I'm rebuilding my '87 Giant Sedona. I bought new bearing races but they seem a bit off and I can't get the adjustment correct. I lost one bearing from the old race so using that is out. Can I use loose bearings for this and how would I figure out the number needed? I can go to a place that sells loose bearings.

Do you mean bearing race or caged bearings?
Bike shops sell caged bearings or you can use loose ball bearings of the same size as what is in the caged bearings. Fill the race with ball bearings, then remove one. (Someone will correct me if the best practice is to remove two.)

Yes. I meant caged bearings.

If you use loose ball bearings, there are more balls to share the load. Just make sure the balls aren't tightly packed in there - that's why you remove one.

The headset is the one place that you remove two balls, sometimes three.

https://icelord.net/bike/BarnettsBicycleManual.pdf

There's no reason why headsets should be handled any differently than other bearings when it comes to how many balls to use. (the entire assembly is the bearing, consisting of inner and outer races aka cups and cones, the balls are just that - balls) . The object is to fill the outer race to capacity, while making sure that the balls aren't forced against each other.

Generally when filled, there will be a space of about half a ball. Since it's harmless to be 1 ball short, and totally destructive to have one ball extra, folks who are unsure may remove one.

One way to be sure is to line the cup with grease, pack in the balls, and place it against the cone and spin them together a second or two then pull them apart and look. If the balls are in a nice neat circle with grease swirls from the movement, all is well. If you see one ball up or to the side, you overfilled it. You'll also know because with an extra ball the two parts won't want to lay flat against and you feel some wobbling as you rotate them.

"In the next step, some balls will be removed. The
reason this is done is to prevent headset failure from
brinelling. By leaving the balls room to move around
relative each other it guarantees that any brinelling
that occurs is in random locations. When ball retainers
are used or the cup is left full, the brinelling occurs
in the same places over and over again until it reaches
a noticeable depth and causes headset failure."

Barnett's Manual 5th Edition

I'm familiar with this theory, but (with due respect to Mr. Barnett) I don't believe it holds water. First of all, more balls mean more contact area. Secondly there is a gap already and more of a gap doesn't mean the balls will actually move more randomly, as they are trapped and roll back and forth with headset motion.

Lastly, once fretting or false brinelling the balls will tend to find their home base no matter how many or few there are.

The key to preventing fretting, or false brinelling is proper (not excess) preload, and a decent grease with adequate film strength and anti-oxidation properties.

With all due respect, Mr. Barnett has far more credibility than you.

Apparently with you, and I'm fine with that. But it depends on who you ask. I enjoy plenty of respect from many sectors of the bike industry, and in many quarters Mr. Barnett is not as well respected. But we're not in competition, so there's no need to debate the point.

The fact is that Mr. Barnett and I have a different approach to bicycle mechanics, and that's perfectly fine, with neither being right nor wrong in any absolute sense. Like most things that have any degree of subtlety, there's room for disagreement.

Look at chain lubrication. There are many possible approaches, and I chose to take a very different one. It doesn't make me right and the others wrong (or the other way around). But as they say, "the proof of the pudding is in the eating" and many who've tried mine will confirm that I make a damn good pudding (though, naturally, it isn't to everybody's taste)

To the OP, removing one or two, or even 3 balls from a full complement won't cause any serious problem (though I don't think it does any good either), but there's no disagreement that one too many will, so if in doubt go ahead and remove one more if you prefer.

Fbin +1

Pack in all the loose bearing that you can put in---yet have a small gap of bearings in the outer race. Simple mechanical fact, the more balls you have the more the load is spread.

Personally I think this business of removing 1-2-or 3 bearings is an old wives tail that has been told so many times it has become "the thing to do".

Thanks guys for all the input. what I ended up doing was taling a loose bearing I had from and old european headset 5/32 added that to the lod cage and repacked the headset. Works fine. Not the ideal solution but I just wanted to get this bike together so I could use it for commuting.

Not when he writes crap like that. I've never seen a brinelled headset race. I've seen lots and lots with fretting damage, which is a lubrication failure. Using fewer balls won't improve the lubrication problem, or the fundamental problem with most headsets, which is they have bearings that aren't capable of supporting the loads they're carrying.

-3 balls for the weight weenie win!

J/K. I go one ball short of a load on any bearing pack anywhere on the bike. It's been working fine. Haven't wrecked a headset since 1990 or so, and that was due to not keeping things tight.

A variation on this theme: can I get away with having a couple less loose bearings on the bottom race than on the top race? I was being a bit careless when servicing the headset on an old Peugeot and lost count. It feels quite smooth and rides fine at the moment, just want to make sure I'm not breaking a major rule. I'd take it apart and do it again, but the threads on the fork steerer tube are very rough and it's quite difficult to tighten anything without dislodging the bearings.