Thanks, David. I've managed to find the relevant thread on SubTalk:

http://talk.nycsubway.org/perl/read?subtalk=713864

:-( I go away for two weeks, then SubTalk's down when I get back!!!

"David Fairthorne" wrote in message .cable.rogers.com...
Excellent answer, James.

The station that you describe in Sao Paolo is Cornithians-Itaquera on line
3.

"James" wrote in message
m...
Depends on where the bottleneck is. Is it the station, or is it the
incoming lines? (though the Isle of Dogs line is mostly new and will
provide extra line capacity).

Most likely the station. Lines can carry X tph (where X is quite high
and depends on signalling and junctions) - but a terminus station can
only accept and despatch a certain number of trains due to its layout.

Look at the Victoria line. Presumably signalling can take the frequency
up to a high level (I believe 40tph was once achieved on the Central
Line a long time ago?) but terminal capacity (specifically the scissors
crossover) will limit this frequency - adding a loop past Brixton would
enable a higher Victoria line frequency because the terminus is
essentially eliminated.

There are essentially three terminus arrangements which maximise
frequency:
(1) The single track loop (eg Kennington and Heathrow in London, South
Ferry and City Hall on the IRT in NYC) - as you say, abolishing the
terminus - the problem is that this removes the layover point and
therefore doubles the length of the route, making the line more
susceptible to disruption.
(2) The multiple track loop (eg WTC station on PATH, NYC) - this
allows a certain amount of layover time to even out outbound service.
(3) A three track, two island platform terminal, where the following
moves happen (where track A is from the inbound track, track B is a
bay in between the other two, and track C leads to the outbound
track):
(i) train arrives on A
(ii) train arrives on B, train on A proceeds to turnback siding beyond
station
(iii) train departs from B, train in turnback siding pulls into C,
another train arrives on A
(iv) train departs from C, train arrives on B, train on A proceeds to
turnback siding beyond station... etc...
This third layout is only found at a station in Sao Paulo whose name
totally escapes me at the moment. A similar layout used to exist at
the BMT's Park Row El terminal in Manhattan (demolished in the 1940s -
so much for progress...)

Actually the highest frequency on a two-track line I've ever come
across is the original cable-hauled shuttle service (1883-1908) across
the Brooklyn Bridge (Sands St, Brooklyn, to Park Row). It ran 90 (yes,
ninety) tph peak, and averaged 40tph over a 24 hour period. I think
that if we saw their operational practices in use these days, we'd
have a fit, but nevertheless it goes to show what can be achieved with
decent terminals and without such useless objects as a signal system
;-)

At one time frequencies like 40tph weren't at all uncommon. Now you
only get them in Moscow, Sao Paulo and a miserly 36tph in Paris.
Indeed, the only thing which tends to restrict a 2-track line to 40tph
is the line staying two-track in stations. Once regular el service
commenced over the Brooklyn Bridge in 1902, the Sands St station had
four platform faces in service, with one two track line continuing to
Manhattan and two two/three track lines feeding in from various parts
of Brooklyn (these lines sub-divided further into about ten branches -
you can guess how intensive service would end up being!).

Less trains running into Liverpool St mainline = more platforms and
fewer movements across the approach junctions. This could be used to
improve reliability or add services (and here I suspect
adding services will generate more revenue than improving reliability)

Flat junctions aren't necessarily bad - you just need to time them
right. The junction at Sands St was a flat junction.

Agreed. But how does this suggest that spreading CrossRail out to
five* branches, each of 4 tph, is not the way to maximise CrossRail
benefit?

Five branches means huge operational complexity and extra cost - unless
they all have flying junctions then that adds conflicting movements,
which means reduced reliability.

Flat junctions can work on two track lines. They're only an absolute
disaster on three and four track lines (which is why they're so
necessary on the SWML). The ideal situation is to have the inbound
branch and outbound branch trains passing through the junction at the
same time. On branching systems, however, capacity is rarely an issue
once outside the central area. A flat junction where two lines at 4tph
merge, for instance, is not going to cause many "bangs" (trains will
be on average 7½ minutes apart (front to front) over the section with
conflicting moves - which with 600ft trains at 20mph will have an
occupation time of 20.45 seconds, still leaving over 7 minutes gap).
Even in a situation with lots of "bangs", service reliability can be
maintained by providing a third track just before the flat junction,
so that an outbound branch train waiting for an inbound main train to
pass won't delay an outbound main train (another thing which just so
happened to be a feature at Sands St... these Brooklynites knew what
they were doing...).

How do Grays trains get to Crossrail? There are a number of options but
they are either expensive or reduce reliability on existing lines. For
example, they could run from Stratford to Barking via Woodgrange Park -
but so do a large number of freight trains from Tilbury, which all have
to cross flat junctions. Or they could surface near Bromley-by-Bow
instead - but that means extra tunnelling and an underground flying
junction. Similar arguments may apply to any other branches.

Freight trains are something to avoid running on metro systems. They
are an unmitigated disaster (remember the Southern Electric's rush
hour freight embargo - they had their priorities straight there). If
they must be run, they should have a separate track or run between
2000 and 0600 (and on more lightly used sections also between 1000 and
1500).

*This makes 20 tph, compared to tunnel theoretical capacity of 24 tph.
I would then have train waiting at Liverpool St and Paddington to fill
any spare slots caused by late arrivals. These would only go Liverpool
St to Paddington, and make up the numbers to 24 tph.

This would require turnback capacity at Liverpool St (which will be more
expensive) and extra platforms at both Paddington and Liverpool St
(which will also be more expensive - particularly at Liverpool St which
I think will be bored).

A clear argument for making them Westbourne Park to Bethnal Green or
something instead. However, I see the flaw in this idea as being that
the delays are most likely to happen between these points.

Although I like the idea of standby trains, the more I think about it,
the more it seems unfeasible for a high-frequency service.

It has been done before on the IRT in NYC. It was the subject of a
budget cut.