Friday, February 15, 2013

ANT+ Bicycle Power V3.0

As mentioned in my previous post on the information I found on the new Rotor Power Meter, it became official yesterday that the Bicycle Power Profile is upgraded to V3.0.

What is neat is that it maintains backwards compatibility. What is different is that it adds page 0x13(hex) that defines the metrics for torque efficiency and pedal smoothness. This new page is sent every 1.25 seconds though which is about 1.5 rotations at 90rpm.

So what is in the new 8 byte packet

  1. Event Count
  2. Left Torque Effectiveness
  3. Right Torque Effectiveness
  4. Left (or combined) Pedal Smoothness
  5. Right Pedal Smoothness
All metrics are given in 0-100% and you can be sure Vector will also implement this when it arrives.

My guess at using ratios was wrong so I'll have to fix my calculations:
TE = 100 x (Sum of all power both positive and negative)/Positive Power = 100 x(sum(p))/sum(p_pos)
PS = 100 x Pavg/Pmax

If you read the document you'll notice that if the Right Pedal Smoothness is set such that it's value is invalid the Left Pedal Smoothness value is a combined value. I kind of dissagree unless this is reserved for L/R meters only. I whipped up a new Excel example with 32 divisions per rotation. Left and Right Legs are similar but with a slight imbalance to the right. It's a combination of an exponential decay and a sine wave function. 

Calculating the new metrics gives the following:

Power Meter Measure Percent
Rotor / Vector / Keith V3 Prototype / Left - Right Measuring TE L 53.0
TE R 61.7
TE Combined 57.4
PS L 16.6
PS R 19.2
PS Average 17.9
Quarq, SRM, Power2Max TE non L/R 100.0
PS non L/R 56.3
For a L/R based power meter a head unit would display a torque efficiency of 57.4 average between both legs. On a crank spider based unit it would measure 100 percent torque efficiency because as one leg is dropping the other is rising and vice versa. This means the torque never does go negative.

Pedal smoothness, from graphs I've seen, will likely be in the 20 - 30% (Pioneer, but this is a guess). A Crank spider version would always be saying higher. However, there could be a better way to calculate pedal smoothness for a spider based version. By doing what Quarq is doing to fake L/R measure by comparing the 180 degree average's they could make a guess at the negative torque. However inferior the balance assumption is, this one would be even more difficult to swallow. I don't know which I would rather have: made up data, or if the data wasn't available.

This is the problem with people doing clever engineering. The people who spend time designing, executing, and reviewing the data might put faith in these new measures without realizing what engineering folk have done and therefore not realize that they might be guesses leading them down an incorrect road. They could be asking an athlete to balance their legs better when it's really a single leg performing better in a part of a stroke. It could be that the spider based DFM makers implement pedal smoothness as is and then anyone with the new L/R power meter looks to be way less "smooth" but actually has more accurate numbers.

Lots of difficulties in implementing this new data. I think I'm in the camp that would suggest Quarq stop faking L/R balance, and no non-L/R measuring meter should be able to use the new metrics. Coaches are relying on power meters. They need good data, not "Engineer Magic".

PS: Don't get me wrong, I love Engineer Magic - See Kalman Filter


  1. First, I think your assumption of the "negative torque" is a bit overstated based on pedal plots I've seen...

    Plus, doesn't the fact that the descending mass of one leg "lifts" the rising mass of the other (i.e. it's a balanced, coupled system) imply that the metrics as described don't really mean anything unless the mass effects are removed? Sounds to me like TE is a non-starter right out of the chute.

  2. You're right that my plots are exaggerated, though I still think they illustrate a point on how they need to be calculated individually.

    I'm not sure what you mean about mass effects. If you are pushing 100N-M of torque on your downstroke with your right leg, and your left is being a laggard on the upstroke by 20N-M, then your left measurement goes negative and your right is still 100N-M, but your chain ring into the drivetrain will only see 80N-M. Fighting your left leg in this instance is still an overall loss. The "mass" off a leg is either fighting the opposite leg, or helping. So in that way it's independent of mass effect.

    If you maintain positive torque throughout the entire rotation for each leg then you'll get 100% even if it's very varied. I think that's a good metric in theory. However I'm not a coach and have yet to implement this on my own prototype (aiming for 2 - 3 weeks from now). Nothing will record this yet, and only Rotor and Garmin will know what test data looks.

    That being said, I've been yelled at in spin classes to "do full circles", the TE metric should be the numeric equivalent. If you're pushing near 100% that means your legs aren't fighting you during your full rotation. The PS metric will then tell you how smooth that is.

    I think that the metrics are probably the best that can be done without information overload. I think it would be neat to have stroke measurements 1/16 of a rotation but on a bike you'll never figure out what that means on a tiny cycle computer. You either need a person or "clever" maths to disseminate that data down to something more easily digested. The more familiar I am with the metrics the more I think that they make sense. It's up to the pro's and people like yourself to say if they are useful. Vote with your wallet! SRM, Quarq, Power2Max or Vector or Rotor Power... or in a few months, whatever my beta test units will be called.