aaargha wrote:Tallinu wrote:I found that I had to double the length of the split tracks leading to the paired train deletion stops, otherwise the exit tracks were backing up and throttling the max throughput of the junction. Any train more than a car or two longer than the defaults might experience a similar issue, and a train longer than 14 cars would undoubtedly require even more space on the deletion stops.
I've been thinking of replacing them with instant stop stations but I don't think that the performance hit from all the extra path changing is worth it. For now, extending them is probably good enough.
Well, even the extended length I gave the deletion stations has been overwhelmed!
This upgraded spawner I've been working on now should be able to handle variable length trains, and automatically resets if traffic backs up and causes a spawned train to stop before the merge point. I've been testing it with 4-10-0 trains so far and it handles them beautifully, so well in fact that the amount of traffic put out across a straight left-to-right connection is more than the pair of deletion stops can handle, even though they're about 25% longer than the trains are. Traffic across the line becomes gradually slower, until trains leaving the spawner aren't even able to hit top speed at all, eventually causing the spawners to begin occasionally resetting just because traffic can't quite exit fast enough to keep up!
I've now tested it with 2-4-0 trains as well, on default length deletion tracks, and the same problem occurs.
So, those "instant stop stations"... How does that work?
I don't know for sure, but the only alternative I can see is to make three or four copies of each deletion stop instead of two and put them much,
much further down each of the forks so trains don't start slowing down until they have fully entered one fork...
Further Edit: I decided to find out exactly how many duplicate deletion stops would be required to handle saturated high-speed traffic without causing slowdowns, so I tore out one of the regular set, took the tracks about 50-60 tiles past the template area, and had each of the four tracks make a branching right angle turn into eight more 140 tile tracks, ending with the deletion stops. 2-4-0 trains being launched by my fast spawner appear to only require five of them... but no harm in leaving in extras, just in case!
Aaand a few more tweaks made, splitting the branches into several smaller blocks instead of a big one leading to all eight tracks. While it improved results, it didn't improve it as much as I expected. But I think I've figured out why.
Trains try to reserve longer sections of track as their speed increases, and when they can't reserve as much track as they want (due to a train ahead of them) they won't accelerate until they can. Or to be more precise, they enter braking mode until the block becomes free, then reserve it and, if that gives them enough extra room, they switch back to accelerating... but when blocks are small, this switching back and forth happens rapidly and they maintain approximately the same speed.
When the distance between a series of trains is more or less steady, the amount of track each train can reserve is also steady, meaning that each train cannot reserve any more track than it already is, which in turn means that they won't go any faster. No tailgating allowed!
It's like trains effectively
grow in length the faster they are traveling!
It also seems like the distance a train tries to reserve may not have a linear relation to its speed. I'm not sure how to go about accurately measuring this to determine the exact relationship. Wiring up an entire length of track with minimally spaced signals and measuring how many yellows it counts wouldn't be hard to set up, but it would be a little tricky to associate yellow counts with exact speeds except for the three top speeds granted by different fuel types. Finding a way to calculate the speed of a train based on the pulses of signals turning red or green might also be possible... Getting correct conversions between rate of signal pulses per, say, kilotick and the displayed top speed values would require some testing.
Oh, and some speed measurements from the straight-through spawner to deleter tests I've been running between the east and west stations...
Turning onto straight track right after the spawner/deleter side-swap grid: About 200kph
Passing first (opposing) counter, heading toward middle: About 225-230kph
Middle of map: 235kph
Passing far counters (and being counted): 255-260kph
They continue accelerating to top speed as they approach the branches to the deletion tracks. The moment the train ahead of them clears the part of the branches they want to use, their path is no longer blocked and they can reserve as much as they want.
It's reading about 263 TPM for eight lanes of 2-4-0 traffic doing this (all four lanes from each of two spawners). 4-10-0 is getting about 164 TPM. If it was a four-way instead of two-way test, that would be the equivalent of 526 TPM and 328 TPM, but without tunnels or bridges...
Not gonna happen.
Some more interesting results...
Train size, tiles physically occupied, tiles reserved at top speed (not counting whatever other tiles are part of the last block reserved), total track required at top speed:
2-4-0: 20-21, 58-59, 79
4-10-0: 48-49, 59-60, 108
Since the number of tiles reserved is virtually identical, the larger train makes far better use of the track by dramatically increasing the cargo to track ratio...