Register Latest Topics
 
 
 


Reply
  Author   Comment  
juergfeldmann

Development Team Member
Registered:
Posts: 1,501
 #1 
Hallo Roger. We  discuss a 2 MOXY 5/1/  rowing on the forum  and as you can see there was  some  questions  concerning overlap. It  turns  out  that one unit  seem  to have  lost   data collection in different  areas, but it does  no look like  a connection or  light  signal problem  as the datas onGC  seem t  be   not missing. .

Ryinc. 
Yes  great thanks as  I checked that last night  by looking all the raw  datas of  VL and yes  there are a  lot of  gaps  there  for example by 18.32.36  than 4  seconds   jumps  and many more to come.
 So  question to Roger . how  can  we  fix  that?
is that a one moxy model problem of this  person  or  is  that   happening for any  one ?
 ?
xcskier

Development Team Member
Registered:
Posts: 44
 #2 
I have two Moxys and I have been experiencing gaps with both.
Sometimes the gaps are only for a few seconds, sometimes
longer.

I will look at the Garmin FIT files and send them to Roger.

I think that the GoldnCheetah program I used to create plots
does some sort of interpolation and gap filling, because I have
used CSV files to create them (and not FIT files from the watch).
Roger

Moderator
Registered:
Posts: 245
 #3 
It looks like there are a couple of things going on here.

First is that we're seeing a lot of data gaps and the SmO2 and THb values bounce around a lot when the gaps occur.  We should be able to do the entire test without a single missed reading.  This problem is most likely due either to ambient light or due to the sensor not staying in contact with the skin.  Halogen lights or bright incandescent lights tend to be more of a  problem for ambient light than fluorescents.  With fluorescents, you typically don't need any light shielding.  With incandescents, a pair of shorts over the sensor usually is sufficient.  In direct sunlight, you need something like the Moxy light shield.

Sometimes the sensor can come off the skin just due to the changing shape of the muscles especially if the sensor is mounted close to a joint like the elbow or knee.  Moving closer to the belly of the muscle can fix that.  It doesn't need to be pressed in to the skin, just touching it.  Sometimes more aggressive taping or strapping that encircles the limb can actually aggravate the issue.  The adhesive attachments work well if you have to be near a joint or if the muscle is changing shape a lot.  The Cover Roll Stretch tape works well too.

The second issue is in how Golden Cheetah is smoothing the data.  This appeared to work correctly when I imported the data from the .csv file.  The .csv file skips over any missing data.  My guess (correct me if I'm wrong) is that this data was imported via the .FIT file instead.  The FIT file typically has a row of data every 1 second.  When there is a missed reading, it enters a 0 in for THb and SmO2.  It looks like it is including the 0's in the smoothing.  That's why the THb values on the LV dropped down to about 4.  It was mixing a bunch of 0's in with the rolling average of the valid 11's.  The Golden Cheetah smoothing algorithm appears to have features to deal with gaps in time properly, but I didn't find an option to exclude 0's

For right now, this Golden Cheetah issue can be worked around by fixing the source of the missed readings or by setting the smoothing down to 0.  (Hover you mouse pointer near the top of the graph and slide the slider all the way to the left)  I will look into a better longer term fix.  I can make it write the most recent SmO2 and THb values instead of 0's in the .FIT file when a no-read occurs.  I'm not sure if there is an option to leave the reading "empty" in the .FIT file.  I will also poke around in Golden Cheetah to see if they have an option to exclude 0's from the smoothing calculation.

Please email or post the FIT files so I can look a little closer at how Golden Cheatah is handling them.
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.

HTML hit counter - Quick-counter.net