Reservations about GPS Watches (Read 1795 times)

kcam


    My dad, now retired from farming and other stuff, sometimes helps out other farmers by driving their tractors. He's told me about how using GPS they can drive up and down through a field w/out touching the steering wheel. The "system" takes care of ensuring no over-lap and no gaps.  Knowing the error in the watches, I figured there had to be something else accounting for the error.  Makes sense...

     

    What you said about cycling and their speeds intrigues me for this reason. My "running track" is on a bike path loop around a local park. I've measure the distance with a wheel so I can do fairly accurate 200s, 400s, 800's, 1K's, etc.  Well, when I use my GPS watch at this park, the GPS watch is closest to the wheeled markings when I walk slowly. When I run or jog, the distance is considerably off, similar to what you'd experience running on a regular track. In both cases, my watch reads farther than what I actually go every time... which makes sense based on other information posted here. 

     

    Joe Scott can probably explain this much better than I (or correct me where I'm wrong) but Garmin has a proprietary software algorithm called Smart Recording that, supposedly, does a better job of guessing your actual position rather than relying on the one current recorded point.  It tries to decide if you really 'turned' and went 10 feet to the right or if that was just a one-data-point inaccuracy.  It must do something like looking at the last several datapoints and from those points plus the current point decides the most likely path.  It also discards data points when it thinks you're not changing direction.  Anyway, I'm guessing that the slower you move the better this algorithm works.  My 305 has an option for 'Smart Recording' or '1 reading per second' recording.  I choose the former because I find it to be more accurate as well as smaller data size.

      Joe Scott can probably explain this much better than I (or correct me where I'm wrong) but Garmin has a proprietary software algorithm called Smart Recording that, supposedly, does a better job of guessing your actual position rather than relying on the one current recorded point.  It tries to decide if you really 'turned' and went 10 feet to the right or if that was just a one-data-point inaccuracy.  It must do something like looking at the last several datapoints and from those points plus the current point decides the most likely path.  It also discards data points when it thinks you're not changing direction.  Anyway, I'm guessing that the slower you move the better this algorithm works.  My 305 has an option for 'Smart Recording' or '1 reading per second' recording.  I choose the former because I find it to be more accurate as well as smaller data size.

       

      I believe I have smart recording enabled.  If I get bored next week while recovering from my marathon I think I'll try it w/ Smart Recording and w/ the "1 reading per second" and compare the results.

       

      I've bugged Joe enough with Garmin questions, and gratefully has answered a handful of questions.. 

      joescott


        Hey guys, I didn't read the whole thread by a long shot, but I noticed the end of this discussing smart recording.  Smart recording only affects actual recording, i.e., what gets written to the FIT files.  My understanding (99% certainty) is that under the hood the distance calculation (and I am absolutely certain the pace calculation) is still taking place on every GPS sample so that accumulated distance should not differ at all between smart recorded and 1/s recorded data.  But now you guys have me curious and I'll double check this statement tomorrow just to make absolutely sure.  If there is an advantage to 1/s recording it is that it typically makes for a smoother track log by rounding out the corners in the appropriate places.  The big disadvantage is that the FIT files are a lot bigger.  I use 1/s recording, but that is because I often do a lot of per-second-based post processing of various things at the office on my own data (since I know what the ground truth was!).  But from just a normal user's perspective I would have no problem using just smart recording.

         

        So, there are other reasons why getting the distance right can be challenging, especially on curved paths, but smart recording is not your bogey man.  Principally, the reason for the difficulty on curved paths has to do with the fact that every time you turn your watch sees a different constellation of satellites.  I'm sure many of you have noticed that this is often particularly -- shall we say, "not good" on a track where you are turning 180 degrees every 200 meters and your body is blocking satellites x, y, and z and now has a good lock on satellites a, b, and c.  I find that on my "straighter" runs GPS accuracy is pretty darn good.  On the track it is sadly often off by as much as 2-3% (and as others have noted, always "long", i.e., the watch says 1.02 when you are crossing the 1600 meter mark).  

         

        I have a feeling I've missed out on a lot of the rest of the conversation above, and I might have raised more questions than I've answered.  But hopefully this helps some of you.  I'll try to pay more attention to the thread going forward.

        - Joe

        We are fragile creatures on collision with our judgment day.

        joescott


          GPS units used for excavation and agriculture....ones that are very precise compared to running watches...use a land-based system to remove most the the error in GPS.  It's an add-on that adds considerable size to the unit and you won't find that in today's watches.

           

          GPS used for these applications uses another piece of equipment as spaniel has said that enables the system to actually compare the exact phases of the signals between the known reference point (the fixed "land-based system") and the tractor's GPS receiver, allowing extremely precise position determination.  This would be entirely possible, by the way, with running watches, say if you were running around the track and  you had the reference receiver sitting out there at the 50-yard line.  I personally don't think we will ever see this in consumer running watches.

          - Joe

          We are fragile creatures on collision with our judgment day.


          Feeling the growl again

            GPS used for these applications uses another piece of equipment as spaniel has said that enables the system to actually compare the exact phases of the signals between the known reference point (the fixed "land-based system") and the tractor's GPS receiver, allowing extremely precise position determination.  This would be entirely possible, by the way, with running watches, say if you were running around the track and  you had the reference receiver sitting out there at the 50-yard line.  I personally don't think we will ever see this in consumer running watches.

             

            The closest I have seen is a backpack-filling unit carried by scientists in Death Valley to track mysteriously moving rocks.  I think it would negate the advantage of wearing flats.  Big grin

             

            It works amazingly on my dad's tractor though.

            "If you want to be a bad a$s, then do what a bad a$s does.  There's your pep talk for today.  Go Run." -- Slo_Hand

             

            I am spaniel - Crusher of Treadmills

             

            xhristopher


              Rather than starting a new thread, I thought I'd add on to this one with this question, since it's somewhat related.

               

              Are the Garmin cycling computers more accurate than the Garmin watches?  Or is the accuracy primarily a function of the accuracy of the data? 

               

              BT, I'm not sure I gave you the answer to the question you were asking? That's what I get for attempting to mentally process and respond to a thread in about 45 seconds.

               

              BTW, I don't get this GPS watch trend anyhoo. These work great for me.

               

                No problem! I have a similar measuring wheel for when I want to be accurate for "track" workouts.  Who makes those?

                 

                BT, I'm not sure I gave you the answer to the question you were asking? That's what I get for attempting to mentally process and respond to a thread in about 45 seconds.

                 

                BTW, I don't get this GPS watch trend anyhoo. These work great for me.

                 

                joescott


                  Rather than starting a new thread, I thought I'd add on to this one with this question, since it's somewhat related.

                   

                  Are the Garmin cycling computers more accurate than the Garmin watches?  Or is the accuracy primarily a function of the accuracy of the data? 

                   

                  No.  Smile

                   

                  There is one *critical* difference though between the cycling computers and the running watches.  The antennas in the running watches are optimized for the loading seen by the antenna in close proximity to the wrist, whereas the antennas for the bike computers are basically optimized for a "free space" environment hanging out on your stem or handlebars.  If you (and I have done this) strap a cycling computer to your wrist its antenna performance and thus likely overall GPS performance will be worse than that of a running watch.

                  - Joe

                  We are fragile creatures on collision with our judgment day.

                    No.  Smile

                     

                    There is one *critical* difference though between the cycling computers and the running watches.  The antennas in the running watches are optimized for the loading seen by the antenna in close proximity to the wrist, whereas the antennas for the bike computers are basically optimized for a "free space" environment hanging out on your stem or handlebars.  If you (and I have done this) strap a cycling computer to your wrist its antenna performance and thus likely overall GPS performance will be worse than that of a running watch.

                     

                    Thanks! I'll just abandon my thought of running while pushing or pulling a bike :-)

                    LedLincoln


                    not bad for mile 25

                      The closest I have seen is a backpack-filling unit carried by scientists in Death Valley to track mysteriously moving rocks.  I think it would negate the advantage of wearing flats.  Big grin

                       

                      It works amazingly on my dad's tractor though.

                       

                      Even better:

                      Google in the Grand Canyon

                       

                      GPS is probably an integral part of these Google Street View camera assemblies.

                      kcam


                        Hey guys, I didn't read the whole thread by a long shot, but I noticed the end of this discussing smart recording.  Smart recording only affects actual recording, i.e., what gets written to the FIT files.  My understanding (99% certainty) is that under the hood the distance calculation (and I am absolutely certain the pace calculation) is still taking place on every GPS sample so that accumulated distance should not differ at all between smart recorded and 1/s recorded data.  But now you guys have me curious and I'll double check this statement tomorrow just to make absolutely sure.  If there is an advantage to 1/s recording it is that it typically makes for a smoother track log by rounding out the corners in the appropriate places.  The big disadvantage is that the FIT files are a lot bigger.  I use 1/s recording, but that is because I often do a lot of per-second-based post processing of various things at the office on my own data (since I know what the ground truth was!).  But from just a normal user's perspective I would have no problem using just smart recording.

                         

                        So, there are other reasons why getting the distance right can be challenging, especially on curved paths, but smart recording is not your bogey man.  Principally, the reason for the difficulty on curved paths has to do with the fact that every time you turn your watch sees a different constellation of satellites.  I'm sure many of you have noticed that this is often particularly -- shall we say, "not good" on a track where you are turning 180 degrees every 200 meters and your body is blocking satellites x, y, and z and now has a good lock on satellites a, b, and c.  I find that on my "straighter" runs GPS accuracy is pretty darn good.  On the track it is sadly often off by as much as 2-3% (and as others have noted, always "long", i.e., the watch says 1.02 when you are crossing the 1600 meter mark).  

                         

                        I have a feeling I've missed out on a lot of the rest of the conversation above, and I might have raised more questions than I've answered.  But hopefully this helps some of you.  I'll try to pay more attention to the thread going forward.

                         

                        "1-second recording doesn't measure any more frequently then smart recording. It just records more data, often data that are unnecessary to reconstruct the run. For example, if you run a constant speed for 10 seconds in a straight line only the start and end points and times are needed for a complete picture.

                        The technique is similar to compression algorithms used in other applications."

                         

                        This is what I believe Garmin is doing with Smart recording. Can you verify, Joe?  

                         

                        MTA:  I assume they are doing this type of compression which is why they call it Smart Recording rather than just recording a datapoint every, say, four seconds.  I would say that the Smart recording has the potential to be more accurate IF you are not running a lot of curves on your run ... say an out and back on a straight road.  

                        joescott


                          "1-second recording doesn't measure any more frequently then smart recording. It just records more data, often data that are unnecessary to reconstruct the run. For example, if you run a constant speed for 10 seconds in a straight line only the start and end points and times are needed for a complete picture.

                          The technique is similar to compression algorithms used in other applications."

                           

                          This is what I believe Garmin is doing with Smart recording. Can you verify, Joe?  

                           

                          MTA:  I assume they are doing this type of compression which is why they call it Smart Recording rather than just recording a datapoint every, say, four seconds.  I would say that the Smart recording has the potential to be more accurate IF you are not running a lot of curves on your run ... say an out and back on a straight road.  

                           

                          Yes.  The difference is ONLY in what gets written to the file, not what is actually measured and calculated. 

                           

                          And Yes.  It is called "Smart" recording because it is adaptive to changes in the path of the run.  On a very straight path it will record fewer points than on a very curvy one or one with a lot of tight turns or corners.  I would agree that Smart recording is more accurate (in terms of reconstructing the track) than if we just willy nilly decimated everything down to a 4-second recording interval.

                          - Joe

                          We are fragile creatures on collision with our judgment day.