Walking pace difference - Sense vs Charge 4

[wicks21]

I recently upgraded from a Charge 4 to a Sense - finding a noticeable increase in time/decrease in recorded pace. As I repeatedly walk same circuits, I find the minutes/km is approx 10% higher with Sense compared to Charge 4. It is strange as I have not adjusted any settings.. Any ideas please?

[SOlewine]

I noticed the same in moving from the Versa 3 to the Sense. 

[Rich_Laue]

Is the distance and total time between the watches the same? 

[wicks21]

Excellent point - just checked... The elapsed time is the same and no of steps the same. However, the km is lower with Sense.

eg.  Sense 7.47km / Charge 4 8.47km.. So quite a difference in km which obviously calculates a slower pace. I am not sure why, I have had stride length set at auto always (both devices)

[LZeeW]

@wicks21   Are you using GPS when you track these walks?  If you are using GPS, Fitbit uses the GPS measured distance.  GPS isn't pinpoint.  You are somewhere within a certain radius.  My husband and I can walk side by side and get different distances.

[eezeepee]

@wicks21 Couldn't it be possible that the Charge 4 is the one that's been off?  One way to be sure which is more accurate than the other would be to wear both, make sure you have GPS on for both, the same stride length, and walk an exact measured path. I think, but I'm not 100% certain, that walk uses stride length for the distance. GPS is just for mapping.

[t.parker · Accepted solution]

@eezeepee "I think, but I'm not 100% certain, that walk uses stride length for the distance. GPS is just for mapping." - this is true.

@wicks21 GPS is also used (afterwards) to adjust automatically stride length. I'm not sure whether it's still an issue but with my first Sense I had plenty of GPS related issues (not connecting, disconnecting, dropping accuracy) and bad GPS could affect the next run with bad distance (due to automatic stride miscalculation). However, at some point, those issues disappeared (I also got another Sense, maybe the previous watch wasn't working properly, I will never know) and the distance calculation is pretty accurate (and GPS works well, too).

If you track your activities with GPS, then from Dashboard export the TCX file and upload it to https://www.mygpsfiles.com/app/. See, what is the length of the actual GPS track. It will be different from Fitbit distance but you will know how much is off (also, see whether the GPS path has been tracked accurately).

See examples:

5k, distance short by 100m

Half-marathon (+a bit), distance short 430m

The second case it's quite a long- istance and the difference is almost 0.5km (which makes difference to the pace). The first case isn't so bad (just 100m) but annoys me more because it cuts short the race distance but in general, it's an acceptable margin. Try to compare the GPS to Fitbit distance using the link I gave you and see the real difference (no point to compare Charge 4 and Sense if your point of reference is the GPS map).

[LZeeW]

@t.parker   Fitbit uses steps x stride length for non-GPS activities and measured GPS when it is used.  Help article  If you start your activity before the GPS is connected, you get a steps x stride length for that short portion.  Fitbit also has a Run Detect feature in the Run exercise shortcut in many models.  If you start running, because the mood hits you, your Fitbit will vibrate and ask if you want the GPS on.  You get a steps x stride length distance for that beginning portion.  

[t.parker]

@LZeeW help article is wrong unless it can explain why I can see distance increasing when a) GPS is connected and b) I am moving in place.

[LZeeW]

@t.parker   I wear a Versa 3.  When you see your distance increase, are you referring to the distance on your screen that is one of the real time stats or something else?  As a scientist, I love a good experiment.  So I did one.  When I walk, I try to keep a 15'00" pace (4 mph).  So I walked in place for five minutes, a time that should give me 0.33 mi.  It didn't.  It gave me 0.04 mi and a terrible scattergram GPS map that had me walking through a second story wall, landing in my front yard and ending in my back yard after walking through a fence.

[t.parker]

@LZeeW Sure, let's speak scientifically. It's more right up my street. You assume that distance, when no GPS is involved, is computed by simple stride x steps but it isn't that simple. The accelerometer data is used in order to know how much of the length of stride the step should use. Otherwise, it would be a very simple equation distance = steps x stride but if you try to divide the distance by a number of steps or stride length it never adds up. This is because you are missing the accelerometer input and/or (I don't know whether Fitbit uses it) gyro. This is the reason why while stepping in place you see a rather slow pace - because you don't have much forward acceleration. If you ever analyzed accelerometer input during a run or walk, it involves acceleration that increases the body's speed and on top of that is a loop made of acceleration and deceleration which is caused by arm motion. So how the watch knows we are moving? Because it computes speed based on that acceleration and speed will be non-zero. When we step in place we are not gaining speed (in theory) but the hand still accelerates and decelerates. We gain some speed when the arm accelerates and lose speed when the arm decelerates. The problem is that the final result of acceleration and deceleration may be often non-zero (unless somebody has a robotic perfectly accurate arm). This slight 'non-zero' contributes to the distance.

Now, the distance. I meant the one on the display and the one in the TCX file. The TCX file contains the "DistanceMeters" tag so it is easy to see how distance changes with each data point. There's also latitude and longitude for those data points and using the haversine formula we can very easily compute the actual distance between those waypoints as well as the sum of them. 

So I took the liberty of testing it myself. The GPS connected and I started "running" in the spot. There is lots of scattering as you have noticed but if you take all the coordinates, the distance doesn't match. In my case, the GPS distance was 440m while the one recorded in the TCX file (and seen on the watch) was 176m.

Now, see what's happening when I stop waving my arm:

Total distance (by waving arm)	Difference between waypoints in km	Total distance for selected GPS waypoints
135.49	0.001884411859217	SUM OF GPS
135.49	0.001964036716142	0.036417837733928
135.49	0.001501058859597
135.49	0.002603374448964
135.49	0.001086584149887
135.49	0.001884411859217
135.49	0.002315720822106
135.49	0.001074070287452
135.49	0.002641181205279
135.49	0.001708835363388
135.49	0
135.49	0.00279051646315
135.49	0.000189870595932
135.49	0.000982018358071
135.49	0.000585220480158
135.49	0.001312031921403
135.49	0.00050235037812
135.49	0.001170440960315
135.49	0.002230474470049
135.49	0.002999114098912
135.49	0.00093981150152
135.49	0.004052302935049

The distance registered by Fitbit doesn't change (constant 135.49m) but you can see that the difference between consecutive waypoints is non-zero and the actual GPS distance for this block of data is 36m. It may not look like much but if the GPS was in use here I would see something like 171m in Fitbit distance but I don't. Then, I started waving my hand slower but "deeper" and meters started to kick in much faster (in fact, the way you wave your arm and frequency of that affects that distance but this is because of the acceleration readings I mentioned above). I'm not going to paste the rest of the data, you can easily do the test by yourself. The fact is that the GPS distance doesn't match the exercise distance because it doesn't use the GPS to measure that distance. Yes, there is that scatter of waypoints but ironically, thanks to not using GPS to calculate distance, Fitbit avoids probably quite a serious issue. My test lasted 3 minutes (it's cold outside 🙂 ). Imagine, you're waiting on rail crossing for a few minutes and it would add 500m despite you're not moving.

One more thing to add, the GPS hasn't disconnected during that test. It is easy to detect as in case of disconnection two consecutive waypoints have the same coordinates. In my file, there was no such case.

[N8teGee]

I think a simple way to test this is to get into your car, start a walk exercise and then drive a mile or so. If the Fitbit does in fact use GPS for distance you should have somewhere close to what you actually drove, if it uses the accelerometer you will have close to zero. I already know the answer though 😉 

[wicks21]

Well, thanks for that link - it does clarify the issue. The GPS files shows negligible difference. I uploaded identical walks using the 2 devices:

Charge4 - 7.5km 1:20:47 5.6km/h

Sense - 7.3km 1:19:57 5.5km/h

compared to what is shown in app:

Charge4 - 8.47km

Sense - 7.47km

Should we conclude the Sense is definitely more accurate? (and I am not as quick as I thought!)

[SunsetRunner]

You can actually also obverlap the two TCX files and see where the differences are.

Long ago I stumble upon this post by @Giampi71 where I see evident issue of GPS tracking by Charge 4. Indeed I think Sense GPS is more accuarate ( also tested it as my wife owns it but the terrible HR sensor was a total letdown) but still the strange algorithm used by Fitbit to measure the distance still bothers me.