Explain intersection based vs. loop based rail networks

[Zourin]

what you're actually discussing is a loop rail system (one-direction loops) vs a shuttle system (bi-directional). For a shuttle system running multiple trains, you don't need to run double-tracks the whole length. As long as there's a cut-around for passing large enough to hold the 'waiting' train(s), you should be fine without jams.

I find shuttle systems easier to manage for space, resources, and general sanity. Goods go from point A to point B and back typically. Loops always take up awkwardly large amounts of space in my opinion. If I need more throughput, larger trains are typically a better answer compared to more trains.

The drawback is when you have to 'cross' train tracks. Then you really have to pay attention to how the segments are broken up or you hit awkard deadlocks.

[ratchetfreak]

what you're actually discussing is a loop rail system (one-direction loops) vs a shuttle system (bi-directional). For a shuttle system running multiple trains, you don't need to run double-tracks the whole length. As long as there's a cut-around for passing large enough to hold the 'waiting' train(s), you should be fine without jams.

I find shuttle systems easier to manage for space, resources, and general sanity. Goods go from point A to point B and back typically. Loops always take up awkwardly large amounts of space in my opinion. If I need more throughput, larger trains are typically a better answer compared to more trains.

The drawback is when you have to 'cross' train tracks. Then you really have to pay attention to how the segments are broken up or you hit awkard deadlocks.

with chainsignals crossing bi-directional track can be done.

just put chains at the entrance and before and after the crossing

[Linosaurus]

What I like about double headed trains is how easy it is to build a loading station that doesn't block the main line. When I try to build a station for a single headed train it always needs a *lot* more space. This is worth a little slower acceleration.

Also shown below the station is a T-junction, not my design. I like it because it takes up slightly less space than a circle, and before pre-signals it could let several trains pass at the same time while still having a very low chance of deadlock. All the tracks except the stations are still one-way.

Station and T junction

Factorio_double_headed_station_and_T_junction2.jpg (68.89 KiB) Viewed 6296 times

[Gus_Smedstad]

what you're actually discussing is a loop rail system (one-direction loops) vs a shuttle system (bi-directional).

No, I'm not. I can see how you might think that, but I'm talking about a rail system that uses one-way track but avoids loops. It's difficult because it's not really my idea I'm discussing, it's something I encountered in the Show Your Creations board, in this thread.

In that thread, MadZuri summarily dismisses loop-based rail networks in favor T-junctions, without really explaining why other than to assert that they're much better. If that's the received wisdom, I want to know why.

Loops are completely unnecessary, as are the loop-based junctions. There are no non-loop tutorials out there though, so everyone uses loops.... Well, except those of us that spend a lot of time on twitch.tv

It's clear that he's using one-way track except at stations. Given the above, you'd think that making it possible to turn around at a station and using single-locomotive trains would make more sense, but that seems secondary to the larger question of loops vs. crossing track.

[SpeedDaemon]

In that thread, MadZuri summarily dismisses loop-based rail networks in favor T-junctions, without really explaining why other than to assert that they're much better. If that's the received wisdom, I want to know why.

Loops are completely unnecessary, as are the loop-based junctions. There are no non-loop tutorials out there though, so everyone uses loops.... Well, except those of us that spend a lot of time on twitch.tv

It's clear that he's using one-way track except at stations. Given the above, you'd think that making it possible to turn around at a station and using single-locomotive trains would make more sense, but that seems secondary to the larger question of loops vs. crossing track.

Well, before we had chain signals, the main problem with those roundabouts was the "one train per intersection" limitation. IMO, the roundabouts were still useful when you were starting out, and generally had few trains, and were cramped into small areas behind your defenses. Once the number of trains increased, a system of three-way T intersections was better due to multiple trains being able to use it at once.

Now WITH chain signals, I think the answer is "it depends." For a three-way, properly signaled Ts and roundabouts are basically equivalent, except in space usage. For a four-way, a + would probably be superior, since a train making a left (given right-hand driving) would block more potential traffic at the roundabout than a good cross.

[DaveMcW]

For a four-way, a + would probably be superior, since a train making a left (given right-hand driving) would block more potential traffic at the roundabout than a good cross.

The roundabout does not allow simultaneous left and right turns from opposite directions. But in practice, trains never try to do that anyway.

Most train networks emphasize traffic moving to and from a single base. There is no reason for trains to approach from the far side and then turn onto a side track.

[MeduSalem]

The roundabout does not allow simultaneous left and right turns from opposite directions. But in practice, trains never try to do that anyway.

Most train networks emphasize traffic moving to and from a single base. There is no reason for trains to approach from the far side and then turn onto a side track.

Which is why a simple Y-fork where there's only one track intersecting another would actually suffice as well in 90% of the cases. It wouldn't even be necessary to build a fully fleshed-out 3-way interchange (nor anything beyond that) where every train is able to go in any direction at the interchange due to how in most designs all the trains end up in one main station anyways.

At least it is basically never required for a train to go from one mining outpost to another. If they end up at the mining outpost and no trainstop being free, well then they have to wait there because in most cases it would take much longer to get somewhere else anyways, causing more unnecessary traffic on the entire system.

And by the way... the Y-fork could be handled without chain signals already due to how there's only one intersection of two tracks and thereby only 2 trains who might want to enter that shared block and the one who gets there faster will automatically reserve it and will be able to leave without a deadlock. So the efficient deadlock-free intersection people are looking for already existed before 0.12.

The only system where that doesn't work due to traffic volume and layout complexity is in factories where everything, not only resources is transported with trains rather than by belt/bots, which is in my opinion like killing a fly with a nuke anyways.

[Zourin]

I'm going to go with the "K.I.S.S." philosophy on this one. If you're thinking about it too hard, it's too complex a solution.

If you're simply trying to 'intersect' (or fork) two railways, a basic ring (or T-junction) with controlled ins/outs work best as long as you can block additional trains from entering the loop to jam it. Making any rail layout in this scenario bi-directional is also asking for trouble. Funky shapes and micro-blocking makes my head want to explode, and I have enough of that already.

[Gus_Smedstad]

So I've spent more time with a significant rail network in my current game.

I started with rotaries, and soon realized they take a lot of space compared to a simpler T or Y intersection. A rotary is similar in size to a 4-way, but it's very unusual to actually need a 4-way intersection. It's not just the space for the rotary itself, the rail lines have to diverge on entering the rotary, which creates another Y-shape (though obviously one where the two rails don't cross).

Space matters near my main base. Not for intersections a significant distance from my main base, but for areas near the actual processing plants, they took up a lot of space. So I converted to intersections, but kept single-locomotive trains. Because locomotives are actually somewhat cost-intensive in the early game, and a rail loop at a station, while somewhat space-consuming, is much cheaper.

Now I've got a really large rail net, covering a lot of space, but only 4 regular trains. One each for iron, copper, oil, and stone. And yet, even with a lot of rail and few, short trains, they've still managed to get into deadlocks at intersections, usually doing something as basic as just passing in opposite directions. I wouldn't have thought this would cause a lock, but if they're both turning, it does, crossing the opposite rail locks that rail. Even in as basic Y-junction this happens. So far I've managed to resolve such problems with additional chain signals, but it's not as trivial as it seems.

MeduSalem is incorrect. It turns out that Y-intersections are not enough for many situations. At least in part because the oil train generally makes the rounds to many different wells, picking up a few barrels from each, rather than shuttling between a mine and a smelting station the way the iron and copper trains do. Once depleted, oils wells produce a trickle, but that trickle never completely vanishes the way ore mines do, so it's worth keeping the stations. So while ore mines can get away with Y-intersections, oil depots and branches of the main line need full T intersections. Else trains up end traveling through odd stations just to turn around.

I may convert to Madzuri's two-headed train setup. In the late game, locomotives are no longer expensive, and I could reclaim some space used for turnaround near my main base. Little turn-around loops in remote depots are unimportant, but space near the processing plants is crowded.

[DaveMcW]

Little turn-around loops in remote depots are unimportant, but space near the processing plants is crowded.

Build a giant loop outside your base, and carry the ore on belts the last few tiles.

[SpeedDaemon]

MeduSalem is incorrect. It turns out that Y-intersections are not enough for many situations. At least in part because the oil train generally makes the rounds to many different wells, picking up a few barrels from each, rather than shuttling between a mine and a smelting station the way the iron and copper trains do. Once depleted, oils wells produce a trickle, but that trickle never completely vanishes the way ore mines do, so it's worth keeping the stations. So while ore mines can get away with Y-intersections, oil depots and branches of the main line need full T intersections. Else trains up end traveling through odd stations just to turn around.

I may convert to Madzuri's two-headed train setup. In the late game, locomotives are no longer expensive, and I could reclaim some space used for turnaround near my main base. Little turn-around loops in remote depots are unimportant, but space near the processing plants is crowded.

I play with Bob's Mods, so a late-game single main bus setup is impractical (it'd be like 40 tiles wide). I generally end up with lots of outpost factories and ore dumps, with trains going from everywhere to almost everywhere else. Basically the rail network is a "smart" main bus at that point.

The main drawback to double-headed trains is that you can't generally guarantee which end will be which when it comes into a station (unless you micro-manage its route, which isn't possible on a large network that has loops). That's fine for bulk cargo like ore, but doesn't work at all when you need to deliver 14 different materials to an outpost and have the cars match up with the appropriate smart inserters to unload.

Fortunately, there's no rule that says you have to limit yourself to one or the other, so I run a mixed system and use whatever makes sense for a given purpose.

[hitzu]

The main drawback to double-headed trains is that you can't generally guarantee which end will be which when it comes into a station (unless you micro-manage its route, which isn't possible on a large network that has loops). That's fine for bulk cargo like ore, but doesn't work at all when you need to deliver 14 different materials to an outpost and have the cars match up with the appropriate smart inserters to unload.

But you can always unload trains into active povider chests sitting in the local logistic system. It is quite expencive for large trains, but it is much less bulky and very simple to build.

[SpeedDaemon]

The main drawback to double-headed trains is that you can't generally guarantee which end will be which when it comes into a station (unless you micro-manage its route, which isn't possible on a large network that has loops). That's fine for bulk cargo like ore, but doesn't work at all when you need to deliver 14 different materials to an outpost and have the cars match up with the appropriate smart inserters to unload.

But you can always unload trains into active povider chests sitting in the local logistic system. It is quite expencive for large trains, but it is much less bulky and very simple to build.

Unfortunately, that doesn't work, either, because you end up clogging the unload chests with one item or another. You could have the smart inserter check to make sure the item it's unloading is < than a certain amount, but then you're right back to needing the right train car in front of the right smart inserter again. Ultimately, the "K.I.S.S." solution is just to have a single-ended train.

I do use the "just dump everything into active provider chests" method for ore trains at the central smelter, but those storage chests are wired to all the mining outposts, which don't load the train in the first place if that ore is full. Can't do that for factory outposts, or every outpost would need a dedicated pickup stop at every intermediate item factory.

[hitzu]

Unfortunately, that doesn't work, either, because you end up clogging the unload chests with one item or another. You could have the smart inserter check to make sure the item it's unloading is < than a certain amount, but then you're right back to needing the right train car in front of the right smart inserter again. Ultimately, the "K.I.S.S." solution is just to have a single-ended train.

Well, then you can store items in dedicated requester chests and then empty them via smart inserters that check the fullness of the storage. Just be sure not to use storage chests in that logistic network.

[OBAMA MCLAMA]

@Gus you mention stuff like.. Y and T junctions. Did you link the Y junction so every input has 2 outputs?

and whats with this trains being late game.. i build trains as soon as possible double head x 5 trains maybe a 2-6-2 or what.
My next world im probably going to through out a 3-12-3 from the start for no reason at all. Double unload 6 wagons at a time booya! ;p I don't feel like starting a new SP world for a long time though

[Gus_Smedstad]

Build a giant loop outside your base, and carry the ore on belts the last few tiles.

Well, originally the processing plants were on the outskirts, but the base grew. I could build new ones further out and scrap the old ones I guess, but in terms of revision, going with push-pull trains is less work, and a more permanent solution. The main drawbacks are slight excess mass per train, which doesn't seem like a big deal, and using another 10 engine units per train to construct additional locomotives, which isn't a huge cost in the late game.

Two-way tracks are normally a danger, but the idea is that you construct stations that only have one track in and one track out, even if they do reverse direction. The exit is one-way, like the majority of the network, so it's shouldn't be possible to create a deadlock.

In the real world, of course, two-way tracks are commonplace. My nearby commuter rail station, for example, has just a single track used for both eastbound and westbound trains, and two lane track only starts showing up downtown. Presumably careful scheduling and appropriate sidings handle the issue.

[Gus_Smedstad]

@Gus you mention stuff like.. Y and T junctions. Did you link the Y junction so every input has 2 outputs?

and whats with this trains being late game.. i build trains as soon as possible double head x 5 trains maybe a 2-6-2 or what.
My next world im probably going to through out a 3-12-3 from the start for no reason at all. Double unload 6 wagons at a time booya! ;p I don't feel like starting a new SP world for a long time though

By Y junctions and T junctions, I mean 3-way junctions. The difference being that a T junction allows turns in any direction, where a Y junction only allows the train to exit or enter the main line, but doesn't allow a turn from one branch to the other.

Example Y junction from my current map:



Example T junction:



Note that neither of these is has quite as much signaling as I found necessary elsewhere to break a deadlock.

[Gus_Smedstad]

The main drawback to double-headed trains is that you can't generally guarantee which end will be which when it comes into a station (unless you micro-manage its route, which isn't possible on a large network that has loops).

That's a good point, which I hadn't considered. My ore and stone trains are monolithic, but the oil train has separate cars for full and empty barrels. With the two-headed train and the associated station design, the train reversing car order every time it enters a station. You can't willy-nilly change the schedule from shuttling back-and-forth from oil well to refinery station to doing a grand loop around marginal oil wells, because the train will face the wrong way in half the stations.

You can work around that now by using cargo filters, so each cargo car contains a mix of stuff instead of a single item type, and load and unload with smart inserters. Which could get hairy quickly if your train carries more than 2 cargo types. Which my trains don't, but I don't play with mods that add a bajillion extra ores.

The other solution is monolithic trains, no matter how many different ores you have, which means a lot more platforms and shorter, individual trains.

[Gus_Smedstad]

and whats with this trains being late game..

This depends on your map, but I've often found that I've progressed a fair bit before I exhaust my initial iron and copper deposits, which obviously predate train technology. I guess it depends on what you mean by "late game," but for me in my most recent game, I built my first train to pull in oil from my second well.

[ssilk]

The main drawback to double-headed trains is that you can't generally guarantee which end will be which when it comes into a station (unless you micro-manage its route, which isn't possible on a large network that has loops). That's fine for bulk cargo like ore, but doesn't work at all when you need to deliver 14 different materials to an outpost and have the cars match up with the appropriate smart inserters to unload.

Ahh, yes, I have made this kind of experience lately. This is a nice problem. I found an easy solution: double headed trains needs to be always an odd number of wagons. 1, 3, 5 wagons (more don't make sense). In that way, you can keep them symmetrical!

While single headed trains (one direction only) make more sense with even numbers of wagons: 2, 4, 6 (more makes also no sense).