Buried in this article on $100 billion in capital spending on the New York City subway system is the following:
the capital plan invests $2.75 billion to upgrade signals
If there are already signalling cables along the tracks capable of supporting TCP/IP, how is it possible to spend $2.75 billion on a set of lights driven by a few microprocessors? That’s enough to build two of the world’s largest passenger ships (see Symphony of the Seas), complete with water park, skating rink, thousands of cabins, accommodations for 2,175 crew members, etc. I don’t think the answer is “stuff purchased by the government costs a lot” because, according to the same article, they are going to get 1,000 subway cars for almost the same price ($3.2 billion).
[Separately, the Moscow Metro system was amazing by American standards. It is huge and growing every year. Trains on every line run every minute on weekdays and every two minutes on weekends. A 23-year-old native said that the longest she’d ever waited for a train was 12 minutes. That was at 1:00 am and the delay excited a lot of discussion among customers. The artwork is justly famous. I loved the aviation emphasis of the ceiling mosaics, which are designed to show what you might see if you looked skyward (some photos that I took a few hours after staggering out of the Moscow airport).]
What would that $100 billion spent by New York City buy in China? The World Bank says it costs China roughly $20 million per km to build a high-speed rail system. So $100 billion would suffice for 5000 km of track, stations, and trains(?). That’s 3,100 miles, or Boston to San Francisco.
Related:
If there is a way to inconvenience riders that the MTA hasn’t found yet, rest assured, they are spending lots of money looking for it.
If it’s like San Francisco, which is spending about $110M on under a thousand or so radios for buses, it’s because they are required to comply with the Project 25 trunked public safety radio standard, which specifies proprietary codecs from Digital Voice Inc. (DVSI), a MIT spin-off.The standards group never bothered to negotiate a capped license fee, so DVSI can and does charge as much as it wants.
TL:DR version: price gouging by your alma mater.
Modern signalling systems have variable distances between the trains, with no signals at the trackside. When there are delays the trains can get very close to each other to speed recovery. No driver is required. They usually cost a lot because the tracks need to run throughout the transformation. Trains have to work on both systems until all the lines are transformed. When the job is done the system that switches back and forward between the old and new systems has to be removed. None of the companies delivering these systems is making out of the ordinary profits. I work for one of them (Thales).
@Dave, since you work for one of them, can you tell us how long such a system will last? 20 years? 50 years? How about yearly maintains cost?
At $2.75 billion, I’m guessing it would be cheaper to install low tech system such as hiring people (even 2 drivers per train at all times to provide extra safety) to get the job done manually.
I’m trying to justify the rational behind spending $2.75 billion, what are we getting back for it.
Dave: Thanks for the insider’s perspective. What are they actually installing as part of this $2.75 billion? There are already cables along the track to carry information, correct? So the installation is of sensors to figure out where trains are? At what intervals are the sensors placed? Why are these more costly than other RFID sensors?
How much harder would it be and how much more would it cost to make the trains self-driving? It looks like the drivers each cost at least $200,000 per year, including pension and benefits (see http://gothamist.com/2010/06/04/mta_salaries_like_239148_for_a_cond.php for some 2010 data). If they’re buying new cars and installing new signals, why can’t robots drive?
[Separately, in Moscow, it seems to be the case that the main signal is a timer in each station that starts running when a train leaves the station. This is visible to the driver of the next train in sequence. He or she departs the station when the time is about to hit 1 minute (weekdays) or 2 minutes (weekends). This can keep the entire system running at a fixed interval even with no coordination among signals or drivers.]
The technology is driverless (see http://www.railjournal.com/index.php/middle-east/thales-nets-dubai-cbtc-contract.html). In London the system runs driverless on the DLR lines and with driver on the Jubilee and Northern lines. Usually comms systems are quite old and the new signalling system comes with a new integrated comms system. As the new service has to be installed and tested while the old one continues they are usually retained as two working systems that can be swapped back and forward.
I can foresee cost savings in the future from competition with autonomous vehicles in general. Most old signalling systems comply with a very small set of design patterns and hardware elements. We need technology that will acquire information from 3D video/laser track surveys, maps, etc. then produce all the designs that deliver the new system and remove the old. I guess Google may have broken the back of this research but have shown no inclination to manage guided vehicles.
On the infrastructure issues the train knows where it is at all times. A mixture of on train sensors (like wheel rotation) and communications with fixed track infrastructure (like transponders). The signalling system controls how much distance the train can travel and maximum speed allowed. Other factors like weather outside that affects adhesion can also be added. The trains will automatically slow and stop safely if comms are lost.
I don’t know anything about the $2.75 billion. I expect they will have bundled in everything associated with developing and introducing the system and a fair bit of contingency too. For example if you want to run completely driverless there will be considerable redundancy, early retirement and redeployment costs.
American subway signal equipment needs specially engineered bum urine resistance properties.
I’m sure it’s all allocated to a denver-airport or air traffic control rebuild-style large software engineering project fiasco
Not strictly on topic, but re transit costs: When MARTA was under construction in Atlanta, my neighbor the lawyer reported at cocktails that he knew lawyers (not him of course) who often won compensation or sold easements to the system for property their clients did not own. Behind my church the cut-and-cover encountered a big tree on a lawn; instead of offering the owner $10K (which she would have accepted), the system hired her a lawyer, the system a lawyer, her a botanist, the system a botanist, and spent months negotiating, whereupon they paid her $10K.
I was involved with procuring the underground trains at Atlanta airport at the time. We were also engaging local engineering firms, and heard from them of contract padding by MARTA’s out-of-town umbrella consultant, who collected a 10% markup. Coincidentally, the airport system was almost an analog of the MARTA in terms of station count (except the airport trains are driverless and the platforms have coordinated doors, thus more complex).
First phase MARTA was almost $2 Billion in 70’s, the airport system was $23 MILLION. Apparently the cost is in the right-of-way and external construction abuses, not the technology. It is definitely not about running a railroad.
Does not sound like a rocket science and as described the system relies on old technologies and computational algorithms. Wheel circumference length is a known variable 🙂 If there are 1,000 trains that need to be upgraded it is hard to see how it could cost $2.5 million per train. I am pretty sure it is not just a vendor costs, vendors are getting just a fraction of total cost.
Maybe it’s not “Stuff purchased by the government costs a lot”, because things like rail cars are basically fungible and transportable, allowing companies worldwide to bid on them competitively. (I imagine you build these things to some kind of standard, so they should be able to be built anywhere and shipped to New York)
Maybe what you’re looking for is that “Stuff built by the government costs a lot”. I’m sure that the industrial sensor suites’ hardware isn’t cheap, but it’s also probably not ludicrously expensive, but the installation and integration probably cost a fortune. I’m sure the project includes set asides for small business, minority owned, and local sourcing. All of it done at prevailing union wages. See: http://www.coyoteblog.com/coyote_blog/2017/05/your-labor-regulation-fact-of-the-day.html In that example, the traffic cone guy costs $88k/year. I can’t imagine what they pay a software architect or master electrician.
To an outsider, the NY subway system is a complete city under the ground so it’s like replacing every elevator button in the city. It has to be done by workers who owe $10,000/month in rent, what Jenet Yellen deemed the necessary price to ensure total employment.
I say if you like your 2.75 billion dollar signal system you can keep your 2.75 billion dollar signal system.
jack #15, from Wikipedia: “As of November 2016, the New York City Subway has 6418 cars on the roster.[152][note 11] A typical New York City Subway train consists of 8 to 11 cars, although shuttles can have as few as two, and the train can range from 150 to 600 feet (46 to 183 m) in length.[153]”
If we add 1000 new cars it still comes to under a thousand trains that need to be modified. Per train cost seems bloated even with 2 or 3 times worker or engineer pay then what is expected by some. City workers tend not to live in Manhattan and usually live in outer boroughs or even out of state, as far as I know. Their rents or mortgages not even close to $10,000/month. I would say that for a major world business center and the largest US city NYC subway has to few trains.