Eliminating power spikes

Sometimes power meters screw up and output power numbers out of range, typically for 1 or 2 seconds. I have seen other sites. (like intervals.icu) successfully addressing the issue. In xert I had to flag a bunch of activities for on e of may athletes because the power spikes would affect the model, in particular PMAX.
Could you implement some sort of data filtering, maybe even with user input, so that we could tell the algorithm to ignore any datapoint above a certain power?
Detecting power spikes is usually pretty simple, very often there is no cadence associated or the power is zero before and after the large spike (very hard to generate 2400W from 0 and back in 2 seconds.) This would be a great improvement.

Xert does use a 5s filter for the signature extraction. Can you make the activities with a spike public, and then share the links with us here so we can take a look?

Hi Marco, we do filter for data spikes but look to leave in data that could be valid. Flagging is one way to address these. Alternatively, you can using something like www.fitfiletools.com to look for and remove spikes that make it through our filter, particularly if your activities have breakthrough efforts.

Scott and Armando, thanks for the replies. I recently moved one of my athletes to xert as a pilot and when we imported 3-4 years. of Strava history, her signature was significantly off, with PP, FTP and HIE twice as high. In the past I have cleared her data in WKO4+, but not feasible here. Typically I would see a large spike (up to 2400W) for 1 or 2 seconds, with all zero before and after. Stages had issues related to battery disconnect during bumps (this is on MTB). So what I did was to look at the table format of her activities and flagged out the ones with peak power above 700W. Even if you look at 5s average, a large spike (say 2400W x 2sec) is still close to 1000W avg @ 5s, would make sense to set a limit above which the sample is altogether disregarded.
Also, TP has a “fix function” that allows to select a section of the workout and ignore the data. There’s basically no way to edit data once a file is on xert.

I’ll follow Scott suggestion and share some of the activities (but there are dozens I had to flag, which would be nice to include when looking at the season)





I figured out a workaround! It’s a pain, but I found I can delete the workout from Strava, delete it from Xert, download the fixed file from Training Peaks, then upload it to Strava, then Xert syncs it up. Phew. I will do that with these old files, and see if I get an updated, more accurate fitness signature. It’s a real shame that Stages power meters still do this power spiking thing on bumpy downhills with no pedaling, even with a brand new unit. Not everyone who trains with power is a roadie!

1 Like

Maybe GoldenCheetah helps. It is a opensource software which crashes from time to time but you have the possibility to edit your .fit files even to a second by second resolution.

1 Like