Thursday, February 27, 2014

Garmin Vector: LR balance update after 3 rides

Recently I've been experimenting with the Garmin Vector pedals, which measure L-R power balance. I've long been curious about L-R balance. How symmetric am I? How does the symmetry vary with different conditions?

What I've not done yet is to validate the pedals against my Powertap. One reason for this is I'm out of 357 batteries. I ordered a pack of 20 via Amazon for $3. $3??? That's cheaper than a pair at Walgreens. I'm not sure what I'm going to do with twenty 357 batteries.

Anyway, on the Vector, I've had a few issues. One was an anomolous 1-second right-foot-only power spike of 900 watts. I've not looked into whether this was a torque or cadence anomaly. The other was when I was carrying my bike off the train and accidentally slammed the pedal into a metal column. That wasn't a good thing, but it's neither the first or the last time I'll do such a thing. For that ride, the Edge 800 was reporting no left pedal data. But after power cycling the Edge after the ride, left pedal data were back again.

The last time, I observed that my L-R balance tended to be slightly right-foot biased at lower power, then close to 50-50 at higher power. Here's an update on that after 3 significant rides:

L-R balance

In the plot, I binned power into 10-watt partitions. In each partition I averaged the L-R power data from the ride for L-R data falling in the range 1% to 99% inclusive (I don't want to bias the result with one-legged pedaling).

The result? There appear indeed to be systematic shifts in L-R balance from one ride to the next. But then there's differences in the rides themselves. The ride of 16 Feb was a solo ride around the city serving as a recovery ride from a big block of run training. The 22-24 Feb rides both occurred after my 30 km trail race this past Saturday, which left my legs fried to a nice brown crisp. So none were on fresh legs, but racing is harsher than any training runs.

But each ride contains a similar signature with respect to power: primarily right leg at lower powers, shifting to close to 49-51 in favor of the left leg at higher powers.

Is this real? I'm not 100% sure. I really need to compare this to the Powertap wheel. Even if there is an error in the Powertap, the Powertap has no idea if I'm pedaling with my left or right foot, so comparing one-legged pedaling data to powertap data will give an idea of whether the L-R balance is biased.

In any case, the message to me with regard to Stages or Ergomo, which measure only one leg and double it, is that L-R balance is not a fixed quantity for a given rider (given = me), and there's a clear systematic error in that assumption. For example, if I'm 52% right at one power and 49% right at another power, that will result in the first power being reported 4% low and the second power being reported 2% high if the left leg only is measured. The other message is I appear to have no gross assymmetry in my pedalstroke. That's encouraging.

The bin averaging code I wrote also spits out bin counts. It's interesting how those worked out for the ride:

L-R balance

The distribution is clearly non-Gaussian due to the prominent exponential tails. I fit a simple asymmetric form of a hyperbolic secant function to the counts. That worked out surprisingly well. But I'm almost embarrased to show it because my power numbers were so lame. 30 km trail runs will do that, I suppose.

No comments: