Efficiency of Roundabouts vs Crossings for T-Junctions
Posted: Thu Mar 31, 2016 1:40 pm
I've heard this has been contentious, and it was something I wanted to measure.
tl;dr: A single T-junction seems to be 13% more efficient than a roundabout (283 trips vs 251 in my testing.)
First off, tremendous thanks to viewtopic.php?f=18&t=18621 for making me realize my roundabout structure was wrong. (if you try to build a roundabout going into a 6-wide straight lines which is the "obvious" way with a circle, you end up with trains that have to go all the way around the loop to make the shortest turn).
So I set up a test scenario and have spent a couple of days now turning the counting and making sure my signal spacing are "optimal" at least as well as I can make them. I saved some blueprints of my circuits in case anyone wants to critique them. I can't actually fit them into a single screenshot, but I can make a couple of screenshots to show the basic layout.
Essentially, I create a T-junction, with each direction having 3 same-named stations with a loop back on the end (North, South, West). I then set up 2 trains going N-S, 2 going N-W, 2 going S-W, and then 2 going N-S-W and 2 going N-W-S. Essentially 2 trains going each pairing, and 2 going clockwise, 2 counter-clockwise. 10 trains, 9 stations, all utilizing the same T intersection.
The blueprints have a fair amount of unessential items (extra roboports and electrical stations), just because it is hard to carve out exactly what you need, but they should be functional, just add trains and materials for counting)
I've tweaked the signal positions quite a few times, it is pretty easy to end up missing a signal and throughput on a given line goes way down. (Easy to make it so a train sitting in one station accidentally blocks a train wanting to go to a different station.) So it is possible that the measurement isn't due to roundabouts vs crossing junctions, but just because I have a signal wrong. But I have spent time trying to make them as efficient as I could. (or some detail of exactly the ordering my trains are running).
I then run it for 10 minutes (putting a storage tank into a logistic storage chest and then removing it to start and stop the counters). I then counted the how many times a train from one station made it to another station, and ended up with this mapping:
I don't fully understand why some directions are skewed on the Roundabout but the T junction is very even. Possibly because the "average" length of some turns is different.
Note that I'm running 1-2-0 trains, and length would certainly have an effect, as it changes amount of time in the intersection.
I definitely did see at least 2 trains going through the intersection at the same time (2 trains turning left in left-handed traffic W=>N, and S=>W). I don't think I've seen 3 though certainly the Crossing intersection should be capable of it, I'm not sure if the signalling on the Roundabout would allow it. It should be possible, but I'm not an expert.
So ultimately, a roundabout isn't much slower than a regular crossing junction, though if you ever wanted to scale it from 2-parallel tracks to 4/6/8 parallel tracks the Crossing version probably starts to have a big advantage.
tl;dr: A single T-junction seems to be 13% more efficient than a roundabout (283 trips vs 251 in my testing.)
First off, tremendous thanks to viewtopic.php?f=18&t=18621 for making me realize my roundabout structure was wrong. (if you try to build a roundabout going into a 6-wide straight lines which is the "obvious" way with a circle, you end up with trains that have to go all the way around the loop to make the shortest turn).
So I set up a test scenario and have spent a couple of days now turning the counting and making sure my signal spacing are "optimal" at least as well as I can make them. I saved some blueprints of my circuits in case anyone wants to critique them. I can't actually fit them into a single screenshot, but I can make a couple of screenshots to show the basic layout.
Essentially, I create a T-junction, with each direction having 3 same-named stations with a loop back on the end (North, South, West). I then set up 2 trains going N-S, 2 going N-W, 2 going S-W, and then 2 going N-S-W and 2 going N-W-S. Essentially 2 trains going each pairing, and 2 going clockwise, 2 counter-clockwise. 10 trains, 9 stations, all utilizing the same T intersection.
The blueprints have a fair amount of unessential items (extra roboports and electrical stations), just because it is hard to carve out exactly what you need, but they should be functional, just add trains and materials for counting)
I've tweaked the signal positions quite a few times, it is pretty easy to end up missing a signal and throughput on a given line goes way down. (Easy to make it so a train sitting in one station accidentally blocks a train wanting to go to a different station.) So it is possible that the measurement isn't due to roundabouts vs crossing junctions, but just because I have a signal wrong. But I have spent time trying to make them as efficient as I could. (or some detail of exactly the ordering my trains are running).
I then run it for 10 minutes (putting a storage tank into a logistic storage chest and then removing it to start and stop the counters). I then counted the how many times a train from one station made it to another station, and ended up with this mapping:
Code: Select all
Crossing Source Roundabout Source
N S W N S W
N 0 47 47 N 0 39 36
S 47 0 47 S 40 0 49
W 48 47 0 W 36 51 0
Total 283 251
Note that I'm running 1-2-0 trains, and length would certainly have an effect, as it changes amount of time in the intersection.
I definitely did see at least 2 trains going through the intersection at the same time (2 trains turning left in left-handed traffic W=>N, and S=>W). I don't think I've seen 3 though certainly the Crossing intersection should be capable of it, I'm not sure if the signalling on the Roundabout would allow it. It should be possible, but I'm not an expert.
So ultimately, a roundabout isn't much slower than a regular crossing junction, though if you ever wanted to scale it from 2-parallel tracks to 4/6/8 parallel tracks the Crossing version probably starts to have a big advantage.