One of our scientists here at LSU has a bunch of el-cheapo GPS receivers that are mounted in several dozen automobiles for traffic studies. The things are single-board computers with some sort of Asian GPS chips ("GPS-LOGGER-J-FMPO4-C"). When people park their cars overnight, the chips start recording positions (L1 only) in a series of "steps" both north-south and east-west that seem to be equally-spaced in a grid-like pattern, regardless of the satellite geometry throughout the night.
Anyone care to offer bright ideas why this curious behavior happens?
How big are the steps? How far does the position move? Are the N-S and E-W movements correlated or does it go just any direction?
My older Garmin 76s has an internal resolution of about 2 or 3 feet. Although that is less than a least digit of the lat-lon display, there are ways to see the resolution through distance between waypoints and other tricks.
So I can imagine a cheapo design having some truncation or rounding in the computations or trig functions that gives steps of several feet. If the steps are bigger than several meters/a few 10's of feet, then it's a piece of junk design.
L1 signals only? Without a base unit?
I think that is normal for single point L1 deviation. This is the reason why a base unit is need for PP to get smaller error radius.
How Many Satellites Can They Actually Receive?
First consideration, with limited satellites, as the visible satellites change their software may not actually pick the best combination to solve from.
Secondly they may be defaulting to specific grid limits.
I can see my L1/L2 receivers telling me they are continually on the move as they keep averaging greater and greater numbers of positions. However post processing to other fixed points proves that they never moved.
Most of that error is in the unprecision of broadcast orbits, on top of atmospheric variation. With care you could actually determine how weather fronts move across your position.
End result, don't just rely on raw positions.
Paul in PA
I don't think he was asking about the error radius, which would be most likely as large as, or larger than, that of a recreational unit that tracks the code chips on L1. I think it was the quantized, stepped nature of the error that puzzled him.
My $0.02 worth:
If these are tracking routines and the unit is static; I suspect some algorithmic-least-squirrel voodoo that attempts to predict (or narrow) the next solution by utilizing the relationship of the previous solutions.
Remaining static throws a wrench in its solutions by having a value of 0 in the d/dx and it resorts to a predetermined but possibly random 'grid' matrix.
Chuck Norris probably couldn't explain it either...but he could get them to 'stop it' by just narrowing his eyes.
Not really related, but GPS World ran a side by side comparison of survey grade receivers and el-cheapo ones, specifically in urban canyons and as tools to track real time position of cars on the road and in traffic.
The el-cheapo ones were significantly better.
