fresh public FIT files for decoder testing/benchmark from DC Rainmaker Enduro review

That analyser is pretty cool. I'd like one for the 3D sensors I'm dealing with... Is "FIT-o-holic" a thing? Because slowly becoming one, devouring every FIT-file I can get my grubby hands on. Thanks for the links!

Please check fitplotter

https://github.com/karaul/fitplotter

it's amazing people are already producing tools with that javascript parser

HRE seems similar to TRIMP (and garmin/firstbeat's training load which is metmax based)

https://fellrnr.com/wiki/TRIMP

HRE would be limited in that you cannot compare heartbeats for two different athletes because heart pump volume can vary widely depending on body size and training - in fact you could not compare the same athlete at two different ages, TRIMP "normalizes" that somewhat

I am not sure if my guessing is correct but after studying global message id 140  for days and weeks even, I started to believe 140:5 is what I call "met-capacity" and 140:6 is "met-work" for the activity and you can get something like the daily load by  metwork / metcapacity, then the watch keeps a rolling average for 7 days

What's also neat is message id 78 which is HRV data when using a HRM strap, I threw the values into uplot to see what they look like in a graph, this is the timing between heartbeats, not sure what is up with the start and end unless maybe the strap was not working right

ΗRΕ = HR*pace is derived for runners and is of clear sence = number of heart rate beats per kilometer. Example of the analysis is given here https://drive.google.com/file/d/17wK0y5p7rYlMRBogpZ9yicdnV191E1qx/view?usp=sharing  you can find their graphs. The metrics like TRIMP are useful, but these are just empiric correlatuions, and their meaning is not compltely clear for me

> producing tools with that javascript parser

For fitplotter, I used parser published by Dimitrios Kanellopoulos while ago in github. Actulally, the friend of mine who used Suunto, asked me about the utility, http://maraforum-2.ru/viewtopic.php?f=6&t=3477 (in Russion)

I did, looks nice! But If you meant for the 3d sensor data, this exists in separate fit messages (164/gyroscope_data, 165/accelerometer_data, 208/magnetometer_data) and must be augmented with calibration data (167/three_d_sensor_calibration) on top. I don't really make use of Record/20 right now. For GPS on the VIRB it's 160/gps_metadata, for example.

Thank you for testing.  Now online fitplotter version is ready 

https://karaul.github.io/fitplotter/index.html 

As to fit messages, the structure of js fit-parser is clear, so new definitions can be added. Right now, I do not know whether 164/gyroscope_data, 165/accelerometer_data, 208/magnetometer_data, etc messahes are taken into the fit-parser outcome - must look in the code. The fitplotter uses js fit parser outcome as it is, and gives a droplist for all messages to plot the supplied data. As runner, I am interested in few of them, i.e. heart_rate, cadence, etc;  as well there are data what I have no idea about 

Yep, that's the one I tried.

As for data and parsing, I guess I do it in two steps. The first and most important step to me is to have a parser that's as generalised as possible that outputs non-interpreted data (i.e. longitude will just be a sint32 in semicircles still. That way I can pick and choose for the second step, processing. Trying to create a fit parsing library in Rust out of the parse code for a tool I built, but it's not quite there yet.

Of course, in the processing step I still need to re-shape into more specific data structures, similar to what you seem to suggest for the js parser.

FIT-files come in so many varieties. Specific messages may contain a set of fields for one device that another does not log. 160/gps_metadata has only two fields for the watch data I've looked at - if it's even there in the first place - while the VIRB logs 10 known fields and 5 unkown for that message type.

>  as possible that outputs non-interpreted data (i.e. longitude will just be a sint32 in semicircles still. 

it's easy. file src/bf-backfit.js, line 730 - just change there the const 

scConst: 180 / Math.pow(2, 31) 

https://github.com/karaul/fitplotter/blob/51e6b5cb2fa767338976493eeab2485bd87d46d5/src/bf-backfit.js

(other mentioned below files are in the same src/ location )

I have checked quickly the parser, the loop starting at Line 115 src/bf-fit-parser.js is trasparent, and messages are defined in src/bf-fit.js e.g. there is at line 933 - 164/ 'gyroscope_data',  so the above loop treats messages in the abstract way

If you share an example of a fit file to analyse and specify what do you want to have a plot  (which parameters as a function of timestamp, etc ), I can try

Sorry I was unclear. All conversions are working fine (including calibrating 3d sensor values hopefully). I just meant that my first step intentionally outputs something raw-is so that in the next step I can always predict the input data structure.

But yep, it's the abstract interpretation I mean.

I posted a small virb fit file here.

To see what I mean you can try the tool I linked (I have some upcoming bug-fixes + changes not yet added to the linked version):

"geoelan check -f FITFILE.fit"

to see an overview of what's in the file.

Then

"geoelan check -f FITFILE.fit -g 20" to print all 20/Record data for example.