Construction Costs: Signaling
I launched a Patreon poll about construction cost posts, offering three options: signaling and electrification, rolling stock, and historical costs. Signaling and electrification won with 29 votes to historical costs’ 20 and rolling stock’s 6. This post covers signaling, and a subsequent post will cover electrification.
I was hoping to have a good database of the cost of installing train protection systems. Instead, I only have a few observations. Most metro lines in the world have searchable construction costs given a few minutes on Google, and a fair number of rolling stock orders are reported alongside their costs on Railway Gazette and other trade publications. In contrast, recent numbers for signaling are hard to get.
The gold standard for mainline rail signaling is European Train Control System, or ETCS; together with a specified GSM communications frequency it forms the European Rail Traffic Management System, or ERTMS. It’s a system designed to replace incompatible national standards that are often nearing the end of their lives (e.g. Germany expects that every person qualified to maintain its legacy LZB system will retire by 2026). It’s of especial interest to high-speed lines, since they are new and must be signaled from scratch based on the highest available standard, and to freight lines, since freight rail competes best over long distances, crossing national borders within Europe. Incompatible standards between countries are one reason why Europe’s freight rail mode share is weaker than that of the US, China, or Russia (which is Eurasian rather than European when it comes to freight rail).
As with every complex IT project, installation has fallen behind expectations. The case of Denmark is instructive. In 2008, Denmark announced that it would install ETCS Level 2 on its entire 2,667-km network by 2020, at the cost of €3.2 billion, or about $1.5 million per route-km. This was because, unlike both of its neighbors, Denmark has a weak legacy rail network outside of the Copenhagen S-tog, with little electrification and less advanced preexisting signaling than LZB. Unfortunately, the project has been plagued with delays, and the most recent timetable calls for completion by 2030. The state has had to additionally subsidize equipping locomotives with ETCS, but the cost is so far low, around $100,000 per locomotive or a little more.
That said, costs in Denmark seem steady, if anything slightly lower than budgeted, thanks to a cheap bid in 2011-2. The reason given for the delay is that Banedanmark changed its priorities and is now focusing on electrification. But contracts for equipping the tracks for ETCS are being let, and the cost per kilometer is about €400,000, or $500,000. The higher cost quoted above, $1.5 million per km, includes some fixed development costs and rolling stock costs.
Outside Denmark, ETCS Level 2 installation continues, but not at a nationwide scale, even in small countries. In 2010, SNCB rejected the idea of near-term nationwide installation, saying that the cost would be prohibitive: €4.68 billion for a network of 3,607 km, about $1.6 million per route-km. This cost would have covered not just signaling the tracks but also modifying interlockings; it’s not purely electronics but also concrete.
The Netherlands is planning extensive installation as well. As per Annex V of an EU audit from last year (PDF-pp. 58-59), the projected cost is around $2 million per route-km; the same document also endorses Denmark’s original budget, minus a small reduction as detailed above due to unexpectedly favorable bids. Locomotive costs are said to be not about $100,000 but €300,000 for new trainsets or €500,000 for retrofitting older trainsets.
A cheaper version, ETCS Level 1, is also available. I do not know its cost. Switzerland is about to complete the process of a nationwide installation. It permits a trainset equipped with just ETCS equipment and no other signaling to use the tracks, improving interoperability. However, it is an overlay on preexisting systems, so it is only a good fit in places with good preexisting signaling. This includes Switzerland, Germany, and France, but not Denmark or other countries with weak legacy rail networks, including the US. The Northeast Corridor’s ACSES system is similar to ETCS Level 1, but it’s an overlay on top of a cab signaling system installed by the Pennsylvania Railroad in the 1930s.
Comparing this with American costs is difficult. American positive train control, or PTC, uses lower-capacity overlay signaling, nothing like ETCS Level 2. One article claims that the cost per track-km (not route-km) on US commuter rail is about $260,000. On the MBTA, the projected cost is $517 million for 641 km, or $800,000 per route-km; on the LIRR it’s $1 billion for 513 route-km, or $1.9 million per route-km. Observe that the LIRR is spending about as much on a legacy tweak as Denmark and the Netherlands are on a high-capacity system built from scratch.
Pedestrian Observations 
FWIW, the SBB Neubaustrecke Mattstetten – Rothrist has been operating under ETCS L2 for a good 10 years now. Back then, the total cost for ETCS L2 was around 1.35 Million CHF per kilometer. This number is relatively high because of the short length of the line equipped (an RBC (Radio Block Center) can handle more than just 45 km line), as well as the line itself (several tunnels, hilly landscape), requiring a somewhat higher number of GSM-R transmitters). FWIW, an interesting factlet is that the line was originally equipped with regular lineside signalling, which was used as fall-back when ETCS had problems. Last year, that lineside signalling was removed, as required in the building permit for the line. Another factlet is that on this line, the trains run at 200 km/h, and may follow each other in 2 minute-intervals; that’s only possible with ETCS L2.
On the rolling stock side, conventional signalling equipment in a locomotive for the Netherlands – Germany – Switzerland – Italy corridor costs more than 1 Million CHF (15% or so of the total price of the unit), including ETCS; ETCS only is about a quarter of that. This reduced cost is one of the reasons why the SBB is switching to ETCS L1LS (essentially the existing system, but the trackside-vehicle interface is using Balises only, instead of Signum, ZUB, etc.)
Yeah, I was looking for information about the cost of a locomotive equipped with multiple legacy signaling systems but couldn’t find anything comparable. I saw 500,000 euros at one point but the same source also said ETCS would cut it to 100,000.
I got this number from a presentation given at the introduction of ETCS on the Neubaustrecke, 2007 (https://www.gdi-adi.ch/fileadmin/user_upload/downloads/ortsgruppen/mittelland_jura/de/060510_etcs-vortrag.pdf). There they say that the total cost of signal equipment in a Re482, aka BR 185 (at DB) for Germany, Switzerland and Italy, seven systems have to be supported, and all that cost more than 1 million CHF (actually one quarter of the price of the locomotive). This opposed to ETCS only at 250000 to 350000 CHF. I don’t know for sure, but I think the prices for equipment came down, and that would explain your numbers.
Actually, the presentation was given in 2006, not 2007.
How much was ETCS on Thameslink? They threw in ATO as well so I can imagine the signalling will be more expensive compared to ETCS alone but will be interested to find out.
This is speculation: ATO is implemented on vehicle side, and not on the signalling side. With ATO, it is the on-board computer interpreting the information from the signalling system (instead of the driver). So, the signalling cost would not be much different, but the on board equipment would be more expensive.
ETCS Level 1 and Level 2 are also overlay systems in a sense: they overlay on top of existing occupancy detection systems. And I’m not convinced that Amtrak/LIRR style cab signals with ACSES are all that much worse than ETCS L2 in terms of capacity, given that 24 tph operate through the North River Tunnels with cab signals with no problem, and given that the highest volume ETCS corridor is also going to be 24 tph. In particular, if they update the braking rates from the old PRR standards, I suspect capacity could improve even further.
ETCS L2 is not an overlay; it is on its own. What it does need lineside are balises which update the exact position, but these balises are not controlled, and always send out the same message (you are here).
Right, but it’s still overlaid on top of some existing occupancy detection system like axle counters or track circuits. Level 3 is the one that is basically CBTC.
An example from double-byte side of the world…
ATC system on Tokaido Shinkansen (515.4 km, double-track) was converted from legacy analog system (ATC-1) to digital system (ATC-NS) between 1999 and 2005 for 41 billion yen:
Click to access 000019879.pdf
Now JR Central is now in the middle (from 2014 to 2038) of an 11-phase ATC-NS upgrade project. The estimated cost (as of 2014): 54.6 billion yen
http://www.jrea.or.jp/search/?p=132576
http://jr-central.co.jp/news/release/nws001580.html
Do you know the costs of ATC installation on commuter line segments that only have ATS?
Denver’s RTD construction costs would be a good baseline for a new system built with PTC and electrification from scratch in the US.
Partially. RTD does follow some UP and BNSF corridors that were previously signaled and mandated for PTC anyway, so minority % of their signaling costs were defrayed by signal upgrades rather than ground-up signal installs and monetary contribution + interoperability work with an already-committed Class I installation rather than total ground-up PTC installs. The overall finished product ends up ballpark-accurate, but because the system is a blended mix of some all-new/some borrowed-n’-upgraded infrastructure it’ll take a deeper dive through the individual line segments to truly hang a precise formula applicable for widespread comparison.
https://www.eurailpress.de/news/technik/single-view/news/norwegen-siemens-liefert-etcs-level-2-fuer-gesamtnetz.html
ETCS-2 for all of Norway (4200 track-km. progressively over 16 years) for EUR800m.
First 700km route (Europe is so tiny, you know, unlike the wide open spaces of New Jersey or San Mateo County in California) by 2022.
Norway’s one those famously cheap parts of the world, like Switzerland.