- 3 way.png (1015.96 KiB) Viewed 2906 times
- 3 way 2-4-0 trains.png (40.61 KiB) Viewed 2906 times
3 and 4 way intersections
Re: 3 and 4 way intersections
Three way, LHT, one plane, spacing 4. Less fancy than previous wiggly idea, it returns slightly better results with careful signal placement.
Last edited by Nerby on Thu Oct 31, 2024 2:04 pm, edited 1 time in total.
Re: 3 and 4 way intersections
I tend to use zero, one, or two rail spacing. Three seems exorbitant, and more is nonsense. But I see y'all love your three, four, more rail spacing, so I took this detour in my quest for minimal size intersection. This design started with the question "what if left and right turns both happened inside the intersection?", and that dictated the spacing.
80x50 footprint (new personal best for me)
22(!) rail spacing
Score 102.65 with default test settings (2-4-0 trains on nuclear fuel)
I'd call this an inverted cloverleaf.
80x50 footprint (new personal best for me)
22(!) rail spacing
Score 102.65 with default test settings (2-4-0 trains on nuclear fuel)
I'd call this an inverted cloverleaf.
- 10-31-2024, 08-57-15.png (705.58 KiB) Viewed 2897 times
Last edited by sparr on Thu Oct 31, 2024 2:25 pm, edited 1 time in total.
Re: 3 and 4 way intersections
68x68, 12-rail spacing.
N/S is LHD, E/W is RHD.
Score 99.57
Size is dictated by the elevated left turns.
N/S is LHD, E/W is RHD.
Score 99.57
Size is dictated by the elevated left turns.
- 10-31-2024, 09-58-37.png (416.46 KiB) Viewed 2881 times
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
So many awesome intersections! Man they look nice=) Added them all that are normal 4 ways and 3 ways=)
Re: 3 and 4 way intersections
Welp. Can't afford games right now, and was planning not to spend time on Factorio until I could get SA, but... Can't do it. Here I am redesigning my rail line blueprints for the new 2.0 rails and longer big pole wire length, and then I thought, "hey, what's going on in this thread?"
I know two-level interchanges are always going to beat those with crossings due to relying only on splits and merges, but as others have noted, there are places even in SA when the entire rail system needs to be suspended in the air, so single-level still has a place. Not that anyone is likely to build fully-elevated max-throughput interchanges, but...
*Cracks knuckles*
I guess it's time to at least fix my MC blueprints to use the 2.0 rail system, and maybe find some aesthetic improvements while I'm at it!
BTW, did anyone else find that their entire blueprint library had all the diagonal and curved rails corrupted after the 2.0 update? Re-importing things that I have blueprint strings for works, and rails on the ground in my saves are still correct, but anything rail-related (except for stuff with only/mostly straight track) that existed solely in my blueprint library is effectively kaput... I was not expecting that.
Last edited by Tallinu on Thu Oct 31, 2024 10:58 pm, edited 1 time in total.
-
coppercoil
- Filter Inserter
- Posts: 503
- Joined: Tue Jun 26, 2018 10:14 am
- Contact:
Re: 3 and 4 way intersections
I designed an intersection that allows U-turns without a roundabout. I need the U-turns for a character train that runs here and there.
For the scoring, I run Testbench Forum Test that does not include U turns as they are not optimized and should be used rarely, though intersection was also tested for U turns with Everything Test.
Lilac+U
For the scoring, I run Testbench Forum Test that does not include U turns as they are not optimized and should be used rarely, though intersection was also tested for U turns with Everything Test.
Lilac+U
- Lilac_U.png (1.15 MiB) Viewed 1774 times
Last edited by coppercoil on Mon Nov 11, 2024 11:10 pm, edited 1 time in total.
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
Looking forward to your designs=)Tallinu wrote: Thu Oct 31, 2024 10:20 pmWelp. Can't afford games right now, and was planning not to spend time on Factorio until I could get SA, but... Can't do it. Here I am redesigning my rail line blueprints for the new 2.0 rails and longer big pole wire length, and then I thought, "hey, what's going on in this thread?"
I know two-level interchanges are always going to beat those with crossings due to relying only on splits and merges, but as others have noted, there are places even in SA when the entire rail system needs to be suspended in the air, so single-level still has a place. Not that anyone is likely to build fully-elevated max-throughput interchanges, but...
*Cracks knuckles*
I guess it's time to at least fix my MC blueprints to use the 2.0 rail system, and maybe find some aesthetic improvements while I'm at it!
BTW, did anyone else find that their entire blueprint library had all the diagonal and curved rails corrupted after the 2.0 update? Re-importing things that I have blueprint strings for works, and rails on the ground in my saves are still correct, but anything rail-related (except for stuff with only/mostly straight track) that existed solely in my blueprint library is effectively kaput... I was not expecting that.
There are entirely new rails and all 1.1 rails wont work in 2.0. So all old rail blueprint are useless.
Not everybody playing factorio will buy SA, so having some great intersections for those would be good!
Re: 3 and 4 way intersections
You can still copy and paste "old rail" if you really have to tweak an existing setup.
Horizontal and vertical rails still match up, but diagonal old rail absolutely will not connect to new rail without first pasting an old curve to make it straight and then transitioning from that to new rail. (The rail planner and copy/pasting are both equally incapable of connecting new to old diagonals.)
I'm still curious if old blueprints getting everything but straight rail corrupted is universal, common, rare, etc. I don't remember seeing any warning about that -- I specifically remember reading that old rail wouldn't just vanish on us and we'd basically be able to modify things to use the new rails at our leisure. It would've been nice to have something to work from when recreating blueprints with new rail, without having to roll back the version to export blueprint strings or search through old saves for the thing I made the BP from (if it even exists in them).
And it looks like I missed something when I said that importing the old BP strings still worked fine. Rails and signals actually end up weirdly offset with some signals overlapping neighboring tracks as a result. It looks like it will be possible to paste it down in two places, decon all the rail ghosts in one place, decon all the signals in the other, and then merge the signals back together with the rails... But I might just resignal everything from scratch anyway. So much extra work, ugh...
Edit: Initial results of redesign! Highly WIP of course.
Each arm has gone from 15 track-widths across to 11. I was able to eliminate the huge waggle on the left-turn entrance lane while maintaining manual driveability, AND making everything look loads nicer:
Horizontal and vertical rails still match up, but diagonal old rail absolutely will not connect to new rail without first pasting an old curve to make it straight and then transitioning from that to new rail. (The rail planner and copy/pasting are both equally incapable of connecting new to old diagonals.)
I'm still curious if old blueprints getting everything but straight rail corrupted is universal, common, rare, etc. I don't remember seeing any warning about that -- I specifically remember reading that old rail wouldn't just vanish on us and we'd basically be able to modify things to use the new rails at our leisure. It would've been nice to have something to work from when recreating blueprints with new rail, without having to roll back the version to export blueprint strings or search through old saves for the thing I made the BP from (if it even exists in them).
And it looks like I missed something when I said that importing the old BP strings still worked fine. Rails and signals actually end up weirdly offset with some signals overlapping neighboring tracks as a result. It looks like it will be possible to paste it down in two places, decon all the rail ghosts in one place, decon all the signals in the other, and then merge the signals back together with the rails... But I might just resignal everything from scratch anyway. So much extra work, ugh...
Edit: Initial results of redesign! Highly WIP of course.
Each arm has gone from 15 track-widths across to 11. I was able to eliminate the huge waggle on the left-turn entrance lane while maintaining manual driveability, AND making everything look loads nicer:
Pictures
It's... so smooth... and pretty... 
Re: 3 and 4 way intersections
64x64, 4-way, 2-lane, wide spacing, offset entrance, LHD, 4-fold rotation symmetry, intersection
It's not the prettiest child but it has no crossover and can handle a little under 94 (on S1) trains per minute.
It's a bit signal spamy. It has wide 20-tile spacing and an offset entrance which I'm not a fan of but it is very small because of that.
A RHD variant could be easily be made by removing the signals and placing a similar signals on the opposite side of the track. (they are all normal signals no need for chains) if anyone wants me to post this please let me know
Edit: I've been told S1 is the important number and S2/3 are only important if they are lower
It's not the prettiest child but it has no crossover and can handle a little under 94 (on S1) trains per minute.
It's a bit signal spamy. It has wide 20-tile spacing and an offset entrance which I'm not a fan of but it is very small because of that.
A RHD variant could be easily be made by removing the signals and placing a similar signals on the opposite side of the track. (they are all normal signals no need for chains) if anyone wants me to post this please let me know
- 11-01-2024, 14-03-52.png (532.56 KiB) Viewed 2696 times
Edit: I've been told S1 is the important number and S2/3 are only important if they are lower
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
Nice=) Added it, if you want another name just tell me!ella wrote: Fri Nov 01, 2024 2:11 pm 64x64, 4-way, 2-lane, wide spacing, offset entrance, LHD, 4-fold rotation symmetry, intersection
It's not the prettiest child but it has no crossover and can handle a little under 94 (on S1) trains per minute.
It's a bit signal spamy. It has wide 20-tile spacing and an offset entrance which I'm not a fan of but it is very small because of that.
A RHD variant could be easily be made by removing the signals and placing a similar signals on the opposite side of the track. (they are all normal signals no need for chains) if anyone wants me to post this please let me know
11-01-2024, 14-03-52.png
Edit: I've been told S1 is the important number and S2/3 are only important if they are lower
Re: 3 and 4 way intersections
MultiCross 2.0!
Standard 6-car LHT and RHT versions tested with defaults get 100TPM in set 1 and 85 overall.
Expanded 6-car versions bump that up to 107 and 98.
This is all just two-lane blueprints for now, since those have always seemed to be the most practical of an already very large style of interchange. Though, like I said, this version is noticeably slimmer than its predecessor, thanks to the rail rework.
I've made "standard" blueprints for up to 18 car-length and "expanded" for up to 12. These already get pretty unwieldy in length, but you can expand the buffers with some judicious copy-pasting if you really want to. (Longer trains get lower TPM scores, of course, but potentially more cargo delivered, depending on exact details of things like locomotive to wagon ratio, etc. For example, going from the Standard 6-car to 8-car blueprint and from 2-4-0 to 2-6-0 trains lowers the TPM by about 20%, but each train could theoretically deliver 50% more, which works out to 20% more cargo.)
Find them all here: https://drive.google.com/drive/folders/ ... RQD_SExbhW (It's also in my signature)
One thing that surprised me is that when I mixed standard with a couple of the unbuffered "stub" variants, like you might do on a high-traffic main line with low-traffic outposts on either side, it only lost about 25% of the TPM -- and judging by the stats and observation of train behavior, the bulk of that appears to be due to trains attempting to enter one stub, go straight through, and exit from the other! Since this involves an unbroken chain of chain signals all the way through, it's blocked by almost any other kind of traffic, so straight-through trains sit thereforever a long time. (But heavy stub-to-stub traffic is also something you shouldn't get in normal use, as that's not what they're designed for.)
(Another surprise was that assembling an interchange out of four stubs, while pointless as there are better compact interchanges, still achieved 49TPM set 1 and overall 42.85.)
Standard 6-car LHT and RHT versions tested with defaults get 100TPM in set 1 and 85 overall.
Expanded 6-car versions bump that up to 107 and 98.
This is all just two-lane blueprints for now, since those have always seemed to be the most practical of an already very large style of interchange. Though, like I said, this version is noticeably slimmer than its predecessor, thanks to the rail rework.
I've made "standard" blueprints for up to 18 car-length and "expanded" for up to 12. These already get pretty unwieldy in length, but you can expand the buffers with some judicious copy-pasting if you really want to. (Longer trains get lower TPM scores, of course, but potentially more cargo delivered, depending on exact details of things like locomotive to wagon ratio, etc. For example, going from the Standard 6-car to 8-car blueprint and from 2-4-0 to 2-6-0 trains lowers the TPM by about 20%, but each train could theoretically deliver 50% more, which works out to 20% more cargo.)
Find them all here: https://drive.google.com/drive/folders/ ... RQD_SExbhW (It's also in my signature)
- 20241101193113_1.jpg (308.19 KiB) Viewed 2063 times
One thing that surprised me is that when I mixed standard with a couple of the unbuffered "stub" variants, like you might do on a high-traffic main line with low-traffic outposts on either side, it only lost about 25% of the TPM -- and judging by the stats and observation of train behavior, the bulk of that appears to be due to trains attempting to enter one stub, go straight through, and exit from the other! Since this involves an unbroken chain of chain signals all the way through, it's blocked by almost any other kind of traffic, so straight-through trains sit there
Mixed with Stubs
(Another surprise was that assembling an interchange out of four stubs, while pointless as there are better compact interchanges, still achieved 49TPM set 1 and overall 42.85.)
Stubby
Last edited by Tallinu on Tue Nov 05, 2024 6:09 pm, edited 2 times in total.
Re: 3 and 4 way intersections
This is my first attempt at submitting an intersection. Hopefully I did it right
What is not obvious from the design, is that the outer lanes on each side would be the enter/exit to a station at a local facility. Also, the main rails are elevated in my factory so the outer ramps are not on the center 2 lanes.
This is rotationally symmetrical, no crossings, and only 1 or 2 merges per path no matter the exit direction. Direct crossings through intersection are straight for max throughput. Only downside, is the Left turn takes a while to go all the way around.
Entire intersection fits in a 4x4 grid (outer ramps and exit lanes excluded)
For the test bench, I had to remove all the signals for the other LHD on each lane, or the trains sometimes wouldn't move.
You can name it whatever you like
- 11-02-2024, 10-45-24.png (2.14 MiB) Viewed 2510 times
This is rotationally symmetrical, no crossings, and only 1 or 2 merges per path no matter the exit direction. Direct crossings through intersection are straight for max throughput. Only downside, is the Left turn takes a while to go all the way around.
Entire intersection fits in a 4x4 grid (outer ramps and exit lanes excluded)
- 11-02-2024, 10-51-53.png (63.78 KiB) Viewed 2510 times
- 11-02-2024, 10-52-11.png (75.83 KiB) Viewed 2510 times
You can name it whatever you like
-
jamsessionein
- Manual Inserter
- Posts: 2
- Joined: Sat Nov 02, 2024 9:13 pm
- Contact:
Re: 3 and 4 way intersections
Silly question - as I'm using the test bench here, I'm noticing a significant drop in throughput if I use 1-4-1 trains (both locomotives pointing the same direction) instead of 2-4 trains, but I was given to believe that they were functionally identical. Did something change, or did I misunderstand the base premise?
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
How can you make the trains face the same direction? i dont know any way on the bench to do that, maybe on the old savefile version. But yes they should be identical. But you are probably doing 1-4-1 where the back is facing backwards.jamsessionein wrote: Sat Nov 02, 2024 9:14 pm Silly question - as I'm using the test bench here, I'm noticing a significant drop in throughput if I use 1-4-1 trains (both locomotives pointing the same direction) instead of 2-4 trains, but I was given to believe that they were functionally identical. Did something change, or did I misunderstand the base premise?
-
jamsessionein
- Manual Inserter
- Posts: 2
- Joined: Sat Nov 02, 2024 9:13 pm
- Contact:
Re: 3 and 4 way intersections
Ah, of course, that makes more sense. I don't know why I assumed the rear locomotive would be facing forward!
-
train_of_fish
- Manual Inserter
- Posts: 1
- Joined: Mon Mar 07, 2016 7:48 pm
- Contact:
Re: 3 and 4 way intersections
Tried to make a clean design, ended up with an 8-space 4x4 chunk with a respectable 100tpm
https://factoriobin.com/post/o95jsw
https://factoriobin.com/post/o95jsw
- 20241104205238_1.jpg (1.35 MiB) Viewed 2144 times
Re: 3 and 4 way intersections
I just realized I missed 1 rail on each corner of Star Loop, just before the down ramp of the inner lane right-turn.
However, after removing them, the testbench score gets around 2 points higher, which is extremely weird as that extra rail does nothing, and shouldn't interfere with the train pathing. Is that a bug or did I miss something?
However, after removing them, the testbench score gets around 2 points higher, which is extremely weird as that extra rail does nothing, and shouldn't interfere with the train pathing. Is that a bug or did I miss something?
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
176.60akulen wrote: Tue Nov 05, 2024 5:40 am I just realized I missed 1 rail on each corner of Star Loop, just before the down ramp of the inner lane right-turn.
However, after removing them, the testbench score gets around 2 points higher, which is extremely weird as that extra rail does nothing, and shouldn't interfere with the train pathing. Is that a bug or did I miss something?
176.6
4
RHT
35
90 min
2-4
4
4
TRUE
nuclear-fuel
Star Loop
https://factoriobin.com/post/vj9dgs
akulen
https://i.imgur.com/ZpKFoRF.png
191 x 191
There is a margin of error. 176-178 isnt too bad. If you want it more accuratly you have to run longer test amd 1000 min tests are the only one that i consider accurate. 90 min isn't good enough to seperate 4 lane intesections if they are close. You can try with multiple runs if you want to control the margin of error
Re: 3 and 4 way intersections
I forgot to include direct blueprint links for you to test/post. And here's one for the Expanded variant as well, since it scores so differently.
MultiCross
Multicross Expanded
-
Factoriointersection
- Long Handed Inserter
- Posts: 59
- Joined: Sat Sep 25, 2021 4:58 pm
- Contact:
Re: 3 and 4 way intersections
Yeah=) I was a bit unsure which to add aswell. Do you think I should add all variants to the set up? Or just link the book? I can test all and add themTallinu wrote: Tue Nov 05, 2024 8:41 pm I forgot to include direct blueprint links for you to test/post. And here's one for the Expanded variant as well, since it scores so differently.
MultiCross
Multicross Expanded
