Saturday, January 12, 2013

What's up with those funky rings...?

With the successes of Bradley Wiggins in 2012, there has been a lot of interest shown lately on the non-round rings he's been riding.  These rings are nothing new, and non-round rings have been dabbled with for a very long time, but the objective data (physical tests, as opposed to mathematical modeling) on whether or not these rings allow for a greater power output than round rings has been mostly in the negative.  Yet, there are typical reports of riders seeing 15-20W (or more) increases in their power output after mounting them on their bikes.  For a while, it's been speculated that some of these reports may be the result of an artificial inflation due to the way that some power meters calculate their power readings, and mathematical analyses along with user reports have suggested that the "real world" matches the artificial inflation theory.

What is this "artificial inflation" theory, you might ask?  Well, to understand the theory, first one has to understand how bicycle power meters measure and calculate power.  Simply put, for the most common power meters on the market (I'm talking SRM, Quarq, and PowerTap) there is a torque measurement that is averaged over a certain "window" that is then multiplied by the angular velocity to get power, i.e.

Power = Torque x Angular Velocity

However, the power output of a cyclist isn't constant torque, like with an electric motor, for example.  Instead, torque input to the crankarms of a bike is pulsed, with the majority of the torque being applied in a fairly narrow range during the pedal downstroke.  As such, if one plots torque vs. pedal position over a single pedal revolution, you'll typically see a sinusoidal-type plot with 2 "peaks" per pedal revolution.  Because of this, in order to better represent the true power output of a cyclist, the SRM and Quarq power meters have been designed to use what's known as an "event-based" power calculation algorithm.  What this means is that the torque is averaged over a complete pedal revolution (no matter how long it takes) and then the average torque is multiplied by the average angular velocity over that single pedal revolution.  This way, the only measure of angular velocity is a simple magnet/reed switch construction whereby each crank trigger signifies that 360 degrees of pedal rotation have been accomplished in the time between magnet pulses on the reed switch.  The over-riding assumption of this architecture is that the crank velocity does not vary significantly from the average during the pedal revolution, and with the high inertia of a cyclist traveling on the road at any reasonable speed, this is a fairly accurate assumption...with round rings that is.  If tension is maintained in the chain of the bicycle during the pedal stroke, then by definition the crank rotational velocity will be fairly steady.  If one where to be able to measure the instantaneous rotational velocity and the instantaneous torque and calculated the instantaneous power, then if you averaged all of those instantaneous power calculations over the complete pedal revolution, the result would be very close to averaging the torque over the complete pedal cycle and multiplying by the average rotational velocity.  This is why PMs have been designed this way.

Now then, how does this factor into non-round rings possibly causing artificially inflated power values?  Well, the way that most non-round rings are designed to "work" (one exception being the old Shimano Biopace rings) is that the effective chainring size is increased during the times that the legs are applying the highest amount of torque (and, since the system needs to be symmetric, also during the time when the leg is rising back into position in the "recovery" portion.  But, I digress..) and the effective chainring size is decreased during the portions of the pedal stroke across the top and bottom of the cycle. This causes the instantaneous rotational speed of the crank to vary in a roughly sinusoidal manner (dependent on the shape of the ring) above and below the average crank rotational velocity.

For a given wheel speed and for equivalent number of teeth on the chainring, the average rotational velocity of a non-round ring will be the same as for a round ring.  This is true because in order for the teeth to be at a spacing of 1/2", that means that the circumferences of a non-round and round ring of the same tooth count are equivalent.  However, with a non-round ring that means that, by design, the rotational velocity of the crank is being manipulated such that the rotational speed of the crank is slowed during the downstroke (and upstroke portions as well due to symmetry) to a value below the average

So, what does that mean?  In short, it means that during those times that the largest torques are being applied, the rotational velocity is lower than the average over the complete pedal revolution.  Since the majority of the power is produced in the downstroke portion of the pedal cycle, this means that power reported will tend to be over-reported, hence the term "artificial power inflation".  Get it?

So, that's the does it end up working in reality?  Well, in short, reality matches the theory fairly well and by my measurements, using Osymetric rings on a Quarq power meter will cause the Quarq to report a power value ~2.7-3.5% higher than reality.

Here's some testing I did near the end of 2012:

The setup: 

Test performed on a LeMond Revolution trainer to maintain a rear "wheel speed" similar to riding on the road (i.e. high flywheel inertia)
54T Round and 54T Osymetric both mounted on the same Quarq with the Osymetric in place of the inner ring. This was done to eliminate changes in drivetrain losses due to using different cogs. Calibration shows the Round ring (mounted on the outer position) reads 1% high and the Osymetric (mounted on the inner position) reads 1% low and the torque slope was set to the average value between the 2 rings. Power values from the Quarq reported below were corrected based on the calibration. Zero offset was checked before the first pair of runs, before the 2nd set of runs, and then at the finish. Offset did not move by more than 5 counts (512, 507, 511) which is equivalent to 1.3W @ 80 rpm. Runs were ~4 minutes in length.

Runs in order of performance:
- 54x15 gear, 80 rpm (36.2 km/hr indicated)

  • Round Ring (Quarq - corrected) = 262.0W, LeMond Power Pilot = 249.5W, HR=160 bpm
  • Osymetric    (Quarq - corrected) =271.3W, LeMond Power Pilot = 250.3W, HR=166 bpm

- 54x16 gear, 80 rpm (34.1 km/hr indicated)

  • Osymetric  (Quarq - corrected) = 227.5W, LeMond Power Pilot = 212.3W, HR=158 bpm
  • Round Ring (Quarq - corrected) =221.6W, LeMond Power Pilot = 212.0W, HR=161 bpm

So, in brief, based on just this limited test, there actually DOES appear to be a "non-round ring inflation factor" of 2.7%-3.5%. (BTW, Dan Connelly estimated a 4% "inflation" based on a 20% within pedal stroke speed variation here ( and by my measurements the Osymetric ring would cause a 20% crank speed variation assuming a constant rear wheel speed.

I followed up the trainer testing rides outside where I ran a PT wheel on the same bike as the Quarq/Osymetric setup with the large ring being the non-round ring and the small ring being roundThe artificial inflation with the non-round ring was observed there as well with the difference between the power as measured by the Quarq as compared to the PT being ~4% higher when in the non-round ring under a continous effort.
So, there you have suspicion is that a good majority of the power "improvements" claimed with non-round rings are merely mis-measurements of the actual power.  It's important to note that this sort of power inflation is not anything that can be "calibrated out".  It will be present even if the PM torque calibration is perfect since it's a result of the power calculation algorithm itself and ANY power meter that employs an "event-based" calculation as described above will suffer from it.

Now then...on the question of whether or not non-round rings have the potential to actually increase pedaling power output...well, I have my opinions on that, but I'll leave that for another blog post....


  1. Speculated since 2006:!msg/wattage/5prBUKY20s0/npZID_tb-5AJ

    1. I'm pretty sure that the Dan Connelly blog post I linked to also links to that very same wattage list thread.

      Robert, yes.

      Alex, thanks!

    2. Correction, I linked to that thread above under "user reports".

  2. Tom did you try the osymetric with the modified position? Just checking as it is very similar to my experience almost no change as measured on my powertap.

    1. Do you mean the "optimum" position from that mathematical study? If so, yes. The inflation was still present. Didn't notice any real power gains either.

  3. what is your opinion when people claim more than 4% power improvement do you think that is real powerimprovement or not ?

    1. Info,
      That's a subject for another blog post ;-)...but, for now, I'll just say that the "theory" of non-round rings is predicated on them being able to change the joint velocity and muscle shortening speeds in the upper leg, and yet you have vastly more capability of doing so just by flexing your ankle joint than can be accomplished with any reasonable ring ovality. Think about that for a bit...

      Oh, and placebo is a pretty powerful thing. Otherwise, it wouldn't need to be controlled for in studies, right?

    2. I tried them for a week with no noticeable difference in Powertap reported power output even considering my lack of one ankle to flex. It wasn't a formal test and plenty of other variables in the mix so nothing I can hang my hat on but honestly I didn't notice any difference when pedalling. Lost in the noise.

  4. I did some trainer testing and recorded HR vs rear wheel speed on a graded power/speed test (I.e., increase speed every 2 minutes until my HR hit 180). I would switch back and forth between rings. My slope of the fitted HRvs.speed graph was no statistically sigificant, but the intercept indicated a 3 bpm lower heart rate for the rotor ring. Any thoughts on the validity of this testing method?

    1. My only thoughts are that HR is only a response to your effort (amongst a myriad of other things) and there's a LOT of variability in that response.

      This is the reason why on board power meters were developed...otherwise, why bother with the expense? Just buy a HR monitor. They're much cheaper.

      I guess I'm just questioning the ability to be able to determine if that 3bpm difference in the intercept you see is "real" or not...Besides, let's say that the rings DO drop your HR for a given effort. Does your ability to do that effort (i.e. power vs. duration) change? If not, big deal.

    2. I have thought about the oval chain ring debate a bit. It seems like definitive proof of improved performance would be very difficult. The one thought I have is to look at the problem from the reverse perspective. Saying oval rings can't improve performance is equivalent to saying that round rings are optimal.

      One idea I mull over is what would happen if you shifted the orientation of the oval ring by 90 degrees. This would slow the velocity of the foot at the top of the pedal stroke and speed it up as you go through the "power portion" of the pedal stroke. I can also imagine some very radically shaped chain rings that would be nearly impossible to go fast on (big rectangle)

      Me thinks that chain rings may be like seats. There is a shape that would be optimal, but it depends on the individual.

    3. Actually...I think that the proper way of looking at it is that even if it is possible that non-round rings may be "optimal", it's going to happen under limited conditions of orientation with respect to the rider...therefore, round rings may be the best compromise all around.

      One thing to remember is that key assumption of non-round rings is that it's possible to change the muscle shortening velocities and joint angle velocities in the upper leg (where the driving muscles are) by varying the crank rotational speed. That assumption relies on the extra degrees of freedom in the lower leg not changing when changing ring shapes. Kinematic data suggests that's not the case.

      Our bodies are REALLY good at figuring out how to operate in their preferred manner (i.e. muscle shortening speeds, etc.) and understanding that there's more potential for changing the upper leg joint velocities and muscle shortening speeds just by flexing of the ankle than is possible with any reasonable chainring ovality give rise to questions about non-round ring effectivity. My suspicion is that's basically what the whole "accomodation" is all's your body "figuring out" how to keep using those upper leg muscles in the way it prefers...thereby effectively defeating the whole point.

  5. "on the question of whether or not non-round rings have the potential to actually increase pedaling power output...well, I have my opinions on that, but I'll leave that for another blog post...."

    Such a tease... Can you at least add a follow or subscription link to the blog so we can keep updated?

    1. can probably guess the content of any future blog post on the subject from my replies above ;-)

      Good point on adding the subscription link thingy...I just did that (at least I think I did). I'm new to all this blog stuff....

  6. interesting blog. It would be great if you can provide more details about it. Thank you...

    Torque Calibration

  7. I believe dorsiflexion of the foot and RPE have implications on cycling performance.

  8. I asked Quarq about this. I obtained a Wahoo Key and calibrated the Quarq with the Oval Chainrings on. Would this result in accurate data?

    1. No, as I explained above, this isn't something that can get calibrated out. Although the torque values will be correct, the assumption of constant rotational velocity through the pedal stroke is still violated. The folks at Quarq should understand this...I know that Jim Meyer (the founder) does.

  9. I know this might be an old thread, however, from the picture above I noticed that you installed the Osymetric chainrings the wrong way. This is a known 'issue' with cranksets with hidden bolts. The correct way to instal them is to rotate them one hole counterclockwise on the crank spider (the chaindrop protection pin doesn't line up correctly). If you made your measurements with the pictured setup all data is invalid since the Osymetric shape is not designed to work in that way. Jean-Louis Talo (inventor of Osymetric) mentions himself that he tested many shapes and setups and none came even remotely close to the current iteration. The biggest diameter of the rings should be when the cranks are at the 3/9 o'clock position (horizontal). In your picture this falls at the 1 o'clock position. Again, I am basing my comment on the picture above. I am surprised nobody noticed that up to now....

    1. Hi Nikola,
      Thanks for reading. The picture at the top was one that was taken when I was trying different configurations of alignment to see if there were any differences. The picture, in fact, shows the position closest to the supposed "theoretical optimum" from the Malfait, et al mathematical studies linked to at the beginning of the blog post.

      As you read in the test setup above, I installed the Osymmetric ring in place of the inner ring, with a standard 54T round ring on the outside. That's obviously not what is show in the picture at the top. The Osymmetric ring was installed for "side-by-side" 54T testing as you described in your comment so that the timing of the ring curvature relative to the crankarms was as designed.

      The picture at the top just happened to be the best picture I had of the Osymmetric ring installed on the bike.

      So, in short...that picture does not invalidate anything :-)