I fully disassembled, cleaned and regreased the bearings of my semi integrated headset. After reinstallation I was unable to adjust the "perfect" preload to eliminate all play and without having stiffness in the steering. Somehow there still is a tiny amount of play in the steering when checking really thoroughly with pulled front brake and moving the bike back and forth. When adjusting the preload this either results in steering stiffness or this tiny amount of play (no real rattling or knocking, more like some subtle play of the fork when paying close attention). The bearings didn't look damaged to me or anything like that. Is this subtle play acceptable/normal or should I investigate this further? Is it possible that some movement in the bearings themselves can transfer to the fork and create this slight noticable movement? I've added some pictures although I'm not sure if they're helpful.

1 2 3 4 5


I think I've finally figured out why there was still slight play despite the top bolt being tightened. It looks as if only the two spacers below my stem slide over each other and have subtle play. Since the spacers are not tight around the tube, some lateral movement is probably noticeable. I assume that this is not an indication of a poorly adjusted headset? If I fully tighten the adjustment bolt, the play in the spacers also disappears. The real bearing play can be eliminated with significantly less torque. I uploaded a video that hopefully shows the phenomenon: https://www.veed.io/view/04f7c2d1-d2ca-4d23-9220-7069fdd15ada

  • I clearly see marks where the bearing balls sat when straight-ahead. Are there indents that you can feel when lightly scraping a pick/fingernail over them? If so, that's brinnelled .
    – Criggie
    Jul 16, 2022 at 23:17
  • 1
    In the last picture it's just some old grease. The container ring (I think it's called retainer?) does not show indents just some superficial abrasions/discoloration if I remember correctly
    – conste
    Jul 16, 2022 at 23:26
  • Are you sure it’s bearing play? If you put your hands around stem and frame while rocking the bike back and forth (with front brake engaged), do you actually feel movement?
    – Michael
    Jul 17, 2022 at 6:32
  • Did the headset have bearing play initially which is what inspired the teardown?
    – Criggie
    Jul 17, 2022 at 9:21
  • Not that I'm aware of. Haven't ridden the bike for almost 10 years so I can't tell. I installed a new stem and just wanted to give it a good cleaning before I'll start using the bike again.
    – conste
    Jul 17, 2022 at 20:00

2 Answers 2


It's unclear as written what you're doing with the front brake to test.

Pulling the front brake and rocking the bike back and fourth with both hands on the bars is an inadequate way of testing headset play on most bikes. Most brakes of all types have some amount of play in them. Almost all road caliper brakes, the exceptions being a very few with adjustable ball bearings themselves, need some amount of free play in the pivots, or else they bind.

What can work is pulling the front brake with one hand and rocking the bike back and fourth while the other hand feels for movement at the transition gap between the upper cup and the "beginning" of the headset, which might be difficult on this bike because in this case that's the soft-looking seal type part below the conical spacer.

Assuming a front hub that's free of lateral play, usually just turning the bars 90 degrees and rocking the bike that way is better than using the front brake, if one is going to do this kind of test. That way play in the brake is completely eliminated as a factor.

Other good tests are, with the wheel on, putting the bike in a stand and grasping the bar in one hand and one of the fork legs in the other and trying to get movement that way, or dropping the front end of the bike from a height of several inches to check for rattle.

If there is definitely play in the headset you can't solve without causing binding, check the cups and crown race are all pressed fully, i.e. no light gaps.

Headsets of this sort, which look a little like modern cartridge bearing headsets but aren't, are low-end parts and yours does appear to have some wear on it, so that could also be the entire problem. However, that wouldn't usually cause it to be either loose or binding, more like either loose or very slightly rough but still functional.

Pay close attention to how well the compression ring mates with the top of the upper race. That's an area where these kinds of issues can sometimes arise. Sometimes what's happening is that by the time you're preloaded the headset enough for the compression ring to mate fully with the race, the top cover or a seal attached to it are pressing down against the top cup, causing binding. If that's the problem there are micro shims you can add between the compression ring and top cover to solve it. You can test whether this is the problem by seeing if the headset adjusts cleanly by substituting a spacer or spacers for the top cover, and have them press directly down on the compression ring.

It's possible for bikes to have headset play that can't be eliminated without introducing binding due to poor prep on the headtube or the crown. Don't jump to that conclusion, but it is there once you've eliminated everything else. (Milling the crown of a bonded fork of any kind isn't really a thing, but you could get the frame reamed and faced).

  • Yes I'm feeling for the movement where the stem sits on top of the steerer tube. I can feel it especially at the moving spacers below the stem. With the bars rotated 90 degrees it's even more noticable. To eliminate all play I have to tighten the top bold with quite a lot torque (if I had to guess between 3 and 4 Nm). Not sure if this is even normal for typical adjustment torque, also I know this can't be measured really. Is there typically a "range" for the correct setting or is this a small sweet spot I would have to find? The top cap definitely presses correctly on the tube for preloading.
    – conste
    Jul 17, 2022 at 20:10
  • that much torque just to eliminate the play is a pretty strong indication that something is wrong. look at whether the microshims can help. they are called part 1.3 in this: canecreekhelp.freshdesk.com/support/solutions/articles/… Jul 17, 2022 at 20:37
  • again, a simple test is take away all parts of the headset above the compression ring and replace then with spacers, and see if it adjusts normally then. if so the micro shims will almost certainly solve it. Jul 17, 2022 at 21:34
  • I think the problem is only present in the spacers below the stem. I updated my question and added a video link to show the phenomenon.
    – conste
    Jul 17, 2022 at 22:46

I finally managed to adjust the headset properly. I think that in the beginning two things deceived me, so I couldn't exactly feel the play and eliminate it. During the test with the front brake pulled, some knocking noises were generated, which I misinterpreted as coming from the headset. Either they came from the brake, a brake cable or something else. Due to an incorrectly set play, the spacers under the stem moved and irritated me even more. I finally managed to set the preload correctly by carrying out the test for play with the handlebars turned 90 degrees and without applying the front brake. I only focused on the transition points between frame and fork as well as frame and stem. In addition I blocked the wheel against a wall so the bike in itself was stable. This way I could easily feel existing play in the headset. Another test I finally did to confirm was lifting the front wheel up so the bike was vertical in the air, holding the fork legs and rocking/moving them back and forth. This will clearly show if there is too much play and at the same time eliminate all disruptive factors such as the brakes. For me this test is much more effective than checking the bike at the ground where multiple things can mislead you.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.