Do the driven flange bearings actually meet any of these criteria requiring bearings of the highest quality or is the quality bearing here one that meets standards much less stringent? The flange bearings do not support the weight of the bike on the axle, the spline loads and rotational speeds are low, ST1300s simply do not wear out splines if kept correctly greased, and there is absolutely no history of dangerous or expensive catastrophic final drive failures even if ST1300s are operated for long periods with toasted driven flange bearings. My OEM bearings served until 96,000 miles and a set of Peer bearings sourced at that time have served the last 70,000 miles. I remove the collar and separate the bearings for inspection every tire change but I suspect no bogey man hiding in the FD requiring anything other than OEM or inexpensive aftermarket bearings.
Yes, highest quality bearings will have effect. And so will proper fit of the mating parts. But sadly, we don't know what the proper dimensions are for every part.
It isn't about whether the 6905 bearings support the bike's weight, it is about load on the bearing. And low speed rotation can be more damaging when loads are high--static load ratings are lower than dynamic load ratings. If the design didn't require the highest quality and if the mating parts didn't really matter, then we wouldn't see *some* driven flange bearing failures, would we? And to be fair, even a highest quality bearing can/will fail prematurely if loads are frequenly very high on it, if the bearing is undersized, and/or if there's an unintended final stackup of dimensions and fit of the bearing in a particular assembly.
'Fortunately' the failures are detectable at tire changes and tire changes are frequent. ;-) And fortunately it us unusual for both bearings to fail simultaneously.
Unfortunately, there's no unbiased third-party tester of bearings. There are poor standards of testing (tests setup differently by different vendors to tout their bearing as better), and there's no uniform application and statement of bearing 'quality'. So, vendor A's 6905 bearing is not as capable as vendor B's 6905 bearing, even though they both claim the same rating (if vendors A & B tested each other's bearings, they'd both claim the other vendor's bearing was inferior to theirs). And then vendor C doesn't even test their bearings, doesn't have process control at the *micron level* that bearings require. If the bearing turns when it is assembled, vendor C ships it! I'll take vendor A or B bearings over vendor C's any day.
Bearing life is driven by bearing quality, by the amount of interference fit, by load, and by many other factors. Cheap bearings do not have the same level of control on tolerances of each ball, the raceway on the innner ring, the raceway on the outer ring, the OD of the outer ring, the ID of the inner ring, and the shape of the cage around each ball. Also, parts mating to those bearings have to be designed well, have to anticpate the load directions, load magnitudes, and temperatures--otherwise bearing life is short. Either/both the bearing and the application need to be well matched.
In studying the ST's wheel, driven flange, rubber isolators, rear ring gear assembly, etc., it is easy to see that with high applied torque (both during the long duration of when maximum power [torque] is applied and reverse torque spikes during engine braking)--given the ST's final drive assembly design, the driven flange tries to twist off-axis of the axle. The only way this load is taken up is by the 6905 bearings on one end, and into the smalley snap ring and rubber isolators on the other end (of the driven flange). Fortunately, most of the time there's not ~100 ft-lb of torque through the ST driveline (or we'd likely see only 100's of miles between 6905 bearing failures).
Obviously, I'm not the Honda engineer who designed all the mating parts--and there are a LOT of mating parts and bearings around the ST's rear axle (6 bearings total!!), and the loads between them is complex. I don't have access to Honda's prints on all the parts to see the tolerancing on every single part, to perform a FEA to see the loads through the driven flange, and then to make final calculations which match those parts' dimensions and tolerances to the bearing's dimensions and tolerances. But certainly, the 6905 is the smallest bearing there, and clearly by our collective experiences that bearing in this application is proven to be marginal.
If you search ISO bearing and shaft fits, and look at some of the online calculators and literature, it is easy to see how complex and confusing this topic is ... and how easy it is to make a design mistake. If you read bearing literature and speak with bearing design companies, you quickly get into the complexities of bearing dimensions and tolerances, and see that there *is* a difference.
In my searching, I see that some 6905 bearing manufacturer's state a static load rating of around 7kN (about 1573 lb-force), and others list that as the dynamic load rating, illustrating how uncontrolled the industry is. BTW, that load rating is for a bearing that has been properly designed, fabricated, and tested, and is based on a perfect application (greasing, cooling, inner and outer race fit, load direction, etc). 7kN is a relatively low number and is tied to the bearing's overall small size. With the ST's ~100 ft-lb torque through the driveline, and with the small dimensions of the ST's rear axle assembly to counteract that torque, I expect that loads applied to the left-most 6905 bearing reach or exceed 1500 lbf. I.e., I suspect 7kN is undersized for the design (and hence why Honda used 2 of the bearings side by side trying to solve the problem without making the bearing larger and growing the whole rear drive assembly's size).
I further suspect Honda has an error in the mating collar's bearing seat dimensions/tolerances (or Honda hasn't checked their collar vendor!). I have a spare collar, and it has different dimensions at the bearing seat than the one that came in my bike (and which was in place at the first bearing failure at 35k miles). Two 6905's don't fit the same on that orignal collar's seat compared to the replacement collar I have installed now.
So, I believe that Honda made a mistake (still has a mistake) in individual part(s) and/or the assembled dimensions/stackups of the rear drive design. This is not such a big mistake that it is needing recall, rather a mistake that allows some bearings to survive, and others not. And putting a higher quality bearing in place compensates to some degree for design mistakes. And as mentioned earlier, cheap bearings do not have the same load rating as well-controlled name-brand more expensive bearings. While cheap bearings aren't *certain* to fail, they are *more likely* to fail in a demanding application (the ST's rear assembly).
To wit: we see *some* ST driven flange bearing failures, and *some* of BMW's final drives fail. There's a lot of loads/torques in either final drive, and individual bearings and assemblies can lead to failure. Even the best engineers make mistakes in design, even the best bearing manufacturers can make mistakes, but it is guaranteed that no-name bearing manufacturers don't even know the dimensional variation that they manufacture into their cheap bearings--those cheaper bearings will on average have shorter life.
This post is too long... sorry all.
:
So, *I* use the best quality bearings (name brand bearings or Honda OEM) to reduce chances of short bearing life. Why test fate by using a $2.50 bearing when the OEM bearing is $15? How much would really be saved?