This is a description of my trail odometer, also known as my Lummis Meter (he walked from Ohio to California and founded the Southwest Indian Museum).
I took about 2 years to develop an odometer that met my accuracy goals. I aimed for ±0.1%.
My first effort was with a 2-wheeled thing that was supposed to trail behind me with very little attention paid to it. Unfortunately, Euclid and Newton didn't agree. The darned thing would tend to crawl up the high side of the trail and bounce over all the rocks it could find.
Then I went to a one-wheeled arrangement, using one of the 1.5 x 12" tires and wheel from the two-wheeler. I found the one-wheeler worked pretty well as long as I was willing to push it in front of me. However, to steer it around turns, I had to tip it, thereby causing the ground track to roll on a different radius than when upright.
Finally I built a wheel out of a 1/2-inch x 12-inch pulley with a 1/2-inch fan belt glued into the pulley groove. I put on a longer handle so that I could comfortably push it in front of me. The flat fan belt had a constant radius no matter how far I tipped it, so I was able to steer it around corners without losing any accuracy.
All models were based on the concept of counting wheel turns. I found that reliable counters are hard to find, and those that appeared reliable were only so until they broke. Some would increment counts when the wheel bounced over rocks. To encourage reliability I mounted the counters up near the handle, where they were not likely to hit rocks and would be less affected by rough ground. Even so I played it safe and used two counters, and used the distance that seemed more plausible. When the counters disagreed, they were off by a LOT.
My final design had a cog-driven bicycle chain that drove an equal-sized 24-tooth cog on a shaft near the handle. I tensioned the chain by mounting the driven shaft on a sprung door hinge. One end of the driven shaft had a small cog pulley that used a small-toothed belt to drive another cog pulley on a turns counter. I sized the ratio of the small cog pulleys to cause the readout to approximate the number of miles to a resolution of 0.001 mile, allowing me to quickly read the approximate distance traveled. The other end of the shaft had a cam to cycle a five-digit event counter. With this arrangement, the only calibration necessary was to determine the number of wheel turns it took to cover a known distance. I used a low-expansion 100-foot surveying tape, properly tensioned, in the shade (under Bldg. 301 at JPL). Resolution of this form of calibration was about 1/8 inch or better, so the calibration supposedly was accurate to 1 part in 9600. To achieve the most accuracy possible I would make about 10 measurements and throw out one, if any, that appeared wild, and average the rest. It seemed as though one significant contribution to calibration error was a lack of repeatability in following the exact ground track each time I walked the 100 feet.
I would guess the final accuracy was 0.1% or better, but never proven to reach 0.01%. I felt that 0.1% was all that could be used, since the errors inherent in the actual distance measurement along a curving rocky trail would introduce errors much greater than 0.01%.
Anyone who buys a meter or builds one for himself should keep in mind that it must tip in order to steer around curves, and therefore the wheel must be flat so that when tipped, the rolling radius does not change.
I hope my experience in developing the Lummis Meter will be useful to somebody.
Copyright © 2001 by James S. Carter.
Permission is freely granted to reproduce any or all of this page as long as credit is given to me at this source:
Comments and feedback: James S. Carter
Last update: 22 February 2001