Discrepancy in power between Garmin connect and Strava/Intervals.icu

I was thinking maybe the use auto-pause, but that is unlikely to be the case during a race. Would you be willing to share the FIT activity file?

Please find the link below:

connect.garmin.com/.../11669401251

What brand of power meter are you using?

The problem appears to be associated with the weak/poor data receipt rate. The RX rate is around 78%, for a power meter it really needs to be up in the 90% range otherwise you will see data issues.

With missed data records it depends on how Garmin/Strava are handling the missing the data. The Edge has back filled 15% of the records using the accumulated power data that is also being sent. If the time gaps are too long then the Edge will not back fill those missed records.

This can lead to some of the differences you are seeing as the accumulated power and individual power records do not sum to the same value.

This missing data has resulted in the difference of 9% between the total work reported and summing the individual power records. This gives an average of power of 293 using the 1 second power data, but 322 using the accumulated power value.

The values will differ depending on if you use the code is using the 1 second power data or the accumulated power data information.

You may want to contact your power meter manufacture and check that it is up to date for software and if they have had other reports of it having a weak signal.

What other electronics are on the bike? It is is possible that something else is causing issues with the signal.

I see the manufacture is 4iiii

Wow there's so much good tech info in this, thanks aweatherall! I didn't realize that "reception rate" was something that was recorded. After seeing your post I found it in Fit File Repair Tool under the battery status info. I see it for both my power meter (Garmin Rally) and HRM (Wahoo Tickr Fit). Interesting it's not recorded for my speed/cado sensor (Trek/Bontrager DuoTrap). I checked a few FIT files and it's in the 60-90% range for my power meter, whereas my HRM is 90%+. I'll need to check more FIT files to see if my power meter is typically low.

How can we tell that there was missing data from the power meter and whether the Edge was interpolating to fill in the gaps? I mean can we see the actual gaps somehow in the FIT data itself? I guess I always assumed that any data loss was simply recorded as zero power. It didn't occur to me that the Edge might be doing interpolation as an "error correction" for missing sensor data (at least up to a certain duration). It makes sense tho, and in hindsight this might explain some things I've sometimes seen in HR data where it gets frozen at a certain value for a few seconds before updating again. So I never knew that the averages etc reported by a Garmin device after an activity might differ significantly from those reported by a post-processing app such as Strava, FFRT, etc because of missing data.  I always assumed any differences were because of with/without zeroes handling. I want to check some of my data now to see if it's actually because of missing data.

I checked the OPs FIT file in FIT File Repair Tool and see that the "Device" avg power and NP match what the OP posted (of course), and the "Calculated" (by FFRT) values match what they posted for Strava values (within 1 watt, so just rounding error I assume). So that's certainly consistent with the Garmin filling gaps on its own in realtime, while other apps getting the FIT file after the fact are stuck with the data at face value (though it would probably be relatively easy to apply a similar dropout-smoothing algorithm if one wanted to).

With each record you get the one second power and an accumulated power value. If the 1 second value = the delta in the accumulated power then there is no back filling.

If there was for example a two second loss in data then when the data did come through the accumulated power value has that missed data in the total. Rather than just dumping all that into the current 1 second value, which would give a power spike, the Edge splits it over the missed records. It will only do this for gaps up to 3 seconds I think. If the gap is longer than three seconds the value is captured in the accumulated power information, but not in the 1 second records. Those 1 second records would show as zero (0)

If you plot accumulated power and the accumulation of 1 second power you can see if some data is not being accounted for in the 1 second records if they lines deviate.

Cool thanks!! I saw the Accumulated Power before in FFRT, but didn't realize it was part of the data stream from the power meter itself. I assumed that it was just calculated within FFRT for convenient reference (because accumulated power = total energy expended in Joules). Sending it from the power meter this way along with instantaneous power to assist with dropout correction makes complete sense, clever really.

I just checked one of my files and about 8% of the records appear to have had this dropout smoothing applied. (I simply looked for records where the Accumulated Power didn't change yet the Power was not zero.) Naturally some of these occurred around pedal starts/stops, but most were during regular pedalling, as they're scattered relatively evenly throughout the file. Skimming thru it I also saw a few cases of missing data for 2 consecutive seconds, but didn't notice any lasting 3 seconds or longer. Kinda humbling seeing this in action and just how regularly dropouts occur and are smoothed without us knowing.