Junction signal not pulling off until right before it? (AndiS scripting)

[albinopigeon]

Hi all,

I have a signalling conundrum. I'm making this post on behalf of one of my routebuilding partners, and I'm sure the problem is a common one.

This is a bog standard two-arm splitting home signal, which operates normally elsewhere on the route but this particular example only pulls off around 50m from the signal itself, so the preceding splitting distant doesn't work at all. We've got both signal types to work fine both in isolation and on other parts of the route.

No wrappers have yet been employed (mostly because either a) we haven't needed to use them or b) we don't know how to yet), but there is a shunt signal looking the other way just within the platform. There is a crossover within the platforms a short while down but I don't know if that would impact it.





Any ideas?

Thanks in advance!

[AndiS]

Idea 1: At least the left-hand track looks like a bay to me. In that case, put a into the route character of line 1. Then the signal will clear despite having no distant ahead. Otherwise, it might bore you with rule 39. If it is a loop, forget idea 1.

Idea 2: The 50 m sound a bit like the signal does not clear when the train is announced or sensed, but a signal ahead starts clearing at some point. However, sensing range is this case is 500 m for stand-alone distants and only 200 m for other signals and the starter will be more than 200 m ahead. Such a thing generally happens because of some other flaw, or a very sick situation in the prototype.

Idea 3: Those isolated crossovers are where you are going to need wrappers. You need something 1T to span across the trailing switch on the through track. A disk is not prototypical (if I dare to generalise) so you need a 1T wrapper which is just a disk without the disk. No magic, you just save sinking the disk underground (if you ignore a few optional extras that you do not need here).

What happens here could be this: The dispatcher does not set the path through the crossover when the home first hears about the train, which can be very early. In this case, any message sent from the home forward along the through track just gets lost. The home decides that the through track is a dead end and plans for rule 39. As the train approaches, it either starts a second call to the signal ahead which now gets through to the starter since the crossover will be set by now, or - more likely - it still believes there are buffers ahead and does rule 39.

If you did not slow down in your test, this theory is falsified.

If the engine was standing within (err) 200 m (or so) from the home, then the condition to clear was fulfilled.

No idea how far away you started and how fast you approach but 50 m is a bit close for the signal to complete clearing.

But I have seen cases where people extended the sensing range because of the delays. There are two cases: Signal just waits for the train to slow down - no delay. Signal for some reason starts to call out "prepare for train" which can be forwarded a few signals (until it meets a section signal), then they clear in reverse order, which can take several seconds, and the nearest signal clears last.

[albinopigeon]

This is a detailed look at what happens.




This is after a wrapper has been added to that trailing point on Platform 3, which (for some reason) stops the gantry signal working altogether. Do switch wrappers need something adding to it other than just putting it down and placing the nodes? I added it in the route editor instead of the scenario editor but I'm not sure how that would impact it.



The starter (just visible under the footbridge) doesn't pull off until the train is actually going through the station. I had it at the preceding station (Corfe Castle) where both the home and starter signals were green at the same time with a clear path through. We also tried adding a lower-case "a" to the first line of the gantry signal which does nothing. Another quirk is that as soon as the Wareham starter pulls off, all the signals for at least two miles down the line work normally. It's just this one gantry signal that seems to be blocking messages. What you theorise to be happening sounds right to me; it's just figuring out a way of making TS behave itself.

You also mentioned something about augmenting the sensing range of a signal. How would we do this in this context? Because it might also be a handy addition further down the line.

[AndiS]

Note 1: the first two pictures don't work. I get something about not having permission to access the cloud when I copy the link in a browser address line.

Note 2: always a good idea to give the place name as I can then look up the diagram and give an assessment of the whole situation. The problem with the Kuju approach to signalling was that it would be very popular if you could just stick in signals as animated scenery and flip the arm when the train comes but it just does not work that way - as we all know.

Now my signals take full blame of taking the mess a bit further and having them send 3 time as many messages to and fro in an attempt to make them appear as if controlled by one guy with a plan and his interlocking. So let me give you a full walk-through as only all the pieces together will become a working whole even if not all the comments are immediately linked to your observations.

Distant: Coming from Worgret Jcn, I see a single arm distant. This will only clear for the through route (which is a very good idea, safetywise). In this case, you should make the right-hand doll of the home route 1 and the bay route 2. Then the default configuration ensures that the distant clears when the arm for the up through route clears and not for the up bay.

Home bracket: Independent of the doll numbering, you ought to have an inner link beyond the crossover. This means you make it a 3-link signal and link 3 goes between the trailing switch of the crossover and the facing switch into the bay. Route character is the minus sign. No further fuss needed. Now a train that uses the crossover does not emanate out of plain air and one shunting (reversing) across it will not disappear inexplicably and therefore blocking the signal forever.

Level crossing: Argh. I forgot all my preliminary research on them. All I remember is this: They don't talk to signals and an crossing open to road traffic appears as a "broken path", i.e., a trailing switch set against the train. I am not too sure about the last bit. The situation gets blurred by the fact that crossings tend to close soon, so they are closed when the crucial train messages arrive - most of the time. To be on the safe side, I would make the platform starter a 1T signal with link 0 beyond the level crossing (i.e., beyond its link(s)).

Of course, this is a place where the link placement needs extra care. From the platform, you hit
link 0 of the wrapper
the end of the curved track piece of the crossover
link 1 of the wrapper
link 0 of the platform starter (*)
the level crossing
link 1 of the platform starter

(*) Since the signal reads to trains coming across the crossover, too, you need to put link 0 here even though the stopping place of AI might look better if link 0 was at the platform end. Stopping with the stack below the footbridge was certainly not recommended practice.

Besides the up siding east I see another trailing crossover. Without a wrapper there, no joy is to be expected.

Then there is the trailing connection with the this same up siding which needs another wrapper. You can use one 2T wrapper instead of the two 1T wrappers. There may be minimal benefit in message passing efficiency.

The advanced starter will be a 1T signal, or else you need another wrapper at the trailing connection with Sandford siding.

The ensemble is completed by Holton Heath distant. After placing that, take a sip of your preferred drink and cross your fingers.


Extra caveat about the level crossing: I guess that people make clones of the script and alter the distance at which it closes to their needs. You cannot chose between stopping trains and through trains. Although you could check whether GetTrainType of whatever it was works in OnConsistPassing in that script. If it does, you could order it to close at a big enough distance for train types that do not tend to stop at Wareham and a shorter distance for those that do.

At any rate, to not expect the platform starter to clear while the crossing is not open for the train and thus do not expect the distant to clear even in a correct setup, unless the closing distance of the crossing is set to such a value that the closing is complete several seconds before you pass the distant.


Footnote on wrappers: You are right in placing them in route editor. Scenario editor is only for the scenario specific Movement Predictors.

Footnote on sensing range: If you append $=400 to the signal id, the signal will observe your speed starting 400 m from link 0 instead of the default 200. As always, $ is the separator between the shown signal id and what follows is summarised on the last page of the manual, which is merely a cheat sheet with the full stuff distributed across the manual.

Footnote on Holton Heath: The platform starter there must be 1T with link 1 beyond the crossing. Beware of stupid behaviour when shunting between the down line and sidings while the up train already runs into the platform. There is an issue with such diamond crossings in that they sometimes send their "path broken"/"path set" messages in a way that I did not foresee. This only matters if the path is not already set for the train when it arrives.

[albinopigeon]

I tried a couple things (adding $=1400 to the signal ID and making it a three-link signal with the third dummy node) and it worked like an absolute charm! I'll relay the good news to Monty (routebuilder) so he can fix it on the master copy. You might be right on the signalling diagram having only a single distant instead of a splitting distant but it's up to Monty as to which one to pick.

Thanks a bunch Andi! Lord knows that I wouldn't have got half as far with signalling without your expert help.

[AndiS]

You might be right on the signalling diagram having only a single distant instead of a splitting distant but it's up to Monty as to which one to pick.

Everyone knows that all those surviving diagrams are just a snapshot of the 100+ years of semaphore history. Actually, the diagram I have shows a bracket for the single arm that may well be the surviving half of a splitting distant.

I was just a bit worried that with the single arm announcing the route into the bay and the home enacting rule 39, you can tear out all your hair and there would still be no clear aspect on that distant.

Did I mention that the route suits the signals quite well?

[albinopigeon]

Thank you for your kind words! I'm very happy with them visually.

However, there is an unfortunate epilogue to this case - the home signal won't work when routed into the bay. It works perfectly when routed into the main, but for some reason it refuses to work for the bay.

A recap of my setup:

The home signal has three links: the bay, the through platform beyond the trailing point, and inbetween the facing and trailing points, labelled 1, 2 and - respectively. The signal ID is $=1400 and there is a wrapper covering both trailing points for the bay and main.




The bay has a buffer lamp at the end which in actuality is coded as a distant signal. I've tried making the home signal a 4T and adding another dummy link inbetween the trailing and facing points on the bay but this didn't seem to work very well? It might be because of the testing scenario being slightly bugged so it may warrant a re-test.

[AndiS]

In a previous picture, link 1 of this signal was beyond the trailing switch, at the barrow crossing. This was fine. In this case, the new wrapper, the one in the funny skewed orientation, in the second picture, just has to go. It would be a signal inside a signal which never must happen. The whole message passing circus only ever works if we got link 0 - numbered link - maybe some other numbered link from a signal in a parallel track - plain track (no switches or crossings) - next link 0.

If you changed link 1 of the home signal to be placed just beyond the facing switch (about where the guide rails are) and well before the train would reach link 0 of the new wrapper, then I say no more (and remain clueless).

[albinopigeon]

I may have to remain as equally clueless as you! I've binned the second wrapper and moved the link 1 to be either just after the trailing point or inbetween the trailing and facing points and neither seem to be working. I had this signal working perfectly well in isolation but it seems to be in this one spot where it refuses to work (for reasons unknown). I can have another play around but I am steadily getting into the fabled grasp-at-straws stage! Perhaps it might be best to leave this one until Ken can see it, chances are he's probably had a similar situation (unless since your last message you've come up with a potential solution!)

[AndiS]

You face two challenges here that can be separated, a trivial one and a big one.

The trivial one is to ensure that there is no stupid little mistake in running into the bay. If that fails, there is some stupid thing with the home signal. But that did work. So I would, first of all see that link 1 is beyond the red or yellow frame of the converging track ribbon and not just beyond the rendered switch points. Only where the outline of the track ribbons meet is the abstract location of the switch and this is the only thing that matters for signalling.

You already checked your buffer stop lamp cum distant signal in a test layout. If you add a three switches there, you can depict the full situation for running into the bay. If it works in the separate test and not in the route, you got wonky trackwork in the route which I do not suppose. But I don't know which features and follies the track editor has recently. When I last laid track, he put placed upright when he found that my double slips or switches would not suit he geometric capabilities. Anyway, I have seen signalling fail because of duplicated track pieces and gaps and malformed switches, but I see no hints at such a thing in the screenshots you showed. (All the red frames made me nervous at first glance but they all pertain to point rodding and telegraph lines and fences.)


For the other branch of track, I would start a new test layout which initially contains splitting distant, home and Holton Heath distant - and a mile of empty space between the two. The bay would just be the lamp-distant to be sure that 100% of everything works here. Then you put in the platform starter and then the level crossing. Then one after the other each of those trailing switches is inserted, each with its 1T wrapper, or some 1T signal. You can certainly have 1T home signals here if that makes you comfortable, ruling out all wrapper uncertainty. I am sure there will be some small detail to find here as you combine quite a few firsts here - wrappers, multiple starters, level crossing.

I find these test bed exercises quite thought purifying because you easily get carried away by the beauty and/or complexity of the scenery and it is important to think nothing but "what can go wrong here". Its just a series of programmed items that send messages and the question is where they are lost.