Is there a way to improve Sense GPS accuracy?

[mmalach]

I'm unfortunately finding the Fitbit Sense GPS to be very inaccurate.  I've accurately measured a couple of my running routes using a couple of different mechanisms (and they are all in agreement).  The Fitbit Sense is consistently measuring my routes as short by approximately .1 miles per mile (i.e., my 4 mile route is being measured as 3.6 miles).

Is GPS accuracy dependent on where you live?  Is there a way to improve GPS accuracy?

Moderator Edit: Clarified subject

[t.parker]

@Rich_Laue are you sure your stride is the same every time? Mine isn't (the white line is stride length, and purple is cadence):

Every run results in a different stride length and even a 5cm difference may give a significant difference in total distance. If I use stride 1.18m (cadence 195) and run an easy flat run (the one with 0.98m, cadence 190) and will use famous steps x stride (still, I don't think Fitbit can't be that dumb to implement it this way) at 40min mark I would overshoot distance by 720m. That's a lot. So the question is still open - how do I know the correct stride that will work for all my runs?

There are two major inaccuracies coming from this assumption:

- a total distance of the activity, which sometimes will come close enough and probably won't matter that much for walkers (will matter for runners).

- local distance inaccuracy. Let's say you want to run 1km flat at n easy pace and then 0.5km uphill, then 0.5km downhill. With a fixed stride every distance marker will fall in a random place which would break distance based intervals. The elevation is just one thing affecting it.

See what it looks like on the treadmill:

The stride length is quite consistent until something changes. A treadmill forces a certain way of running and this is what keeps metrics more stable but a slight change in the gradient or pace will also change the stride length. In this screen, I included power (yellow line) that shows how hard I was running. 'Harder' means either faster or with more incline (or both). There is also clear evidence that cadence affects stride length: the purple line goes up, and the stride length goes down. This behaviour is specific to a treadmill, I find it hard to be observed outdoors as there are more variables caused by terrain and surface types.

And last but not least. To detect stride length you don't need GPS. Here's an example of a Fartlek workout:

For this one, I forgot to turn the GPS on. It didn't stop the device to detect elevation and stride. I would expect Fitbit to do something similar and maybe use stored stride length as data correction (so any inaccuracies from the sensors would be flagged as errors if for example stride will rise to 3m which is unrealistic).

In all the examples, the stride is not an input. It's an output, read-only value that is not used for distance computation. The stride length is a result of distance computation.

As I said, it may work for walkers as long there is a negligible variation in stride length (although lots of walkers report issues with distance, too) but that would not work for runners and for walkers who change stride length a lot (steep hills, soft or unstable terrain etc. will affect walking stride).

[Rich_Laue]

You are right, my stride isn't always the same. 

I measured the stride on multiple days, and took an average. 

[GAIslandGirl]

Hi, Parker,

Please forgive my ignorance, but I don't understand what stride length has to do with a GPS measurement. Doesn't GPS just measure your distance, irrespective of one's stride length?

Thank you

[t.parker]

@GAIslandGirl no, the GPS distance is used only for non-step based GPS activities like cycling. You can see it for example when stepping in place and distance increases. If you compare GPS distance of the route and Fitbit distance, they will never match. If you used GPS only then realtime pace showed by watch would be all over the place. Fitbit in that matter doesn't do anything unusual. Garmin and Polar also use inertial measuring methods in tandem with GPS but it kind of doesn't work so well.

[Rich_Laue]

I'm not sure how they put it together, as. @t.parker says, all none step activity needs GPS for distance. 

Step based activity has the option to use both, and with the small antenna, low signals, causing dropouts, clouds, trees, to close to a building, adds timing errors. Balancing the 2 dats, hopefully will get better results. 

Does it work? I don't know, I used 2 weeks of walks, over a known, 3 mile course and averaged the stride. 

Total distance for all walks / total steps for all walks. 

This way, my changes in how I feel and daily changed get factored in. This seemed to help in distance accuracy. 

It would be nice to have the option to turn off the stride factoring.