[Retric]

Train speed is not the limiting factor. Most tunnels have very low utilization in terms of % of time with trains in them.

If you have 4x as many trains but don't load and unload on the same platform you could easily get by with 1/4th as many passengers per train.

[jcranmer]

Fun fact: a subway trunk line is more frequently physically occupied by a train than a highway lane is by a car.

The factors that limit train frequency are station dwell times and switching time. A subway line can generally hit 26TPH, and the top speed of most subways is usually about 70mph, with average speeds generally being in the realm of 30mph. Making trains faster actually reduces capacity; a HSR that goes 220mph is considered to have a capacity of around 4-6TPH. You can also improve throughput by cutting out all branching; Moscow gets about 40TPH as a result, which is about the feasible limit of rail systems.

[peatmoss]

I’ve been to Moscow twice in the early 2000s. The metro system there is still my mental benchmark for a metro system. The scale of everything is impressive. The number of people moved as well as the grand scale art. Say what you will about the soviet system, but you have to give them style points for one hell of a metro.

[Retric]

I am going to call BS on your train occupancy.

30 MPH which is low = 5280 * 30 = 158,400 feet per hour / 26 TPH = 6092 feet per train. Actual subway trains are 600 feet or less long ~= 10% utilization. At 30MPH cars don't keep 126 feet between each other. Bump that to 60MPH and the trains are at 5% or less utilization and again cars don't keep 266 feet between each other even if they should. And again this is very long 600 feet subway trains most are significantly shorter than that.

Further trains have a stopping distance @ 62MPH of 820 ft with (0.15 g) deceleration. High speed trains can get an extra 0.3 m/s2 deceleration which could also be added to normal subway trains but would be an emergency situation as they knock people over.

PS: At 60MPH the theoretical limit is over 220 trains per hour assuming all trains can stop before hitting the train in front of them. But you can only approach that with full automation and multiple lines for acceleration.

[spiralx]

There are 40 trains on the Victoria line at peak, each about 125m in length, which gives a total length of 5km. The length of the line is about 21km, so that's 5 / (2 * 21) = 12% occupancy.

[Retric]

Trains travel in each direction so 20 trains per direction or 6%, unless this is entirely single tracked.

We can also exclude any car stopped at a station, as you can have multiple cars unloading at the same station or have a train bypass a station without making a wider tunnel.

[spiralx]

I included both directions in my calculation, hence the denominator being (2 * 21km). And for the Underground there are no passing trains or multiple trains unloading at a station, platforms are generally only a few metres longer than the train itself.

[spiralx]

The Victoria Line in London runs at 36 trains per hour at peak, one every 100 seconds. Station dwell times are already a limiting factor at this point, and having separate load/unload platforms would provide only a small decrease in dwell time as you'd need to open the unload doors before the loading doors to prevent people just exiting the wrong side.

[Retric]

You can have multiple platforms and send trains to each not just separate load/unload platforms. This was very common with passenger trains as boarding times could extend for long periods. Even when it was a single line for loading and unloading they would normally have a bypass line to avoid the station.

[spiralx]

> You can have multiple platforms and send trains to each not just separate load/unload platforms.

Sure, but we're talking about subway networks underneath major metropolises, and adding extra tunnels and platforms is basically as expensive as digging a whole new line - when it is even possible due to space or geological constraints.