Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Erik Magnuson <mekire@xxxxxxxxxxxxx>
- Date: Sun, 16 Apr 2006 21:39:12 -0700
Terje Mathisen wrote:
Erik Magnuson wrote:
Stephen Fuld wrote:
Last week I attended a lecture on atomic clocks. You can now get them within a few cc, but there is work to get one to be essentially a chip type package. This is being driven by the military. The idea is to include one within a soldier's GPS receiver in order to be able to get your location accurately with one fewer satellite in view, for example, if you are in a a steep canyon (natural or man made - i.e. urban skyscrapers). And yes, they are planning to use Rb instead of Cs. The loss in accuracy is made up for by the smaller size and lower power requirement.
A likely civilian use of that capability is in using GPS for positive train control - where the receivers may spend a significant amount of time in cuts or other areas with restricted sky view. One problem is that GPS accuracy is insufficient to tell what track the train is on in multiple track territory.
My wife is project leader for speeding up public transport in Oslo, Norway.
She tells me that rail-borne equipment really doesn't need much in the way of GPS capability to know exactly where it is at all times:
The distances between stations are known at the sub-m level (thermal expansion doesn't add/subtract that much), and it is easy to recalibrate the system each time a carriage passes a control sensor.
I.e. if it took 567 wheel revolutions to get from station A to station B yesterday, chances are pretty good that it's going to be pretty close to that today as well.
If such a train ever slides, or lock the wheels when braking (probably because someone pulled the emergency stop handle), it has to go into the workshop to get rid of the flat spots on every single wheel.
Until that happens, you have a pretty easy way to measure distance. :-)
The issue with GPS accuracy is not so much knowing that the loco is at mp239.1 vs mp239.11 (where mp = milepost), but whether the loco is on track 1, track 2 , track 3, etc. The advantage of a GPS overlay is being able to tell where the train is in a 5,000 to 11,000 foot block - whereas the track circuit just tells you that the block is occupied. Could be a big difference between the train being next to the block entrance or block exit.
Modern freight loco's use a technique developed by GM, where the wheels are allowed to have a slight slip. The amount of slipping would be very sensitive to rail conditions (dry or covered with dew), so wouldn't be that reliable for measuring distance traveled.
Speaking of grinding, noticed that on my last trip on BART in 2003, that the rails looked in desparate need of grinding - and suspect the wheels weren't that great either.
Back to the original topic that got this sub-thread started - IIRC, you were the one who first reversed engineered the appendix H for the pentium - seems to me those timers would be just the ticket for getting sub-microsecond resolution.
- Erik
.
- Follow-Ups:
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Terje Mathisen
- Re: RFC - New Object-Oriented Method of Parallel Programming
- References:
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: rambam
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: anonymous
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Slawomir J.
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Andreas Buschmann
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: anonymous
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Jan Vorbrüggen
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: JJ
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: anonymous
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Jan Vorbrüggen
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Terje Mathisen
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Erik Magnuson
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Stephen Fuld
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Erik Magnuson
- Re: RFC - New Object-Oriented Method of Parallel Programming
- From: Terje Mathisen
- Re: RFC - New Object-Oriented Method of Parallel Programming
- Prev by Date: Re: RFC - New Object-Oriented Method of Parallel Programming
- Next by Date: multiple TLBs
- Previous by thread: Re: RFC - New Object-Oriented Method of Parallel Programming
- Next by thread: Re: RFC - New Object-Oriented Method of Parallel Programming
- Index(es):
Relevant Pages
|