04 June 2014

The Great Platform Construction Extravaganza

Redwood City in August 2000
One of the arguments against level boarding is that it would involve a massive reconstruction of just about every single station platform along the entire peninsula rail corridor without interrupting rail service.  On the face of it, that sounds like a very expensive logistical challenge, simply too hard to take on with all the other modernization efforts currently underway.

Would you believe that Caltrain has already done it?  Over the last 15 years, no fewer than 37 station platforms have been built from the ground up.

Following Caltrain's forward-thinking strategic plan, these new platforms were all built for future compatibility with level boarding rolling stock, after a hard-fought waiver of CPUC General Order 26D.  This regulation had previously limited platform heights to 8 inches, causing lengthy station dwells and slowing Caltrain trip times, which thankfully have improved by several minutes over the last decade.

Um, never mind, scratch that entire last paragraph!  All of these new platforms will have to be re-built all over again to achieve level boarding.

Station Qty Opening Cost Comments
Bayshore 2 Mar 2004
San Bruno 2 Apr 2014 part of grade separation project
Millbrae 3 Jun 2003 part of BART to SFO project
Burlingame 2 Jun 2008 $20.5M
San Mateo 2 Sep 2000
Hayward Park 2 Nov 1999
built for future third track
Hillsdale 1 Oct 2005 $2+M new northbound only, with southbound improvements
Belmont 1 Oct 1999 part of grade separation project
San Carlos 2 Oct 1999 part of grade separation project
Redwood City 2 4Q 2000
Menlo Park 2 Aug 2000 $3.3M
Palo Alto 2 Feb 2009 $35M (cost shared with Cal Ave)
California Ave 2 Feb 2009 $35M (cost shared with Palo Alto)
San Antonio 2 Apr 1999
Mountain View 2 Dec 1999
Sunnyvale 2 May 2003
Lawrence 2 Mar 2004
Santa Clara 2 Dec 2011 $40M (cost shared with San Jose)
San Jose 2 4Q 2012 $40M (cost shared with Santa Clara)

Caltrain has demonstrated that they know how to build a large number of new platforms over a time span of a decade or so, which is why it's particularly important that the new EMU rolling stock (which will last three decades) be pre-configured for a future level boarding platform interface.  It would truly be a shame if an effective level boarding solution--fully compatible with high-speed rail--were to be precluded through bad procurement choices made today in 2014.

Here's hoping that Caltrain's 2014 strategic plan will reflect a new vision for level boarding and a new understanding of its advantages for reducing trip times, improving service punctuality, increasing train average speeds, and increasing the overall throughput capacity of the peninsula rail corridor.


  1. Psst. I think the last two "cost shared with..." in the table are backwards.

  2. If Caltrain-HSR goes to continental European model, Caltrain and HSR should have compatible platform with shared tracks/platform. If they goes Asian model (where HSR is separated from conventional rail by faregate), platform compatibility is not important.
    Caltrain management have no idea how European/Asian model works. So, they stick to FRA commuter rail model.

  3. Caltrain should build 3rd track in Hayward park for increase track capacity in peak hours. (Actually, this station can add 4th tracks). Currently, northbound #211, #221 are bypassed by following express at Bayshore. If this changed to Hayward Park, we can add one more train 10 min earlier than #313, #323.
    Caltrain is looking for longer platform to handle longer train. I suggest more frequent train by adding more by-pass tracks.
    Which is cost effective? Extending platform in several stations, or add 3/4th track in Hayward Park?
    Longer train benefits only existing customer if they can find the seat. More frequent train benefit both existing and new customer.

    1. More trains require additional locomotives and more unnecessarily large train crews. Longer trains require new platforms. It's difficult to say which will cost more in the long run, but the way CalTrain operates, I'm guessing that ongoing operating costs are going to be more of a concern.

      There's some justification for more trains to run on CalTrain, but before that happens the service patterns really need to be regularized - trying to put more trains into the horrible morass of skip-stop patterns we have today isn't going to work well for anyone.

    2. Don't need additional locomotive! Just reduce the trun-around time! (Need time table overhaul)
      In addition, Caltrain have at least 2 spare set of 5-car consist currently waiting in SF and SJ terminal. Those trainset are utilized when BART did the strike and Giants game day.

      If we slow down the Baby Bullet from current 57/59 min to 62~64 min, 6 train/h is archive without additional by-passing trucks.

    3. Under the existing union agreements, one extra car (for a total of 6 cars) can be handled by 2 conductors, but a third conductor is required for a 7-car train.

    4. @Anonymous: are you seriously expecting of Caltrain
      a) short turn-around times
      b) punctual operations
      c) efficient fleet utilization
      These qualities do not just suddenly erupt overnight, and they're going to take far more than a timetable overhaul.

    5. Clem. I also don't expect Caltrain to change those things in short period of time.
      One example: Train 329 (1st or 2nd highest load NB train) depart from Tamien 7:56. This is turn around from #104 which arrives 7:03. So turn around time is 53 minutes.
      Additional car from Metrolink, platform extension all requires capital investment and more maintenance cost. Utilizing idling train need additional crew, fuel and maintenance cost but these cost should be cheper.

    6. I sometimes see the consist with two locomotive. I don't know why Caltrain do this but they have one or two extra locomotive. So, Caltrain can add one or two more train consists without changing their discipline.

    7. I think this might be for reliability reasons. Some of the locomotives are likely more trouble-prone than others, and if you stick two bad ones together they will have a better system reliability (assuming one can fail without affecting the other) than a single good one. That's my speculation anyway...

    8. Reality Check28 June, 2014 13:18

      I believe Caltrain also runs double-headers when it needs to move a locomotive from one end of the line to the other for maintenance or other reasons.

      Of course, If it's not running -- or not running well -- dead-heading it as a "light engine" (RR lingo for locomotive w/o cars) doesn't make much sense anyway.

      And, as Clem mentioned, using two locos on extra long or special trains helps boost reliability and acceleration performance. Yes, if one loco craps out, you can still continue on with it DIT (dead-in-train).

  4. Six car trains is the best for the near future. An additional train during the peak, the capacity will effectively increase by 11 cars per hour (44%). I think turnaround times should be about 15 minutes, but that's up to the union to decide.

    A potential post-electrification/overtake, pre-Transbay/HSR schedule can be found here. (Northbound Schedule under construction.) The normal (off-peak) pattern is 2tph between SJ and SF (local to RWC, express in San Mateo County), and 2 tph making all stops between RWC and SF. During the peak, the local trains are extended into Santa Clara County, and the express trains are Baby Bullets that coincide at Hillsdale. An additional 2tph making limited stops during peak hours ensures 6 tph at the peak, which should be more than enough capacity for the next couple of decades.

    1. Caelestor, very good schedule! Concept of this Redwood local and Santa Clara express are quite make sense and this can be implement even before electrification. So, longer train can be archieved without buying new cars. (I don't oppose Caltrain's buying new car)
      Redwood local can be shorter (2~3 car) while Santa Clara expree can be 6 car. Platfrom extension need to be built only express stop, which justify the capital investment.
      Shorter local train have higher accerelation and breaking which is also makes sense for shorter station to station distance. Rebuilding hold-out station (South SF and Boradway) can be cheaper as shorter platform requirement.

      Turnaround time of crew and equipment should be different. Trainset should be turnaround with minimum time to generate more revenue. For Tamien bound train, crew can be exchange at Diridon if they need some rest.

    2. Nice one. It should be noted this schedule requires construction of a mid-line overtake from Redwood City to San Mateo. Not sure if you'd need the "long" overtake all the way through Redwood Junction.

    3. Just require 4 track in Hilsdale station. Add each one more track both northbound and southbound, if they can use space of parking lot. 2 island platform will provide local <-> express transfer.
      4 track section is just needed at station only but need some signal adjustment for smooth overtake.

    4. Look again, that's incorrect. With just the extra platform tracks at Hillsdale, the overtaken train would have to wait about 7 minutes at Hillsdale (3 minutes for overtaking train to catch up, 1 minute dwell/pad, 3 minutes for overtaking train to get ahead). Caelestor's timetable does not reflect this. Off-peak, his local and express leave RWC simultaneously on parallel tracks. During the peak, his express is 3 minutes behind the local as the latter leaves San Carlos, and 1 minute ahead as the latter leaves Hillsdale. Taken together, these timings require overtake tracks at least from Redwood City station to just south of downtown San Mateo.

  5. My comment got deleted so I'm going to write a shorter reply. The current schedule assumes 4 tracks between Redwood City and approximately where Hayward Park is located today. It's okay to add another minute of waiting time at Hillsdale to make the schedules work out. Hillsdale would become a super station with all local transit routes converging on it, and Hayward Park would be permanently closed to straighten the tracks there. The few existing riders can be accommodated by 6 tph in both directions at Hillsdale or 4 tph at San Mateo plus shuttle service if necessary.

    The local trains would run on the inner pair of tracks, so San Carlos would be rebuilt with an island platform. Redwood City is going to be tricky to engineer. This post-electrification/overtake schedule actually favors a three-track station (the inner track served by both platforms) so that off-peak locals can quickly turnaround while still enabling same-platform transfers in both directions. However, the ideal situation is still the 4 track station, with the local tracks going into a trench underneath the station, turning towards East Palo Alto, and heading over a Dumbarton Bridge into Fremont and beyond. The compromise would be a short extension to a temporary terminal at Willow Road, though it's likely to receive anemic ridership.

    This schedule would be adequate until the arrival of HSR and the opening of the extension to Transbay Transit Center, neither of which is likely to happen for at least 15 years. In the meantime, some routes should be really reconsidered. There's a lot of discussion regarding HSR elsewhere, but the DTX tunnel could be a lot straighter if it headed under 3rd/4th Sts. An East Alternative route would allow the 22nd St Station to be replaced by a more centrally located 16th St Station serving UCSF and a future basketball arena.

    1. Reality Check18 June, 2014 21:41

      Dumbarton rail @ Willow Road = Facebook!

      Facebook's new 'hacker cave' to house 2800 engineers, 3400 staff in ONE room
      Exclusive Photos Of Facebook's Sprawling New Frank Ghery-designed HQ

      And that's not all, in the news just last week:
      Menlo Park: Facebook to buy 59 acres next door to new campus

      "Facebook's domination of Menlo Park is not yet over: After the social media company moved into a new 57-acre headquarters off Willow Road, it announced plans to redevelop 22 acres within walking distance into a "west campus" off Constitution Drive designed by renowned architect Frank Gehry. And now, it plans to buy an additional 59 acres right next door from TE Connectivity, as first reported by the Silicon Valley Business Journal on June 6."

      Here's a map showing all 138-acres of Facebook land ... all hard up against the Dumbarton rail line between Willow and Chilco.

    2. Will the Zuck drop a few 100 million of his personal fortune on Dumbarton rail?

    3. Reality Check19 June, 2014 07:42

      I doubt it ... but he is quite the giver!

      $100m to NJ schools in 2010; $1b to Silicon Valley Foundation last year; $120m to local schools this year; etc.

      I could easily see where FB might kick in for a very nice Facebook station, however.

    4. Caelestor,

      San Carlos will be rebuilt as an island platform? Samtrans is selling a 99-year lease for devepment on the very same land where the shoo-fly tracks fused to build the current San Carlos station. How are you going to quad-track San Carlos? Eminent-domain the development, or shut down the entire Caltrain for a few years?

    5. Island platforms exist only in the minds of us transit advocates. What you'll see built is a mini NEC with four tracks and side platforms. The San Carlos station entrance will be moved to El Camino @ Cherry under the current plan, with parking stretching all the way south to Arroyo. It is conceivable that a new four-track station would be built a few hundred feet south of the current station, but you have to keep in mind the number of neurons dedicated to long-range strategic planning at Samtrans and Caltrain: zero.

  6. EMU train with good accerelation (2.7~3.3Km/h/s) will not require 7 min waiting for local train. Arrive 1 min before express arrival, then 30~45 sec for express train discharge and boarding. Local train can depart 1 min after express departure. So, total waiting time for local train is about 3 min.
    Good signal system and crew's disciplinemay not be expensive as buliding long 4-track sections.
    This type of express-local overtaking and platform transfer is very common in Japan.
    See example here.

    1. Can you find the timetable for this Japanese overtake? The video has multiple cuts and gives no clue as to how long the local waited. I bet it's longer than 5 minutes.

    2. Cannot be 5min. In the midday, they runs 18 train/h
      Here is timetable of this station. (Weekday)


      See line map. This movie taken at KK29 (Keikyu Tsurumi) which serves local and express train.


    3. One of unofficial stringline diagram of Keikyu Main Line:


    4. So, 3.5 to 4 minute dwells. That's very impressive. It does require level boarding: otherwise dwell times are too long and unpredictable. A single wheel chair could ruin the entire rush hour, through no fault of the wheel chair user.

    5. They still need small bridge between door to platform. However, all the station are staffed and having fare gate. So, station staff prepare the small bridge for wheelchair customer.

    6. Can people in wheelchairs not board on their own?

    7. Look at the video linked above. There is a significant gap and a significant step up that one would have to mind.

    8. In Japan, the step up is intentional- apparently, having the car floor level with platform actually increases the chances of people tripping or falling into the gap, as they tend to ignore the risk- having a step up or down (when exiting the train) makes people more aware of where they plant their feet. Of course, this makes wheelchair boarding w/o assistance impossible- but this may also be intentional- to reduce railway liability, and possible delays in boarding, especially in the peak- as you all know, dwell times are timed to the second in Japan.

  7. Anonymous: This has nothing to do with acceleration. You're limited by the minimum headway between trains, which is typically 2-3 minutes with modern signaling (for safety reasons). So even with 30s dwell, good acceleration, and modern signaling, you're still looking at the local sitting in the station for 5 minutes unless you extend the 4 track section.

    1. Minimum headway behind a train leaving a station is determined partly by the acceleration of that train, though. Which makes sense: the faster the train ahead of you accelerates, the sooner you'll see a green signal, assuming fixed block lengths (determined by stopping distance from full speed).

    2. It's better, yes, but you're still limited by the signaling system and safe deceleration time. You're not going to get 1 minute headways.

    3. Why can't you have 1-minute headways for local trains following express trains? The minimum headway on an all-local line is really a combination of stopping distance and station dwell, and if the train that's ahead is not making a stop anytime soon whereas the train that's behind is, it allows much shorter headways than the usual.

    4. eventually the express is going to catch up with the local ahead of it. One minute headway and 20 seconds for the local to slow down, stop and accelerate how many stops does the local ahead of the express get before they are both in the same place as the same time.

    5. Caltrain's capacity analysis considered this very question. They showed a 90-second headway on successive departing trains eventually converged to about 180 seconds for two successive locals. If the first train is an express, however, there is no delay to the following local when the local departs 90 seconds after the express. So 90 seconds seems feasible *after* the overtake. It's the part *before* the overtake that gets dicey, especially when you throw in random delays. As the Japanese know, punctuality (to levels never achieved on Caltrain) is the key to successful overtakes.

  8. 7 min may be long for some rider who do not need express train. However, many rider will be happy if they can transfer to express or vice versa. Rider who is traveling from San Jose to Hayward Park, Burlingame, San Bruno...,they can use express to Hillsdale and then use local train in front of express.
    7 min waiting is even better than current 5 min waiting in Lawrence or Bayshore, which you just need to wait.
    Building long 4-track section will be financially justify if then run 20 or more train per hours.

    1. 20 trains/h can be processed without long 4-track section.

      Example: Keio Line, 21 trains per hour (24 trains per hour during evening peak) west of Sasazuka without long continuous 4-track section or intergalactic station at end of line (There are only 3 stub-end tracks at Shinjuku). Unofficial stringline diagrams available at:

  9. To elaborate on what Joey wrote... I believe Caltrains current signal system needs about 3-4 minute spacing for the following train to run at full speed. If you look at Caltrain schedule the penalty for each stop is about 2 mins. This includes deceleration, dwell and acceleration time. In order for both local and bullet trains to meet a common station (say Hillsdale), you'd need a 4 track segment of 2 stations before and 2 stations after hillsdale, or a total span of 5 stations. And we're assuming that the local station will stop at all the stations.

    Looking at the stations, the cost of making 4 tracks through the 5 stations in increasing cost is:
    * Hayward Park - station is not elevated, ROW is big enough, CHEAP!
    * Hillsdale - Little more expensive, but still easy
    * Belmont - Moderately more difficult as it's grade seperated, and the station is not symmetrically between the two tracks.
    * San Carlos - More difficult due to the existing historic station building. Platforms would have to be shifted North if you want to save cost (and PR headaches) related to the historic station building
    * Redwood City - Multiple grade crossings surrounding the station would need to get cleaned up. Track would have to be elevated. Also, you'll need to do some eminent domain near the UP spur to Port of RWC. As you look out the right side side of NB train, you'll note that the 4 track section ends to make room for the freight spur. You probably don't need more than about 5 ft to make things fit, but Eminent Domain comes with PR head-aches, so it's worth highlighting it.
    * San Mateo - ugh...

  10. I'll just leave a note;
    Came looking for renderings of new Caltrans equipment.
    Please direct to or provide link. I'm looking for answers as to
    platform height compatability Talgo or Talgo-type.
    Yes, it's THAT guy again. I'd like to see cross-country trains compatable, why not? Anyway. I've mostly been fair enough until you guys hammer me as a dumbass. I'm actually accomplished:
    San DiegoTrolley Mission Valley Line, West Coast LRT. I'm not accepting blame for Seattle worst case examples of Streetcar, Sounder & Link LRT. Silicon Valley, meet Microsoft. BOTH doing GREAT rail, sorta, not really. Oh but MAX is so slow the complaint sounds, BUT, traffic is actually managed. I fought 1995-98 against MAX S/N while guys like you defended it. I built support for replacements I-MAX Yellow Line, Green Line Transit Mall, and now Milwaukie MAX for guys like you. Impact reduction, major improvements, costs savings. I worked on Denver and SLC light rail. Honolulu. Portland Streetcar. Maybe I just don't speak your language, but, as Einstein said: Faster is slower and slower is faster. If you want to go faster, go slower. And if you do NOT want to go slower, do not go faster.
    Okay, it was just me that said that. Einstein could've said it.

  11. This comment has been removed by the author.