page 47
Stewart's SEA-Horse Corral

A Study of the Accuracy of GPS


"The more satellites in view, and locked in, the better the accuracy."

That's not good enough. You want to know more than what the Garmin manual says.
What does it take to get accuracy of 1 ft?
Surveyors add a base station, differential GPS, to increase accuracy to the centimeter.

But how much variation and drift is there?

Here is one answer, and an example. The setup is one Garmin 62s, with external hi-gain antenna (gain range is 32dB up to 42dB) plugged in, and externally powered. The Track settings are to save one waypoint every 5 seconds. The antenna is placed on top of a vehicle, outside with clear view of the sky. No trees. Surrounding hills and ridgelines are at or below about 10 degrees elevation.

 
12 satellites, strong reception. Let the test begin.
Click to view animation. 10 hours in 4 minutes Once with the trail following each actual point, then again with the trail following each 6 minute averaged point. The usual time it takes to get one averaged waypoint.


One screenshot from the garmin per hour. See the filename for the time stamp.
 

The unit reports accuracy within 7 or 8 ft. How did it Really do?

Wobble and jitter of the actual track, from the average waypoint over the entire 10 hours.
Each hour has an individual color matching the animation.
There were a few times when accuracy really deviated from "actual location." 10:20 and 16:20 really stand out as times when the number of satellites in view was quite low. The reported location drifted up to 7.8 meters. Not feet.

When using the Averaged Waypoint function, how long does it take to get a Good data point?

Distance over time, from final center average.
Waiting 10 minutes gets the accuracy below one meter. This time waiting 40 minutes did not improve confidence in the waypoint.
The longer the sample size, the lower and smoother the wobble gets.
A sine wave can be seen the longer you wait. Slowly flattening.

Depending on what time of day, what satellite orientation there is, what solar conditions are affecting the ionosphere path, and how many satellites are overhead with strong reception, accuracy may be over 2.5 meters off. Just what the device was saying. 8 feet accuracy on average.
When there were 2 to 4 satellites too low on the horizon to contribute to the "triangulation" (except this is in more than 3 directions), then the accuracy suffered noticably. The position can really drift around at these times.

The longer the waypoint averaging time, the higher the confidence in the accuracy. But it looks like anything less than 5 hours, and your "confidence" in the result may be less than desired.

I hope it was educational to see just how much the gps location jumps around, using consumer grade equipment, and only capable of locking up to 12 GPS satellites. Newer equipment that can handle more sats, and more constellations (Glonass, Beidou, Galileo), will likely give different, and better test results.

Update: 2018.Apr.09
Due to popular demand, I am repeating the test with a newer GPS receiver that can receive signals from more constellations, and track more than 12 satellites.

Samsung Galaxy 5

Much smaller jitter tracking more satellites, but I am still recording a drift of a couple meters, and an interesting warm up period of 30 minutes, but much much smoother. I have one recording done, and the early drift suggests I need to repeat this, and try to get a longer sample period.
Click to view animation. 7 hours in 3 minutes Once with the trail following each actual point.




[Home] [Republic photos] [Horse and goat photos] [Electrical Hyper Sensitivity]
[e-Mail me]

© 2014   All Rights Reserved.
Stewart Andreason
Republic, Washington 99166
visitor #

gps-study.html   Updated 2018.Apr.12