[dorkbotpdx-blabber] Fwd: [freegeek-social] Bike Counter for WNBR -- ideas? volunteers?

[Mykle Hansen]

On May 28, 2008, at 7:42 AM, Jason Plumb wrote:

> I've always wondered how those things work and what the response  
> rate is like.  Do they really just give an on/off or are they using  
> a summing pressure system of some kind?  Are they *really* capable  
> of detecting when two cars/bikes pass over very *very* close  
> together?  Just given the number of cars on the road, isn't there a  
> pretty high statistical probability that two will hit in the same  
> millisecond (or lower)...so how do *they* handle it?

i don't know exactly how, but here is a paper from 2002, in which
NZ traffic engineers tested some of the tube devices on the market
for counting both cars and bikes:

    http://www.ipenz.org.nz/ipenztg/papers/2002_pdf/34_MacBeth.pdf

i worry, tho, that with their sample size of 50 bikes they're
not close to hitting the level of traffic we're expecting.  on the
other hand, it's all about the sample rate ... not just the chip,
tho, it's also limited by the properties of air in a tube.  but
if the rate is high enough, it ought to be able to reduce this
problem to an acceptably low level.

> At 15mph, bikes wheels spaced 1.3cm apart will hit the sensor  
> within 2ms of each other (hitting Paul's 500MHz example).  1.3cm  
> ain't big, but it also ain't tiny.  I could see this happening in  
> reality....so the question is, what do you do with the missing/bad  
> counts/anomalies?  Is it ok to drop somebody?

one option is to funnel the ride through some narrow counting
point where the possibility of side-by-side riders is lower,
and lay the tube there.  another is to use multiple counters
in multiple locations and average the results.  (i was going
to suggest running a group of tubes at different angles, until
i imagined riding over that naked. =)

-m-