The following fairly simple argument shows why Hillsdale is the only feasible overtake location on the Caltrain line, and why Caltrain's major civil works should be focussed very, very stringly around (possible phased) of construction a mid-line amplification around a four-track, two-island-platform Hillsdale station.

The final suggested operations pattern/infrastructure I describe below is shown in this diagram.

My starting assumptions are always that:

• Clockface schedule is non-negotiable.
  30 minute period should be the baseline, given billions invested to be in the line; 15 is aspirational; 20 is possibly an intermediate stage (20 peaks, 30 off-peak, though it results in BART-like messy irregularity.)
  Caltrain planning should aspire to an 18-hour-a-day 7-day-a-week base service level that frees its customers from complex timetable consultations and that facilitiates reliable and regular connecting feeder bus services at any station.
• A minimum number of simple stopping patterns is key.
  Apparently arbitrary messes (like today's Caltrain clever-but-flawed compromise-of-necessity) actively confuse and repel potential riders, and larger than 30 minute headways for stopping patterns are simply unacceptable.
• There are two classes of stations: major and minor.
  All regular expresses stop at all major stations.
  All local trains stop at all minor stations.
  Caltrain is not a large enough or varied enough system to require or support significantly more complexity than this (with the possible exception of supplemental peak-of-peak services.).
• Same-direction stationary overtakes of locals by expresses while both are stopped at major stations, using cross-platform transfers to effect minimum-time station-to-station trip times for a rich selection of O+D pairs, is the most desirable base scheduling solution for Caltrain, given the above.

In essence, most of the above points reduce to "if it's good enough for the Dutch and the Swiss and the Germans it's good enough for me", or, "copy what has been shown to work", or, "emulate, don't innovate".

Timetables should be memorized ("express at 17 and 47 past the hour, every hour"), not carried around and constantly consulted. Deciding whether to board a train should be a trivial yes/no decision ("is my station served by express trains or only local trains?"), not an error-prone and intimidating consultation process.

Slight variations to strict service pattern regularity are possible in special cases. Adding 22nd Street stops for expresses in one direction only in AM/PM is the only one that springs to mind.

Supplemental peak trains in addition to the clockface might have customized stops (driven either by market demand, infrastructure inadequacies or a combination), but the base service regularity is non-negotiable.

Again, if it works for the people for whom things demonstrably work, it's good enough for me.

With only two train types, both operating predictably and regularly, we can provide scores of competitive trip times, without operating a mess of different customized train patterns.
This is simple, it's clear to riders, it's easy to operate, and it's much more efficient to operate (more people served better by fewer trains, combined with "good" trains for a given O+D operating minimum every 30 minutes rather than every hour for today's ad hoc collection of stopping patterns = better loadings = better revenue/cost.) It's also pretty clever. Lastly, people already know how to change trains at MacArthur station on clockface-scheduled BART with zero time penalty and near-100% reliability, one is not attempting to impose something foreign or suspiciously newfangled on the Sophisticated Bay Area Transit System.

OK! So those are the assumptions. I've never seen anything to indicate they're not the right ones!

So, building on that, the Major Stops are -- and I think this is pretty non-controversial, especially since we're guaranteeing far better service to all stations, including Burlingame and Belmont, with this plan:

• MISSION BAY (aka Fourth and Townsend, aka Fourth and King)
• MILLBRAE (for political, not "intermodal" or ridership reasons)
• HILLSDALE (for logistical reasons)
• REDWOOD CITY
• PALO ALTO
• MOUNTAIN VIEW
• SAN JOSE

The overall major/minor stop pattern is:

So, here's the simple argument, at last!

It comes down to this simple table, explained more below:

Now I claim we can model the run-time penalty for making a local stop versus running through a station as an express acceptably accurately as a constant per stop, given Caltrain's comparatively flat and comparatively straight route, and assuming station spacing of 1.5 miles or better. This penalty is somewhere around 2.0 minutes at present, and should decrease to somewhere around 1.5 minutes in the future under even unambitious operations/rolling stock/platform improvement scenarios. (1ms^-2 braking/acceleration = 2.2mph and a top speed of 35ms^-1 = 78mph, combined with 30 second dwells, both easy goals, gives a 60 second penalty. So we're can imagine being in the 90 second ballpark. The same direct calculation with present-esque circa 0.5ms^-2 "performance" and circa 1 minute dwells, pencils out correctly at 110 seconds stop penalty.)

So let's sketch out what happens when different major stations are chosen as the overtake location.

If, for example, Millbrae is the overtake station, then at the time a southbound local starts to decelerate for San Bruno should be 1 stop penalty period ahead of the overtaking express; 2 penalty periods ahead at South SF, 3 at Bayshore; 4 at 22nd Street; and 4 at Mission Bay (where the express also stops).
Heading past the overtake, by the time it slows for Hillsdale it will be 2 penalty periods (corresponding to Burlingame and San Mateo) behind the express. By the time it reaches San Jose it will be 10 local penalty periods behind.

Putting this all together in tabular form, with entries corresponding to penalty periods that an express ought to be behind or ahead of a same-direction local train at a particular stop, we arrive at the table above.

Now, to be realistic, given the Egregious Right-of-Way Sabotage Campaign of the period 1992-1999, it is today quite infeasible to construct express/local coordinated-overtake (four tracks, two island platforms) at either Millbrae (ROW given away to BART) or Mountain View (ROW given away to VTA) at any time in the forseeable future.

So let us consider only the three cases of Hillsdale, Redwood City and Palo Alto from here on.

For a stop penalty of 1.5 minutes, the above translates to:

With such (optimistic) train performance, the above shows that no additional scheduled express/local overtakes should take place on the Caltrain line for local headways of greater than 10 minutes.

Buy high-performance trains and operate with a lot of discipline (two corners of the famous Swiss "rolling stock"/"service plan"/"infrastructure" Magic Planning Triangle) and one need construct only one overtake location line-wide, at a choice of one of three locations.

For a stop penalty of 2.0 minutes:

• This shows that if the coordinated local/express overtake point were at Hillsdale, a southbound express would catch up to a second local train (assuming 15 minute local headway) one stop before San Jose, ie around Santa Clara. So passing tracks would be necessary at that point. (Because of the inherent symmetry of the service pattern, this also means that a northbound express would pass a second local train around the same point.)
  Fortunately, such passing tracks are feasible to construct around this point, and appear to be part of Caltrain's longer-term plans.
  No additional overtakes are scheduled north of Hillsdale, so further extension of the existing short Bayshore quadruplication is not necessary.
• If the local/express overtake were at Redwood City, an overtake would occur one stop before Mission Bay, ie around 22nd Street. This is an infeasible construction project, so a timetable featuring 2 minute local station stop penalties, 15 minute local train headways, and coordindated overtakes at Redwood City is not a candidate.
• If the local/express overtake were at Palo Alto, an overtake would occur three stops before Mission Bay, ie around South San Francisco.
  Providing the amplification for this overtake is a feasible, if expensive construction project, involving extension of existing quadruple track from CP Brisbane, though South San Francisco and San Bruno stations, as well as quadruplication around Palo Alto (at least south past California Avenue, and ideally north also across San Francisquito Creek and through Menlo Park.)
  Extension of the Lawrence quadruplication would not be necessary to provde this service pattern (in isolation).

For a stop penalty of 2.5 minutes:

• This shows that if the coordinated local/express overtake point were at Hillsdale, additional overtakes of 15-minute local service would take place around Bayshore and Sunnyvale.
• If the local/express overtake were at Redwood City, additional 15-minute overtakes would take place around South San Francisco and Santa Clara.
• If the local/express overtake were at Palo Alto, an additional overtake would occur around San Bruno.

This shows that a 2.5 minute local stop penalty combined with reasonably frequent service results in excessive capital costs (a bad 1940s choice of "rolling stock" compromising the "infrastructure" and "service plan" corners of the Magic Planning Triangle.) I claim that it is self-defeating to make any plans for significant expansion of Caltrain service or major investment in Caltrain infrastructure under such a train performance assumption -- the wasted cash involved would be better spent building BART to Santa Barbara or something.

I claim that overall the Hillsdale overtake pattern is the optimum realistic and feasible service pattern for the Caltrain corridor, for the following reasons:

• A 2.0 minute local (minor) stop penalty is a realistic planning and operations goal, allowing a great deal of recoverability, especially when combined with higher-performance trains and level train boarding.
• A timed-transfer cross-platform overtaking move near the midpoint of the line results in the most rider-friendly regular service pattern.
• Construction of a new and suitable Hillsdale station is already a medium-term (or even near-term) Caltrain goal, is supported by the host city, and appears in official city plans.
• Quadruplication to support this service pattern is only necessary at locations
  1. associated with the coordinated overtake point (Hillsdale and adjacent stations) itself; and
  2. as a second phase to support of 15-minute local service headway at a location (CP Bowers to CP Coast) where is both fairly technically simple to construct (wide right of way and only three significant rail bridges), where there is little lineside residential (NIMBY) development, and where it ties in simply with existing and planned quadruplication at either end of the necessary amplification to provide a usefully long and flexible overtaking stretch (existing CP Hendy to CP Bowers quaduple track linked to future CP Coast to San Jose amplification by others.)
• Incremental extension of the extent of the quadruplication is possible (to the south, placing the Downtown San Mateo Grade Separation and Right of Way Sabotage Outcome Disaster in the "too hard" basket for now) in a small number of steps which are technically managable, operationally useful, and perhaps even politically feasible:
  1. Quadruple Ninth Avenue/CP Palm (MP18.3) to Forty Second Avenue underpass (MP21.0), constructing the already-planned new four-track two-island Hillsdale Station;
  2. Quadruple MP21.0 to CP Ralston (MP21.6) (ideally combined with first phase; this is technically separable, but at greater mobilization expense).
  3. Triplicate south from CP Ralston by building an express track on the east (southbound) side of the tracks around Belmont Station from MP21.6 just north of San Carlos Sation MP21.0. This involves no reconstruction of the new-ish Belmont station, and should involve no or minimal ROW takes on the east side.
  4. As ridership builds (as it inevitably does based on predictable and faster timetables), and the local train service interval is decreased below 30 or 20 minutes, add express tracks south of CP Bowers to meet with quadruple track at CP Coast (by others).
  5. At a much later date (the above infrastructure is adequate to operate the service pattern with acceptable reliability), complete quadruplication through Belmont, San Carlos and Redwood City stations to Redwood Junction.
     San Carlos would be reconfigured as a single island platform station with the existing eastern platform widened, the western platform abandoned, and two new tracks (local and express) constructed to the west of the existing ROW, in conjunction with planned Transit Oriented Development and relocation of the historic station building.
     Redwood City station would be relocated to a location between Brewter Avenue and Broadway (circa MP25.3) and configured with two island platforms, with scope for eventual expansion to full HSR (400m, 1300') platform length.
• Even though an ideal mid-point overtake would involve two or more local stations both north and south of the major station overtake point (providing more reliable on-the-fly catch-up overtake approaching the major station and and pull-ahead following it, and thus less likelihood of late lead trains delaying followers), the choice of Hillsdale is an acceptable service plan versus infrastructure compromise.
  Slightly delaying northbound local departure from the Hillsdale station platform (1 to 1.5 minutes) allows the overtaking express to depart and start to clear (shorter, closely-spaced) track blocks through two-track downtown San Mateo; and scheduling southbound locals to arrive 2 minutes early and dwell at Hillsdale awaiting the southbound express' cross-platform arrival enables low-average-delay entry of the latter into the station.
  Alternatives to this small (circa 2%) increase in local train trip time include:
  • Extending track amplification north of CP Palm through San Mateo station, building many highly desirable grade separations, though in a highly constrained ROW with enormous construction and staging challenges
  • Retaining a station at Hayward Park, and using the local stop penalty it induces reduce scheduled extra dwell at Hillsdale. However, given that the Hayward Park station site is literally within sight of the proposed relocated Hillsdale station's platforms (risking politically dangerous analogy with Burlingame/Broadway) and that the capital costs of building a centre-island (local tracks) station at Hayward Park will be significant due to the need to build a pedestrian underpass to reach the platform and due to the constrained right-of-way width at that location, the compromise of slightly delaying local trains at the Hillsdale station while avoiding the capital costs of a Hayward Park station seems to be the correct solution.
  • Designating Redwood City as the overtake point, and quadrupling from north of San Carlos to south of Menlo Park -- a solution which may even superior from a service planning perspective due to the potential importance of the Redwood City station, but which involves great political challenge (San Carlos station ripped up; Atherton hysterics; much Menlo Park uncertertainty) beyond the comparatively settled and approved situation at Hillsdale and the opportunity for a comparatively low-impact incremental triplication though Belmont station
  • Abandoning the concept of regular interval service and cross-platform, same-direction local/express overtakes.

-- Richard