Measuring train throughput

[vanatteveldt]

It's quite easy to determine the maximum unloading speed of a station, and fairly easy to measure the total throughput of a station including time between trains. However, it's much more difficult to measure the maximum throughput of a "train link", ie the combination of unloading, loading, tracks, and #trains, especially since it depends on so many factors. For that reason, the aim of this post is not so much to give definitive statistics, but mostly to test out some configurations I'm actually using and to get an idea of the approximate numbers..

So, I wanted to make a test setup where I can answer questions like: how much copper will a single train be able to move to my circuit plant? How much can I do with 3 trains (with two unloading stops and a waiting bay)? Is it worth it to add more waiting bays, make trains longer, go to rocket fuel, research better braking pads, etc?

Assumptions:
- Working with LCCL double headers with full (24) stack unloaders and unlimited capacity on either side.
- Medium distance between loading / unloading (see screenshot below; about 600 tiles or so).

test setup

This is the unloading station, similar design (but with stacker) for the loading station.

A train activates the inserter inserting into the "memory chest", and the train leaving enables the inserter removing the train. The inserter is counted "pulse", i.e. 1 for each arriving train.

The leftmost decider is a counter (output=input) until it gets the reset signal from the left inserter (rotate=reset). The combinator can be toggled to emit yellow signal, which will pause the experiment by stopping the counter and setting a number of signals to red (to prevent unloading after the experiment is paused). The arithmetics are simply to convert ticks to minutes and seconds.

Map view of stations (with stacker for unloading, this is only added for final 2 tests). Loop is only used for manual moving around of trains. Tracks connect further to the right (see minimap above)

I count the time for 30 trains to arrive. I count from all trains empty in the loading bays (ready to depart) until the counter is at 30, ie train 30 enters the unloading bay. This is divided by the time to get trains per minute (tpm).

Outcomes: [in progress]

Single train
coal, stack inserter bonus 2, no brake pads: 64m41 / .46 tpm [trip time about 30 seconds, (un)loading about 40 seconds]
coal, full stack inserter bonus, no brake pads: 40m55 / .73tpm [(un)loading time about 12 seconds]
coal, all techs: 37m50 / 0.79 tpm [trip time about 25 seconds...]
solid fuel, all techs: 36m22 / 0.82 tpm [trips not much faster]
rocket fuel, all techs: 31m34 / .95tpm [trips about 20 seconds]

Conclusion for single-train improvements: stack inserter is hands down the best improvement. Solid fuel is probably not worth the hassle (I generally have coal lying around in a lot of places..). Under normal circumstances you can expect about 1 tpm at medium distances from a single train, or 8k plates/m.

Rocket fuel and braking improvements will be more worthwhile if the trains have to stop more often.

multiple trains [all full tech and rocket fuel]
2 trains, 2 unloading bays: 15m38 / 1.9 tpm
3 trains, 2 unloading bays, 1 waiting bay: 11m15 / 2.67 tpm

Conclusion: the second train+stop doubles throughput (as expected). Adding a third (which is easy in my station design because you get a 'free' waiting bay for each pair of identical stops) gives an extra 40% throughput as the trains start bumping into each other. Maximum throughput in this setup (without a stacker) is about 20k plates/m

6 trains, 2 unloading bays, 1 waiting bay plus stacker: 12m25 for 60 trains -> 4.8 tpm
12 trains, 2 unloading bays, 1 waiting bay plus stacker: 9m27 for 60 trains -> 6.3 tpm

(This is the maximum capacity for 2 unloading bays as trains were continuously waiting)

To test: longer trains (LCCCCL and LLCCCCLL) and single header trains (LCC, LCCCC and LLCCCC). But not tonight