Topic: Time interval with telegraph

[Ves]

I can see from the commits that there appears to be a new working method for signals: the time interval with telegraph.
Could you write a few lines about this new signal type and what makes it different to the 'normal' time interval?

[jamespetts]

I shall in due course, but I have not quite finished implementing it yet, and precisely how it will end up working is potentially subject to change. In general, however, it is intended to treat junctions differently to ordinary time interval signalling: sections in which there are junctions (divergent track between one signal and the next) will be treated akin to something between absolute block and token block, whereas a plain track section will work in the same way as standard time interval signalling.

[Ves]

I have been playing around with the time interval telegraph signals and I really like the idea of them!
I just get a little confused as to how it is working and how Im supposed to use it:

I see that after a while (this might be a bug in one way or the other) the track up to the next signal will be direction reserved in blue. I write "after a while" because for the first 3-4 runs of the trains, it will be a completely red reservation up to the next signal.
This (the blue reservation) suggests that the signal can be used to protect single tracked lines, to prohibit trains going in oposite directions, but yet allow multiple trains in the same direction.

However, Im unsure how to correctly provoke such a blue reservation. If placing the signal at the platform ends or the end of a passing loop, the signal will only book red reservations due to the junctions, and so there would need to be another signal outside the station/junctions, but this can create deadlock when the train gets released from the platform but the line is occupied (in the wrong direction).

I know you wrote they are not finished yet, but will there be some mechanic to prevent some of these deadlocks?

Compilation version e2c88d5, have not acces to compile fresher at the moment...

[jamespetts]

It is indeed intended in due course to be usable with single lines, although this is not implemented at present. On the basis of the Eastern Counties Railway's 1848 rulebook, a copy of which I have managed to download, this looks as though it should work, in Simutrans terms, more like token block and not use the the directional reservation system. The directional reservation that you currently see is a leftover of when I had used directional reservations over junctions, not knowing how junctions were handled at all in time interval signalling. They are likely to be removed and replaced with a block reservation that clears only when the train reaches the end of the section for sections that are not plain track (i.e., those that have junctions, including single line sections).

[Ves]

I have been exploring how to use these signals and might have found some issues.

This is how I have interpretted the use of them. If i am doing it wrong, then please tell me

Code: [Select]

_____->_                                            _<-_______
_____->_\__<C-_->__________________________<-_-C>__/_<-_______

->  = Time interval signal with telegraph in the pointed direction
-C> = Time interval choose signal with telegraph in the pointed direction

My tests shows that, when the train leaves a platform, it will reserve a path through the points to the time interval signal at the line.
The line it self will be unreserved, but a path from the destination choose signal until an available platform will be reserved.

This all seems sensible, but there are some issues:

(1) When a second train is following the first train, that train will not reserve any path from the destination choose signal, with the result that when the first train arrives at its destination, the points out of the station are unreserved, allowing for a train in the oposite direction to head out, resulting in a crash and a deadlock.

(2) When a train attempts to depart from a time interval signal with telegraph trough points and there already exist another reservation made by another train, the departing train will depart from the signal, but "crash" into the reservation. Both the train crashing and the train that owns the reservation will stop and report Emergency Stop. After a while the train owning the reservation will start drive again.

(3) Occasionally, trains will reserve the entire line between the stations.

(4) Also occasionally, trains will only reserve the tile with the choose signal and not reserve the path from the choose signal to the platform.

Savegame:
https://www.dropbox.com/s/8ulohiw6grhuumw/Time%20interval%20test.sve?dl=0
All 6 trains are at the station. Fast forward and the trains will depart. After a minute or two, the trains will start crashing into each other at the single tracked line.
Everytime I load the game, it will behave differently.

[jamespetts]

There are two things happening here. Firstly, when a train first traverses a line, the signals do not know whether there are any junctions ahead, so they assume that there are until the train passes. This will result in reservations being made the first time that a train passes but not on subsequent occasions.

Secondly, the line is incorrectly signalled. (I have been looking only at Meningport to Sackfield for the present). See here for the line correctly signalled, which behaves consistently and does not result in deadlocks.

Specifically, some of the platforms at Meningport had distant rather than stop signals, and one had no signal at all. Further, the signals beyond the station throats at Meningport and Sackfield allowed a train to leave the station and stop at that point, without any way of stopping a train in the opposing direction from reaching that point (since it allows the train leaving the station to enter an area without junctions which the train in the opposite direction will therefore not try to reserve). Removing these signals should end the deadlocks.

[Ves]

It was indeed only the part between Meningport to Sackfield that I too was mentioning!

But in what way is this then different to absolute block signalling?
I did somehow expect trains being able to travel after each other in the same direction, but in the savegame, there is ever only one train traveling the line.

[jamespetts]

It was indeed only the part between Meningport to Sackfield that I too was mentioning!

But in what way is this then different to absolute block signalling?
I did somehow expect trains being able to travel after each other in the same direction, but in the savegame, there is ever only one train traveling the line.

The behaviour seems consistent with what I have read from the Eastern Counties Railway rule book of 1854,

"323. No two trains travelling in opposite directions must be permitted on the Line between Bury and Newmarket at the same time.

"324. Prior to any Train leaving Bury for Newmarket the Station Master must ascertain from Newmarket that no Train, which has not arrived at Bury, has left that Station for Bury, which will be effected as follows: -

"325. The question, 'Is the Line Clear?' must be asked; to which Newmarket will reply, 'Last Goods (or Passenger) Train left at 12," or otherwise as the case may be. The Station Master, then knowing whether that Train has arrived or not, will act accordingly in starting the Train he has to send, - first inquiring of Saxham whether the preceding Up Train has passed that Station, which must be done as follows: - The question, 'Is the Line Clear?' must be put; if such be the case the answer 'All Clear, to follow,' must be returned by Saxham; if the preceding Train has not arrived the answer must be 'Stop.' If the Line be clear the Train may be dispatched, and a telegraph message sent to Newmarket - 'Goods (or Passenger) Train left at 1 p.m' or as the case may be.

"326. Saxham, after giving the answer 'All clear, to follow,' must Telegraph Higham, 'Is all Clear?' and Higham must return the answer, 'All Clear, to follow,' if the preceding Train has passed there before the train from Bury is allowed to pass Saxham. In a like manner Higham must communicate with Kennet, and Kennet with Newmarket, before an Engine or Train is allowed to pass those Stations respectively.

"327. Trains from Newmarket to Bury must be worked in a manner corresponding with the above, except that the order of telegraphing will be reversed.

"328. No Train or light Engine must leave Bury for Newmarket, or Newmarket for Bury, until the Driver has received from the Station Master a written order to proceed.

"329. In the event of the telegraph communication failing, from injury to the wires or any other cause, the Trains must be worked strictly in accordance with the Staff Time Tables, with regard to their order of precedence in arriving and departing from Newmarket and Bury, and all Special Trains must be strictly prohibited."

As to differences with absolute block, this implementation of time interval with telegraph requires half line speed through any junction section, whereas it works at full speed (and with normal time interval rules) on any non-junction section.

[Vladki]

James, what you describe for the Bury and Newmarket, is IMHO absolute block on single line with telegraph/telephone communication. It's exactly the same as I was trying to describe recently as Czechoslovak (austro-hungarian) method of signalling single lines. As I understand Bury and Newmarket are stations where trains can pass, while Saxham, Higham and Kennet, are jsut signalboxes (perhaps with a stop), without passing loop.

In time interval, I would expect something like this:
stationmasters at Bury and Newmarket negotiate if the line is clear as per 325. The intermediate signalboxes will only be informed (by whatever means - e.g. optical, bell, needle telegraph) that from now on, trains will go from Bury to Newmarket. Bury will then dispatch trains in 10 minute intervals without any further communications necessary, just as with time interval on double tracked line. Perhaps the station masters will have to keep track how many trains departed and arrived, to be able to negotiate if the line is clear again, so that they can send trains in opposite direction.

[jamespetts]

The rulebook describes time interval signalling as the default method of signalling, and gives the above instructions for specific lines (which can be inferred were at the time single track). The inference is, therefore, that this is the method of signalling adopted for single lines when time interval signalling was used on double track sections.

Remember, the same type of signal is needed for junctions. Earlier on in the rule book, we have this about junctions:

"JUNCTION SIGNALS
...
"42. The Semaphore Signals must always be kept at Danger, except when required to allow an Engine OR train to pass, the Caution Signal must then be shewn; and in the event of Trains from two or more lines approaching a Junction at the same time, the Danger Signal must be kept on, until each Train comes to a stand, when the Branch Train must take precedence in passing the Junction points, except when specially ordered to the contrary."

However, the instructions for the Halifax and Ipswitch dock junctions (under "instructions for switchmen") are intriguing:

"59. The Telegraphic Signal communications between these points is worked as follows, and which must be strictly adhered to until further orders:- A single ring of the bell from one Junction to the other, if replied to by a single ring signifies "All Right," Train to proceed; if replied to by a double ring it signifies "Caution," Train to proceed very slowly; but if replied to by three rings, it signifies that hte Line is blocked and that all Trains must be stopped. In like manner, if one Signalman sends a signal to the other of one, two, or three rings, as the case may require, it signifies "All Right," "Caution," or "Danger," and it must be replied to by the same number of rings, to let it be understood that hte Signal is received; and when the Danger Signal of the rings is given, and until counermanded by the All Right Signal of a single ring, nothing must be allowed to pass between the Junctions. Also when no reply is given to a Signal, it must be considered as dangerous, and no Train be allowed to proceed until the All Right Signal is received; or in the event of no Signal being received after a lapse of five minutes, the Trains may then be allowed to go very slowly indeed..."

This does strike me as unclear as to the extent to which time intervals were actually used at junction locations. I could replace a full block reservation with a directional reservation for time interval with telegraph signals, allowing trains to proceed in the same direction, but it is unclear how this actually worked.

[Vladki]

I had the general idea that the basic principle in time interval is that you can dispatch trains (in the same direction) solely on the fact that some time has passed since the first train departed. Especially that there is no confirmation that the train has passed the next signalbox. In contrast absolute block is based on confirmation (by whatever means), that the train has passed and the block is clear. From that I conclude that in UK, absolute block was first implemented on single track lines, while for double track lines, the time interval was considered sufficient.

[Vladki]

I'm trying to understand the junction rules, and translate to simutrans working. We have to distinguish a junction of lines without station, that would be (in simutrans) signalled by plain signals, and a station, where a choose signal would be used.
Choose signal should work just the same as in absolute or circuit block. Make reservation up to the platform or end-of-choose. In old czechoslovak (and austro hungarian) rulebooks, there is a sort of old entry signal, that was probably used to protect the station already in time-interval era. It was electro-mechanical, and operated by the station master - so he knew exactly if the points are set correctly and tracks are clear, just like in absolute block, with no need for telecommunications.

For junctions with simple signals it is a bit more complicated. Simple solution would be to switch to absolute block - default danger, make reservation through the junctions up to the next signal. The rule 59. seems to be some sort of transition between time-interval and absolute block. Otherwise for pure time interval at junction (42) I would expect something like this:
- if there are no junctions ahead of signal (up to the next signal), normal working applies - default clear, < 5 min danger, 5-10 min caution, >10 min clear
- if there are junctions ahead - default danger. All signals around one junction must be connected to the same signalbox. Only one of them can be pulled off (at caution) at any time. Next train (in whatewer direction) can follow after 5 minutes. Clear aspect is not used at all.
- I know this makes quite a bottleneck. It could be improved if the 5 minute interval is kept for each branch (next signal) separately, but I don't know how hard would that be to implement.

However, the idea that trains pass junctions at caution, takes me back to modern (absolute/circuit block) choose signals, especially those with has_selective_choose=1. Usually when train has to take alternative (not-straight) path, it has to traverse switches at lower speed. CZ/SK speed signaling has aspects for 40/60/80/100 km/h, and I used the 40 km/h aspect for choose signal showing the alternative route. I know that the speed would be limited by curvature of the track, but it could be further limited in similar way as in time-interval's caution - half of tracks or signals max speed. And if the correspondent distant signal could show 3rd aspect if alternative route is selected, it would be even better. But consider this as low priority wishlist item

[Junna]

I had the general idea that the basic principle in time interval is that you can dispatch trains (in the same direction) solely on the fact that some time has passed since the first train departed. Especially that there is no confirmation that the train has passed the next signalbox. In contrast absolute block is based on confirmation (by whatever means), that the train has passed and the block is clear. From that I conclude that in UK, absolute block was first implemented on single track lines, while for double track lines, the time interval was considered sufficient.

As I remember time interval signalling was used on early Swedish railways (single track), but was quickly phased out after a severe accident (in the 1860's).

Also I think chose signals can be used for non-station tiles? They will chose at a junction, I think-- If not currently, this could be useful, if there are parallel tracks and you might want to distribute the trains over them if they go the same way.

[jamespetts]

I have implemented longblock signals in time interval with telegraph. In the Github repository for Pak128.Britain-Ex, I have used the station signal/"T signal" graphics for them, although this is only cosmetic: only the normal directional arm is operational. (It would take a lot of work to make these operative in both directions and would not add anything significant to the actual signalling logic).

These work by making a directional only reservation to the next longblock signal or signal protecting a junction, whichever is sooner. They are less flexible than token block signals, in that they do not reserve through stations and less flexible than track circuit block junction signals in that they give a caution aspect and require half speed (until the next plain signal in any event). The directional reservation will, however, continue past non-junction signals of the time interval or time interval with telegraph type, which will continue to work normally (although I have carried out only basic testing of this). These signals can be used on plain track (i.e. not in front of a junction) to create a directional reservation from that point onwards, although I am not sure how useful that this is.

They are mainly intended to be used as starting signals from stations ("exit signals"). They are not suitable to be used facing into stations ("entry signals") as they will not check whether there is a stationery train already in the station, leading to possible rear-end near misses (which will cause a lengthy countdown before either train may move). One of the main reasons for not using the directional reservation for all time interval with telegraph (and plain time interval) signals protecting junctions is the station issue. Hopefully, this will be an acceptable compromise.

I should be grateful for the results of any testing, as I have had time to do only basic testing of this myself this evening.

[Ves]

I have been away for a time but tried to keep up to date with reading posts etc, so pardon me if I have missed something!

I am testing time interval signals with telegraph and is_longblocksignal=1 now.
There is nicely created a directional reservation behind such a signal and the train is happily departing. However, when a second train approaches the signal, it again creates a directional reservation, but it will never be sent out on the stretch, until it is clear (the first train passing another signal). You would expect the train to depart after a certain amount of time.
In other words, the signal seems to behave like it does without is_longblocksignal=1 defined, only the reservation is blue instead of red. If a signal also is put outside the station (as the infamous "Exit" signals it will also create a deadlock, since the directional reservation from the other end will stop at the "exit" signal, even though that is faced the wrong way (it should be ignored from that direction).

Putting is_longblocksignal=1 on a normal time interval signal does work, although it creates some odd directional reservations and the trains are restricted to only travel at 35kmh.

Also, the signals tend to disappear when they should be showing CLEAR.

[Junna]

Disappearing when green for four-lamp colour-light chose signals too... I mentioned that in a bug report, but there was no answer.

[jamespetts]

When you both refer to the signals disappearing, do you mean the entire graphic disappearing, or just the arm? If the latter, this is intentional, as this is how slotted post signals behaved in real life: they had three positions: danger (with the arm at 90 degrees to the post and a red lamp); caution (with the arm at 45 degrees to the post and a green lamp) and clear (with the arm slotted inside the post and a white lamp). The use of a green lamp and the arm arm 45 degrees for clear only emerged later.

As to the issue relating to not releasing a train, this seems to be a bug, and I will look into it when I have a chance after properly getting back from my holiday.

[Junna]

The entire graphic disappears. See the bug report thread.

[Ves]

Junna is right, for me it is also the entire graphic that dissappears.

[jamespetts]

I think that I have fixed the issue with not releaseing the train; do you care to re-test? There is something of a complication with the fix to this, however, which I have not yet thought of a satisfactory way of dealing with, in that the signal will release the train if at least 5 minutes (by default) have elapsed since a train last passed that signal and if there are no contrary directional reservations. This does not maintain the correct interval when another train had recently departed from another platform in the same direction. The trouble is that it is extremely hard to code a means of detecting for relatively abstract things such as a train departing from the same station along the same line, so players may have to make sure that they maintain the correct interval manually with schedules unless and until I find a way of dealing with this. (In reality, railways sometimes were a little flexible with time intervals for trains departing from stations).

Edit: I cannot find a separate bug report thread for the disappearing signal graphic, and I cannot seem to reproduce this issue; can anyone upload a saved game in which it can reliably be reproduced?

[Ves]

I think that I have fixed the issue with not releaseing the train; do you care to re-test? There is something of a complication with the fix to this, however, which I have not yet thought of a satisfactory way of dealing with, in that the signal will release the train if at least 5 minutes (by default) have elapsed since a train last passed that signal and if there are no contrary directional reservations. This does not maintain the correct interval when another train had recently departed from another platform in the same direction. The trouble is that it is extremely hard to code a means of detecting for relatively abstract things such as a train departing from the same station along the same line, so players may have to make sure that they maintain the correct interval manually with schedules unless and until I find a way of dealing with this. (In reality, railways sometimes were a little flexible with time intervals for trains departing from stations).

Hehehe, here comes the famous Exit Signal in very handing

joke aside, I get your point, it might feel strange that the signals dont communicate. Is there no way they could communicate via a signal box? I think I recall you explaining that two one train staff cabinetts could be placed next to each other on two tracks, and then they would communicate together (ie a passing loop, so the staff gets returned). Would it be possible to reuse some of that?

I do not have time to compile right now, so I cannot test your fix. Will try to make it in some of the next days.

Edit: I cannot find a separate bug report thread for the disappearing signal graphic, and I cannot seem to reproduce this issue; can anyone upload a saved game in which it can reliably be reproduced?

The signals that disappears for me are from the swedish pak. I cannot guarantee that I have not made a mistake in the dat-files, but I dont think I have. They (the dat-files) are anyway on the github server.

[jamespetts]

I did wonder after our discussion on exit signals whether they originated because of the need for a single point to keep a record of the time in a time interval system: the platform signal would show that there was a clear route to the exit signal, and the exit signal would show that the a train had not passed within 5 minutes along the same line. This logic makes sense in time interval in a way that it does not make sense in absolute block and related systems.

Indeed, if enough space is given after the junctions and before the "exit signal" to fit an entire train, this logic can be used now, only the trains will now wait outside the station for the "exit signal" to clear rather than waiting in the platform.

As to the disappearing signals, can you link me to the .dat files of the specific signals that disappear in the clear state?

[Ves]

I read that the Exit signals originated as a safety precausion on lines with a top speed of more than 90 kmh. The record of which train had left when was solely in the head (and papers) of the station manager inside the station. The exit signal was his tool to block any faulty released trains I guess.

As to the disappearing signals, can you link me to the .dat files of the specific signals that disappear in the clear state?

It occurs to all time interval with telegraph choose signals in the pak, here the choose signal in this document (the second signal entry):
https://github.com/VictorErik/Pak128.Sweden-Ex/blob/half-height/Way/WayObjects/Rail%20Signals/Hand%20Signalling.dat
If you would wish to compile it, it goes together with this picture:
https://github.com/VictorErik/Pak128.Sweden-Ex/blob/half-height/Way/WayObjects/Rail%20Signals/images/Hand_choose.png

On a different manner, where can I adjust the time between the states for time interval w/out telegraph, and where can I adjust the speeds that the different states enforces?

[jamespetts]

The problem is, I think, that you are lacking a caution aspect. This should be taken from the second row of graphics, those with the green flag.

Edit: To answer your other question, see here from the base simuconf.tab

Code: [Select]

# The following settings apply to time interval signalling only.
#
# time_interval_seconds_to_clear is the time, in seconds, after a train has completely passed a time interval
# signal that it will reset its aspect to clear.
#
# time_interval_seconds_to_caution is the time, in seconds, after a train has completely passed a time interval
# signal that it will reset its aspect to caution.
#
# Default: clear - 600 (10 minutes); caution - 300 (5 minutes)

time_interval_seconds_to_clear = 600
time_interval_seconds_to_caution = 300

The maximum speed is hard-coded to twice the lower of the track's maximum speed or the signal's maximum speed.

[Ves]

The problem is, I think, that you are lacking a caution aspect. This should be taken from the second row of graphics, those with the green flag.

I thought that the choose signals would not need a caution aspect? Note that the normal stop signals works without any caution aspects defined.

Silly me, I have been looking and looking without finding that passage, but realize I have been looking in an old simuconf.tab!

[jamespetts]

Time interval signals always need a caution aspect: time interval signals on straight track display a caution aspect if a train has passed between 5 and 10 minutes ago; time interval signals protecting a junction (currently) display only danger and caution aspects, as this was how these signals actually functioned, I understand, albeit only from information concerning two connected railways in the South-East of England (the Great Eastern/Eastern Counties and the London, Tilbury & Southend). Information about time interval signalling is limited, so I am not sure how universal that this was, and I am also unsure of how well that this will work with these time interval "longblock" signals given their intended function and the fact that trains are restricted to half speed at caution (or drive by sight speed when passing through junctions in plain time interval, without telegraph).

[Ves]

I recompiled the signals with three aspects and then they work without disappearing!

The trains can now also travel nicely behind each other in the same direction, however, the second train seems to have no directional reservation. It creates a directional reservation up until the preceding train, but that reservation is not updated (extended) when the trains move along. At one point the train is catching up on its own directional reservation and so has no one after that. This might result in a deadlock when terminus stations are involved.

[jamespetts]

This issue is extremely difficult to resolve, as the reservations are cleared by the train as it goes along, which has no way of knowing that there is a train in rear in the same "section" chasing it with a directional reservation. I suspect, however, that this will not be able to cause any difficulties in practice, as a new signal will always be needed before any junctions in any event, so nothing can ever interpoose. As to terminus stations, it is probably better to have separate signals (preferably choose signals in most cases) immediately before the entrance to a terminus station in any event.

[Ves]

Ok, I understand and I notice the similar behaviour in the moving block method (the directional reservation of the second train is not "chasing" the prior train). It would however be awesome if that could be achieved as it would feel very consistent.

I notice another thing:
The directional reservation made from a signal will only direction reserve until the next signal. When approaching the next signal, it will then direction reserve until the next one and so on. You would expect it to direction reserve through all signals until at least the choose signal which I also think was the intention?

[jamespetts]

Reserving until a choose signal would not make much sense, since one can have a choose signal not at a station (one of the signalling improvements has been to allow choose signals to specify alternative routes other than when a train is stopping at a station) and one can have stations without choose signals. The idea is for it to reserve until the next time interval with telegraph signal, which is what it would have done in reality. It will not stop reservation at a plain time interval signal.

[jamespetts]

I should note that I have made some pakset modifications to this: the slotted post signals replace the vane signals for the plain time interval signalling after the introduction of time interval with telegraph working in 1848. As far as the game is concerned, the signals connected to the "with telegraph" cottages are not the same as those connected to the plain cottages (and this can be seen with the new working method tooltip), but they have the same graphic and name. The prices of both are the same, as are the maintenance costs (as the actual signal posts themselves would be identical irrespective of the telegraph equipment at the cottage).

This is more realistic as vane signals were not being built new well into the 1860s as was previously possible: the semaphore pattern had been settled upon well before then. It also makes it clearer that the methods work in a very similar way (and are identical on plain track), but that installing telegraph equipment in a signalbox/cottage will affect signal behaviour at junctions. The idea is to allow/encourage players to do what was in fact done routinely throughout the later time interval period, which is to mix telegraph equipped and non-telegraph equipped installations on the same line, using the telegraph equipped installations for more complex or important junctions and single line sections. In such cases, the design of the signals would be identical for both.