The problem of measuring courses has fascinated me for some time. Doodads like
the GPS units and accelerometer units never seemed to cut it.
Problems with accelerometers:
the accelerometers like the Fitsense and Nike SDM were not very reliable (too
much variance in multiple trips, too much dependence on variables that could
change during meaurement, such as pace. The calibration resolution on these
units, 0.5%, is just fine for estimating ones training milage (who cares about
knowing ones milage to within less than a percent ?) but inadequate for
computing pace (1% difference in pace is about 20-30 seconds difference in a
10k).
Problems with GPS units:
GPS units seem to produce more reliable measurements. Unfortunately, these
things are so good at taking tangents (including cutting corners) that they
seem to systematically underestimate distance. Some units also apply smoothing
algorithms to trackpoints, which may lead them to cut corners in a way that the
course does not permit. On winding courses through parks, the GPS unit I used
(Magellan Sportrak) produced odometer readings that were obviously short.
Importing tracklogs into GPS software and summing distances between trackpoints
improved accuracy, but the course measure still appeared short. Moreover, the
documented accuracy of GPS units -- a few percent, or even 1 percent -- still
does not cut it for measuring a course for time trials.
I recently stumbled across a fascinating website about how to set up a bicycle
computer to behave like a Jones counter (used to certify courses).
http://home.earthlink.net/~caverhall/newrevcounter/abstractcontents.htm
Basically, one can achieve excellent accuracy with a bicycle computer (USATF
certification rules require a bicycle odometer called a Jones Counter).
However, as has often been pointed out, the main problem with bicycle computers
are issues with tire pressure, and accurately knowing the wheel circumference--
a calibration issue. A course measure is only as good as the calibration.
So the problem with the typical odometer setup is that one only can measure
distance to 1 revolution accuracy -- about 2m on a bicycle. So a calibration
course of 200m still leaves you with a calibration resolution of 1% --
unacceptably low. While 1% might sound good, consider that it is 100m (about
20-30 seconds) on a 10k course, and that this is in addition to other sources
of error.
So the website proposes an ingenious way to to set up the bicycle computer to
count wheel revolutions instead, so that one can count the fractional wheel
turn. The fractional wheel turn (<2m) is not important in measuring a large
loop course, but it is important in the calibration phase. The way this is
done is that the wheel circumference is set to 2.5m and 4 magnets are placed
on the wheel. The odometer is set to display km, so each revolution of the
wheel reads as 10m or 0.01km. So with this configuration, the bicycle is a
good revolution counter.
Using this method, I was able to measure a 0.9 mile loop within 5 spokes
(5/32 revolutions, about 30cm) on two separate runs. My calibration course
was an athletics track (inside lane, trying to stay 6in from the white line).
I suppose there are possible improvements to this -- riding the straights
multiple times, or trying to ride exactly *on* the white line (that would
make it 399m)
Cheers,
--
Donovan Rebbechi
http://pegasus.rutgers.edu/~elflord/