Originally Posted by
BCRider
Prathman, what you describe sounds like adapting to a pre-existing condition. If so then the usual rules do not apply. I've had to do that too to make do with a part on an old bike that I was just fixing up. But it's not a good situation depite your success so far.
No, in both of the cases I mentioned the headsets were brand new when initially setup with some amount of preload (but my estimate is that it was a bit less than you recommended). Both failed rather quickly and it was only after they started to fail (i.e. became slightly notched) that I did any adapting by removing the cages so I could insert more balls and then adjusting the headsets more loosely.
I may have also been a little overboard on the spec for 2 lbs of push at the fork leg to get it moving. But for lower to mid quality headsets I'd still want to see from 1 to 2 lbs of grab before the fork wanted to move. But the key is that when moving the travel should be easy and smooth with no notchiness or other signs of distress. If there are any then there's no option but to lighten it up a little if that fixes it. If it doesn't get any smoother up to the point of being too loose then I'd sigh and set it with some preload and expect it to either live or die as fate decrees. But under no circumstance would I set it to just barely no play since that just invites the risk of the bearings losing contact with the races and putting all the load on only one spot up top and one below. Bearings that are too loose and suffer from loss of ball or roller contact are a major reason for early bearing failure in cartridge style bearings. And it's no different with our situation.
But headset bearings operate in a very different mode than typical bearings. They do very little rotating at all and are instead loaded with a great deal of vibrational and fore/aft stress while remaining almost motionless with respect to angular rotation. I agree that crank and hub bearings should be setup with noticeable preload for the reason you cite, but my experience indicates that it should be avoided for headsets since they fail for a different reason - which was explained on Sheldon's website as referenced in my previous note.
I'll close that in the last 15'ish years of setting all of around 25 to 30 headsets up that I've done I've always set them for some preload but also for a smooth rotation. And unlike your experience all of mine have survived and those that haven't been sold off are still going strong with no sign of any indexing or other issues. So it would appear to be a case of YMMV. And then there's the thought that when you set it to just no play perhaps you're far enough on the tight side of "just no play" that you're putting in an amount of preload that isn't that far from what I'm suggesting after all.... <shrugs>
No, as I indicated before, the way I set my headsets is for zero preload with minimum play (not "no play"). If, due to manufacturing tolerances, I can only get zero preload by still allowing there to be a detectable amount of play then that's the way I leave it. I've never had that cause a problem and the only issues that I've had with headsets have been those set up by others with preload.