Re: W'bal export contains wrong power numbers

[Ale Martinez]

El jueves, 20 de febrero de 2025 a la(s) 9:38:53 a.m. UTC-3, Aitor Altuna escribió:

Hi,

When exporting the W'bal modelled data to a .csv file, the power data exported seems to be smoothed or wrong. I attach a screenshot with time, W'bal export power data, actual power data, and W'bal. Any thoughts why this might be happening?

It is smoothed power: https://github.com/GoldenCheetah/GoldenCheetah/blob/559c8fd1d2645227e87f8d63e98f9f8241dd2047/src/FileIO/CsvRideFile.cpp#L1980, smoothing window is 25s: https://github.com/GoldenCheetah/GoldenCheetah/blob/559c8fd1d2645227e87f8d63e98f9f8241dd2047/src/Metrics/WPrime.cpp#L65 

[Ale Martinez]

I kind of used to users don't even thanks for an answer, but deleting the question after receiving an answer is new to me... I am curious about this behavior. 

[Steve Mansfield]

For my own interest and to make modelling easier, I'm looking to replicate the W' calculation in a spreadsheet.

Looking at the github files, the power is smoothed exponentially I guess because you define E = 2.71828183, but I can't read the code well enough to get from there to the smoothed power number. Is it something like (Pt*0.0392)+(0.9608*Pewma-1) ?

Also, I'm pretty sure I can export the W' numbers for a ride, but try as I might I can't see the option.

Thanks

Steve

[Ale Martinez]

El miércoles, 10 de diciembre de 2025 a la(s) 2:34:46 a.m. UTC-3, quit...@gmail.com escribió:

For my own interest and to make modelling easier, I'm looking to replicate the W' calculation in a spreadsheet.

Looking at the github files, the power is smoothed exponentially I guess because you define E = 2.71828183, but I can't read the code well enough to get from there to the smoothed power number. Is it something like (Pt*0.0392)+(0.9608*Pewma-1) ?

It is a 25s SMA filter, IIRC.

 

Also, I'm pretty sure I can export the W' numbers for a ride, but try as I might I can't see the option.

See  https://github.com/GoldenCheetah/GoldenCheetah/wiki/UG_Menu-Bar_Activity

[Steve Mansfield]

Ahh got it - thanks!

OK so I exported the W' numbers for a specimen ride, and I can replicate the W' balance when I use the Tau the file has, which is 395

However I can't replicate the Tau. First I calculated the DCP_avg using the formula 

average of all 25s SMA smoothed power > 0 when the 25sSMAPwr < CP gives 184W

and putting that into the Skiba formula for Tau which is 

Tau = 546 x e^(-0.01*DCP_Avg) + 316 I get a Tau of 408

I'm still working on why the difference. If I reverse the calculation ie what would the DCP_avg be with a Tau of 395, I get a DCP_Avg of 192 which is quite a big difference.