I’ll need to go back and watch it again to be honest.
I guess the problem may be that the drive side is quite a bit more asymmetric than the non drive side.
Easy solution to all of this… Stick strain gauges on Cannondale cranks!
Would be nice if 4iiii were to make a comment on here but I seriously doubt that’ll happen.
Mike
I’m now wondering if I should only be using the left side power from my 4iiii Precision Pro. Am I likely to see an improvement in accuracy reporting the left side only vs dual sided measurement? That is to say, does taking the left side power and doubling it provide a more accurate measurement than measuring both sides, given the inherent inaccuracy on the right side?
Really enjoyed this episode and super interesting. I have had my 4iiii left crank 105 for about a year now and works well. Sometimes it is a little slow to wake up but not really an issue.
If or when I get my tri and mtn bike I will be getting another left crank only as in my experience the 4iiii is it just works without issue. The battery life as well is pretty stellar.
I think my only question from the episode would be how did powermeters come to be. How did they start and what was the time between 4iiii inception to get a truly viable product.
Hey Shane!
Great work on this item and I’ve been hesitant to get a new powermeter connected to shimano cranks because of it.
question - is this primarily only happening in the ultegra/DA cranks? or all shimano cranks? I’m specifically curious about the new GRX cranks (but guessing this hasn’t been tested yet).
thanks!
To be honest, there are bits of the explanation I don’t quite understand, e. g. how axial (load?!?) is connected to the “fudge factor”. But as best as I understand the final verdict was that the asymmetry in Shimano’s newer cranks prevents you from getting accurate power measurements, no matter where you place the pods. Either you pay a price on the calibration side (where larger distance to the crank axle improves accuracy) or on the fudge factor side (where larger distance to the crank axle makes things worse).
Am mechanical engineer (although unfortunately not a masters nor am I super educated on mechanics outside of the typical undergrad courses) but my guess is that the issue is general strain on the drive side arm due to forces being exerted on the body that are not in line with the planes, mixed in with issues of asymmetrical crank design and where the strain gauges are located. The whole “left arm” coupling interaction was interesting, not something I really considered as an issue but it makes sense. When you have the asymmetrical drive side crank arm mixed with the coupling, my guess is that you minimize the physical mechanical deformation interactions between the chainring and crank arm by going far from the chainring. This is a trade-off of a larger interaction with the coupling of the left arm and the asymmetry of the actual chainring arms absorbing forces in a non-predictable way causing fluctuating strains based on how the cranks are positioned in the pedal stroke (or what kind of forces are being exerted at that moment in time).
Basically how I understand it, the asymmetrical body of the drive side causes the strains in the crank arm to fluctuate due to how the non-uniform or “un-symetrical” drive side spider deforms/absorbs forces with respect to each force exerted on it and where in the cycle the forces are exerted on it. Combine this with coupling and you have this ugly mess of fluctuating strains that you have no way of counteracting, especially since you have no idea what part of the stroke you are in, so you cannot use general “fudge factors” to counteract your errors. This seems to be an issue as the strain gauges are now reading more variations from an asymmetrical spider than a more symmetrical spider body, hence why the newer shimano cranks are seeing an issue. This compounded with the already flawed “fudge factor” numbers will cause even greater error, as you cannot use a flat fudge factor when it comes to the drive side.
I have a few guesses as to how they might mitigate this but A) I am not a specialist in this so I likely am overlooking things and B) it might work but be much more expensive, such as adding more strain gauges or something.
TLDR; looks like the issue is the deforming body of the spider impacting the strain gauge reading due to the shape of the crank and how it will distribute forces depending on how the forces are positioned. Combine this with variable loading and you get a nightmare of strain fluctuations that will occur based on both how you are loading the crank arm and how the drive side also impacts the strains. Having a more symmetrical crank arm will cut down on the fluctuations due to more predictable interactions but I think this was always a problem, only that with such an asymmetrical design, you cannot “fudge factor” the error out anymore (interactions between the shape of the spider and the forces as a function of where you are in the cycle and how you load become too big of an issue).
Anyways I might be out to lunch but I think this makes sense. I always thought this was figured out but I guess this is a bigger problem than I would have thought. Good thing I only have a single arm power meter!! 
I think this is due to again, asymmetry in the crank arm. You cannot assume that the axial load is evenly distributed through the body like it would be in a solid body case. I believe since you are 1) calculating the strain at some point on the crank, which is offset from where the axis of rotation is and also where the spider is connecting to the chaining, and 2) this is not as simple as a chainring-spider interaction, you have to account for the left-side crank as well impacting the system, and 3) the spider is non-symmetrical, so depending on where the crank arm is positioned, the spider will behave in a different way because of forces exerted on a non-uniform body (even with uniform material properties). The result is a disaster for you to calculate the strain, as you have so many interactions between different factors impacting your strain readings. You cannot fudge factor this because you need to know where you are in the cycle to determine which fudge factor number you should apply, and on top of that, since you spider shaped oddly, even your pedal stroke will impact the reading, so you have no way of easily fudging out the “errors” on the drive side. The non drive side is much easier, as its a simple crank on an axle.
R8000 R9100. 100% every time using my test protocol. GRX are a different design so I’d need to see the data to know. I can test whatever I can get a hold of… which has become slightly more difficult after publishing my report.
I get more accurate data from Left only Shimano based power meters with my testing protocol. This shouldn’t be the case, a dual L/R crank should always be more accurate… that is unless one side reads incorrectly and introduces a problem and screws things up. I’m also monitoring L/R balance for the steady state intervals to ensure 50/50 average on the pedals… pedals I can (mostly) trust.
my understanding is that whenever you free wheel the powermeter will automatically do a zero offset. do you have the latest firmware? i just did a calibration on mine with a wahoo head unit, pressed the left hand button go to sensors, go to power, pedal to wake up powermeter, and the calibrate option came up on my screen. I hope this helps.
Because it’s a fixed gear track hub there’s no freewheeling. I do have latest firmware. But I will try to calibrate with Wahoo as you describe - hopefully that will do the trick. Thanks muchly!
Wut? Manufacturers are sulking in a corner because you have been “mean to them” in your reviews? From where I stand, you have done them a service, actually.
Even then I don’t think you need perfect 50:50 power balance, as long as the power balance is consistent, you can train off of them (sans single leg drills, obviously). That’d still be more accurate (in relative terms) at the expense of a user-dependent systematic error in the form of an offset when the power balance deviates from 50:50.
@Achin
Awesome posts, thanks.
So, bear in mind I’m a bridge engineer and don’t deal with things on this scale very often…
For axial strain under pure axial load the transeverse positions of the gauges is critical to ballancing the strain readings on an assymetric section. This is because the axial load will cause bending due to eccentricity of the load from the neutral axis. If you put the gauges in the correct position for a given cross section the readings will be the same under any pure axial load. (They will actually give equal and opposite readings in the weatstone bridge which cancel eachother out. This is how these bending beam arrangements automatically disregard axail loads.) This is, however, only possible as long as the bending stresses due to eccentricity don’t fully overcome the axial stress within the area that you can position the strain gauges.
In the video the position of the strain gauges close to the spider just happened to give good readings. Moving them away from the spider but keeping them at the same spacing lead to poor readings which could have been rectified by choosing the correct locations. I’m surprised that Keith didn’t rectify this and show us the axial strain plot.
For bending in the plane of the crank it doesn’t matter what location the strain gauges are in as the strains at the given loactions will be proportional to eachother under an given bending load (assuming that both gauges are not on one side of the neutral axis to rectify the axial load problem) and also proportional to the load so that the bending force to total strain be calibrated.
That’s all well and good, but out of plane bending causes another problem. Typically the out of plane bending (force along the line of the pedal axle) causes the strain gauges to see exactly the same strain as they are both on the extreme fibre. Again these strains cancel eachother out in the weathstone bridge and the output will be zero. This is, again, the fundamental principle on which bending beam systems are built. The assymetry in the out of plane direction can cause the strains to read differently, which the system has no way of differentiating from bending in the plane of the cranks.
My advise to anyone wanting to develop a crank based crank power meter that isn’t fundamentally flawed would be to avoid assymetric shapes. Just take a look at the Verve Inforcrank, a crank developed first and foremost to be a bending beam power meter. It’s symetrical!
Mike
One thing I haven’t got my head around with all of this is why active temperature compensation is required in any of the bending beam designs. It isn’t present in the Infocrank but that has been designed from the ground up to be a crank for measuring power.
My understanding is that there are three main temperature effects you have to deal with:
Differential between the coefficient of thermal expansion of gauge and parent material.
Overall strain due to thermal expansion.
Connecting cables resistance changes with temperature.
The mitigations for these are as follows:
Use of a gauge that is temperature corrected for the specific parent material.
Use of gauges in either a half or full wheatstone bridge, which cancels the strains due to thermal expansion in exactly the same way as it deals with pure axial load.
Is dealt with by the circuit arrangement so that changes in resistance of the connections cancel each other out.
There are a few more minor effects like gauge warming due to the voltage but these tend to be very small.
The only thing I can think of is that the gauges aren’t arranged in a wheatstone bridge or that the presence of other material types (steel axles etc) can affect the readings of the gauges, particularly when they are arranged close by. Perhaps it’s just cheaper to install a thermistor and deal with it all in post processing.
Can anyone shed any light on this?
Mike
Really enjoyed this episode. I have a 4iiii left-arm 105 that works great. I am curious though in that when I see what devices are paired in Trainer Road, the app shows two power meters. One is 4iiiii and one is Precision. I only have one power meter though. Are they the same power meter and perhaps one is paired via bluetooth and one is ANT+? But if they are the same power meter, why would the battery life show different.
Any ideas would be greatly appreciated.
Yep, one is bluetooth and one is ANT. They broadcast different names over different protocols. Looks at the symbol next to “Paired” to tell you which one is which.
Thanks @Jayacher That’s very helpful. I failed to notice the international symbols for Bluetooth and Ant+
I’m running oval rings on my MTB which by now I’ve learned is not ideal with a crankarm powermeter.
When I got my 4iii left arm power meter I matched it to the readings of my Kickr Core by setting the offset scale to 0.94 because the 4iii was reading conciderably higher than the Kickr Core. After I put the offset scale to 0.94 it was close enough so I felt I didn’t have to do a Ramp Test again (pre AI-FTP Detection).
Now for the big question, did I steal watts from my FTP and was the Kickr Core reading low? Or was the 4iiii reading high because of the oval ring? 
It’s nog really important, because as long as it’s consistent it will be a good training benchmark. But the ego in me says if the 4iiii is right and I can put the offset scale on 1 I’m a lot closer to my training goal of reaching 4w/kg as I don’t race and have no other short term goals at the moment.
