If you have "use wireless networks" and Gps turned on, things don't always work right. If you have just Gps turned on the experience seems to work better.
I live in Madison, wi and having both turned on makes my phone think I'm in appleton (100 miles away). Using Gps, my location is reported correctly.
When out of Madison, and turning on "use wireless networks", things work correctly.
I'm not experiencing this issue with my own HTC Incredible, for what it's worth. Maybe there's some glitch with Google's wireless network database...
As for the Incredible's GPS performance, I couldn't be happier. It's accurate enough that when I go cycling and let My Tracks record my ride, I can zoom in and see where I headed out on one side of the road and came back on the other. The biggest downside is the elevation inaccuracy, but from what I can tell that's an issue with all devices that gather altitude data from GPS (rather than an internal barometer).
Somewhat unrelated to the original article, but: I got great results with GPS on my droid in north Florida, but I recently moved to Seattle and the performance decreased. It takes me 2-3 times longer to get a GPS fix, and it's often off by as much as a block. Could it be the latitude?
In theory, GPS fix quality is roughly the same in all mid-latitudes[0]; any difference between Florida and Washington should be lost in other factors.
Probably, either your sky is more covered in Seattle (by trees and hills, for example), or there’s more reflection and interference in the radio signal.
0. http://blogs.agi.com/navigationAccuracy/?cat=4 has some nice illustrations. “DOP” is a sort of ideal fix quality, with smaller numbers meaning more accurate fixes (under various assumptions); Wikipedia will explain the math.
Edit: https://gps.afspc.af.mil/gpsoc/PerformanceReports.aspx is a live-ish map. When I check it, FL and WA are both in the lowest (best) category. However, it’s worst-case DOP for only the 4 best satellites; under a clear sky, a good receiver will usually see several more. I often get DOPs of 1.5 in residential neighborhoods.
There's definitely some variability based on location. I've used my phone in various places around the Midwest. Usually, it's pretty good.
Last time I was on I-65 between Chicago and Indianapolis, it went haywire. Regularly claimed I was out in west Illinois near Decatur. (A friend of mine reported similar behavior.)
Other than that, it's pretty solid. I haven't really paid too much attention to how off it is, usually it seems to manage at least 30 meters without a problem.
Last time I was in Chicago with the Droid it was snowing pretty heavily, and I figured bad GPS readings were normal with that kind of weather. Didn't get to test it out in clear weather, unfortunately.
As an iPhone 4 user I find this depressingly disappointing but not surprising. Anecdotally I have found the GPS performance of the 4 to be far superior yo the 3G but, sadly nowhere near sub-street-lvel accuracy.
The problem with the iPhone 4 seems to be low sample size. Honestly, I would not be at all surprised if Apple lowered the sample rate of the GPS to extend battery life. GPS is a terrible power-sucker, and you probably won't know the phone is skimping on GPS readings unless you do comparisons like in the article, since most consumers are pretty happy with ~20-foot accuracy. People recording precision data, like surveyors or geocachers, though, would do well to get a device with a high sample rate and good accuracy.
My N1's GPS accuracy is easily good enough for my purposes; if I pull up Google Maps, it shows my location as about where my desk is on my lot in my neighborhood. I can't really complain.
The iPhone’s built-in Maps.app absolutely culls the GPS hardware to save battery life. Developers can (and do) change the default value for the CLLocationAccuracy method as referenced here:
http://developer.apple.com/iphone/library/iPad/index.html#do...
I haven't done any iPhone development, but can't the GPS sample rate be overridden in software? I know it can on Android -- My Tracks will let you tweak the sample rate if you want, for example.
I'm not an iPhone developer either, but I can't find anything in the docs that would seem to allow you to set a sampling rate. What you do is tell the OS that you want location data, and give it a desired accuracy (AccuracyBestForNavigation, AccuracyBest, AccuracyNearestTenMeters, AccuracyHundredMeters, AccuracyKilometer, AccuracyThreeKilometers) and the OS decides when to give you updates. It could be that Motion-X isn't using an appropriate resolution, but I'm somewhat dubious that that's the case, given its intended purpose would need sampling at more than a 30-foot resolution.
It's better if you have cellular reception (and sometimes you need your carrier to enable that feature for you).
Unfortunately, if you are using non-assisted GPS on these newer phones, you get poor performance. They were designed for AGPS and lack the necessary hardware that dedicated GPS units use.
Poor performance meaning usually very long acquisition time and less accurate reception (unable to handle many satellites, etc.)
The original iPhone did NOT have GPS at all. It had cell tower triangulation and Skyhook. Assisted GPS is a superset of GPS using other technology to assist GPS by doing things such as getting almanac data from a server instead of from the satellites or getting the time and approximate location which allows faster computation of position from the satellites, or offloading some of the computation to a server.
I'm suffering from confusion of terms. Folks have been referring to cell tower triangulation as "assisted GPS" in relation to cell phones, and the term has become somewhat muddled. It's not A-GPS as we have in modern hardware, but people tend to use the term "assisted GPS" to refer to cell triangulation/Skyhook systems.
I was thinking that, too, but I'd presume that an app designed specifically to do GPS measurements like that (for use in plotting a pedestrian route) would be using true GPS, since you have to be moving pretty quickly (like, traveling in a car) for assisted GPS to be useful. Additionally, on the Motion-X website, they say that Motion-X doesn't work on the original iPhone or the iPhone 2G, because it doesn't have the necessary hardware to get a true GPS signal. While it is entirely possible that true GPS was turned off (it'd be great if the author could verify/retest if needed!), I'm somewhat doubtful that those tests were done with only cell tower position estimations.
True GPS was definitely turned on. Motion X doesn't tell you how many satellites it's tracking, but handy runner was tracking 6 at the time (iPhone4 vs N1), and MotionX was showing an acquired signal so it must have been tracking at least three. I also tried to hold the phone as loosely as possible (although I doubt this made a difference).
There is a comment above on Apple choosing a lower sampling rate in exchange for better battery life. I think this is probably right on. Most users will only be using the GPS to mark driving directions or to record something like running. In these cases the GPS is definitely good enough (you're just hugging streets anyway). Unfortunately, it's not good enough for me : ).
I would definitely be more than willing to run more tests or try different programs if people throw out ideas.
Honestly, I think that's Apple being Apple. Cutting GPS quality means better reviews and a better user perception. No reviewer is going to write "Yeah, my GPS was only accurate to 14 feet", but you can be damned sure that every reviewer is going to write about the battery lifetime under normal usage conditions (including GPS). I'm not meaning to suggest that stacking reviews is the only purpose behind it - if it doesn't affect the user experience, then the battery savings is justified - but it is typical Apple.
By way of another example, their approach to being caught with their hands in the signal cookie jar with regards to the iPhone 4 signal problems (specifically, how they manipulated the bar calculation algorithms to show a stronger connection than other devices for a given signal level)? No worries, we'll fix that, but we'll make bars 1-3 taller so that psychologically, you feel like your signal is better. Apple is masterful at manipulating what people see to get them to believe that they're getting the best possible experience, but the laws of physics dictate that they have to cheat somewhere sometimes. I don't mean that to sound perjorative, but just to point out that part of Apple's ability to do user experience so well is their ability to manipulate the user into thinking the positive things they want them to.
I think the best way to deal with this is to offer choice - fine-grained GPS (+ associated battery cost) if the user chooses it; but I suppose that isn't the Apple way ...
I have similar experience on the iPhone 3GS right after I start some GPS-using application (like TomTom). It is sometimes really terrible in the first 10-15 minutes. Though after that, it becomes quite accurate. Often also earlier. Also it is often better when I start it already when being outside of the car and not when being in the car.
I guess that it searches each time for new satellite signals if the last usage of GPS was too long ago.
If a GPS receiver has no current satellite orbit data, it'll need to listen for 12.5 minutes to determine the current set of satellite orbit information.
There's a useful explanation of the various possiblilities of GPS startup (specifically talking about Garmin models and terminology, but more broadly applicable too) here:
How did you process/present that data after collection? Those look like Matlab plots - did you use it analyze the raw GPS data, or just use some GPS visualizer altogether?
If it would be helpful to you, I'd gladly repeat this experiment using my HTC Hero (CDMA) + Handy Runner, and a Garmin for reference.
Hi kgroll. Yes I used matlab to process the raw GPS data (lat and long). It'd be really helpful if you could try and replicate this test with a Hero. Do you mind emailing me please? Email: me [at] davidlokshin [dot] com?
I think the test should've been performed with more care, by using multiple GPS loggers instead of just one, to see if the result holds true. It might be an application issue at show here.
I used Run Keeper for the iPhone and it gave the same results as MotionX, which is why I continued using just one of them to collect data (and then write up the blog post). There's nothing to say, though, that both programs may just be crap...
If you have "use wireless networks" and Gps turned on, things don't always work right. If you have just Gps turned on the experience seems to work better.
I live in Madison, wi and having both turned on makes my phone think I'm in appleton (100 miles away). Using Gps, my location is reported correctly.
When out of Madison, and turning on "use wireless networks", things work correctly.
I have no idea why!