21 October 2018

Thinking Big in Redwood City

The architecture of Amsterdam Bijlmer
(photo by tataAnne) could represent
the future Redwood City station.
In a seamless transportation network that runs on a regular clockface schedule with timed, well-coordinated transfers, connecting nodes play a key role. Redwood City has natural potential as a connecting node, being located approximately at the midpoint of the peninsula rail corridor, serving as a logical transfer point between local and express trains, serving as the entry point to the peninsula from the future Dumbarton rail corridor, and being in of itself a significant destination with extensive connecting bus service and a willingness to grow.

With Redwood City currently renewing its interest in grade separations, it's important to think big and to re-imagine the station as a key node in the Bay Area's transportation network.

Start with a good timetable

Using our handy service pattern generator, let's see what we could do if we organized a blended system that made Redwood City a key transfer node. When you make a business plan, the first thing to be crystal clear about is: what is your product? In Caltrain's case, the timetable is the product, and all these stations and tracks should only be built as long as they contribute directly to delivering a quantifiably better timetable for the ordinary rider. Building a major new station in Redwood City isn't about trite superlatives like "Grand Central of the West," but simply about efficient and seamless coordination of timely and reliable ways to get from point A to point B.

Let's set some ground rules for our timetable:
  • Caltrain expresses will operate every 10 minutes on a regular clockface schedule. A base 'takt' of 10 minutes reduces gracefully to 20 minutes or 30 minutes in the off-peak.
  • In Silicon Valley, there will be no skip-stop service because the population and jobs are evenly sprawled. Every station in Silicon Valley needs to be served frequently, doing away with the ridership distortions induced by the Baby Bullet effect.
  • In San Mateo county, where stop spacing is closer, slower local trains will operate every 20 minutes. These local trains will meet the express at Redwood City, before turning back north.
  • Dumbarton service will operate every 20 minutes, meeting the express at Redwood City with little or no wait to transfer to trains on the peninsula corridor, before turning back towards the East Bay.
  • Because the overall pattern repeats every 20 minutes, HSR will operate 3 trains per hour rather than the planned 4. Otherwise, there is a harmonic mismatch between the HSR frequency and the Caltrain frequency. 4 HSR trains per hour in a clockface timetable forces the base 'takt' to increase to 15 minutes, which is not desired.
  • If we're going to make Redwood City a major node, it certainly rates HSR service, so we will create a new mid-peninsula stop for HSR.
This is the resulting timetable (see also additional data on service pattern), shown here for one hour in the southbound direction only (the northbound side is symmetrical). Colors denote the 10-minute Caltrain express, the San Mateo local, Dumbarton service, and HSR.

Notice the express arriving at Redwood City at 7:43 meets the Dumbarton train departing at 7:44, and the local arriving at Redwood City at 7:52 meets the next express at 7:53. Every ten minutes there is a cross-platform transfer, alternating between express-to-Dumbarton and local-to-express. Counting both directions, a cross-platform transfer occurs at Redwood City every five minutes!

Implicit in this timetable are a number of other capital improvements besides a new Redwood City station, such as overtake tracks in various locations along the corridor (highlighted in yellow in this view of the timetable... and while we're here, look how much less yellow is needed if HSR uses the Dumbarton corridor via Altamont Pass). It's important to remember that there is no formulation of the blended system that avoids the need for overtake tracks, unless one is willing to push slower trains into station sidings to sit for at least five minutes while a faster train catches up and pulls ahead. If you are a Caltrain rider, you should be wary of the cheapskates at the HSR authority who want to do this to your commute.

Deriving the functional requirements for the Redwood City node

To enable this timetable, we need the Redwood City station to have the following attributes:
  1. Four platform tracks serving two 400-meter long island platforms to facilitate both northbound and southbound cross-platform transfers of very long, high-capacity trains.
  2. Platforms centered on the best cross-town corridor, namely Broadway, for convenient access to and from local destinations on foot, by bike or scooter, by bus, or using the planned Broadway Streetcar.
  3. A turnback track that enables certain Dumbarton corridor trains to originate and terminate in Redwood City, without fouling other train traffic, long enough for an EMU-8 train.
  4. A turnback track that enables the San Mateo local to turn back in Redwood City, without fouling other train traffic, long enough for an EMU-8 train.
  5. Elevated grade separation of all downtown Redwood City crossings, enabling free flow of pedestrians, bikes and vehicles under the rail corridor and including the re-connection of streets currently cut off by the existing configuration (e.g. Hopkins and James).
  6. Bus facilities placed directly under the train platforms for seamless connections without the need for an umbrella. Same for an eventual Broadway Streetcar.
  7. No mezzanine level. Mezzanines needlessly drive up the size and cost of stations, and impede and complicate vertical circulation. Street level can fulfill all the functions of a mezzanine, including ticket sales, wayfinding, waiting, retail, and dining.
  8. The shortest and fastest possible vertical circulation (stairs, escalators, ramps, and elevators) using a U-shape viaduct cross section to avoid deep and vertical-space-wasting bridge structure. This helps with transferring quickly between the two island platforms, as would be needed for example to continue from the Dumbarton corridor south to Silicon Valley.
The footprint of such a station is not small. However, Redwood City has plentiful available railroad and transit district land, and the street level interface of such a station can be integrated into the city's street grid, opening up cross-corridor access and avoiding a wall effect. The aging Sequoia Station shopping center, with its wasteful surface parking, can be demolished and redeveloped to make room for an expanded station. Station parking can be moved underneath the approach structures, protected from the elements.

One possible station layout
An optimal station layout has four tracks, with the outer tracks for HSR and express commuter trains. The middle tracks are for commuter trains, and allow both northbound (Dumbarton) trains and southbound (San Mateo local) trains the opportunity to turn at Redwood City without impeding the flow of express traffic. The width of the structure is about 130 feet, as shown in the cross section below:
The northbound express track (Track 3) is tangent. The northbound island platform is 400 x 10 m. The center commuter tracks (Tracks 1 and 2) have curves that are not laid out in detail; this detail does not matter since any train that uses these tracks would slow and stop at Redwood City, using standard trackwork and turnouts. The southbound express track (Track 4) is the tricky one: it wows around the station, passing the southbound island platform on a 7500 m radius curve with approximately 1.5 inches of superelevation (not enough to matter for platform lateral tolerances). This track consists of a double reverse curve with six spiral transitions (tangent, spiral, curve, spiral, tangent, spiral, platform curve, spiral, tangent, spiral, curve, spiral, tangent). The curve is necessary to fit a pair of 400-meter island platforms (long enough to berth a double-length high-speed train) without bulldozing too much real estate.

Here is how this all fits (admittedly just barely) in downtown Redwood City:

The sacrificial victim is the Sequoia Station shopping center and associated surface parking crater, which can be redeveloped as part of the station complex with direct access from El Camino Real. Access for high-rise fire apparatus around the viaduct structure might also be a concern for the new condo buildings to the south, although this can be mitigated.

The station includes two pocket sidings to turn commuter trains. The siding south of the station can turn Caltrain locals at Redwood City, while the siding north of the station can turn Dumbarton service. Each siding is sized to store an eight-car EMU. Track center spacing is 15 feet throughout, and platform setback is 6 feet from track center. All viaducts are made from low-profile U-shaped sections that minimize the required height of the tracks and also double as sound walls, reducing the noise of up to 30 trains that would serve the station every peak hour.

Redwood City's slogan, "climate best by government test" would also become "transfer best" with timed, well-coordinated transfers to a variety of destinations. The impending start of designs for grade separations in Redwood City needs to factor in this future, and the city ought to think big.


  1. This is excellent work, but I think that in the end, the only mid-peninsula HSR station that makes sense is Palo Alto. If I want to connect to HSR from the east bay, BART to Diridon make much more sense. I still think Dumbarton rail is a good idea, but I just don't see it justifying a HSR station in Redwood City, local politics aside.

    I'd love to see a timetable analysis of extending the SM local down as far as University in PA. It seems like there is plenty of space there for a transfer station, and the whole University Ave grade crossing needs a complete rebuild in any case.

    1. This is the sort of thing that is really easy to explore using the service pattern generator. You can read the help file here, with thanks to Richard Mlynarik, on whose server this runs.

      The issue with moving the connection point further down the line is that you end up needing a second express overtake, since the local trip time grows to 20 minutes of difference with the express, i.e. two express headways. That's ignoring HSR, which would be a third overtake. That starts to get very risky from the perspective of resiliency to small delays.

      To make a cross-platform connection in Palo Alto, the overtakes end up in awkward places on the corridor: San Francisco with difficult-to-expand tunnels, downtown San Mateo and Menlo / Atherton. These are three of the worst places to add tracks, each for their own reasons. The timetable looks like this.

      The Redwood City plan works better because the local runs only about 15 minutes slower than the express, so you can get away with a single express overtake north of San Mateo, and the HSR overtake south of San Mateo. Showing only the southbound side it looks like this. To space trains out I had to get a bit fancy and make the local skip 22nd and Oakdale, while I forced every express to stop there. Lots of trade-offs to get the overtakes to happen in the right times at the right places.

      This whole exercise gives you an appreciation for just how difficult it is to run an efficient and dense Caltrain service with HSR "blended" on top. If you don't want to put Caltrain in the hole (on a siding waiting for the fast train to pass, wasting countless passenger-minutes) then you need expensive overtake tracks. This is why the current HSR plan is for zero overtake tracks and lots of Caltrain passengers sitting in sidings waiting to be overtaken.

    2. Thanks for the detailed response. It makes sense that the travel time difference to the transfer point needs to be minimized to reduce required overtakes.
      However, it looks like your proposal needs 4 tracks pretty much everywhere south of San Bruno, including downtown San Mateo, with an overtake in Burlingame, so I'm not sure there's that big an advantage.
      I guess it's time for me to try playing with the service pattern generator and see if I can come up with anything interesting. Thanks for the pointers.

  2. Given the fact that the San Mateo Local and the Dumbarton Express have the same base frequency, and same suggested 8-car trains, why doesn't it make sense for Dumbarton trains to _become_ the San Mateo Locals? They could just wait on the local platform at RWC for an Express to catch up before proceeding. Yes having trains wait at platforms is not ideal mid corridor but here there is nothing lost in terms of schedule for passengers, there is something gained for them by not having to change train between San Mateo locals and Dumbarton trains, and furthermore, the double pocket track would be unnecessary.

    I must be missing something for why this isn't the suggestion?

    1. It’s about resiliency. To passengers, overtakes and cross platform transfers are about speed and convenience. To train operators, they are a terrifying source of risk because you have to absolutely nail the timing, or everything unravels in a domino cascade of delays. To protect against this, you build a storage buffer at key nodes in the system to maintain the cadence and absorb minor delays, given the occasional train being late. That way, a stuck door on the Dumbarton express in Newark doesn’t result in a total meltdown of the peninsula commute.

    2. You don't suppose it would cause a capacity bottleneck to have (potentially full) Dumbarton trains unloading all their passengers at RWC and having them cram onto every second Caltrain Express making its way up from San Jose, which would, likewise, already be mostly full by that point? And the reverse, where every other Caltrain Express is met by southbound Local and its passengers are then loaded onto the (mostly full) southbound express?

      If the Dumbarton Express and the Caltrain Local were set up to meet at the same 20 minute interval rather than alternating with each other every 10 minutes, then the total capacity in at RWC would match the total capacity out at every interval. Only Caltrain Express at :00 :20 :40; Both Caltrain Express and Caltrain Local==Dumbarton Express at :10, :30, :50. No bottlenecks.

      Really if you leave in the pocket tracks (For better resiliency thanks to the ability to turn in either direction at RWC) and follow the string diagram for this that you linked to, you could double the frequency of Dumbarton service to 10 minute intervals, and have half of them meet expresses at RWC and turn (as you propose) and half of them continue as Caltrain locals. This reinstates the bottleneck at every other 20 minute interval but oh well.

    3. Do you envision three trains entering the station simultaneously? Sounds iffy in this small a footprint.

      Keep in mind (referring to the full bi-directional string diagram and timetable) that Dumbarton arrivals are timed with ~4 minutes to departure of the southbound express, so the Dumbarton service already connects to two Caltrain expresses (the northbound right away across the same platform, and the southbound a short down-across-up to the other platform.) Vice versa for Dumbarton departures.

    4. I would envision 3 arrivals and 3 departures centered around those 20 minute intervals, instead of three arrivals and two departures (with the third train going out of service to the pocket track). To me, three trains going in and two trains going out has the makings of a bottleneck, although it might not be that big of a deal if there's enough spare capacity, if none of the inbound trains are full when they reach RWC.

      I really don't know how to use that Takt tool, but something like this at least at RWC.

    5. At any rate this is really great work, I think it all makes sense, and the above comments of mine are really just a nitpick of the first degree. The origin of this nitpick is how it's just so frustrating that we Americans are so poor at operations and maintenance that we have to build redundancy and resiliency in all over the place. Meanwhile in Japan, they avoid crossovers and bidirectional signals almost entirely unless they're needed to enable a specific planned service pattern and carry somewhere around five million passengers per day on a single (circular) regional rail line.

  3. Move the station to RWJ and we are all set: https://www.google.com/maps/place/Ebbsfleet+International/@51.4436464,0.319386,488m/data=!3m1!1e3!4m5!3m4!1s0x47d8b68f10b0ee51:0x7acf75bdde51bcc4!8m2!3d51.4431765!4d0.3215084

    1. https://www.google.com/search?q=ebbsfleet+international&tbm=isch&source=lnms&sa=X&ved=0ahUKEwi6u6_U5preAhXPrVMKHXiZAPEQ_AUIDCgD&biw=1536&bih=714&dpr=1.25

    2. The future: https://youtu.be/w9uYgnMrIjs

    3. What is the reason for moving the Redwood City train station out of the city center? If there was no room I’d understand, but there is room. Central stations work better than peripheral stations.

    4. 1) There is no room.
      2) There is a seamless Broadway Streetcar connection at RWJ.
      3) Dragging DBR all the way to "downtown" RWC is nearly as intelligent as transferring from the Santa Teresa line to the Winchester line at the San Jose Convention Center (are you related to Rod Diridon?)

    5. 1) I'm sure readers will be able to decide for themselves how much room there is, using the map I provided. Perhaps you'd like to elaborate on your interpretation of the constraints.
      2) As you know, Broadway goes nowhere near Redwood Junction.
      3) The downtown station is less than a mile from Redwood Junction and is a far more relevant origin and destination.

      Perhaps this would be a good time for you to explain what problem this Redwood Junction station placement is solving, because it still isn't clear to me. Sounds to me like another British solution looking for a problem.

  4. I'm a bit confused why you need the pocket sidings to turn around the trains. Is there anything preventing them from reboarding at the same platform and switching tracks once underway?

    1. (1) The pocket siding allows minimum 15 minutes turn time to absorb delays, i.e. the northbound SM local is not the same train as the southbound SM local that just arrived. (The way I diagrammed it, it’s a utilization-killing 29 minutes... room for improvement!)

      (2) to ensure cross platform transfer with the expresses, the local needs to switch platforms

      (3) the “slow” platform edges are re-occupied in less time than the turn period, so not OK to park there

    2. "Notice the express arriving at Redwood City at 7:43 meets the Dumbarton train departing at 7:44, and the local arriving at Redwood City at 7:52 meets the next express at 7:53."

      What could possibly go wrong?

    3. Ah yeah, that makes sense. I was thinking they would be turning around ASAP and hadn't looked too closely at the timetable. Should've known you put more work into this than me. :)

    4. @Anon, "What could possibly go wrong?"

      I'll tell you what could go wrong: the train could be run by people who don't care about providing a mobility service to actual living breathing riders, and care only about running choo-choo trains back and forth in spite of all attempts by the self-loading freight to disrupt the regular movement of trains.

      Or we could imagine a world where riders are served with efficient and seamless mobility that minimizes transfer friction, with rail operations taking on the challenge of punctuality with renewed purpose. In a world like that, two trains reliably pulling into the station at the same time to exchange passengers across the same platform is an unremarkable occurrence. You'd have to be deeply steeped in British and American practice to believe this to be risky or impossible!

  5. 1) Why clog a badly congested line with a stupid "San Mateo Local"?
    2) WTF happened to the Baby Bullets???

    1. What I call a "San Mateo local" is a train that serves the stations not served by what you might call a "Baby Bullet" in your outdated terminology. It is not "clogging" anything; it is providing a valued service throughout San Mateo county.

      As for "Baby Bullets," they're finished, gone, forgotten! There is no longer a need for them in a world with trains that don't accelerate like snails. A "Baby Bullet" is an artifact of the 79 mph diesel freight commuter rail world, and can really just be defined as a train that makes the SF - SJ run in about one hour, just like the 110 mph Caltrain express in the timetable described in this post. So there's your "Baby Bullet," except it's better because it comes every ten minutes and stops near you.

      Where today's "Baby Bullet" concept fails is in Silicon Valley where people and jobs are sprawled out: we have designated a few stops (e.g. Mountain View) as special stops with abundant service, at the expense of others (e.g. Cal Ave, San Antonio or Lawrence) that are vastly under-served with one or two trains an hour. Those other stops are no less populated or jobs-rich than Mountain View. It's just that a 79 mph diesel can't reach them without busting the one-hour SF-SJ mark.

      The "Baby Bullet" branding was a stroke of genius, turning a slow diesel train's shortcomings into a badge of honor and prestige.

      So, in my scenario, the "Baby Bullet" is an irrelevant and antiquated idea, and you should not feel the slightest outrage about it!

  6. Pretty dense schedule, indeed; I like it.

    How does the track occupancy look at Transbay? Would there be non-revenue runs in and out of Transbay?

    1. Good question... as diagrammed the local has a 23 minute turn and the express a 21 minute turn at Transbay, based on a generous 15 minute minimum in the input form. That clearly won’t work, as it requires too much train parking at Transbay. If you pull in the minimum turn times to 10 minutes it works (express turns in 11 minutes). That means 4 platforms for Caltrain and 2 for HSR, which seems about right for 9 tph Caltrain / 3 tph HSR. Thanks for raising this important point.

    2. Well, in Transbay, 4 / 6 platforms can hold two Caltrain EMU-7s, so as long as crew can be refreshed and some rush hour trains cleaned in San Jose, there is potential for increasing capacity or adding more buffer there to turn trains.

    3. Thanks, Clem.

      Those turnaround times look to me like eternities… Because of its integrated systems, turning around of a KISS can easily be done within 6 minutes. In the proximity where I live, there is a service using FLIRTs (which have pretty much the same control system), where the driver shuts down on the incoming side, walks to the outgoing side, restarts and gets ready within 7 scheduled minutes, and there are rarely delays.

      So, one should be able to turn around in 11 minutes. During the day, could it even be possible that a local switches over to an express and vice versa?

      Those turn-around times also should give sufficient time to do a quick cleaning. It would not even have to be as intense as the Japanese do it with their highspeed trains…

      Martin: In case of a very quick turning around, it would be organizational, but the outgoing driver could already wat at the end of the platform, and take over as soon as the incoming driver has shut down. The incoming driver could then either have a break, or leasurely get ready to take over the next outgoing train.

    4. That makes me wonder if the procedure for turning around a KISS EMU is faster than one required for the current diesel set.

    5. I’m sure they can work on their operations discipline to get turns down to 10 minutes. This has a large benefit on the capital side as well, since the required fleet size to operate the same service drops by two trains, saving about $100 million up front.

  7. Caltrain facing deadline on PTC and possibly large fines, if not stoppage of service.

    From: "Morning Transportation"

    PTC TLC: FRA Administrator Ron Batory met last week with positive train control system suppliers tied to the eight railroads at risk of not meeting the deadline to implement PTC or qualify for an extension, Progressive Railroading reported . After Dec. 31, the railroads will face fines of up to nearly $28,000 per day. "Railroads, PTC system suppliers, and other key stakeholders must seize the remaining fourth quarter with finely coordinated efforts in order to meet the statutory prerequisites of Dec. 31," Batory said, according to the publication.

  8. As a corollary, a Dumbarton train brings into questions of doing new stations at Newark (mentioned in the Samtrans paper) or Shin Park (mentioned in the minutae of a larger BART funding bill). With the suggested Facebook station and existing Fremont station, this would give transbay planners five core stations to chose from.

    Also if RWC were to ever actually do a streetcar and a ferry terminal, it'd be natural to extend it down Middlefield to connect with VTA at Mountain View. Political feasibility aside, it would give VTA riders an alternative to CC transfers at San Jose and Fairfield freeing those trains up for Sacramento area riders. I'd also pull off local riders, freeing up more space for longer distance commuters.

  9. https://www.mercurynews.com/2018/10/23/european-experts-will-craft-big-san-jose-train-station-revamp/

    1. ...they had me until the "Grand Central of the West" trope. This has to be the third "Grand Central of the West" project in the Bay Area alone. Architectural form is important, but transportation function is crucial.

    2. The hired architects designed Rotterdam Central Station which was actually quite smart in it's planning,
      EXCEPT that Holland went to ticketed platform access which ruins a lot of the urban connectivity. Most of the station which seemed intended to link the city on either side of the tracks is only accessible through ticket gates. I think it was changed after the station was designed.

    3. Clem - Obviously you need to call your Redwood City Station Idea the "Grand Central of the Peninsula" if you want to convince the politicians and press.

    4. I'm not asking for much, but if a person going from BART to Caltrain or HSR in San Jose has to cross a street or driveway, the design is a failure. Let's see what happens. My bar to success is low.

  10. @Randy, true. Though there is still a ticketless ped/bike tunnel next to the station (actually the passageway to the platorms of the old station). And using your transit card to go through the station is free if you tap out again within 60 mins.

    1. ... not disastrous but there are alot of shops and restaurants on the wrong side of the gate. I (as a tourist) was excluded. Love how all these Dutch stations have parking for literally thousands of bicycles.

  11. The State Auditor has just released a very critical report from its audit of the High Speed Rail project.

    A summary of the report is at:


    The full report is at:


    and a one page Fact Sheet is at:


    1. I'm not seeing anything new here. Looks to be more of the same-old...

    2. You're not reading closely enough, then. The level of incompetence is breathtaking, and I hope the new administration will shake things up. CARRD has the details on Twitter.

    3. Full disclosure, I've only read the summary so far. I'll be reading the full report tonight.