on a three-hour stroll in Kyoto, Japan
In the comments of a recent post about GPS receivers, it was suggested that the GPS receiver in the iPhone was useful for keeping tracklogs. I had bad experiences with iPhone location services when I tested in 2009, but perhaps my test wasn't good, or perhaps the old iPhone 3 wasn't good, so I thought I'd give it another try.
So, the other day I took three GPS receivers with me while I did some errands. As I'm apt to do lately, I walked.
This post is a comparison of the resulting tracklogs.
Garmin eTrex Legend HCx · Garmin eTrex 20 · black is iPhone 4s
The three devices I took along on my unscientific test:
Garmin eTrex 20 that I bought in Fall 2011 (product page at Garmin). Garmin has adjusted the hardware/software to make it even more troublesome to use than its predecessor, which is quite a feat, but on the plus side it not only contains a GPS/WAAS receiver, but also a GLONASS receiver, to tap in to Russia's own independent version of America's GPS. The theory is that more satellites in the sky means more satellites likely visible to the unit in any situation, and so better accuracy. I got the orange one, so its tracklog's plot is orange.
iPhone 4s with the MotionX GPS app. This is a GPS (update: and GLONASS!) receiver, likely(?) without WAAS. My iPhone is black, so its tracklog gets a black line.
The results from the iPhone was much better than I expected, and the others were worse than I expected. In the short plot segment above, which shows both the start and end of my walk, the iPhone is the closest to where I actually walked, including crossing the street toward the top of the image.
Garmin eTrex 20 is best along here
I walked along the north side of the street
iPhone Least Bad of the Three
I had planned to treat myself to the lunch delights at CHEZ LUC, but they were closed, so I wandered around until I found a place, at the center red dot in the next plot:
While there, I sat about two yards from a floor-ceiling window, and the two Garmin units reported accuracy as good as 10 meters, but the scatterplot above shows that they had no clue.
The more modern eTrex 20 perhaps had a better clue, because while in the restaurant it generally reported less accuracy than the older eTrex Legend HCx:
both wrong, but the eTrex 20 less so
Again the Legend HCx is off
I spent a few minutes inside a building where I had poor-to-no reception, but instead of realizing the lack of reception and pausing the tracklog, the Legend HCx gave incorrect readings scattered across a span of 310 meters. The iPhone mostly realized that it had no signal, and its scatter was limited to 80 meters. The eTrex 20's scatter was about 60 meters.
For while the Legend HCx just drifted off, at one point more than 200m off, as illustrated by the outward bow in the blue line at the left of the overall plot, which I'll repeat here:
Legend HCx Going Crazy
I've never seen it show a suspected accuracy that bad, so it wouldn't surprise me if it was showing its normal accuracy of 3m-10m while it was actually off by 200m, but unfortunately I wasn't looking at it while walking, so I don't know what it thought its accuracy was.
In the plot above, I crossed the street from north to south, then doubled back a bit on the pedestrian overpass (just to the left of center), to check out the view from above the street. Again, the iPhone seems the most accurate here, though the eTrex 20 at least makes it look like I used the crosswalk.
Not Too Bad Down This Stretch
but again, the iPhone is best
again the iPhone is best
Leaving Yodobashi Camera
it all looks crazy, but the iPhone plot is by far the best
More iPhone Bestness
I walked along the south side of the street
Heading Down to the River
eTrex 20 seems best
Walking Under a Bridge
The iPhone track looks best, but in this case it might be a side effect of laying down a plot point less often. The two Garmin units were set to log locations every second, but the app I used on the iPhone seems to have logged a location every six or seven seconds, which is just about the time it took to pass under the bridge.
the iPhone's less frequent updates smooths out the jaggies
iPhone wins again
I don't know how far I actually walked, but I can tell you what each tracklog tells me about my three-hour stroll:
The iPhone 4s tracklog contains 1,163 data points, and the distances add up to 11.8km.
The eTrex 20 tracklog contains 9,973 data points, and the distances add up to 12km.
The eTrex Legend HCx tracklog contains 10,428 data points, and the distances add up to 13.8km.
Remember, the older eTrex Legend HCx kept reporting a location even when it should have realized it had no idea where it was, jumping all over the place while I was in a building, and all that jumping added up to more “distance” and more data points. The other units were smarter in that they knew better when the accuracy was not good enough to count on.
Both Garmin units were powered on for the same amount of time and had the same 1-second logging frequency, so they should in theory contain the same number of data points, but the eTrex 20 recorded 455 fewer, representing seven and a half minutes where it choose to not record a location (because I was inside and it didn't have a good signal) while the Legend HCx continued recording its incorrect location.
Due to the nature of its design, the GPS system is not as accurate with altitude as with latitude and longitude, and on some Garmin units this is exacerbated with a worthless barometric barometer. I've learned to avoid those, so none of my units now have that, so the altitude in today's three plots is all from GPS (or, in the case of the eTrex 20 (and the iPhone!), both GPS and GLONASS).
It's almost comical how random it seems. Here are the elevation plots from each tracklog, along with an extra one for the elevation that Google Earth has on file:
Garmin eTrex 20 (GPS + GLONASS, 2011)
Garmin eTrex HCx (GPS, circa 2007)
iPhone 4s (GPS, circa 2011)
The Google Earth data seems closest to reality, except for the odd hump in the center (Google thinks there's a hill just north of Kyoto Station). The elevation in Kyoto slopes down from the north, and you can see in the Google Earth plot that I'm mostly level as I walk east to west, then slopes down as I approach Kyoto Station, then at the end climbs steadily as I walk north along the river.
The other plots are just ridiculous. (According to the eTrex 20, at one point I was at 225m elevation, and according to the Legend HCx I was at one point walking 0.68 mach!)
In the end, I'm dismayed at how bad my old Legend HCx was, and how not-much-better the eTrex 20 is, especially for having twice satellites available to it.
But most of all I'm shocked at how good the iPhone 4s results are... even judging from this one simple test, it's clearly good enough in outdoor use for geoencoding photos.