defined as small stations without any points). A signal may as well be placed just after a platform as just before it, although usually not on top of the platform. So I'd say that trains should revert to drive by sight only when reversing, leaving a depot or passing a drive by sight sign or a signal showing a drive by sight aspect (also known as call on). In reality they also revert to drive by sight when changing driver.
Quote from: jamespetts on August 13, 2015, 10:11:26 AM
As to the number of tiles reserved, I think that this is probably accurate, especially since t is not really practical for a train/tram to keep speeding up/slowing down depending on whether there are any bends. In reality, it would be hills, undulations, slight corners not simulated in Simutrans combined with lineside obstructions, etc. that would also interfere with sighting, so a sighting distance of over 1km will not be realistic. However, I do plan for the sighting distance to be set in the pakset eventually.
Furthermore, in reality there is also weather. Timetables have to be designed with a speed and therefore visibility in mind that can be expected at least 95% of the time. There may be mist, heavy rain, snow etc. If you design a timetable for trains running on sight, you cannot assume they will be able to look more than about 200 m ahead, even when most of the time they can. So the train will be slow.

jamespetts

The reason that trains generally revert to drive by sight in stations is that, in most working methods, trains only reserve up to the next stop and not beyond, which would be necessary were they to remain in their current working method after starting from a station. However, in token block and one train staff working method, trains do remain in their current working method at stops, as the line is in any event reserved up to the next signal and beyond stations.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

I must agree with octavius in regard that trains should not revert to drive-by-sight on stops. Here is a bit longer explanation why:

I had researched a bit about signalling on Czechoslovak railways.
- Drive by sight is used only for shunting (leaving depot), trams, permissive and call-on signals.
- One-train staff is probably not used at all. Maybe some very early or museum railways do/did use it. It was in consideration for single tracked tram lines (like Kirnitzstalbahn in Bad Schandau), but a more complex and flexible circuit block system is in use. It is used as a temporary measure during repairs of otherwise double tracked tram lines.
- Timetable based - single tracked trams (replaced with circuit block after fatal accidents), some less used local railways, most narrow-gauge railways. Sometimes used in addition with direct traffic control. https://en.wikipedia.org/wiki/Direct_traffic_control
- Token block - (with telegraph based token machine with multiple tokens) - probably never used at all. However the simulated behaviour could be the same as for direct traffic control. Just the token is not physical, but verbal.
- Absolute block - still used in multiple variations - early with only telegraph / telephone verbal communication between signalboxes and visual check of tracks, points and trains. Later improved by half automatic equipment reducing the potential of human error. Used with semaphore signals as well as colour light signals.
- Circuit block - fully automatic permissive signals (on track between stations), improved safety on stations, used on main double tracked lines, more expensive equipment than absolute block, but less man-power (and signalboxes) needed.
- in cab signalling - used alongside to circuit block, mostly for comfort and safety - better visibility in fog, automatic braking if driver falls asleep, etc.

The important thing is that there are many stops/halts that have no sidings, no signals and no personel. Cargo trains just pass through them, and passenger trains continue at full speed after stopping there. The track is reserved through the stop(s) up to the next signal. Also the block is freed only when the train is completely out of the block. Thus there is no need for tokens at single tracked railways. So I think the reservation system for absolute and circuit block should be changed in this way:

At any time the train should have reserved track up to next signal facing in BOTH directions. Any signal counts except for end-of-choose and minimum speed signs. Single-way signal and drive-by-sight signal counts as end of block. Track is cleared when the train passes a signal facing in opposite direction. Track is reserved when the train approaches next signal (facing in the right direction), or its distant (pre-) signal. An attempt is made to reserve aspect-1 blocks ahead, unless there is a scheduled stop closer than that. For in-cab or moving block signalling a reservation equal to braking distance is attempted. The track reservation always ends at signal, even if there is a scheduled stop (or more stops) before the signal. This removes the need for token block.

Further, a directional (blue) reservation is made beyond the next stop up to the next junction (the signal in front of it), choose signal, or signal of other signalling system. Directional reservation is restored when the train leaves the block. Directional reservation can be only on plain blocks that do not contain any junctions. Directional reservation can be removed (reversed), only if the whole track to the next junction is free. This applies for both absolute block and track circuit block. This allows to split single track in multiple blocks, allowing for more trains in the same direction.

Czechoslovak railways distinguish between signalboxes at stations (with sidings) and those along the line, just dividing the line in blocks. I'm not sure if that distinction is worth implementing. E.g. the directional reservation would then be made from one "station" signalbox to another, ignoring the "block" signalboxes on the way. Also choose signals would be allowed only with station signalbox. This could allow for some simple junctions or sidigns without passing loops on single track. But it might be overcomplicating things.

The use of single-way signs would change a little. They should be placed at the end of the track (instead of entry) that is to be used only in one direction, so that they can act as end of block markers.

I'll prepare some screenshots to illustrate if necessary.


jamespetts

That is an interesting overview of Czech signalling - thank you. Direct traffic control can probably (I have only looked briefly at the Wikipedia article) be simulated using the token block working method in Experimental; I do this with Radio Electronic Token Block signalling in Pak128.Britain-Ex, which system also has no physical tokens and seems quite similar in principle (at least in general terms) to Direct Traffic Control.

As to the stations without any signalling, are these single track or double track? If single track, token block signalling can be used, as this is designed to work with unsignalled stations (but incurs additional complexity as a result).

I do not think that it is desirable to use directional reservations with absolute block signalling: this type of signalling is intended to represent signalling technology when the electric telegraph was the only way of communicating a train's presence, and this was not sophisticated enough to accomplish what amounts to directional reservations. Can two aspect track circuit block signals not be used for this purpose, or are they problematic in some other way? If you would use two aspect track circuit block signals but for the fact that the signals that you are trying to simulate are normal danger rather than normal clear, the much easier answer would be to allow track circuit block signals to be normal danger by a .dat file setting, which would be quite straightforward to implement.

Can you tell me more about how end of block markers work in Czech signalling?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

Quote from: jamespetts on November 09, 2015, 12:04:40 AM
As to the stations without any signalling, are these single track or double track?
Both.

Quote
If single track, token block signalling can be used, as this is designed to work with unsignalled stations (but incurs additional complexity as a result).
I'm not sure if simutrans trains are required to stop or slow down to collect the token. Is it possible to have choose signal with token block method?

Quote
I do not think that it is desirable to use directional reservations with absolute block signalling: this type of signalling is intended to represent signalling technology when the electric telegraph was the only way of communicating a train's presence, and this was not sophisticated enough to accomplish what amounts to directional reservations.
Unfortunately I found only czech documentation, but the system that is closest to absolute block signalling has provision for specifying the direction on single track. The signalman on station (with passing loop), has indication not only whether the next block is clear, but also whether he is allowed to depart trains. If he is not, he has first to ask the other station to give up the right to depart trains. There may be intermediate "block" signalboxes, which report only if the train has passed, know the negotiated direction, but cannot change it.

Quote
Can two aspect track circuit block signals not be used for this purpose, or are they problematic in some other way? If you would use two aspect track circuit block signals but for the fact that the signals that you are trying to simulate are normal danger rather than normal clear, the much easier answer would be to allow track circuit block signals to be normal danger by a .dat file setting, which would be quite straightforward to implement.
Probably yes. Anyway the difference between absolute and circuit block is quite cosmetic - default clear, more signals per signalbox (or even without signalbox), no mechanical semaphores.

Quote
Can you tell me more about how end of block markers work in Czech signalling?
Uh, there is nothing like that. Blocks are delimited by signals. On real railways there are almost no one-way tracks (thus no single-way signs). Even on double tracked lines (except for metro and tram), both tracks can be used in both directions. If you imagine a station, you have a departure signal on both ends of platform (or passing loop), before the switches. Thus the whole block between departure signals can be reserved for the train. In simutrans you often want to make a passing loop with use of single-way signs to avoid deadlocks. Thus you need to put something on one end of passing loop instead of signal. A single-way sign is a logical thing to put there. A train could enter such a block in the wrong direction, but could not leave it.

See the attached screenshots. One is regular station with 3 bidirectional tracks and choose signals at entry. The other has 2 single-directional tracks for and no choose signals. In both cases when the train is in the station the reservation will be between signals (either regular or single-way signs). When the train is on track, whole track will be reserved - from one choose signal to another, or even up to the passing loop if there's no choose signal.

Junna

Three-track (or greater) stations of that sort were always rare in the UK, where the norm is passing loop with stations on it and a siding (often only accessible from one direction). I assume that most of those stations with several tracks like that were built such during the CSSR times or prior because of high volume of goods traffic necessitated frequent sidings for overtaking, whereas in the UK goods traffic was in free-fall from an early age and besides, a lot of lines were double-track already and goods trains were generally short. Also the general european practice generally favoured bidirectional lines (almost all Swedish lines have some bidirectional capacity), but UK did not quite follow this. It seems hard to signal these effectively though, because trains could get stuck easier since there is no actual despatching.

Vladki

I would say that here would be a three track station considered a bare minimum. Two track passing loops (though they exist) do not deserve the name "station"  :police:   I think most of those stations were built before WWII, when local cargo trains were common - calling at every small town or even village to drop a few cars and pick up other cars. Mixed (pax+cargo) trains were also common on branch lines. Then you needed quite some more tracks not only for passing but also for shunting and loading. Of course times have changed and many stations have been reduced to passing loops or even to simple stops.

I was thinking more about the reservation algorithm - the releasing the reservation could be as it is now (every tile), but I wanted to keep the block reserved up to the next signal behind the train, for the case when train reverses. So at station, it makes sense to reserve up to the next signal in both directions. However on the running line it makes more sense to reserve blocks between two signals facing in the same direction. So perhaps a signal in any direction could be considered. But more important is IMHO to reserve beyond the stop up to the next signal.

Directional reservation on absolute block is perhaps not very important. If the track between stations is split only in two blocks (one intermediate signalbox), and the blocks overlap (or a bidirectional signal is allowed), then it would work fine even without directional reservation. And I think it is very uncommon to have single track divided in more than 2 blocks.

Is it possible to have an absolute block signal that is signal, presignal and choose signal at the same time? like: aspect=3 and freeroute=1. What about choose signals in token block system?

jamespetts

I will have to give some thought to the question of reserving beyond a station stop: it may be that it is an important thing for the purposes of priority reservations (which already has about half of its code: these will be coloured in yellow) that allow some trains to take priority over another by making a special type of reservation ahead of the normal reservation, which can be overridden by a train with a higher priority level (or the same priority level if a block reservation is requested over a priority reservation). The plan is for both the number of blocks ahead and the priority level to be customisable by line/convoy in the schedule. However, I worry a little that this will increase computational load in circumstances where some users are already reporting a higher computational load than before I implemented the signalling changes.

The optional normal danger for track circuit block should be fairly straightforward, and this seems worthwhile if this will assist paksets with realistic signalling for particular countries. Incidentally, from what pakset are the screenshots above taken? Are there any plans for a Czech pakset? It would be a rather splendid thing to have a Czech as well as Swedish pakset for Experimental.

Quote
Anyway the difference between absolute and circuit block is quite cosmetic - default clear, more signals per signalbox (or even without signalbox), no mechanical semaphores.

Actually, apart from normal clear and whether it is possible to have signals without a signalbox, the things listed, including the number of signals per signalbox and whether semaphore signals are permitted are things that can be set in the pakset (or did you mean to distinguish the real life differences? If so, normal clear is not necessarily a feature of track circuit block, although it usually is; but note that normal clear applies only to signals controlling a section of track with no junctions, even in modern times). However, it is perfectly possible to have track circuit block semaphore signals - I have recently implemented some in Pak128.Britain (to simulate the very early automatic signalling from around 1902, before lamps were bright and reliable enough for colour light signalling to work effectively).

Bearing that in mind, was there a particular reason to have a 3-aspect choose signal in absolute block mode, or did you really just want a combined junction semaphore signal? (These have been implemented as track circuit block automatic semaphores in Pak128.Britain-Ex). I have not tested a combined choose signal in absolute block mode, but I cannot immediately think of a reason why having is_presignal=1, is_signal=1, aspects=3 and free_route=1 should not create a working signal that is a combination of a choose signal and a combined signal, provided that the usual combined signal rules (will only show a clear aspect and reserve through if the next signal's signalbox is in range) are obeyed. Because this has not been tested, it is possible that anomalies may arise in operation. If you need to code this type of signal for a Czech or Swedish pakset (or otherwise) and you find anomalies, I will try to fix them if you upload a copy of the pakset and a saved game in which any issues can be reproduced reliably.

I do not think it sensible to have one way signs as end of block markers, as this would be difficult for users to understand and is not intuitive: one would expect that a one way sign would do no more and no less than determine the direction in which convoys may pass them.

As Junna says, in the UK, unidirectional track is very common even now (and was near universal for stretches of double track main line in previous centuries).

Thank you again for your feedback, which is appreciated.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves


Quote from: jamespetts on November 09, 2015, 11:31:30 PM
I do not think it sensible to have one way signs as end of block markers, as this would be difficult for users to understand and is not intuitive: one would expect that a one way sign would do no more and no less than determine the direction in which convoys may pass them.
Would it make sense to create special sign that makes what Vladki suggest? An "observation point" to put on the tracks, when the entire train has passed this "point" the block behind is cleared. Could be used together with signal boxes as there need to be a guy standing there, observing the train and then releasing the signals.

Vladki

James: the screenshot ideed is from pak128.cs (czechoslovak). You can get it from simutrans SVN @SF, or from the forum. It is a standard pakset, but I'd like to add at least some exp features to it.

Czech semaphore signal (on station entry) has these aspects: danger, clear, caution and caution+40 km. The last I wanted to use as choose aspect.

Swedish 3 arm semaphore could be also used in such way although the meaning of its aspects is different.

Ves: observation point must be implicit, otherwise players will be confused.

I think the unreserving of track behind the train could be left as is, if the reservation ahead continues to the next signal beyond station. A reversing train then may revert to drive by sight (shunting)

jamespetts

Ahh, I do have a recollection of a Standard Czechoslovakian pakset - it would be splendid to have an Experimental version of that. I shall look forward to trying it one day.

As to the Czech semaphore signal, this could indeed be simulated by a choose signal in the track circuit block working method; you could either have both the clear (subsidiary) aspect and the caution (subsidiary) aspect having the same "caution + 40km/h" indication as one another, or have the both the clear (main) and clear (subsidiary) indications being the same as one another, and the caution + 40km/h indication being the caution (subsidiary) aspect only. Choose signals are normal danger in any event (as are all signals facing junctions), but plain signals on straight line would be normal clear in the current system, but this would be customisable if I introduce optional normal danger for track circuit block signalling.

Currently, a choose signal (and, indeed, any signal immediately before a train's next destination) will display a clear aspect. However, if I change the behaviour as you suggest and have it reserving beyond the station unless the station is a reversing point, it may well display a caution aspect depending on (1) the implementation (whether it reserves the normal number of blocks ahead or is limited to reserving just up to the signal immediately after the station - not an easy choice when one thinks about all the implications of both in many situations; in reality, this will vary depending on the situation); and (2) whether the section beyond the station is clear.

As to observation point signs, do these have any precedent in reality?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves

I don't think there is explicitly observation points, but in real life there are tools to know wether the train has passed a certain point.
It would be like a tool for the player to simulate logic on the railways.
"When entire train is on track 1, the line behind is considered free and may be booked by another train"

Vladki

#12
That combined semaphore will show caution + 40 even for clear + choose. Thats what it shows in reality. Theoretically it could have clear + 40 but that is not used.

If you decide to touch that code, may I ask that multi aspect signals show caution if the train has reversing stop in the next block?

Czech rail has a marker for track circuit border. But it is used only if there is no other indication (signal) that this is not a good place to leave parked waggons.
No 975 at: http://mujweb.cz/mikulda/d1/htm/jedenac2.htm

I found that czech railways have interesting maps about used signalling systems:
http://provoz.szdc.cz/PORTAL/Show.aspx?path=/Data/Mapy/zabezpecovac.pdf
http://provoz.szdc.cz/PORTAL/Show.aspx?path=/Data/Mapy/TZZ.pdf

I'm still not sure about the main difference between absolute block and circuit block. I just have the feeling that semaphores+curcuit block are nonsense.
Czech rules distinguish 3 levels of security (but signals are mostly the same):
1 - everything is checked by human - telephone or telegraph is used to communicate between signalboxes
2 - partial level of protection - electromechanic or electric relay systems to prevent human error (clearing signal at occupied block), still requires human interaction, track clearance is checked by humans (end of train signs) and short circuits that say - something has passed through.
3 - fully automatic - track circuits or axle counters that make sure the block is clear. No human interaction required.
Would 1 and 2 fit in "absolute block" system ?

jamespetts

I will give consideration to the onward reservation beyond non-reversing stops and caution aspect approaching a terminus station issues; on the latter, real life practice has varied over the years.

In real life, there have indeed been semaphore signals on the track circuit block principle in the early 20th century (the Liverpool Overhead Railway, the District Railway, the London and South-Western Railway and the North-Eastern Railway all being pioneers in the UK); there can also be colour light signals using the absolute block system.

The main feature of track circuit block in reality is the automatic signal: a signal which, on a straight piece of track (i.e., with no junctions between it and the next signal) will automatically revert to danger on sensing a train in the next block, and will automatically clear when there is no train in the next block (either all the time for normal clear signals or on the approach of a train for normal danger signals). Signals at junctions in track circuit block are semi-automatic: they automatically revert to danger when they detect the passing of a train, but remain at danger until a route which is sensed by the track circuit (or axle counter or similar) to be clear is set from them to another signal. Another important (but not universal) feature of track circuit block is the multiple aspect signal: a single signal which contains indications not just as to whether the block ahead is free, but also as to whether the next signal (or the second or third next signal in four or five aspect signalling) is clear. Again, there have been three aspect semaphore signals: in the US, this was done by having three position arms similar to the old days of time interval signalling, but, in the UK, this was generally done by having a stop and distant arm on the same post. Significantly, the distant arm in this context referred only to the state of the very next signal, not all the stop signals controlled by the same signalbox as had hitherto been the case with absolute block signalling.

Absolute block signalling, in UK practice in any event, is largely in conformity with your descriptions at both (1) and (2) (what you describe at (2) being called in the UK "Welwyn Control" after an accident at that location that caused the introduction of that safety feature). It has the particular feature of generally having only two aspect signals (there were combined signals, with a stop and distant arm on the same post, but the distant arm was in this case worked by the next signal box along the line compared to the stop arm, and specially interlocked to prevent the distant arm showing clear when the stop arm showed danger) and distant signals indicating, not the aspect of just the next signal, but the aspects of all the signals controlled by the same signalbox. In effect, in absolute block, a "section" is the line controlled by a single signalbox, whereas in track circuit block, a "section" is the line between a pair of signals. (This latter principle was moderated in later years by the introduction of "intermediate block" signals, which allowed a section to be split into subsections, each with its own distant signal; but these were introduced only after automatic signalling was invented, and was just a cheaper means of increasing capacity on a line without a full resignalling to track circuit block; the London, Tilbury and Southend used this in the early 1910s). Absolute block working (in the UK at least) also generally assumed running lines to be unidirectional (there being no system for determining the direction of a line), bidirectional lines generally being worked by the token block system, or, in earlier times before the telegraph, by the staff and ticket method (not implemented in Experimental because of its propensity to result in deadlocks).

In Simutrans-Experimental, the differences between track circuit block and absolute block can be summarised as follows:

(1) in absolute block, a distant signal (or a caution aspect on the distant arm of a combined signal) will show whether all the signals controlled by the same signalbox as controls the distant signal are clear, whereas in track circuit block, a distant signal (or a caution aspect on a multiple aspect signal) gives only an indication of the very next signal;

(2) in absolute block, it is not possible to have signals of more than three aspects, and the only three aspect signal allowed is a combined signal, the distant arm of which will only work if the signal is within range of the next signalbox (and will work as described above); in track circuit block, signals can have up to 5 aspects showing the state of the succeeding 1-3 signals, and they do not need to be in range of more than one signalbox to work;

(3) absolute block signals are normal danger, whereas track circuit block signals on plain line (i.e. with no junctions to the next signal) are normal clear, although I am considering allowing them to be normal danger if so specified in the .dat files; and

(4) bidirectional signals are not permitted in absolute block, whereas they are permitted in track circuit block; bidirectional signals will in track circuit block engender a directional reservation until the next one way sign or reversing point when they are passed.

What applies to track circuit block above also applies to cab signalling, although this need not be more than 2 aspect because caution aspects are redundant in cab signalling. Thus, if a particular country's signalling system did not have the rule about distant signals referring to all signals controlled by the same signal box and did allow for directional reservations, even signals of type (1) in your list could be simulated by the track circuit block method in Experimental (possibly with the addition of normal danger).

Does this all make sense?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

Thanks for nice explanation. I think that for cz signalling I'll just ignore absolute block - it seems too UK specific. I'll just use circuit block. And perhaps the same for swedish signals.

Default clear is in cz shown only on permissive 3 aspect signals. 2 aspect signals even in curcuit block are default danger. Could the number of aspects be used to set the default? What about UK and sweden?

At 4, by one way signal you mean a signal pointing only in one direction or single way sign? In which direction it has to be to stop the directional reservation?

jamespetts

It does sound as though Experimental's implementation of track circuit block is closer to what you want for these purposes. I do not think it sensible for the number of aspects to alter the default position as to whether signals are normal danger or normal clear because this would be confusing and it varies by country in any event, but having an option for this in the .dat files (default: normal clear) would leave full flexibility for pakset authors.

As to (4), bidirectional signals are those that point in both directions along a line. This starts a directional reservation. It is ended not by a signal but by a one way sign (which will have to allow passage in the direction in which the train is set to travel, or else a route will not be found past it).
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves

Quote(3) absolute block signals are normal danger, whereas track circuit block signals on plain line (i.e. with no junctions to the next signal) are normal clear, although I am considering allowing them to be normal danger if so specified in the .dat files;
Apparently, the first versions of track circuit block signals in Sweden (from 1920 and onwards) got turned completely of when no route is booked.

Octavius

Interesting discussion on Czech signalling.

In the Netherlands, from an operational point of view, there are 2 kinds of signals:
- Controlled signals. The signaller can set it to show stop, proceed (but it will still show red if the track is not clear or a conflicting route is already set) and drive by sight, for call on. By default they are set to stop. Since the 1990s, they are computer controlled, so if a train is expected according to the time table and no conflicting routes by earlier (but delayed) trains are expected, the computer reserves the track and sets the signal to proceed. After the first axle of the train has passed the signal, it reverts to stop.
- Automatic signals. The signaller has no control over the signal. It just shows whether the track is clear to the next signal. By default these show clear. However, when a train stops at a platform just before the automatic signal and there is a level crossing directly beyond it, the signal may show stop until shortly before departure of the train to prevent the level crossing from closing for too long.

Both kinds come in different versions, depending on the speed aspects they have to show, or for the distant signals, dwarf signals and some special signals applicable only to some (heavy, underpowered) trains. Controlled and automatic signals look identically, apart from an ID plate.

When a controlled signal is set to proceed, it makes a directional reservation of all track until the next controlled signal. Points must always be protected by a controlled signal. When showing proceed, the signal also indicates the maximum safe speed at which the next signal can be approached. The end of the track is considered a red signal, there also exists a fixed sign to be interpreted as a distant signal at caution.

A special version is the locally controlled safety system, used on some (but not all) single track railways. When the driver or guard of the train operates the button, key or infrared remote control at one station with a passing loop, then the decentral safety system checks whether the track is available, reserves the track up to the signal at the next passing loop, sets the points, closes the level crossing (if any right after the signal) and clears the colour light signal. This is somewhat similar to a token system in the sense that it can be operated by the train staff.

All other lines use centralised traffic control, mostly with absolute or permissive colour light signals, but a few lines have been equipped with ETCS-2 (cab signalling only) or dual systems: two different systems on the same line, which never give conflicting information, so that trains can choose which system to use depending on top speed and installed equipment.

Behind a train the reservation is freed immediately. So, when a train has passed some points, then another route over those points can be reserved immediately for another train, speeding up operations. There is no need to wait until the first train has passed the next signal. This means that when a train reverses, the driver can't know whether the track has been reserved for him. It may be unreserved or even reserved for a different train. Therefore he has to drive by sight and check alignment of the points, until he can see a signal telling him to proceed. This arrangement is used in Belgium too. At every location where it's likely that trains reverse, controlled signals have been placed, which in practice means at the ends of all platforms of major (junction) stations, the ends of most passing loops and in yards, but not at minor stations.

I don't really know about signalling in the Netherlands before 1930. Practices were largely similar to British practices I think, although I'm not aware of any large-scale token block working. I think it was mostly flag or semaphore signalled single track with electrical (or optical) telegraph, or signalmen walking from one signal box to the next to make sure the track was free. There were only a few trains per day anyway. Development of the dutch rail network was slow until 1860.

I have found one painting though, made around 1845, showing a train on the then single track main line from Amsterdam to Haarlem. The signaller waves a white flag to instruct the train to proceed. An electrical telegraph had just been installed and an optical telegraph was still present. It looks a bit like a semaphore signal with tilting discs, but it was for communication with the next block post, not with the train driver.

In 1933 automatic block signalling was introduced, along with colour light signalling. This also allowed bidirectional working, so that on a double track line trains could either use the left or the right hand line. Before that, trains of one company always had to keep left and trains of the other company always had to keep right. Nowadays, bidirectional working is possible almost everywhere.

Signals and points where mostly power operated by 1950, maybe before that. From then on the mechanical interlocking mechanisms where replaced by electrical relays. Every station and yard still had to be staffed, as no cables for remote control had been installed past yard limits. In the 1970s there was a brief experiment with radio controlled signalling with spring-loaded points, which depended on verbal instructions by radio (some of the last mechanical signals were removed), but after a large rail crash in 1980, locally controlled colour light signals were installed after all. From then on things have been centralised. After 1990 everything has been computerised. Signallers only have to act to set signals to drive by sight (aka call on) or to change the routes or chronological order of trains to something different than scheduled. Nowadays there are only 13 control centres left, but some people think that's too few. If there is a problem in one of them, not only an important junction is disabled, but all bypass routes along with it, stranding half a million passengers. People are trying to get some flexibility, so that stations or complete lines can be transferred from one control centre to another when the need arises, or just during the night.

Concerning Simutrans, this means that
- Track is reserved up to the next signal, past the next stop. If the train reverses you can't really reserve past the platform (to which signal?), so although realistically it should be reserved to the next signal, you can't in Simutrans. That doesn't really matter as that only allows few additional movements and only if no signal is build at the end of the platform, but at stations where trains reverse there should normally be a signal at the end of the platform.
- Except when using a token block system (which is not used in the Netherlands), trains revert to drive by sight when reversing. Other countries may have different practices though.
- Tied to the statement above, except when using a token block system, track is released as soon as the train has passed.
- There would be two kinds of signals, the controlled signals (or absolute signals, or whatever you call them) and the automatic signals (or permissive signals, or whatever you call them). Track reservations run from one signal to another, direction reservations from one controlled signal to another controlled signal. A controlled signal can only be cleared when there is a reservation to the next signal (or end of line, or reversing stop) and a directional reservation to the next controlled signal. A directional reservation cannot be made if a tile already has a directional reservation from a different tile, but it may have a directional reservation to a different tile, that is, directional reservations may branch, but not merge.
But directional reservations are not extremely important. Single track sections with only a single block signal can already be made and single track operations with more intermediate blocks are rare, except for emergency use, which doesn't happen in Simutrans.

Quote from: Ves on November 12, 2015, 10:15:17 PM
Apparently, the first versions of track circuit block signals in Sweden (from 1920 and onwards) got turned completely of when no route is booked.
A number of years ago I saw something similar at a block signal on a single track line of the Rhätische Bahn in Switzerland. When the block was free, it showed green, when occupied, it showed red, when the direction was reversed, the signal was dark.

Quote from: jamespetts on November 09, 2015, 11:31:30 PM
I will have to give some thought to the question of reserving beyond a station stop: it may be that it is an important thing for the purposes of priority reservations (which already has about half of its code: these will be coloured in yellow) that allow some trains to take priority over another by making a special type of reservation ahead of the normal reservation, which can be overridden by a train with a higher priority level (or the same priority level if a block reservation is requested over a priority reservation). The plan is for both the number of blocks ahead and the priority level to be customisable by line/convoy in the schedule. However, I worry a little that this will increase computational load in circumstances where some users are already reporting a higher computational load than before I implemented the signalling changes.
Interesting idea to allow high-priority trains overtake low-priority trains. Until now we needed either low speed tracks on loops with some weirdly placed signals and then hope the fast train would be so close behind the slow one that it would overtake (which meant it was already slowed down by the slow train), or use specially crafted timetables. I tried that on a few occasions, but the downside was that I needed a simple station, which looks silly, only works on straight track and forces trains to wait at least the minimum loading time, which with 15 minutes was far longer than the typical waiting time of 5 minutes.

jamespetts

Thank you for that interesting insight into Dutch signalling: what you describe, including the difference between automatic and controlled signals, is extremely similar to UK track circuit block practice (and is simulated in Experimental by a setting which detects whether there are any points to the next signal, which determines whether the signals are normal clear or normal danger, the former for automatic, the latter for controlled signals).
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

Thanks for nice description of netherland's signalling.
I think it is not in conflict with what I wanted. We could sum these things up.

- Reservation goes up to the next signal (or more if aspects > 2), or to reversing stop.
- Train goes to drive-by-sight only after reversing (or leaving depot)
- Multi aspect signals show caution if the next stop is reversing.
- Track is freed immediately. (as is now)
- Directional reservation is only in circuit-block, and ends at single-way sign. (as is now)
- Absolute block can have only one block signal between stations on single track, so it does not need directional reservation. Single track split in more block is extremely rare anyway. (as is now)

The first two points, make token-block (radio token, one-train staff, direct control, etc) not necessary in most cases. I think that in real life tokens are used only in places, where building proper signals and signal boxes would not be economical. I think that from the game point of view the difference is really small. Perhaps only that the track is freed only when the train returns the token (at next signal). Thus it won't revert to drive-by-sight even if reversing.

jamespetts

Ahh, the importance of token block signalling is that it is necessary when track circuit block signalling is unavailable. Since, in Pak128.Britain-Ex, track circuit block signalling is available only from 1902 (and then, when the balance is completed at least, at a cost to make it uneconomic for lightly used single track lines), token block will be necessary for single lines for many decades.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

With reservations modified as above suggested, you can signal single track lines with absolute block. But absolute block should be more expensive than token block.

Octavius

Quote from: Vladki on November 15, 2015, 10:38:44 PM
Thanks for nice description of netherland's signalling.
I think it is not in conflict with what I wanted. We could sum these things up.

- Reservation goes up to the next signal (or more if aspects > 2), or to reversing stop.
To make that one even clearer, in multiple aspect signalling reservation goes multiple signals ahead (maybe not necessarily aspects-1 signals ahead), but if a train has a non-reversing stop only to the first signal after the stop. Reserving multiple signals ahead happens no earlier than when the train is ready to depart from the station.
Quote
- Train goes to drive-by-sight only after reversing (or leaving depot)
Maybe add a revert to drive-by-sight sign to that. I'm thinking of trams that run some distance on a railway at high speed, using railway signals, and then continue through city streets driving by sight. These trams exist in The Hague and other cities.
Quote from: jamespetts on November 15, 2015, 10:55:16 PM
Ahh, the importance of token block signalling is that it is necessary when track circuit block signalling is unavailable. Since, in Pak128.Britain-Ex, track circuit block signalling is available only from 1902 (and then, when the balance is completed at least, at a cost to make it uneconomic for lightly used single track lines), token block will be necessary for single lines for many decades.
The painting I mentioned (this one; yes, it's broad gauge, the Dutch were inspired by Brunel) seems to indicate that there was absolute block signalling in use around 1845 on single track lines. Back then there was one train every 3 hours in each direction and the primary task of the signalmen was to check the tracks for defects after each train. Now it's 8 trains every hour in each direction, but the line has been doubled.

That doesn't mean such signalling ought to be available in every pakset of course.

Ves


Quote from: Octavius on November 16, 2015, 07:32:34 PMMaybe add a revert to drive-by-sight sign to that. I'm thinking of trams that run some distance on a railway at high speed, using railway signals, and then continue through city streets driving by sight. These trams exist in The Hague and other cities.
This is already possible by creating a "signal" with working_method=drive_by_sight. The signal will in effect become exactly such a sign.

jamespetts

Ves is correct about the end of signalling signs: these are already implemented in Pak128.Britain-Ex.

As to that lovely Dutch painting, I see telegraph wires, so there was some way of communicating between adjoining signal cabins/huts. Would this signalling system, however, not have behaved in the same way as the token block working method, at least at the level of abstraction in which Simutrans-Experimental simulates these things, even if not actually using physical tokens?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

jamespetts

I have now implemented an option for normal danger: if a signal of track circuit block or cab signalling type has "normal_danger=1" defined, the signal will not reset to clear after a train has passed it. I have not tested this because there are not any signals of this type in Pak128.Britain-Ex; would one of the authors of the Swedish pakset care to test it to make sure that it works?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Vladki

I'll check that. In the meantime I have a few questions:

Should/must drive_by_sight sign have is_singal=1 ?  Can it have two aspects?

Is there a possibility of choose sign that would work similar to road choose sign for drive_by_sight, one_train_staff and token block?

jamespetts

The drive by sight/end of signalling sign should have is_signal=1. The "aspects" parameter is ignored, as the signal's states will not be used and it will only display the one base graphic (it is intended to represent a fixed sign: nothing else makes sense in the context of the logic of drive by sight).

I have not tested choose signals with those working methods; it should work for token block, but I am not quite sure whether and if so how it might work for the other methods. It does not really make sense in one train staff mode, as that is intended to reserve the entire route from beginning to end in advance and work only when there is only one train in an entire and separate section of line. In drive by sight mode, it could in theory work in the same way as the choose signs for road vehicles, but I have not tested this. Is there a real precedent or demand for this?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves


Quote from: jamespetts on November 19, 2015, 12:11:10 PM
I have not tested choose signals with those working methods; it should work for token block, but I am not quite sure whether and if so how it might work for the other methods. It does not really make sense in one train staff mode, as that is intended to reserve the entire route from beginning to end in advance and work only when there is only one train in an entire and separate section of line. In drive by sight mode, it could in theory work in the same way as the choose signs for road vehicles, but I have not tested this. Is there a real precedent or demand for this?
It could be handsome to have choose functionality in drive by sight. Maybe an automated choose function (no sign or signal is needed) but the train automatically find an empty platform? The end of choose sign could then be used to reserve platforms from choose function.

Alternatively a choose sign?

jamespetts

I am not sure what you mean by, "The end of choose sign could then be used to reserve platforms from choose function;" can you rephrase?
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves

I'm just meaning that the end of choose sign would work as it does currently, and a way to get rid of unwanted trains on wrong platforms would be to place such a sign on the specific platform (as it does already today)

jamespetts

Ahh, I see, yes. I am not sure that having an automatic choose system (without any sign) would be sensible, but having a choose sign for drive by sight should work. It would be helpful if you could test whether this works already so that I know whether I need to alter the code for this.
Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves

Quote from: jamespetts on November 19, 2015, 06:18:04 PM
Ahh, I see, yes. I am not sure that having an automatic choose system (without any sign) would be sensible, but having a choose sign for drive by sight should work. It would be helpful if you could test whether this works already so that I know whether I need to alter the code for this.
Do you mean a signal with:
working_method=drive_by_sight and free_route=1 together?

jamespetts

Download Simutrans-Extended.

Want to help with development? See here for things to do for coding, and here for information on how to make graphics/objects.

Follow Simutrans-Extended on Facebook.

Ves

Quote from: jamespetts on November 19, 2015, 09:52:09 PM
Yes.
Putting working_method=drive_by_sight and free_route=1 together doesnt make it behave different than a normal "end of signalling"-sign