hub/"entrepot" based factory; station designs?

[vanatteveldt]

[Note: this idea is mostly based on an earlier forum thread, but I can't find it so I'm not sure who to give credit. So if it was your idea, I'm not stealing it, I just, em, borrowed it ]

My previous big factories are all rail-based, with small scale single-purpose plants (e.g. smelting, green circuit) linked by railroads.

I wanted to take this a step further in terms of scalability by using rail 'entrepots' (hubs is probably a better term) as central distribution / multiplexing points, so e.g. 4 iron smelters all deliver plates to the iron plate entrepot, and different iron consumers all draw from the entrepot. I'm doing a bot-based game so I can use bots to distribute input (provider) evenly over output (requester) stations.

I will then not have a single main/trunk line, but a number of layers, at least [ore] - [plates] - [intermediates] - [final (science/rocket)]. Trains will mostly stay in their own layer, except for the plates needed by the final production. Each layer is connected by entrepots. For illustration, this is a first version of the setup in my current base:



So there are trains bringing ore from the outposts to the ore entrepot; from the ore entrepot to the smelter; trains from the smelter to the plate entrepot; and trains from the plate entrepot to the consumers. The point would be that at some point it is easier to add a second plant (e.g. iron smelter) than to keep expanding one.

(of course I could scale by duplicating the whole ore -> science setup, but that's not what I'm trying to do)

My main question here is good station design for the entrepots. An entrepot needs to give access to two train lines (unloading and loading) and allow for high-throughput operations and preferably be easily scalable. Compactness is nice (reduces bot flying time) but I guess 'more bots' can solve the problem as well.

I had three main ideas:

(1) Dual header interleaving combs:

This is what I am currently using:

ingame screenshot

Input and output are interleaved, but because it's dual header they don't intersect. Of course, trains coming and going do intersect, so that's a disadvantage; and dual header trains are by definition less efficient.

(2) Single header adjacent siding stations

This would simply place two single header (siding) stations next to each other, minimizing intersections. Space can be added after each station (to allow them to vacate the station even if the intersection is blocked, or to pick up speed before the intersection) and a stacker can be added before. Two designs:

zoomed on stations

Right hand side has 4 tiles between stations to allow for inserter+chest on each side. Left hand side has only 2 tiles between lines, and interleaves the stations so they can still (un)load from both sides. This creates a longer but more narrow station, but for e.g. 8 stops I feel it is more compact. The front stations are missing one chest but I don't think that's a big deal.

Left hand side also shows a stacker and space in front, right hand side is barebones. Of course I would probably stick to one type, this is just to show the possibility.

(3) Single header adjacent loop stations

zoom

This uses the loop station design I used in an earlier base. It is more compact and has an integrated stacker but not quite as easily expandable. Possible modifications are placing the stations at 2-tile distance (like the left hand side station above), adding waiting room after the station, and improving the stacker.

Questions / ideas:
- General critique on the separation / entrepot setup?
- What train composition do you like for good throughput? 1-2-1? 2-4? 2-6?
- Which stations do you like? Which are stupid? How can they be improved? How would you design the entrepot station?
- Is it worth it to use separate trunk lines per ore type, plate type. I suspect this is more efficient than 4-track trunk lines since there are fewer intersections?
- It is tempting to do without stackers and simply make sure there are enough (un)loading bays. However, that can quickly grow out of control if you need e.g. 24 trains, which is not that much if you really want to scale up?
- I would like to add a waiting area just before the stop, but in my experience that can lead to trains picking an occupied stations and waiting in front of it instead of picking an empty station. Is there a good way to do this?

I want to do some throughput testing for the different designs. I guess the easiest way to do this is the train creation mod (viewtopic.php?f=93&t=49689) and using a circuit connected to the train stop to count the trains?

[Greybeard_LXI]

Questions / ideas:
- General critique on the separation / entrepot setup?

Looks like an interesting challenge. A bit more than I am interested in, but an interesting challenge.

If you make an entrepot station for each item you will have a confusing mess. I would recommend only making green circuits (for example) at one place and using that as the entrepot station.

- What train composition do you like for good throughput? 1-2-1? 2-4? 2-6?

I use 1-1-1 trains, and build the stations big enough for 1-2-1. But I am still building fairly small (4RPH on my last map). I have seen youtube videos using 3-10-3 or 2-6-2 trains for bigger maps.

- Which stations do you like? Which are stupid? How can they be improved? How would you design the entrepot station?

Not experienced enough to answer.

- Is it worth it to use separate trunk lines per ore type, plate type. I suspect this is more efficient than 4-track trunk lines since there are fewer intersections?

I think you will have just as many intersections as a combined trunk line. If your plate line runs east to west and the Iron is north of the copper every time you need copper you will have to cross the iron trunk.

- It is tempting to do without stackers and simply make sure there are enough (un)loading bays. However, that can quickly grow out of control if you need e.g. 24 trains, which is not that much if you really want to scale up?

You understand this as well as I do.

- I would like to add a waiting area just before the stop, but in my experience that can lead to trains picking an occupied stations and waiting in front of it instead of picking an empty station. Is there a good way to do this?

You can put the waiting bay after the stacker. Then the train can pick a station closer to when it wants to use it. You might need to put a station in the stacker so the train re-evaluates stations. This would not prevent a train from pulling behind an almost full train while the train at the other station is almost empty.

[Xtrafresh]

Questions / ideas:
- General critique on the separation / entrepot setup?

I thought about this, but switched to LTN before this became a necessity. LTN basically makes this obsolete. A big downside of a setup like this is the large amount of trains it needs (bad for UPS) and the large amount of planning.
A possible boon is the splitting of supply and demand networks or rails, but that doesn't hold up on closer inspection. Belts, for example require gears and iron, meaning the iron trains need to be "connected" on both the input and output side of the gears station. There's examples of this problem all the way up the chain of products, making the [intermediates] layer you mention pretty much impossible. You are left, then with doing this for ores, plates, and end products.
For end products, it doesn't make that much sense either, because why not just store it at the labs or silo facilities? So you are left with a list of things to build these entrepots for: Ores (Iron, Copper, Coal, Uranium, Crude Oil), and all the items that can be produced from just one of the ores (Iron Plates, Copper Plates, Steel, Gears. I'm assuming here that you are not producing and transporting Copper Coils, and source water locally as needed per production plant, like a sane person.
So, nine entrepots in all.

- What train composition do you like for good throughput? 1-2-1? 2-4? 2-6?

I'm using 1-4-1's with the locs pointed the same direction. This allows for smaller station design than a 2-4, as the last loc can stand in a bend, and even be refuelled.
In a tiered setup like you are describing, I'd probably go for longer trains for the ores, something like 3-12-3 or even 4-16-4. With the buffer of the entrepot, it no longer matters how huge your loads get, and less bigger trains enable a higher total throughput for the rails, and less stress on the pathfinder.
To haul plates, gears, steel and such, i'd use a mix of 1-4-1 and 1-2-1. This allows flexibility with a reasonable throughput.

- Which stations do you like? Which are stupid? How can they be improved? How would you design the entrepot station?

I'd use your 3rd design of station, but instead of placing the heads next to eachother, I'd place them opposite eachother:

If one of the systems is LHD and the other is RHD, the ore transfer will align quite perfectly, and the difference in train length won't matter either. If given enough space during initial setup, it'll be expandable ad nauseum as well

- Is it worth it to use separate trunk lines per ore type, plate type. I suspect this is more efficient than 4-track trunk lines since there are fewer intersections?
- I would like to add a waiting area just before the stop, but in my experience that can lead to trains picking an occupied stations and waiting in front of it instead of picking an empty station. Is there a good way to do this?

Trains repath every 5 seconds when waiting for a red light. What you can do is make sure there's always trains waiting in your stacker. Hook it up to an audible alarm when the stacker is empty, and just add more trains when the alarm goes off.

[vanatteveldt]

I thought about this, but switched to LTN before this became a necessity. LTN basically makes this obsolete. A big downside of a setup like this is the large amount of trains it needs (bad for UPS) and the large amount of planning.

Thanks for your comments!

For me, the amount of trains and planning are what makes it attractive. I looked at LTN but I wanted to stay in (mostly) vanilla and tackle the challenge with the tools we have.

[..] So you are left with a list of things to build these entrepots for: Ores (Iron, Copper, Coal, Uranium, Crude Oil), and all the items that can be produced from just one of the ores (Iron Plates, Copper Plates, Steel, Gears. I'm assuming here that you are not producing and transporting Copper Coils, and source water locally as needed per production plant, like a sane person.
So, nine entrepots in all.

I was planning to make entrepots for circuits as well, at least green circuits; and for plastic (and maybe sulfur). I realize that will give some crossing wires, but I think that will still be managable. We'll see

I'm using 1-4-1's with the locs pointed the same direction. This allows for smaller station design than a 2-4, as the last loc can stand in a bend, and even be refuelled.
In a tiered setup like you are describing, I'd probably go for longer trains for the ores, something like 3-12-3 or even 4-16-4. With the buffer of the entrepot, it no longer matters how huge your loads get, and less bigger trains enable a higher total throughput for the rails, and less stress on the pathfinder.
To haul plates, gears, steel and such, i'd use a mix of 1-4-1 and 1-2-1. This allows flexibility with a reasonable throughput.

Agree on the 1<-4-1<, see last screenshot . I'll see if I can do with smaller trains, presumably simple 1-2 or 1-1 trains for small-quantity items like steel.

I'd use your 3rd design of station, but instead of placing the heads next to eachother, I'd place them opposite eachother:

If one of the systems is LHD and the other is RHD, the ore transfer will align quite perfectly, and the difference in train length won't matter either. If given enough space during initial setup, it'll be expandable ad nauseum as well

Right. The downside I guess is longer bot flying times, but the solution is then more bots . So I think the scalability argument wins the day. For the non-single intermediates I might not be able to mix LHD and RHD (I never thought of that idea!), but in that case the only problem is bot flying times increase with scaling up.

(Note that the direct image link doesn't work, but following the link does: http://imgur.com/a/rPLgK)

-
Trains repath every 5 seconds when waiting for a red light. What you can do is make sure there's always trains waiting in your stacker. Hook it up to an audible alarm when the stacker is empty, and just add more trains when the alarm goes off.

Didn't know about the 5 seconds. Maybe keeping the signal red until a train is at least halfway (un)loaded, or even using pre-stations that only enable when a station is empty or about to be vacated?

[Xtrafresh]

Thanks for your comments!
For me, the amount of trains and planning are what makes it attractive. I looked at LTN but I wanted to stay in (mostly) vanilla and tackle the challenge with the tools we have.

I agree with that sentiment, I love myself a good challenge. However, I'm playing this multiplayer map with a friend, we use LTN there and I'm utterly hooked now. It's just so damned satisfying!

I was planning to make entrepots for circuits as well, at least green circuits; and for plastic (and maybe sulfur). I realize that will give some crossing wires, but I think that will still be managable. We'll see

I can see how you'd want them for greens and reds even. Well, good luck untangling those wires!

Right. The downside I guess is longer bot flying times, but the solution is then more bots . So I think the scalability argument wins the day. For the non-single intermediates I might not be able to mix LHD and RHD (I never thought of that idea!), but in that case the only problem is bot flying times increase with scaling up.

No idea why the image didn't work. I gave it a little more thought now, and realized that you can optimize the design further by interleaving the two stations, as demonstrated in this criminally terrible paint doodle:

That way, your bots are flying nice short flights again

Didn't know about the 5 seconds. Maybe keeping the signal red until a train is at least halfway (un)loaded, or even using pre-stations that only enable when a station is empty or about to be vacated?

Best to not over-complicate things IMO. A decently filled stacker will always keep all stations behind it filled. If you find you need more than 5 stations or so, use longer trains

[vanatteveldt]

Wouldn't interleaving the input/output tracks simply create more intersections and reduce throughput, while any distance problems in the non-interleaved can be solved with 'more bots'?

[impetus maximus]

while i'm a huge fan of terminus stations for picking up ore etc, i'm not a fan for areas with multiple trains/stations.
the criss-cross design in your stations is a bottle neck. a U shape design will have them entering from one side
and exiting another. if you want to expand just make the U taller.

[vanatteveldt]

while i'm a huge fan of terminus stations for picking up ore etc, i'm not a fan for areas with multiple trains/stations.
the criss-cross design in your stations is a bottle neck. a U shape design will have them entering from one side
and exiting another. if you want to expand just make the U taller.

Do I understand you correctly that you prefer the siding stations (design 2 in the OP) above the loop stations? If not, can you give a rough sketch of what you mean?

The problem with making the U taller is that if you want two 'matching stations', you make the ends of the U touch, so there is a limit to how tall you can make it.

I am doing some testing now and it seems that the junction from the trunk to the station is indeed the limiting factor, it seems to allow about 1 L<CCCCL< train every 6 seconds or so, which limits the capacity of the whole station.

[Xtrafresh]

I think, if designed well, the intersections won't be that much of a problem. Not before you have maxed out the station for other reasons anyway, and are better off just building a second station to handle the same task. The reason for this, is the trains will be stopped for a while anyway for loading/unloading. You'll need to use some pretty well-configured signalling, but I don't see why it couldn't work.

[vanatteveldt]

I think, if designed well, the intersections won't be that much of a problem. Not before you have maxed out the station for other reasons anyway, and are better off just building a second station to handle the same task. The reason for this, is the trains will be stopped for a while anyway for loading/unloading. You'll need to use some pretty well-configured signalling, but I don't see why it couldn't work.

Doesn't a (simple) intersection always contain a point where only a single train can be at one moment, and hence a global lock / bottleneck? I will have a look at the recent thread about testing 3-way intersection throughput speeds, but I have the feeling that there is a pretty fixed limit on how many trains can cross paths regardless of design.

Creating a new station in this case doesn't really work because these are supposed to be my entrepots, i.e. they do the multiplexing / load balancing. Of course the train will be stopped, but you can easily expand to 4, 8, 16 parallel (un)loading bays. So, the waiting time at the station should not be a bottleneck, right?

[vanatteveldt]

First test results are in!

test setup

Train deploy mod linked above saturates the incoming track with trains going to loading [until full], unloading [until empty], deletion. Trains are counted when they reach deletion. I wait until equilibrium, then start/reset train and tick counter, tick counter stops when 100 trains reach deletion. Chests are creative provider / void chests, so bot speed is not a factor in these tests

Time to process 100 trains:
[A] dual header interleaved, 4+4 stops for loading+unloading: 547
[C] dual header interleaved, 8+8 stops for loading+unloading: 568
single header separate loop terminus stations, 4+4 stops for loading+unloading: 555
[D] single header separate loop terminus stations, 8+8 stops for loading+unloading: 580

Tentative conclusion: all configurations have the same result of about one train per 5.5 seconds. The junction connecting stations to mainline seems to be the bottleneck indeed, as station design, train length, and number of stations seems to have no effect. Throughput with design B is 16k/5.5=3k plates per second or 180k plates per minute. This used to be about 6rpm, did anyone calculate costs per rocket or per 7 science packs for marathon with prod modules?

[Xtrafresh]

Well, once you hit a certain width, the "loop" station has it's own chokepoints anyway, so you'll need to build a second one. Chockepoints here highlighted in yellow:

Perhaps my interleaved design from above needs a second station slightly earlier (let's say ater 6-8 bays instead of after 10-12), but I doubt the impact is that much, and it doesn't change the fundamental core tenet that no design is infinitely scalable, everything HAS a limit. You'll just have to live with that

As for the solution of just throwing more bots at the problem, that runs into problems pretty fast also. Bots travelling longer doesn't just mean that they take longer and you need more of them, each bot is also going to move less resources per charge, meaning you need more roboports to keep up with the charging demand. That can get out of hand real quick too, and severely limit your design.

[vanatteveldt]

Well, once you hit a certain width, the "loop" station has it's own chokepoints anyway, so you'll need to build a second one.

Yeah, exactly, and the tests above suggest that the limit is about 1 train per 6 seconds, depending on train length of course, and assuming I don't have a trick to make them go through the global locks at full speed instead of needing to accelerate first.

Of course, the trunk line is also going to be a problem pretty quickly, and even if doubled to 4 tracks the bottom tracks still need to cross tracks to reach the station, creating bottlenecks. viewtopic.php?f=194&t=51450 seems to suggest some three way intersections can handle about 40 trains per minute, meaning a train going to the station every 3-4 seconds (and the same amount leaving the station), so there is some room for improvement.

Next I will test the siding-based designs.

[impetus maximus]

here is what i mean about U shaped stops.
i've used the criss-cross method and it's slow.
especially if trains on the main line are waiting to enter.

enter.exit.png (296.84 KiB) Viewed 7058 times

[vanatteveldt]

Cool thanks. That's the same as design (2) in the first part, right? I'll test that next.

[Frightning]

It may have been me that planted the seed of that idea on these boards, I've mentioned it in a few places (usually where the topic of rail throughput limitations came up and I happened to respond). I was considering it for my kilobase once it eventually was scaled up enough that I would have too many trains trying to get in and unload ore, causing traffic inefficiencies to become a limiter on throughput. The solution was to 'pre-centralize' ore sourced from many deposits at hub stations and then run longer trains from the hubs to the main base (the idea can be iterated on too, creating a multi-tiered system when extreme volumes of raw material are required at the main base). I never did scale up my kilobase anywhere near the size scales that would warrant the idea, but I may end up needing it for my new railworld I have begun in 0.15 (I won't have that up to scale for a long time I expect). Your idea of having dedicated rail networks between raw-smelting, smelting-intermediates, etc. is actually similar to how I decided to place the processes in successive rows in my kilobase (I tried to keep the processes near the ingredients that fed into them so as to minimize average robot travel distances. It also does you the favor of modularizing your rail traffic, thereby eliminated cross-traffic inefficiencies.

I like the concept, and look forward to seeing the design come to fruition.

[Xtrafresh]

Well, once you hit a certain width, the "loop" station has it's own chokepoints anyway, so you'll need to build a second one.

Yeah, exactly, and the tests above suggest that the limit is about 1 train per 6 seconds, depending on train length of course, and assuming I don't have a trick to make them go through the global locks at full speed instead of needing to accelerate first.

Of course, the trunk line is also going to be a problem pretty quickly, and even if doubled to 4 tracks the bottom tracks still need to cross tracks to reach the station, creating bottlenecks. viewtopic.php?f=194&t=51450 seems to suggest some three way intersections can handle about 40 trains per minute, meaning a train going to the station every 3-4 seconds (and the same amount leaving the station), so there is some room for improvement.

Next I will test the siding-based designs.

Well, if you need throughputs that are THAT extreme, I'd say what you want is to go for a completely intersection-free setup. Dedicate each section of the map to a resource, have the trains drive in circles that branch off at a mine and back on the circle to the offload station. Connect enough mines to saturate a rail line, then just build another next to it. Devide the map in sectors, and mine a single resource from each sector, to avoid crossing lines. You'll need quite the factory to chew through THAT much material though :p

[Frightning]

Questions / ideas:
- General critique on the separation / entrepot setup?
- What train composition do you like for good throughput? 1-2-1? 2-4? 2-6?
- Which stations do you like? Which are stupid? How can they be improved? How would you design the entrepot station?
- Is it worth it to use separate trunk lines per ore type, plate type. I suspect this is more efficient than 4-track trunk lines since there are fewer intersections?
- It is tempting to do without stackers and simply make sure there are enough (un)loading bays. However, that can quickly grow out of control if you need e.g. 24 trains, which is not that much if you really want to scale up?
- I would like to add a waiting area just before the stop, but in my experience that can lead to trains picking an occupied stations and waiting in front of it instead of picking an empty station. Is there a good way to do this?

The centralization of products of the same type produced at multiple locations mainly makes sense for those things you cannot centralize production of (e.g. raw resources, which must be mined where the deposits occur on the map, and hence, at high volumes, must be mined at quite a few different locations simultaneously (this and the issues with throughput inefficiencies of running many trains to one place are what motivated me to invent the 'hub' system design, which is more or less what you are proposing here, though there is more to your concept than just a 'hub'-system. You can also use hubs for things that have multiple ways of being made (like oil products and U-235 and U-238) in order to create priority sourcing and ensure all users can draw from all production sources.

For train throughput, the simplest for just moving items by train is 1-1-0, and 1-way is generally much better from throughput than 2-way (2-way's main advantage is more compact station layouts, but the price for that is throughput potential). From a 1-1-0, there are two ways to increase throughput (three if you count adding locos and wagons separately). They are making the trains longer, and adding more trains, both of which have limitations and downsides. Adding more trains increases rail traffic, which in turn causes increasing efficiency losses from traffic congestion. Making the train longer means there is more overhead getting into and out of the station, but in-station throughput is correspondingly increased (a station serving 10-wagon trains has 10 times as much throughput (because of 10x as many inserters unloading at once) as a 1-wagon station, ignoring train transition overhead costs; which are large enough to be relevant though). Train speed and acceleration are also relevant, the former moreso on short distance routes and under high-traffic conditions, the latter, for longer distance routes, especially with low traffic.

As someone else pointed out, station throughput is pretty tightly limited by how many lines in and out of the station you have, a stacker exists mainly to prevent surplus from causing trains to back up into the main line, so you only need a stacker with enough space to hold all the trains serving the associated station(s) not counting the train(s) in said station(s). I'd wager that no more than 3 stations per entrance at the most would be useful (and I intend to go w/ just one per entrance for my own railworld, perhaps 2-3 at my ore hubs at most).

Keeping material moving between unrelated processes from crossing each other will certainly help limit traffic inefficiencies, which good for your throughput, so that's definitely something to try and design for, but there will only be so much you can do towards that end, since many things that use fairly advanced components still use basic materials too like Iron or Steel plates, but it's a good thing to strive for with your macro-level layout of the individual process factory blocks (I too am thinking of the same thing with how I am going about selecting where to put each process in my railworld).

Doing away with stackers is doable if you have enough loading bays to match the number of trains serving that process, in general, it's probably easy to have 1-waiting bay per station which with a U-shaped setup really easily, and that allows 2x(# stations) trains to go there with no risk of backup onto the main line.

I believe the 'waiting on filled station when another station is open' problem has been largely solved by some of the tweaks to train pathing that the devs have done for 0.15, but I can't say that I've actually put that to the test (because I haven't field tested it yet), so...try it and see?

[vanatteveldt]

The centralization of products of the same type produced at multiple locations mainly makes sense for those things you cannot centralize production of (e.g. raw resources, which must be mined where the deposits occur on the map, and hence, at high volumes, must be mined at quite a few different locations simultaneously (this and the issues with throughput inefficiencies of running many trains to one place are what motivated me to invent the 'hub' system design, which is more or less what you are proposing here, though there is more to your concept than just a 'hub'-system. You can also use hubs for things that have multiple ways of being made (like oil products and U-235 and U-238) in order to create priority sourcing and ensure all users can draw from all production sources.

Yeah, hub would also have been a good term, at least less typing than entrepot .

I agree that it might not be useful for things like circuits or plates unless there is an advantage to making e.g. multiple circuit plants over one big circuit plant. I thought station capacity would be this advantage, but maybe I'm only moving the problem from the plant station to the hub/entrepot station.

As someone else pointed out, station throughput is pretty tightly limited by how many lines in and out of the station you have, a stacker exists mainly to prevent surplus from causing trains to back up into the main line, so you only need a stacker with enough space to hold all the trains serving the associated station(s) not counting the train(s) in said station(s). I'd wager that no more than 3 stations per entrance at the most would be useful (and I intend to go w/ just one per entrance for my own railworld, perhaps 2-3 at my ore hubs at most).

Yeah, that corresponds to my first results, although it feels like the fourth station is also still useful.

Doing away with stackers is doable if you have enough loading bays to match the number of trains serving that process, in general, it's probably easy to have 1-waiting bay per station which with a U-shaped setup really easily, and that allows 2x(# stations) trains to go there with no risk of backup onto the main line.

Yes, that should be possible for all hubs except the ore hubs, I think - for ore I generally want 2 trains per outpost (so 1 is always loading if demand is there - more if travel time is such that a train is fully loaded before the next one gets there. So, there the # of trains can get pretty big. But I am planning to have longer ore trains anyway, so I might as well have a different station design there which does use stackers.

I believe the 'waiting on filled station when another station is open' problem has been largely solved by some of the tweaks to train pathing that the devs have done for 0.15, but I can't say that I've actually put that to the test (because I haven't field tested it yet), so...try it and see?

It might be alleviated, but it's not solved yet:



The bottom train decided to path to station 3 (with full waiting bay) while station 1 had an empty bay. So, if you have 2xN trains they can apparently still block the main line.

[vanatteveldt]

here is what i mean about U shaped stops.
i've used the criss-cross method and it's slow.
especially if trains on the main line are waiting to enter.

This still has to connect to the mainline, right, presumably as a "siding" station?

I've tried testing that, but it deadlocks if it is saturated (because the main line acts as a loop):



I'll try to make a deadline free version by creating an extra line for trains leaving the station.