In any transport mode, the basic problem of traffic dispatching is to keep vehicles out of each others' paths. For surface transport with separate paths for opposed directions, there are three basic forms of conflict:
- the crossing of unrelated paths (green vs. all others on the diagram below);
- the intersection of opposed directions at a bifurcation (red vs. blue);
- paths merging at a bifurcation (orange vs. blue).
On roads, the basic solution to all three involves either traffic lights at a level crossing or roundabouts. Excepting light rail, roundabouts are no real option for rail because the minimum curve radius of rail vehicles is much longer (and the desirable minimum even longer: due to lower deceleration/acceleration, you don't want trains to slow down too much). As for crossings, those involve gaps and thus increased wheel and rail wear, and are often substituted with a succession of two opposed bifurcations (switches/points; orange vs. blue or red vs. blue on the diagram). Of course, the basic solutions limit capacity, especially on rail: one direction has to wait for the passage of much longer vehicles with longer braking distances than on roads.
A long freight train is crossing over from the mainline to the track of a diverging bypass line at 40 km/h, blocking the opposite mainline track, thereby adding three minutes to the delay of the passenger train I was riding
Trains are supposed to avoid conflicts at level junctions when running by schedule. However, if a train arrives at the junction with a delay, the next train scheduled to pass on a conflicting path may have to wait, thus the delay will cascade. If you have good information flow and a computer model of tracks and trains, you can predict the travel times of trains already in motion. However, this won't help when you're right next to a station where the first delay in the cascade emerges (a train departs late). That is, the dispatcher couldn't have predicted in time that it would have been better to let the second, non-delayed train pass the junction first.
Location matters for another reason, too, one that makes the rail version of the "intersection of opposed directions at a bifurcation" problem really special: the intersection is most often between parallel tracks, and that in built-up areas. Why?
- Many of the busiest junctions are for special rail installations which are parallel to, and right next to a mainline: freight sidings, marshalling yards, maintenance facilities.
- These are rarely duplicated on both sides for both directions, as is usual for the road equivalents (gas stations, highway rests and service areas).
- Even for junctions of 'truly' diverging lines, stations strike again: separate tracks often continue side-by-side into a station where the actual crossing of paths takes place.
The above-fold and the freight train photo were actual examples of the opposed directions conflict at a junction station. Below is another (at the same station), a situation like in the first diagram above: a late train from a branchline (diesel multiple unit [DMU] left of centre) departed from its station platform and crosses over to the mainline tracks, thereby blocking a train that arrived in the opposite direction on the mainline (bottom right).
Actually, you can have the intersection of opposed directions conflict at stations without a 'proper' junction: the paths of trains to/from two different platforms can cross. This will happen frequently if the station is the terminus for train services and trains have to reverse.
An example of the above is Rekawinkel on Austria's old Westbahn. The situation below is almost exactly like on the drawing: the electric multiple unit on the left just turned back but can't cross over to the Vienna-bound track until the locomotive-hauled train from Vienna passed (itself crossing over to the "wrong side", to let a third, faster train overtake it later).
Finally, there is another special case of conflict between opposed directions: at transitions between left-driving and right-driving networks (for example east of Metz en route from Paris to Germany).
Now that we see the problem, let's look at the general high-capacity solution: junctions with at least one bridge or tunnel (and several curves). It's obvious that in all of the above cases, the nice big curves and perpendicular bridges familiar from road (think cloverleaf junctions on highways) won't cut it. Instead, you squeeze a ramp between the other tracks and then have the tracks pass above/below each other at a shallow angle. (In practice, usually even the bridge looks like a box tunnel.) In railway English, this is covered by the term "flyover". The railway German term Gleisüberwerfung (c. track throw-over) is narrower and more descriptive. The railway French term saut-de-mouton (sheep jump) is not just spot-on but funnier. In railway Hungarian, a term doesn't even exist, because not a single one has been built yet.
I show two examples. The first is an old one at Česká Třebová in the Czech Republic, a city that grew at the big junction of the lines from Prague, Brno (and Vienna) and Ostrava. Its role is to separate freight and depot-bound traffic from passenger traffic. The first photo from the ground is reproduced from my 2008 diary Trains in Moravia; I shot the second from the train window en route to InnoTrans 2012.
Flyovers can be even more expensive than 'normal' bridges, because you can rarely use terrain to limit earthworks and you may need retaining walls if the width of the right-of-way is very limited. Flyovers spread in Europe in recent times only thanks to high-powered passenger trains (multiple units with distributed traction or relatively short double-deck trains hauled by modern locos): such trains can climb steeper grades, thus the approaches can be built shorter, as on my second example below.
A flyover on the 3+2-track east approach of Berlin's Ostbahnhof (East railway station) which currently facilitates the access of different train types to different platforms (in East Germany times, it was also useful for train reversing)
You may have noticed that all of the above considered main traffic routes with lots of traffic and separate paths for opposed directions. On routes with light traffic, however, it's cheaper to have a single path for both directions, in which case the main traffic problem is the meeting of vehicles travelling on the same route but in opposed directions. On roads, drivers solve this by pulling aside at a suitable location. On a fixed guideway, however, you absolutely don't have this option. You need a section where paths separate, a multi-track section with at least the length of the longest trains.
The most common place for train meets on single-track lines is a station with at least one extra track (called siding) where trains have to stop anyway. I show two examples at station Morgó on the Kismaros–Királyrét narrow-gauge railway in Hungary. On the first, a locomotive-hauled train is arriving (left) while a pointsman is waiting to throw the switch for the DMU from the other direction, which is waiting on the siding (right):
The DMU had to wait another minute, because the locomotive-hauled train was followed by the solar-powered narrow-gauge motor car on a test run:
On lines across low population density areas (like much of the USA), or on specialised lines like mining railways, stations might be too far and between. Then you build a 'station' without cargo or passenger facilities or even any buildings: a passing loop (British – and EU – railway English) or passing siding (American railroad English).
Let me note in advance that trains don't have to stop at passing sidings: if the train first to arrive slows down a bit, its leading unit can arrive at the other end of the siding just after the end of the second train passed it. In North American railway English, such a dispatching wizardry is called a rolling meet (or in some regions a mallard). I couldn't find a truly good video, but on the one below, the second train passes the end of the first and the signal that just turned green around 06:05 in.
Let's return to Europe, where single-track lines have relatively frequent stations and you want to run passenger trains on schedule. What happens if two trains are scheduled to meet at a station, but one of them is delayed? If the schedule is tight, then, obviously, the other train has to wait, thus the delay cascades. If the first two trains meet upon further trains as they travel on, then the delay will cascade further. To avoid this, trains are scheduled to meet with buffer times of a few minutes. Usually the train more likely to be delayed is scheduled to arrive earlier and then wait a few minutes (see the green train at Einstadt station and the purple one at Deuxville station on the virtual train diagram below). Sometimes the other train is scheduled to wait after the train meeting before departing (green train at Deuxville).
However, such buffer times aren't an ideal solution:
- they extend travel time by default,
- they don't help if it's the other train that is delayed (purple train at Einstadt and green train at Deuxville on the second diagram below); nor
- if the first train's delay exceeds the buffer time (purple train at Deuxville).
Double-tracking is an expensive solution, justified only if you want high train frequency. However, you can do with less by turning the relationship of infrastructure and schedule on its head: first you prepare a regular-interval (clockface) schedule ignoring train meetings, and then add an extra track only at locations where trains would meet by that schedule:
In English, such a selectively double-tracked section can be called a passing loop/siding, too, and is treated as such from a signalling viewpoint. But, although these European passing loops aren't longer in absolute terms than North American passing sidings and dispatching resembles the "mallard", there is a key difference: they are much longer than the maximum length of the trains they were built for. This way, you can have a buffer time for delays without idling: usually, the time a train takes to run the length of the siding minus train length.
The example I show is the first application in Hungary (and the occasion for the present diary). This 2.3 km passing loop was built as part of the total upgrade of the Budapest–Esztergom line (more on that in a comment), and will be used for train meets in two more years (after the rest of the line is finished).
Of course, American passing sidings can be used for rolling train meets with buffer times against delays, too, if both trains are much shorter than the maximum train length. In the example below, two NJ Transit passenger trains have a rolling meet (the second train appears 1:20 in, the switch is thrown at 1:55).
Passengers crossing tracks
Now here is a problem that most West European rail passengers won't be familiar with any more. Vehicles can block each other without crossing each other's paths: if their passengers cross each other's paths. On roads, you can have bus-loads (or tram-loads) of people crossing at pedestrian crossings, but this is usually synchronised with the management of road traffic at an adjacent road crossing, with traffic lights. On rail, however, this type of conflict can lead to additional delays at busy junction stations, where trains in opposed directions might be scheduled to arrive at almost the same time to catch connections. If one train is delayed, then either it won't be allowed to leave or enter the station until the train in the opposed direction leaves (thus increasing its delay), or the other train will have to wait (the delay cascades).
Both situations occur frequently at the main station in my hometown, a junction for one mainline and two branchlines. Since the introduction of a coordinated regular interval (clockface) timeplan, up to seven passenger trains occupy the platforms simultaneously. On the example below, a three-minute delay is cascading between two trains that arrived from opposed directions on the mainline:
The solution is to prevent passengers from crossing the tracks. This is done with elevated platforms and platform bridges or tunnels above/below the tracks. After much empty talk, my station is finally getting these in a reconstruction (the roofs of the first two island platforms are visible on the right while you see another delayed train left of centre blocked by passengers of the train right of centre).
On the example of an in-service elevated platform below, at station Kelenföld in Budapest, the train on the left is given the signal to depart while passengers are boarding the train on the right – this wouldn't be possible with passengers walking across the tracks.
Of course, delays aren't the only or even primary reason for this infrastructure:
- elevated platforms provide for comfortable and wheelchair-suitable boarding (the main selling point towards the public),
- keeping passengers from the track at all times improves safety,
- safety is improved in particular when non-stop trains pass the platforms, which then can be allowed to do so at higher speeds.
Obviously, the main motivation for elevated platforms with over/underpasses is safety. For example, Prussia started a programme of platform tunnel construction in reaction to the Steglitz Rail Accident, which occurred in the outskirts of Berlin in the evening of 2 September 1883. The calamity arose when the station dispatcher chose to let an express train pass before allowing passengers to board an already arrived, late suburban train. The 300 waiting passengers stormed across the barrier and across the through tracks, and this impatience cost at least 39 of them their lives.
:: :: :: :: ::
Check the Train Blogging index page for a (hopefully) complete list of ET diaries and stories related to railways and trains.