This site uses cookies to offer you a better browsing experience. Find out more about how to manage cookies, or dismiss this message and continue to use cookies.

What is GPS accuracy? Testing the Garmin GLO in the forest

Boyd 1725 Points
edited December 2015 in GPS Discussions
Have been working on a pet project over the past week. I've been mapping the little trails on my property with various GPS devices for almost 10 years now. Have gotten very inconsistent tracks with a Garmin eTrex Legend C, GPSMap 60csx, Oregon 400t and Montana 600 in the past. So I figured it was about time to see how well the Garmin GLO bluetooth GPS receiver performed.

There are about 1.4 km of trails on my land, and over the course of a week I racked up a total of 26.7 km walking them over and over (and over....). :) These trails are very narrow, about 1/2 meter wide. I clipped the GLO to my cap so my body wouldn't interfere with the GPS signals and recorded the track with the Galileo app on my iPhone 6s Plus. The GLO is capable of providing 10 position updates per second, and a quick look at the tracklogs showed that Galileo was recording an average of 6 per second.

The forest canopy is very dense and it's hard to see the trails on aerial imagery. But there's one area where I can just make them out on the New Jersey 2012 orthophotography. So I picked this area to do a little analysis of my data. I have traced the trail positions in red here.

Note - you should be able to see the full sized images by clicking on them. This works fine for me in Firefox on my Windows PC, but for some reason it doesn't seem to work in Safari on my Mac. If the images don't expand when you click them, try right-clicking (or control-clicking the Mac) and you can open them in another window.

image

Now according to the metadata, the NJ 2012 orthophotography has a resolution of 1 foot per pixel and +/- 4 foot horizontal positional accuracy with 95% confidence. Here is the same map with 4 foot buffers around the trails. Theoretically, the actual position of the trails falls somewhere inside the blue shaded area.

image

And here are the raw tracks overlaid on the map.

image

Next, I converted the tracks to points. There are a total of 23,964 track points on the map, each is represented by a one pixel white dot. 50% of the track points fall within the +/- 4 foot margin of error of the aerial imagery. To make it easier to see, I will use a solid color background for these illustrations.

image

I did some further examination, and based on this test I think we can conclude that the GLO achieves an accuracy of better than +/- 4 meters 95% of the time.

image

It appears accurate within +/- 3 meters 89% of the time and within +/- 2 meters 73% of the time.

image

image

Of course I would like to have seen a tighter pattern of track points, but remember these images are zoomed WAY in. These tracks look a lot cleaner than anything I've recorded with my other Garmin devices and it's nice to have 6 position fixes per second (Garmin handhelds only provide one fix per second). All things considered, I think the GLO performs quite well for a $90 device. :)

All of these images were created with GlobalMapper. It was necessary to use double-precision coordinates or the data was really mangled at these high zoom levels. Consumer software like Basecamp is not capable of this kind of precision.

Would be interested to hear your thoughts on my methodology and conclusions. Am I guilty of pseudo-science here? ;)

Comments

  • Tim 1456 Points
    I'm sure many of us could nitpick on some parts of the methodology, but I would be surprised if anyone actually goes out and does this and improves on what you've done. So bravo!

    From what I recall, at least many years ago, Garmin used a 50% benchmark to determine the accuracy estimate coming out of the GPS. For those who haven't heard about this before, basically they were (and maybe still are) tuning their estimated accuracy field to be accurate 50% of the time. So if the GPS says it thinks it is currently accurate to within 5 meters, that will be true 50% of the time and false 50% of the time.

    Given that 50% of your points were within 4 feet of the margin of error from the images themselves, Garmin might therefore suggest that the GLO achieves 4 foot accuracy. Not too shabby.

    While I have no idea what is being used for a processor in the GLO versus other Garmin devices, I wonder if the fact that the processor is dedicated to crunching GPS data only and not doing anything else helps contribute to the great accuracy being displayed here.

    I have a potential mapping project coming up next summer... sounds like GLO it will be.
  • Tim 1456 Points
    Oh, and it should be noted that Garmin claims 3 meter accuracy for the GLO.
  • Boyd 1725 Points
    edited December 2015
    Thanks Tim. :) Do you really think Garmin uses a 50% number? That would be a little deceptive, although I've never had any faith in those "accuracy" numbers. We discussed this way back in 2009: http://forums.gpsreview.net/discussion/comment/97067/#Comment_97067

    Look at the white paper linked to in that thread. I don't know whether or not that is true or whether it applies to newer devices.
    ______________

    "The estimated positional error (EPE) is really HPE (horizontal positional error). On the new Garmins (GPSMap 76, 76s’s and V’s, etc.) the EPE is approximately at a 95% (or 2 sigma) confidence level when you are not using differential corrections (WAAS or NDGPS) this drops to a 68% (or 1 sigma) confidence level when you are using differential corrections."
    ______________

    Interesting theory about the GLO processor. I think it's nice to have so many position updates per second as compared to a handheld. It provides much smoother track lines. AFAIK, all Garmin handhelds still have a limit of 10,000 points in a track. I was recording track at 6 points/second continuously for over an hour on a few outings, producing logs with over 20,000 points. Galileo had no problem with that. Don't know what kind of limits other apps might have.

    Ultimately, my goal is to make a very accurate map of my trail network, and that's difficult down by the creek with heavy coniferous tree cover. The tracks are starting to suggest where the trail is, but I'll have to keep walking to get 1 meter accuracy. :)

    image


  • Tim 1456 Points
    Boyd said:

    Thanks Tim. :) Do you really think Garmin uses a 50% number? That would be a little deceptive, although I've never had any faith in those "accuracy" numbers. We discussed this way back in 2009

    I don't know. As you point out, that was a long time ago. It was first pointed out to me by an engineer at the company of one of Garmin's competitors. I then approached a Garmin engineer about it, and they essentially confirmed it to me. Basically what I was told was that the engineers wanted to implement estimated accuracy at a 95% confidence level, but the marketing team said 50% and the marketing team won.

    Again, many years ago and even it was accurate at the time I wouldn't think it to be accurate now. And I believe the estimated accuracy field has disappeared from a few devices as well. I'm not sure it matters much at the consumer level, and what we are looking at here is pushing the boundaries of what a consumer level device can do.
  • privet01 142 Points
    edited December 2015
    Boyd said:

    Would be interested to hear your thoughts on my methodology and conclusions. Am I guilty of pseudo-science here? ;)

    No you are not guilty. <<grin>> It a good show of what a gps user can expect in the real world.

    My tracks on my Garmin marine gps's show slightly less accuracy when overlaid even though I travel pretty much the same path as I go along several miles of marked channel getting to and from the marina. And I'm in open water with no obstacles other than the mast of my sailboat. Though I've never really measured most the tracks probably fall within 20 feet of each other. I have seen them showing me well out of the channel when I'm dead center in the channel.

    I perceive some of the positional errors as seasonal, though again I've never specifically compared the data to be certain. Perhaps it's the changing positions of the satellites and some are dropping below the radio horizon.
  • Tim 1456 Points
    privet01 said:

    I perceive some of the positional errors as seasonal, though again I've never specifically compared the data to be certain. Perhaps it's the changing positions of the satellites and some are dropping below the radio horizon.

    Each satellite in the GPS constellation orbits the earth twice per day, so satellite position is constantly changing, but is equal through the seasons.
  • privet01 142 Points
    edited December 2015
    I was under the impression that the orbits where more seasonal as the satellites are NOT orbiting about the Earth's axis. Sort of like the Earth being tilted toward the Sun causes our seasons. This puts some at a less than optimum angle to make accurate calculations and/or for radio reception.

    Also, using this site calsky.com/cs.cgi they show anywhere from 6 to 9 gps satellites to be within an optimal position at my location during different times of the year. Of course I'm not looking for every hour of the day so maybe they do in fact balance out, but the not orbiting about the Earth's axis thingy still makes me think some won't be usable.

    But then again I can concede that maybe the number of usable gps sat's still make the calculation accurate enough for any kind of precision that we can determine through measuring our tracks.
  • alanb 347 Points
    edited December 2015
    I believe the article you linked is referring to communications satellites (for television, etc.) and not GPS satellites. I think the WAAS satellites are the only GPS satellites that hold a constant position over the earth. The GPS constellation is constantly changing in fairly short time spans which is easily observable if you watch the Garmin satellite screen for a few minutes.
  • Boyd 1725 Points
    edited December 2015
    Remember, the GLO also receives the GLONASS satellites. You can see the satellites at the site @privet01 linked to, you need to click the GPS/GLONASS link and also check the Russian GLONASS box. Not sure how that site works, I was able to choose my own location but the observer site still shows Northfield. Didn't spend much time there.

    Here is what my GLO sees in my yard right now.

    image

    And this brings me to something I don't like about the GLO and iOS. The screenshot above comes from the free Bluetooth GPS Android app running on my Galaxy Tab. Android doesn't natively support external gps, but this program is part of a simple hack to read data from the GLO and make it available to Android apps.

    iOS natively supports external gps devices, which sounds good on the surface, but the implementation is poor IMO. iOS is supposed to automatically use whatever gps data is more accurate, but there's no way to know if it's actually using the GLO. You can look at the Bluetooth menu and see if the GLO is paired, but that's all. iOS reports an accuracy number in meters, but never shows anything more accurate than 5 meters.

    Now some of the other device makers offer their own propietary apps to monitor their bluetooth receivers, such as dual and Bad Elf

    https://itunes.apple.com/us/app/bluetooth-gps-status-tool/id422661436?mt=8
    https://itunes.apple.com/us/app/bad-elf-gps/id391764718?mt=8

    Nothing like this for the GLO. It's funny though, when I first paired the GLO with my iphone, I got an alert saying something like "this device requires an app to function, would you like to download it?" I thought, "Wow, Garmin finally made a GLO app!"

    But the app turned out to be Garmin Pilot. :P Apparently you get a one month free trial along with the GLO. I downloaded just to have a look, and of course was clueless. It does display the GLO battery level, but no satellite or accuracy data.
  • alanb 347 Points
    @privet01 regarding the orbital period of GPS satellites, and the repeat period of the constellation, I found this short article kind of interesting: http://gpsinformation.net/main/gpsorbit.htm
  • privet01 142 Points
    The site I linked has a calculator to assess the orbits of just about every artificial satellite around earth, including GPS and GLONASS. You have to click the "satellite" button/tab at the top right of the page then select "GPS/GLONASS"
Sign In or Register to comment.
↑ Top