Right-side or dual-side powermeters on R9100 or R8000 cranks are likely inaccurate

[weiwentg]

I'm not sure how many people this affects, but GPLama, an Aussie Youtuber, recently documented (https://gplama.com/2019/06/29/shimano-crankset-power-meters/) that a number of right-side powermeters installed on drive-side, current generation Shimano crankarms produced readings that were 6-8% lower than a known power input. This guy tested Stages, Pioneer, 4iiii, Watteam, and Shimano's own DA powermeter, so it doesn't matter who made the meter. He believes that left crank power is correct, and that the issue is due to the asymmetrical crankarm design.

This means that if you have a dual powermeter setup by anybody on a R8000 or R9100 crank, you likely can't trust your total power or your power balance readings. He seems to think that 6800 cranks should be much closer to correct readings (see comments in Reddit thread (https://www.reddit.com/r/Velo/comments/c8ft2j/shimano_based_crankset_power_meters_not_as/)). Presumably this would affect any Shimano crankset of that era also. The differences may not be as pronounced riding outside; his testing protocol was mainly indoors.

Keith Wakeham, one of the original designers of 4iiii's powermeters, posted (https://titanlab.co/shimano-crank-design-blunder/) on his blog about this. He thinks it's a fundamental engineering issue.

[wallymann]

interesting. thanks for posting!

[skouri1]

precision with a powermeter is arguably more important than their accuracy. As long as values are consistent, it doesnt matter if they are lower than the true numbers. still just as functional as a training tool.

[wallymann]

precision with a powermeter is arguably more important than their accuracy. As long as values are consistent, it doesnt matter if they are lower than the true numbers. still just as functional as a training tool.

in the case of dual-leg measurement, it's complicated by getting incorrect L/R balance readings...which i understand to be a big element of this issue.

can/have powermeter mfgs simply use a fudge-factor to correct for those erroneous right-side readings?

[zmalwo]

Giant having the least diff%. Like always, they make the cheapest yet the best quality stuff out there.

[Mark McM]

precision with a powermeter is arguably more important than their accuracy. As long as values are consistent, it doesnt matter if they are lower than the true numbers. still just as functional as a training tool.

That may be true, but in this case, the right side power meters likely don't give consistent results, either.

Power meters work by measuring the strain caused by loads applied to the crank. These loads are converted to torque, and then multiplied by crank speed to derive power. But strain in the crank can be caused a variety of different loads on different axes. For example, a radial load on a crank (as when standing on pedal with the cranks vertical) may cause a strain, but that load does not generate any torque. So one important aspect of power meter design is make sure that the off-axis loads are cancelled out, and only the loads perpendicular to the crank are used to derive torque.

The problem with right side power meters is that the shape of right side cranks are more complex than left side cranks. This means that strain patterns due to loading are more complex, which makes cancelling off-axis loads is more difficult. Off-axes loading will produce larger errors in torque measurement, and the size of those errors will vary with the distribution of off-axis loads. Off-axis loading will change with pedaling technique - for example, the pattern of off-axis loading when seated and spinning will be different than when standing and grinding. I'd expect that the magnitude of error in right side cranks will vary significantly between seated and standing, or between sprinting and TTing.

The problem of off-axis load can be solved, but it requires a more complex design, with many more strain gauges positioned in different locations on the crank. The calibration also becomes more complex, as the crank will have to be calibrated with loads from a variety of directions. And there will be slightly more complex formulas to convert strain to torque, using more strain measurements to resolve and extract only the drive torque.

(Just for a little background - in a previous life I designed and developed multi-axis load cells for robotic applications, which used arrays of strain gauges applied to specially designed structures to resolve loads into the 3 force and 3 torque components about the X, Y & Z axes.)

[weiwentg]

can/have powermeter mfgs simply use a fudge-factor to correct for those erroneous right-side readings?

Well, I don't make powermeters, and I'm not any sort of engineer, so this answer is off the cuff. The issue there is that they don't know how much the fudge factor is. The amount to fudge the reading probably varies in real-world conditions - GPLama himself said that the differences aren't as pronounced in mixed terrain, and they seemed to get worse with increased power.

precision with a powermeter is arguably more important than their accuracy. As long as values are consistent, it doesnt matter if they are lower than the true numbers. still just as functional as a training tool.

If GPLama is correct that the amount of bias in the drive-side arm powermeters is variable depending on, say, power input or variations in grade, then that does get right to their consistency.

If it were a constant bias, then sure, I totally agree with you. Except that your power balance metrics would be off, but different riders may or may not pay as much attention to that.


...
The problem with right side power meters is that the shape of right side cranks are more complex than left side cranks. This means that strain patterns due to loading are more complex, which makes cancelling off-axis loads is more difficult. Off-axes loading will produce larger errors in torque measurement, and the size of those errors will vary with the distribution of off-axis loads. Off-axis loading will change with pedaling technique - for example, the pattern of off-axis loading when seated and spinning will be different than when standing and grinding. I'd expect that the magnitude of error in right side cranks will vary significantly between seated and standing, or between sprinting and TTing.

The problem of off-axis load can be solved, but it requires a more complex design, with many more strain gauges positioned in different locations on the crank. The calibration also becomes more complex, as the crank will have to be calibrated with loads from a variety of directions. And there will be slightly more complex formulas to convert strain to torque, using more strain measurements to resolve and extract only the drive torque.

...

This is a much more detailed critique than what Wakeham posted on his blog (I understand that Wakeham may have a YT vid that goes into more detail, I haven't watched it). This does raise a question. I don't buy the title of Wakeham's post, that Shimano made a design blunder with their cranks. They are, after all, a bike component manufacturer. However, bikes are headed towards greater integration of parts. Lots of people are putting power meters on their bikes. Should Shimano design their crankarms more symmetrically on account of this? What will pressure them to do so?

[Mark McM]

Well, I don't make powermeters, and I'm not any sort of engineer, so this answer is off the cuff. The issue there is that they don't know how much the fudge factor is. The amount to fudge the reading probably varies in real-world conditions - GPLama himself said that the differences aren't as pronounced in mixed terrain, and they seemed to get worse with increased power.

If the "fudge factor" was a simple offset and/or linear multiplier, than it would be intrinsically included in the calibration. I suspect that it is more complicated than that. I suspect that it is not a matter of non-linearity (strain in metal is very linear below the yield point), but instead is a matter of sensitivity to off-axis loads. And off-axis loads can vary quite a bit with pedaling technique and terrain.








This is a much more detailed critique than what Wakeham posted on his blog (I understand that Wakeham may have a YT vid that goes into more detail, I haven't watched it). This does raise a question. I don't buy the title of Wakeham's post, that Shimano made a design blunder with their cranks. They are, after all, a bike component manufacturer. However, bikes are headed towards greater integration of parts. Lots of people are putting power meters on their bikes. Should Shimano design their crankarms more symmetrically on account of this? What will pressure them to do so?

Left cranks are well suited to torque measurement, because they are long, relatively uniform beams, which will have uniform strain fields in the middle of their spans (which is why you'll find most left crank power meter strain gauges mounted right in the middle of the arm). There will still be some off-axis load sensitivity here, but this can be cancelled with a few extra co-located strain gauges.

A right crank has spider arms sprouting from it, often some distance out the crankarm. Loads are transferred from the crankarm to the spider arms in different proportions, depending on crank position (and off-axis loads). The affect of the spider arms can be mitigated somewhat by moving the strain gauges further out on the crank arm. Unfortunately, you'll then run into space constraints, as there may not be enough clearance between the crankarm and the chainring and front derailleur cage.

Right crank power measurement can be accurate and repeatable, if the crank arm is designed properly. But that would take a great departure in shape from the current Shimano crank design.