Interesting video on crank-arm power accuracy

[jdv]

Found this yesterday on YouTube:

Is technical and detailed, but makes an interesting watch and is probably right up @GPLama’s street. Essentially (my take) the asymmetric shape of a crank arm prevents them from being accurate due to how the positioning of the strain gauges measure relative to the cross section of the arm.

[GPLama]

• This sums up my WattTeam Powerbeat findings last August exactly. See 18:25 onwards https://www.youtube.com/watch?v=s_xgx3cYl4s
• A likely answer to why I see lower readings on every single Shimano right crank fitted with a strain gauge. Likely. It’s way too common across multiple Shimano units. I still need more data on this.
• Within the first minute Keith is pretty much calling out what I, Ray, and others do as bullshit testing. Fair enough from a scientific standpoint… but from a user perspective of “is this thing accurate or not” my methodology and protocol seems to work pretty well. My G3 findings case in point. My side to side bike throw sprint data comparisons also highlight when the gauges might not detect all forces.
• The level of detail is brilliant. The title lets it down a little. “Why your TrainerRoad data is wrong” isn’t as clicky as anything with Zwift in the title. (and yes, I exploit that keyword a LOT)
• I don’t get the take out. What’s the end game here? That everything is shit? A ‘wake up sheeple’ for power training? Is this a job application spread across a few videos? Someone hire Keith and put his talent to use (it’s wasted on YouTube!)

[rocourteau]

Classical engineering demonstration: the strain gauge install cannot differentiate between relevant and irrelevant forces. Ok, good. But then - how important is it? 20%? 0.001%? First, the meter is reliable for a given force distribution - it will always give the same results for the same force applied the same way. Second, the potential variability of the force distribution (in various conditions for a given cyclist, and between cyclists) and its impact on power measurement is unquantified. Third, side-by-side comparisons do demonstrate repeatability and accuracy over a large set of use conditions - which makes one think that the impact is not significant. Fourth, the conclusion is that “your Zwift data is wrong”? Really?

[jdv]

@GPLama I knew you’d find it an interesting video. I think his comments on the current testing being done by you and Ray are unnecessary - you are providing robust testing which has identified accurate units as well as those with their own selection of rabbit accommodation problems. Your testing is a great example of repeatable tests that a typical user could perform (and the level of accuracy and consistency they should expect).

There is certainly quite a lot of commentary about power meter/trainer accuracy with regards to Zwift and e-sports. Is it even plausible to have an international race with verifiably accurate data; and how accurate does it need to be?

Does make me think I might have hastily sold a Kickr which consistently over-read by 50w.

[GPLama]

Does make me think I might have hastily sold a Kickr which consistently over-read by 50w.

Not to worry. There’s a quick way to have most indoor trainers read 50w high. The challenge is to make them read ‘correct’. And I wish they all did. It’d save time.

I’m on day 3 of collecting data from another single sided left power meter… rabbit hole doesn’t even begin to explain this. The upside is I have a ton of data showing the Assioma UNO is just as good as the DUO (assuming 50/50).

Power meters are a solved problem. The problem is there’s more and more companies trying to solve the problem their own way. Speaking of… IQ^2… Hmmmmm. Still waiting.