(by John Allen) Until recently (July 2020), I had kept the same smartphone for nearly 6 years -- a Samsung Galaxy S4 Mini. I liked it because it was compact. Despite my having dropped it numerous times, an Otter Defender case kept it looking and working like new. But recently, the time had come to replace the phone.
Among other problems, GPS reception had become completely unreliable. My bike club uses the RidewithGPS app to map routes and provide navigation. Sometimes the app would display the message "Acquiring GPS Signal" for several minutes. Often, navigation would quit in the middle of a ride.
A member of my bike club bought an old phone to use for GPS, without a connection to the cell network, uploading and downloading data over home wifi or with a wired connection to a computer. I tried this too with my son's old phone -- same problems. What could we do about it? Would a new phone solve the problems?
For answers, I turned to my friend Mark Sevier, who has explored the question in depth. To put it briefly:
Following here is Mark's technical explanation, in case you care for the details.
Global positioning systems work by measuring how long it takes a date and time signal to go from 4+ satellites to a GPS receiver. GPS is an extremely expensive overall system with an individual receiver which is remarkable in that:
Three satellites could provide a position fix, while the fourth satellite is apparently needed to avoid the need for an expensive atomic clock in the receiver. In the initial implementation of the full GPS system (starting January 1980), satellites sent a week number and the number of seconds into the week, using a 10 bit week number, meaning every 1,024 weeks the week number would roll-over from 1023 to 0 and start counting up again, as it did in August 1999 and April 2019; a 1,024 week cycle is called an epoch. Newer satellites use a 13 bit week number, so will roll-over every 4,096 weeks or 157 years, even though the satellites only last 15-20 years (probably contributing to the original 1,024 week decision) Depending on when the GPS receiver was made and how long the manufacturer expects the GPS to be used, the receiver can support expected changes in week number. See https://en.wikipedia.org/wiki/GPS_Week_Number_Rollover
After spending considerable time learning about how GPS works, and trying different software upgrades, reboots, and GPS-repair applications, I've discovered that my Moto G 1st gen (circa 2013) smartphone GPS will still work reliably if I turn on 'airplane' mode when using location services. This disables Assisted GPS, (AGPS), which uses the cell network and wifi in addition to satellite signals. It appeared to me that something about cell connectivity was leading to GPS drop-out. (A cell phone connects so that that the network can find it, even when not sending or receiving data. The connection can be lost, regained, switch to a different cell tower, and these operations burden the phone's processor.) I suspect that the problem was due to processing needs when re-connecting to the cell network after dropping the connection.
I happened upon this nugget by noticing that my smartphone GPS had many fewer drop-outs cycling in upstate New York than around home in Sudbury, Massachusetts. My cell phone service is good around Sudbury, but not in upstate New York, and there was none where we stayed. I usually turn on airplane mode when we arrive there. I didn't always remember to turn off airplane mode when we went out cycling. I was getting very good or nearly perfect GPS tracks during upstate New York rides, vs. maybe 95% GPS tracks around Sudbury. After a few weekends observing the difference in GPS performance, I figured that cell service might be a factor, and to try airplane mode in Sudbury. This change has yielded nearly perfect GPS performance since then.
AGPS can use wifi for precise location in your house/driveway, potentially leading to a GPS drop when you ride away, until the phone has a GPS fix from the satellites. I've read that this can take minutes, though I haven't tried measuring. I haven't found a place that describes how frequently the satellites send their signal, but I suspect it might be every 10 seconds, since that would allow 10 bits to cover a week's worth of seconds, with another 10 bits as the week number. I suspect the phone might use its accelerometer to fill in the blanks.
Last October, my smartphone GPS was happily recording my rides on Strava (an online mapping and navigation service), but in November it started drawing straight lines between a few points where it seemed to 'find itself'. Strava said this was due to a satellite 'week number' roll-over and to get any and all updates for my GPS / smartphone, which I reluctantly did. I liked my phone the way it had been in October and didn't want new 'bloat' to reduce battery life and slow it down. After I upgraded the Android operating system, re-installed the most recent versions of Strava, RidewithGPS and other apps and factory reset the phone, I had gotten to maybe 95% GPS reliability in the spring, but there were still times when Strava would draw a straight line between points where apparently the GPS information wasn't available. Also odd, in RidewithGPS, the elapsed time of a ride would be over 173,000 hours, nearly 20 years.
I couldn't understand how the smartphone GPS could work as well as it did most of the time, but then not work at other times. I came to theorize that my phone's internal directory of satellite tracks was accurate for all but one or two of the 24+ GPS satellites (hence GPS was lost when that satellite was part of my GPS fix), but my subsequent experiment with airplane mode suggests the main issue is that my phone's brain just isn't big/fast enough to manage cell and GPS tracking at the same time -- and/or possibly Strava and RidewithGPS have since fixed how they manage GPS-week epoch errors.
When I first started having a problem with Strava, some of my 2019 rides saved with 2000 dates. At one point, I started recording rides with RidewithGPS and uploading ride data to Strava, and soon after, Strava didn't accept some RidewithGPS ride data due to 'date inconsistencies'. Luckily, I had used a heart rate monitor on one of the rides, and could go looking in the TCX file (downloadable from RidewithGPS) for my heart rate near one of the GPS drops, where I found RidewithGPS had a data point with a 2039 date instead of 2020, off by the 19.7 year / 1024 week GPS epoch difference. I replaced the few errant dates (maybe 12 out of 2,000 total logged points in the TCX file) with the correct ones, and was then able to upload the ride to Strava - a demonstration of the epoch problem and a solution, if the information is accessible. RidewithGPS still shows ride times over 173,000 hours. Somehow not all the satellite track week epochs are correct, but probably both RidewithGPS and Strava are now self-correcting. My GPS tracks no longer have straight-line segments and are correct as long as my phone doesn't get over-busy dealing with the cell network. I expect continued success at least until some new smartphone bloat comes along.
I mentioned my success to Jack (another club member), who then had success for a ride (or two?) with his similar-vintage Moto E. For some reason, his phone GPS seemed to stop working, unlike mine, which continues to work well. So, I wouldn't say that I thought I knew about all old smartphone GPS's, and 'your mileage may vary'. One thing I dislike about electronics that phone home is that they can change without your knowing / understanding that they change, so they can go from working to not working for reasons you may never figure out. I realize that I got lucky to have figured out how to be able to continue using the GPS in my old phone, which otherwise meets my needs just fine.
On one over-the-air 'upgrade' of the software in my Tesla, the car stopped unlocking the car when I approach with the key in my pocket, a feature I would have preferred to keep -- but I haven't spent much time trying to get it back - goes to show that Strava segments keep me more interested than the hassle of double-clicking my Tesla key fob, though it is also true that I don't drive the Tesla as often as I ride :-)
This e-mail and the house it came from are powered by sunlight.
(by John Allen) I have visited the solar-powered house where Mark lives with his wife. It is remarkable and beautiful, and he built it largely by himself. So, please don't think ill of him for having a fancy car -- it is also ahead of its time. Also note that he has economized with an ancient (in Internet time) phone thanks to his tech savvy and inquisitiveness.
I have a less fancy house and car, but on the other hand, I recently replaced my phone with a spanking new Apple iPhone SE. Along with it, I bought the Apple Watch which a doctor had recommended for its health-tracking features. I find these features sometimes useful, sometimes hilarious and and sometimes annoying. Among other things:
Clearly, I have some work to do to tame the watch. But on the other hand -- the iPhone's GPS performance is excellent!
Last Updated: by John Allen