I’m pretty sure there are short overlaps all over the system because that’s the way the signal system was originally designed. What makes you think “it’s only a small percentage of signals on some lines”? Do you know something I don’t (which is very possible)?

The issue is due to three factors. One is instances where the signals are spaced more closely than usual. Sometimes this is done to increase track capacity. Other times it’s just a vestige of an era when equipment performance was lower, hence the older trains couldn’t reach speeds exceeding the limits of the signal system but the new ones can. The latter would actually be mostly a problem on the IRT. Most of the IND was built post WWII, when trains were only marginally slower, or the same as now. Also, the IND was built with faster running in mind compared to the IRT, so signals were spaced further apart. In any case, there aren’t that many places where the signals are problem. Remember until the WB collision we were running unneutered trains system wide without any issues. If there were so many places where the signal system was flawed, we would have had a lot more collisions.

The second cause when trains were overhauled in the late 1980s through early 1990s. 100 HP motors were replaced with 115 HP motors. This increased the acceleration capabilities of the trains slightly. In most cases the signal system could still cope with it because signal systems have margins of safety built in. In a few instances, like on the WB, it couldn’t.

The third cause was changing the composition of the brake shoes. This reduced emergency braking capabilities slightly.

In the end, it was the unhappy collision of all three factors at the same time which resulted in the WB accident. Could such a thing happen in other parts of the system? Certainly it could. However, a proper solution might have been to see where the problems were, then if need be install timers to limit train speeds only in those areas. Still not an ideal solution, but at least you would only have a slowdown in a fraction of the system, not everywhere.

Grade timers operate 24/7. The MTA doesn’t really have the option to turn them off when there aren’t many trains.

I’m really can’t see the rhyme or reason behind many of NYCTA’s timers.

Neither can I which is why I’m complaining about them. If they were only installed in places where the trains really need to slow for curves, then that’s fine. This doesn’t seem to be the case. Also, as I mentioned elsewhere, these timers often force trains into the range of speeds where there is greater oscillation, increasing track wear, plus detracting from passenger comfort. They largely make no sense. No other system uses them to the extent the MTA does.

One is instances where the signals are spaced more closely than usual. Sometimes this is done to increase track capacity.

It has never been the practice to space signals more closely than is safe. So that’s not really a factor.

Other times it’s just a vestige of an era when equipment performance was lower, hence the older trains couldn’t reach speeds exceeding the limits of the signal system but the new ones can.

This is the main issue.

The latter would actually be mostly a problem on the IRT. Most of the IND was built post WWII, when trains were only marginally slower, or the same as now. Also, the IND was built with faster running in mind compared to the IRT, so signals were spaced further apart.

Your discussion here omits the BMT, on which the 1995 collision occurred. In terms of signal overlap planning and standards, the BMT probably more closely resembles the IRT.

In any case, there aren’t that many places where the signals are problem.

How do you know this, I ask again?

Remember until the WB collision we were running unneutered trains system wide without any issues. If there were so many places where the signal system was flawed, we would have had a lot more collisions.

Of course that isn’t necessarily true, because the other barriers that are in place to prevent collisions (procedures, employee training, etc.) are often effective even when the signal systems may not be 100% effective.

The second cause when trains were overhauled in the late 1980s through early 1990s. 100 HP motors were replaced with 115 HP motors. This increased the acceleration capabilities of the trains slightly. In most cases the signal system could still cope with it because signal systems have margins of safety built in. In a few instances, like on the WB, it couldn’t.

“In most cases”? What’s your source for this?

In the end, it was the unhappy collision of all three factors at the same time which resulted in the WB accident. Could such a things happen in other parts of the system? Certainly it could. However, a proper solution might have been to see where the problems were, then if need be install timers to limit train speeds only in those areas. Still not an ideal solution, but at least you would only have a slowdown in a fraction of the system, not everywhere.

I think there may be many more portions of the system where this is a potential problem than you are willing to acknowledge.

I don’t like the fact that the trains are so much slower today than they used to be, but I applaud the MTA for putting safety ahead of speed.

This article mentions poor manners being a cause of subway delays:

http://nypost.com/2015/02/24/overcrowding-on-subways-leads-to-massive-spike-in-delays/

Yes, off-peak ridership appears to have increased more rapidly than peak. But I’m not sure the difference is huge. I’ve just calculated the “Gini Coefficient” — a measure of peakiness (vs. flatness) in a distribution (originally developed to characterize economic inequality, as you know) — for 2012 subway trips to and from the CBD, and same for 2007. The coefficient fell (indicating greater flatness), but only by 1.5% — from 0.4433 for 2007, to 0.4365 for 2012. That’s noteworthy but not huge.

My source for 7.5 more trains an hour is the RPA report I linked to in my sixth graf.

Perhaps my estimate of a one-third increase in capacity from system-wide CBTC overstated the case. At least I’ve taken a stab at it. I’m convinced that unless we can “concretize” the value of funding this and future MTA capital plans, we’re going to end up with precious little.

A second possible solution is to spread out the load through more of the day by giving employers incentives to allow employees to keep something other than 8 to 4 or 9 to 5 hours (also give incentives for 3 or 4 longer days instead of 5 8-hour days). Furthermore, incentives can be given for telecommuting. The idea to attack this on the demand side, rather than the supply side, makes more sense to me. It’s very expensive to add subway capacity. However, right now we have a lot of potential capacity off-peak which can be had for little cost. That and decreasing the number of trips via either telecommuting, or 3 or 4 day work weeks, seem like eminently sensible ideas.

Just a single year’s growth, at 2.1 percent, amounts to 113,000 rides during that 15-hour peak.
…
total gain in capacity reaches 55,000 — enough to handle a mere six months’ worth of ridership growth.

Isn’t this only true if all the increase is at the peak load point? I assume the MTA has data on peak loading throughout the day for each line, would they care to share it?

The potential downside here might be to lower the maximum track capacity. However, my understanding here is in much of the system the capacity is limited by how fast you can turn trains at terminals, not by the signaling system.

That’s currently true precisely because the signaling system along the intermediate stretches does not limit capacity. If it were rewired to do so — itself an undertaking that is fraught with danger — there’s a good chance it would become a major limiting factor as well.

Indirectly it is. The problem is the trains are capable in some cases of going too fast to stop in the required amount of space if they pass a stop signal. That in turn is caused by T/Os keeping the throttle wide open in response to a clear signal.

To the extent that that may be a problem, that is a signal sighting issue that requires correction in any case. A T/O should always, always, always be able to see a red signal in time to stop without passing it; if that’s not the case, that’s a severe defect.

Here’s another article which mentioned 650 problem areas:

Are you trying to tell me that “650 problem areas” equals only a “small” part of the system?

However, with those speed limits in effect, why the need to neuter equipment such that it kills both acceleration and top speed?

They don’t trust T/Os.

You forgot the Part about the consultant being the commisioner’s Son-in-law 🙂

The facts are: there are many hours on many lines when trains can’t accommodate 100 percent of people who want to board

The facts (even just focusing on the peak hour in the morning) disagree with your assertion: http://newyorkyimby.com/wp-content/uploads/2014/10/amcrowding.jpg

(Note that even 100% in this table is far short of a crush load. Not that loading in excess of 100% is desirable, but even at that undesirable point it’s not true that trains can’t accommodate those who want to board.)

People clog up in the door area long before the middle of the car fills up. Ask your fellow riders to move in.

Since the line was extended from 6th Ave to 8th Avenue in 1931.

Perhaps my question wasn’t clear. You say that the terminal is constraining service to 24 tph. But the MTA says that power is constraining service to 20 tph. If we want more capacity on the line, the first order of business is to upgrade the power system, since that will allow an increase to 22 tph (and, presumably with further investment, to 24 tph). Improving the terminal won’t change a thing right now, since the power constraint at 20 tph will still govern.

What’s the trick? The press release states it will add 2 additional trains for a 10% increase. That means an increase from 20 to 22 tph.

Correct. So why focus on a 24 tph constraint now? There’s no point in lifting a 24 tph constraint while a 20 tph constraint remains in place.

N.B. they operated 24 tph at the 8th Ave terminal in 1954.

Scheduled. Maybe operated in practice, maybe not.

Why don’t you take a stopwatch and time them.

Because I’m comparing the speed of trains on an old signal system which no longer exists to the speed of trains on the current signal system. I’m afraid I don’t have a time machine, so I can’t in 2015 time train movements in 2007.

You need to time both the entrance and exit.

I know how to time trains at terminal interlockings…

Now add up the average of each of the 4 measurements.

…but apparently you don’t, since two of those movements don’t conflict and can take place concurrently, so I want to take the larger of the two rather than add them.

“You’ve glossed over two important differences between fixed block and CBTC systems, as least as far as they exist on the subway.”

You’re too trusting of Siemens and MTA press releases extolling CBTC’s benefits and the conventional block system’s shortcomings. :=)

First, by its very nature, the precision on a fixed block system is quite coarse. The signal system knows whether a block is occupied, but it has no way of knowing which part of the block is occupied. To be safe, it assumes the worst case, but that costs capacity.

Correct. However, blocks can be variable size to make the precision and accuracy finer, where it is required. If headways are 90 seconds (40 tph) and trains are traveling 40 mph (60 ft/sec), then trains are 5400 feet apart. Emergency stopping distance at 45 mph and 3.0 mph/sec emergency braking rate is 500 feet. Precision between stations is not very important.

The minimum approach distance is at station entrances. Block lengths are typically 300 ft down to 100 feet approaching and within a station. This provides adequate precision and accuracy for operating with 90 second headways.

BTW, CBTC location may be precise but it’s accuracy to the train control system is what’s important. There are transmission delays that must be considered. If communications between train and train control system drops out, the message is repeated until 2 seconds have passed (NYCT spec). At that time either the train or the train control system will halt that train. Train spacing must account for this timeout. The accuracy for train location for a train going 30 mph is 90 feet. That’s essentially the length of a block approaching a station.

Second, at least on the subway system, CBTC calculates and dictates the maximum speed on a continuous scale, based not only on track geometry but also on the relative location of the train ahead, while fixed block signals can only indicate stop or go. What passes for speed control in a fixed block system is a series of signals that only clear if a train approaches within allowable speed. And allowable speed is predefined, generally with no more than two distinct options (perhaps one general speed limit for all trains plus perhaps another more restrictive speed limit for trains closing in on their leaders).

The current block system will allow a nominal 40 tph. If the system is operating properly the follower will never see a red or yellow signal aspect due to the presence of a train ahead.

CBTC might slightly improve on this. However, current MTA practice operates far below the what the block system can handle.

Paris is converting to CBTC on routes that currently operate at 33-36 tph. Their goal is to improve it to 38-40 tph. That’s a modest improvement and indicative of CBTC’s benefits.

My suggestion is to wait until the MTA operates closer to what they did in the recent past, before embarking on CBTC.

Wrong incident – I think this is the one you want: http://permanent.access.gpo.gov/websites/www.ntsb.gov/publictn/1996/RAR9603.pdf

The primary issue here is that this is only a problem when/if a T/O runs a red light.

Well, yes, that’s why stop arms exist in the first place – to prevent collisions even when the T/O makes a mistake. They’re a fundamental component of the signal system.

A signal system is designed specifically to provide certain protections. If it’s found that for whatever reason it no longer provides those protections, it is no longer serving its purpose, and it needs to be redesigned in order to provide those protections once again. This isn’t optional.

The TA actually operated 24 tph, as per the insert in their first annual report.

Scheduled. Not necessarily operated. I appear to not be your first skeptic.

I don’t have an explanation for the 32 tph capacity, other than they were assuming short turning half the trains at 6th Ave.

Another possibility is that the numbers given weren’t actually achievable in practice. There’s a lot in that insert that I find hard to believe.

My timi ng studies showed that 24 tph was the capacity for 8th Ave.

Even if correct, that’s not relevant as long as power constraints hold the line back to 20 or 22.

My beef here is not with what the MTA initially did, but rather with it being 19 years later, and we still have what should have been quick, temporary fixes in place. The MTA should have fast-tracked fixing the signaling system to eliminate the problem. The best solution would have been adding another block between the clear and stop aspects. Based on my quick calculations, this wouldn’t have affected capacity. If we assume the signaling system now has a nominal theoretical capacity of 40 tph, we have 90 seconds between trains. Now add in another block. On average blocks are 800 feet. I’ll even allow that the train will traverse this extra block at only 15 mph. That will add 36 seconds to the time between trains, bringing it up to 126 seconds, or ~28 tph rounded down to the nearest integer. In a practical sense, the entire system is limited to less than that by terminal capacity. In practice then, there would be no capacity reduction. The beauty of this solution is that it wouldn’t have required moving signals or other expensive modifications. It probably could have been done systemwide in under 5 years. With the extra block, essentially an extra 800 feet to stop, trains easily could have been going 55 to 60 mph systemwide, at least where the runs between stations were long enough to reach those speeds.

The MTA probably decided not to modify the signaling system because it was overly optimistic about CBTC. They figured why put money into an antiquated system which we’re probably going to rip out within one or two decades. When they saw this wasn’t going to happen, they should have implemented a system-wide signal modification.

Is 24 tph the terminal constraint?
I’ve seen 26 tph quoted with power upgrades.

http://newyorkyimby.com/wp-content/uploads/2014/10/amcrowding.jpg

http://s3.documentcloud.org/documents/254303/l-train-study-by-the-m-t-a.pdf

Note in the second document (page 9) the MTA also says the old signal system was limited to 20 tph, reduced to 17 tph during construction of signal upgrades. What changed after 1954 to cut the capacity from the 24 tph they ran, with a projected capacity of 32 tph according to your link, to 20 tph according to the MTA?

Was there a new rule at some point that slowed trains to avoid any incidents, which had the effect of lowering capacity? Maybe one that isn’t necessary for an automated train? So that a new signal system that allows for an automated train to enter the terminal could increase terminal capacity?

http://www.nyc.gov/html/dcp/pdf/transportation/broadway_junction_pt3.pdf

NYCT also plans to eventually improve 8 th Avenue terminal throughput by installing an Automat ic Train Operation (ATO) system. Right now the terminal?s signaling is set up to have trains enter the station very slowly, since the tunnel ends at a wall at the west end of the station. ATO will allow L trains to automatically enter the station faster at a preset speed, thus improving line capacity. Train operators entering an ATO area are not, strictly speaking, controlling the train?s maximum speed, although they retain the ability to slow or stop the train. Instead, they are required to punch a but ton at regular intervals to acknowledge that they are still lucid and capable of taking control of the train if needed. At the time this report was being written, undisclosed health and safety aspects of this plan were a source of disagreement between NYC T and the unions which operate the transit system.

It was the MTA that decided to first install CBTC before upgrading the power system.

Yes, because increasing capacity was not the objective in installing CBTC on the Canarsie line. In fact, one of the reasons the Canarsie line was selected for the pilot CBTC installation is that ridership was low.

CBTC is installed and running with only 20 tph peak hour. The MTA claims upgrading the power system with CBTC installed will allow them to operate 22 tph peak hour.

26 tph, but who’s counting? (22 is what the currently planned power upgrades will yield.)

Had they not installed CBTC and upgraded the power system, they would be able to operate 24 tph as they had in the past.

As I’ve said, I have difficulty believing that they actually operated 24 tph on a regular basis. But in any case, ridership at the time the Canarsie line was selected for CBTC required only 12 tph.

Suppose patronage on the 14th St line continues to grow. Will 24 tph be sufficient?

If even 22 tph isn’t sufficient, the power system will need to be further upgraded before anything is done with the signals or terminal.

CBTC has not changed the trains’ operating characteristics nor the rule restrictions for speed on the switches or approaching the bumper block.

Of course it has. Trains operate toward much faster toward the bumper block now than before CBTC.

It’s the train’s operating characteristics, the rule book and the station geometry that determine terminal capacity. These have not changed.

And the speed restrictions enforced by the signal system, which certainly have changed, a lot. (Remember what I said before about finer speed control under CBTC?)

I did measure the times with a stopwatch before 2007. My measurements were consistent with 24 tph operation.

Perhaps you should come back now that it’s 2015. Stuff has changed.

Such operation would result in non-uniform headways. This wouldn’t a problem, if there were not any loading problems. There are crowding problems, which is why service level constraints are being discussed. Non-uniform headways are difficult to impossible to maintain, if there is crowding. Therefore, it is wise to calculate terminal capacity for uniform headways.

If you’re worried about the uneven loads that result from uneven headways, hold trains at 6th Avenue to even them out. (If you’re worried about uneven loads, you may want to think twice before proposing short turns.)

Yep, thanks for the correction.

The recovery question is which is faster after the train with the sick passenger begins to move: trains bunched together and proceeding at slow speed or trains spread further apart proceeding at near normal speeds but covering more ground. The answer is the latter.

Except that the train trapped behind the incident train aren’t proceeding at best normal speeds. They’re stopped while the incident train is stopped, and one the incident train begins moving they start creeping forward.

And you’re ignoring the fact that, the closer the following trains can get to each other, the more people can reach their final destinations before the incident train has even begun moving.

The MTA is operating trains in a manner that is not consistent with maximum service level capacity.

The real world unfortunately is not consistent with (your conception of) maximum service level capacity.

The Queens Boulevard express is actually slated to get CBTC after the #7, so that line will be fixed in the relatively near term.

As for fast tracking signal improvements (and other needed maintenance work) it might make sense to completely shut down an entire line or portions of it. I’ve heard it takes many multiples of time trying to run trains while doing improvements or maintenance. It may well be that what would take 2 or 3 years trying to do while keeping the line open could be done in 2 or 3 months if you just shut it down completely. Yes, it’ll be extremely disruptive for those few months but I feel it’s the lesser of two evils. The MTA could run shuttle buses on the portions of the line which were shut down. To keep the buses moving, the street they’re on could be closed to everything except buses.