One question that comes up all the time when we talk about subway capacity is the number of trains we can run per hour. There are several important components to this question:
- Terminal design. How long does it take to turn trains around and how often can we dispatch them?
- Signal design at major stations. How soon after a train leaves a station can the next one pull in, load and be ready to depart?
- Signal design along the line. Are there locations where signals keep trains further apart than they normally need to be for safe operation?
The most important component, and the hardest to change, is the terminal design. The TTC has plans to replace the signal system on the Yonge-University line over the next 10 years, and this would eliminate some of the bottlenecks. Terminals, however, are a major problem on both lines.
Assuming that we are running as frequent a service as possible, the starting point for any operational estimate would be to have a train on both terminal tracks plus one at or close to the station within the approach distance of the signal system (i.e. within the area where the signals will bring the train slowly up to the crossover). For practical purposes, we must assume that an incoming train would be stopped at the crossover before it received clearance to enter the station.
The complete cycle for one train’s departure and arrival is this:
- Signal turns green
- Train guard initiates door closing and this completes
- Train moves off from platform and eventually clears the crossover
- Signal system determines that the crossover is clear and realigns the switches
- Signal system displays clear for the incoming train
- Incoming train starts up and crosses into the station
- Signal system determines that the crossover is clear and realigns the switches
At this point we are back to the beginning of the cycle. Although there are minor variations in timing depending on whether a train crosses over or runs straight in or out of a platform, this cycle takes around two minutes to complete at a typical terminal. (I used Kennedy for my samples because I ride through it every day.) In practice, if we attempt to schedule a two minute headway, this will quickly bog down because of minor delays including operators who are not ready to leave when their signal clears, ad hoc service adjustments by the tower and passengers who rush the doors at the last moment.
Both the Bloor-Danforth and Yonge-University lines run headways of about 140 seconds, and this gives some elbow room at the terminals. Newer terminals like Finch, Downsview, Don Mills and Sheppard-Yonge pose special problems because the crossovers are so long. This stretches the time a crossover is occupied and lengthens the total cycle. Of course, on Sheppard it does not matter because the trains run every 330 seconds.
Reducing the headway to 120 seconds on Yonge or Bloor will be difficult and may result in more erratic service than we have now at 140 seconds. It will certainly cause regular backlogs of trains at terminals.
(Terminal delays were common even at 140 seconds until the TTC moved to step-back crewing where an incoming train’s crew “steps back” one or two trains and so is ready to leave as soon as possible. Without this scheme, the time needed for an outbound crew to walk two carlengths back to their inbound positions, not to mention possible calls of nature, can cause major delays at terminals.)
90 second headways have been achieved on some systems, but these systems were built from day one for that type of operation. This is not the case in Toronto, and moving below a 120 second headway would be almost impossible.
At best, we can operate 30 trains per hour (current operations are about 25 per hour) with a design capacity of 1,000 and a crush capacity of around 1,200 each. This translates to 36,000 crush at minimum headway, or 25,000 as a design load at current headways. One important point: these are average figures over an hour, and short bursts can push the values up by about ten percent for brief periods.
Any discussion of the future of subway operations and surplus capacity needs to take these figures to heart. The subway is not full, but there is a limit to what it can handle. The closer we approach that limit, the less reliable the service will become.
It would probably be more efficient to use the tail tracks behind the station, rather than the crossover ahead of the station, to change tracks. This is what most systems with short headways, such as the Paris Metro, do. This would require that crews be in position at both ends of the train while the maneuver is being performed. However, it would allow trains to enter and leave the station without delay. Unfortunately, this is only possible at Finch and Sheppard-Yonge. Hopefully, the York University extension will incorporate a pocket track at its terminus so that this can be done at both ends of the YUS line.
Steve: According to the service design in the Environmental Assessment, half of the service will turn back at a new high-speed pocket track beyond Downsview Station. This means that the headway and turnaround demands at either terminal (Downsview pocket or York U) will be less stringent. Even so, for any far-side turnaround, it would be preferable if the crew boarded before the train went beyond the terminal station. This avoids the need for any special crew provisions in the tail or pocket track areas.
Has the TTC looked at operating express trains? Skipping stations such as Woodbine, Donlands, or Chester. Stations like Chester have no feeder bus service and so are less busy. Could this approach help to increase capacity?
Steve: The problem is at the terminals where there is a physical constraint on how often a train can leave. Whether it runs local or express, the headway and capacity stay the same.
Also, there are really not that many “light” stations to skip. By the way, there are feeder routes at Woodbine and Donlands.
The layout at the terminals is wrong in most cases. Both tracks should be extended at least one train length beyond the terminal station platform. It should go like this: Train 1 arrives at terminal station and unloads then, it leaves just like any other station and proceeds over a crossover to a dead end stub track. Here it reverses direction and moves to the outbound platform. Meanwhile, Train 2 has arrived and unloaded. It then proceeds straight ahead and stops on another dead end piece of track where it waits for Train 1 to make its move. Train 3 has already arrived and is unloading. When Train 2 makes its move this clears the way for Train 3 to move repeating the cycle begun by train 1. Correcting this layout will be a problem in many cases but, it may well be the answer as it provides flexibility.
If tracks were extended two train lengths beyond it would provide a place to store out of service trains due to breakdown or just to position a train for later addition to the schedule especially if a gap occurs.
Am I right in assuming that these design issues are the cause for the “normal” 10+ minutes of stop-start it takes to get from Wilson to Downsview from about 10am onwards on weekdays? Or is there another problem?
My experience over the last 6 months (Oct 2010-March 2011) is that this travel time from Wilson to Downsview appears to be “normal” operations after the morning rush hour (can’t comment on evening rush hour).
Strangely during the morning rush hour trains seem to be able to make the Wilson to Downsview journey in their scheduled 2-3 minutes.
Steve: This problem occurs at other terminals after the AM peak from a combination of two factors. First, the trains are still on the peak period running times, but do not need all of this time to actually reach the terminal. Second, depending on the pattern of trains dispatched from the terminal staying in service and those going back to the yard, the headway can actually get longer between trains that were close together. Combines these, and you get a queue of trains waiting to reach the terminal.
A similar problem used to beset the BD line and there would be a queue back at least to Warden, sometimes to Victoria Park. It was fixed, somewhat, by running the same number of trains closer together, shortening the headway and the round trip time. This didn’t completely get rid of the queues, but they were a lot shorter. The problem tends not to happen in the PM peak because by then most of the trains are late, and they are not held for their scheduled time at the terminals. This is effectively the same as shortening the running time.
As the TTC tries for shorter and shorter headways, this will become a bigger and bigger problem. On a bus route, you can pad the schedules and store vehicles at terminals that are waiting for their scheduled departure. You can’t do this on a subway line.