Replaying FIT file

I'm writing an app targeting the 645 (my first app) and to test it with "real" data I took a FIT file from a 5K I ran a couple weeks ago that was recorded on my 920XT and ran it in the simulator. I thought I had a computation error since the data it was showing me was well off reality -- I haven't run that fast in over 15 years! Then I looked at the data passed to compute() in the Info structure and it is indeed well off of what I ran that day (simply looking at the meters run vs. the elapsed time was enough to tell).

So am I missing something about how FIT file data is stored, something in the simulator, or ??? The app itself seems to be processing the data correctly, so maybe it will work on the actual device, but I don't have one at present to try it out on.

jim_m_58 over 5 years ago

There was a bug a while back where the distance was wrong when playing back a .fit, but that was fixed. I'm playing back one of my .fits from about 3 years back right now just to check. Ok, I let it run for a while, and it's looking fine. (this one was recorded with smart recording)

One thing about a recording - are you looking at the value when it's complete, or while it's running? If something was recorded by "smart recording" it can look a bit off distance wise in the sim, as the data isn't written at regular intervals (like every second) and therefore not changing every second when read, so you might see distance stop and start every few seconds, while the timer runs normally. This is something that doesn't happen on a real device, as there, the data seen in your app is always every seconds, even with smart recording.

subra over 4 years ago

As OP already stated the values from playback a .fit are way off. I've still the same problem with the sim 3.1.6. Though the .fit is recorded every second but info.elapsedDistance is way off. I'd like to playback a .fit for validating and calibrating my DF and analyzing my real activities. Therefore I've further investigated this issue.

I'm starting the playback of the .fit as follows:
1) Eclipse: starting the sim
2) sim: Simulation > FIT Data > Playback File > selecting the .fit > open
3) sim: (with different time delays) klick on start button in the Fenix 5 picture

By klicking the start button on F5 the timer starts and begins to read the .fit. I don't know exactly what is stored in the .fit but it seems to me that the sim uses its own timer time for info.elapsedTime, not the time stored in the .fit. Then the sim looks in the .fit for the proper time period and calculates/interpolates the info.elapsedDistance.

In my DF function compute(info) I directly print out info.elapsedTime and info.elapsedDistance.

I've done 3 playbacks with the same .fit and different time delays (t = 1s, 5s, 10s) between open the .fit and starting the timer i.e. klicking on button start.

All 3 playbacks differ depending on the time delay between opening the .fit and starting the timer. For info.elapsedTime = 60000ms an info.elapsedDistance = 190m to 208m is output, which corresponds to an average speed of about 11.4km/h to 12.5km/h.

Displaying the same .fit with GC or BaseCamp Graph shows an average speed for the first 60s of roughly 8km/h which is different by nearly 50% from the above 11.4km/h to 12.5km/h from the sim.

Either, I've made something wrong or there is a bug in the sim.

Any hints for solving this problem would be appreciated.

subra over 4 years ago in reply to subra

From the feedback it seems that no one else, but only OP and I have this issue that the values from replaying a .fit with the simulator are way off. I wonder how can DF and apps are validated without using data from a real activity.

Here are some graphs that shows the diffences in speed

1) Simulator
The current speed is calculated as speed = (d2-d1)/(t2-t1) using the following code

	function compute(info) {

		if (info.elapsedDistance != null) {
			var elapsTimeTest = info.elapsedTime;		// [ms]
			var elapsDistTest = info.elapsedDistance;	// [m]
			var speedTest = (elapsDistTest - elapsDistTestOld) / 
				(elapsTimeTest - elapsTimeTestOld) * 3600;
			elapsTimeTestOld = elapsTimeTest;
			elapsDistTestOld = elapsDistTest;
			System.println(elapsTimeTest + "  " + elapsDistTest + "  " + 
				speedTest); 
		}

        ...

2) BaseCamp
BaseCamp Graph directly offers the option for plotting speed from a .fit.

3) Garmin Connect
GC displays only the pace, but with speed = 1/pace it can easy converted.
The pace for the first 0.2km is shown of about 8min/km which relates to a speed of 7.5km/h

BaseCamp and GC shows roughly the correct speed in the range of 7-9km/h, but the info.elapsedDistance from the sim is way off and delivers a speed in the rage of 10-15km/h.

So, the question remains "what is wrong?"

jim_m_58 over 4 years ago in reply to subra

When playing back a fit, you'll probably have the best luck if it was recorded with 1 sec recording and not smart (smart is the default on most devices.).

And when testing you may want to start recording in the sim before you start playing back the fit to avoid missing the first part of the recording.

subra over 4 years ago in reply to jim_m_58

Thanks Jim. The .fit is already recorded every second, because I'm aware of the former bug in the sim regarding the smart recording.

If I first starting the recording in the sim and immediately after that starting the FIT playback I get the following values:

 t[ms]     dist[m]     speed[km/h]
 1218        0.000000   0.000000    // starting recording in sim
 2281        0.000000   0.000000
 3250        0.000000   0.000000
 4281        0.000000   0.000000
 5281        0.000000   0.000000
 6281        0.000000   0.000000    // starting .fit playback
 7281  5834499.500000  21004198.000000    // huge distance ???
 8281  5834500.000000   1.800000
 9281  5834501.000000   3.600000
10281  5834502.000000   3.600000
11281  5834503.500000   5.400000
12281  5834505.000000   5.400000
13281  5834506.500000   5.400000
14281  5834508.500000   7.200000
15281  5834510.500000   7.200000
16281  5834513.500000  10.800000
17281  5834516.500000  10.800000
...
57281  5834655.500000  16.199999
58281  5834659.000000  12.600000
59281  5834662.500000  12.600000
60281  5834666.500000  14.400001
61281  5834670.000000  12.600000
62281  5834673.000000  10.800000
63281  5834676.500000  12.600000
64281  5834679.500000  10.800000
65281  5834683.500000  14.400001
66281  5834687.500000  14.400001
67281  5834691.000000  12.600000

The info.elapsedDistance provides suddenly a huge distance and I've no clue why.

If I starting in reverse order i.e. first starting the FIT playback and immediately after that starting the recording in the sim I get the following values:

 t[ms]    dist[m]  speed[km/h]
  812    0.286220   1.268955    // starting .fit playback and immediately recording in the sim
 1812    1.187548   3.244782
 2812    2.086517   3.236288
 3812    3.751678   5.994579
 4812    5.299979   5.573884
 5812    6.814125   5.450923
 6812    8.894515   7.489406
 7812   11.115501   7.995551
 8812   14.025583  10.476295
 9812   16.960127  10.564357
10812   20.208660  11.694719
...
50812  158.823059  13.268902
51812  162.429001  12.981390
52812  166.528244  14.757276
53812  169.789856  11.741803
54812  172.738480  10.615045
55812  176.005661  11.761853
56812  179.038315  10.917554
57812  182.988708  14.221417
58812  187.025085  14.530958
59812  190.664734  13.102735
60812  194.976593  15.522693

This seems more realistic as the above values but unfortunately still way off.

jim_m_58 over 4 years ago in reply to subra

I'm going to take a wild guess on line 8 in your first record. There was a GPS drop at the time and the sim was confused. Try the same .fit but also display lat/lon (accurancy will always be what you set in the sim and not from the data itself)

subra over 4 years ago in reply to jim_m_58

In the meantime I've further investigated this issue and have found that by first starting the timer in the sim and subsequently the replay of the .fit the first value of info.elapsedDistance is very large, as can be seen in above table line 8. The reason for this issue is due to the initial values in the sim lat=0.0 and lon=0.0 . For my understanding if replaying a .fit the starting position lat/lon in the .fit i.e. the first position should be used instead of 0/0.

This seems now also observed in the thread forums.garmin.com/.../sdk-3-1-7-is-now-out

jim_m_58 over 4 years ago in reply to subra

Make sure there is a bug report about the [0,0] in the bug report forum.

subra over 4 years ago in reply to subra

The original issue described at the start of this thread and shown in the graphs above is still unsolved.

info.elapsedDistance is way off not only due to lat/lon = [0.0, 0.0] but rather it seems there are more bugs in the sim.

info.elapsedDistance should be the sum of the distances between the recorded position points lat/lon. So I've calculated the sum of this distance increments and compared it to the value of info.elapsedDistance. I've only used a fast and easy formula to calculate the distances between 2 position points without any smoothing or interpolation. Thus my calculated elapsed distance may differ somewhat to info.elapsedDistance.

The same .fit is used as in the above graphs and tables and I get the following values
Sum of distance increments = 125.82m
info.elapsedDistance = 194.69m
This is a difference of >54% which is well way off.

In the .tcx file of the .fit for the same lat/lon there is stored the value
<DistanceMeters>123.29000091552734</DistanceMeters>

If I don't use the "Playback File" option but instead use the option "Simulate Data" I get this values
Sum of distance increments = 124.67m
info.elapsedDistance = 129.09m
This is a difference 3.5% that seems to be reasonable to me.

Since "Simulate Data" generates random values the distances are different to the "Playback File" values.

Any ideas about the difference with playback .fit?

Sylvain_PILLOT over 4 years ago in reply to subra

Little come back to this problem with the FIT simulator.

I am working on a datafield that needs to calculate the pace. So I recorded a FIT file using my Forerunner 230 and was planing to use that file as a reference for developing my datafield.

But the simulation seems to return incorrect time data (using info.elapsedTime). My FIT file was recorded every seconds.

As an example, the FIT file contains the following data

time --> distance

0s --> 0m

10s --> 11m

20s --> 25m

30s --> 39m

40s --> 51m

50s --> 64m

60s --> 77m

etc.

while in the simulator, after the same elapsedTime (every 10s) I got the following values :

time --> distance

0s --> 0m

10s --> 16m

20s --> 35m

30s --> 53m

40s --> 68m

50s --> 85m

60s --> 102m

etc.

Seems that time is not one of the FIT file, but something running roughly 25 to 30% quicker than reality.

No needs to say that Pace calculation is totally inaccurate.

Any idea ?

For information, I am using SDK 3.1.8 running under Linux Ubuntu.

Thanks.

Sylvain