Index Home About Blog
From: (Jobst Brandt)
Subject: Re: Indexed headsets
Date: 6 Nov 1999 02:55:47 GMT

Judith Holman writes:

> Its mostly flat spots on the ball bearings and can be fixed by
> loosening the lock nut and top bearing fitting and dropping the
> steering tube down enough to be able to get at, and rotate, each
> individual ball to a new position on both the top and bottom
> bearings.

This is entirely new.  Whare did you see flat spots and how do they
affect the rolling balls in the bearing?  Have you tried this remedy
yourself or are you making this up at the KBD.  I'm sure that many
readers of this piece can tell you that what you propose and your
analysis of the cause is far from fact.

> Rotate them out of plane to their normal travel. Grease the bearings
> and reassemble / tighten. It might be easier to take out the stem
> and disassemble properly. New locations on each ball will restore
> the steering to non indexed.

Please explain how the bearing balls will not find their way back to
the favored positions where they have more clearance than elsewhere.

Even though Mike Iglesias posts the URL once a week, many readers who
should, don't read what is known on a subject or anything elase that
it contains.  I think you might understand the indexed bearing
phenomenon better if you read the FAQ but becasue accessing it seems
to difficult, I'll post the related item:
Subject: 8.23  Indexed Steering
From: Jobst Brandt <>

> In the several years I spent working in a pro shop, I have never
> seen a case of "index steering" (yes, we called it that) that was
> _not_ caused by a "brinelled" headset - one with divots in the
> races.  I am 99.999 percent certain that that is your problem.  What
> are you going to do if you don't fix it?  I suggest that you fix the
> headset even if you sell the bike, as a damaged headset could be
> grounds for a lawsuit if the buyer crashes.

I disagree on two points.  First, because you use the term "Brinell"
that conveys a notion as incorrect as the phrase "my chain stretched
from climbing steep hills" and second, because there is no possibility
of injury or damage from "indexed" steering head bearings.

Damage to head bearings seems to be twofold in this case because
properly adjusted steering, can only get looser from dimples and it
cannot be immobilized by them.  Therefore, the head adjustment was too
tight.  However, dimpling is not caused by impact, but rather by
lubrication failure that occurs while riding straight ahead.  This
occurs more easily with a correctly adjusted bearing than with a loose
one that rattles and clunks.  Rattling replenishes lubricant between
balls and races that would otherwise not be present.  Off road
bicycles suffer less from this malady than road bicycles because it
occurs primarily on long straight descents where no steering motions
that would replenish lubrication occur.

If you believe it comes from hammering the balls into the races, I
suggest you try to cause some dimples by hammering on the underside of
the fork crown of a clunker bike of your choice.  Those who hammered
cotters on steel cranks will recall no such dimpling on the spindle,
even though it has a far smaller diameter than the head bearing
although the blows were more severe and direct.  No dimples were made.

Ball bearings make metal-to-metal contact only under fretting loads
(microscopic oscillations) while the bearing is not turning.  Any
perceptible steering motion will replenish lubricant from the oily
meniscus surrounding the contact patch.  Peering over the bars at the
front hub while coasting down a road at 20+ mph you will notice the
fork ends vibrating fore and aft.  This motion does not arise at the
fork end, but at the fork crown, as it bends the steer tube.  Both
head bearings rotate in fretting motion crosswise to the normal plane
of rotation, as the steer tube bends.  Dimples form in the forward and
rearward quadrant of both upper and lower bearings from this fretting.
That they also form in the upper bearing shows they are not directly
load related.

Lubrication failure from fretting causes metal to metal contact to
form microscopic welds between balls and races.  These welds
repeatedly tear material from the softer of the two causing the
elliptical milky dimples in the races.  Were these brinelling marks
(embossed through force), they would be shiny and smooth.  Various
testimonials for the durability of one bearing over another are more
likely caused by the lubricant than the design of the bearing.  The
rigidly mounted ball bearing has survived longer as a head bearing
than it should, considering its poor performance record.

Roller bearings of various designs have been tried, and it appears
that they were the ones that finally made obvious that fore and aft
motion was the culprit all along; a motion that roller bearings were
even poorer at absorbing than balls.  This recognition lead to putting
spherical seats under the rollers.  Although this stopped most of the
dimpling, these bearings did not work well because the cage of needles
tended to shift off center and drag on the housing while the conical
races also shifted causing the bearing to bind.

It appears that a solution was finally found when Shimano bought a
patent from Wilderness Trail Bikes for a ball bearing that combined
the cup and cone ball bearing with the spherical plain bearing.

Today Shimano offers these bearings model called: LX, XT, 600, STX-RC
and Dura Ace.  They have a full-complement angular-contact ball
bearing, whose races are sufficiently reentrant to snap permanently
together.  They have rubber seals that retain grease for life of the
bearing that is not exposed to weather.  The ball bearing is supported
on a spherical steel ring that forms a plain bearing against the
aluminum housing.  The plain bearing takes up the otherwise damaging
out-of-plane motion while the ball bearing does the steering.  The
bearing is only durable as long as the plain bearing remains properly


From: (Jobst Brandt)
Subject: Re: Indexed Steering
Date: 18 Jan 2000 20:04:49 GMT

Don Gerhart writes:

> Granted the head bearings cannot cause injury or damage, but the
> lack (or restiction) of steering responsiveness can be a very
> serious situation.

The only "restriction" that could occur is that the unwitting user
re-adjusts the bearing clearance, not from the wear.  Therefore, it is
not a hazard, only an annoyance.  There is no loss of control from a
loose head bearing, only a loss of road feedback that should make the
rider realize that he has a mechanical noise between him and the feel
of the road.  That is, if the road is rough.  On a smooth road even
this is not perceptible unless braking at exactly the rate that makes
the bearing float.

Jobst Brandt      <>

From: (Jobst Brandt)
Subject: Re: Indexed Steering
Date: 18 Jan 2000 23:09:36 GMT

Mark Hickey writes:

>>> The reason is that the races inside and outside are different
>>> diameters, so turning the inside race (i.e. the one connected to the
>>> fork) will result in a slightly "smaller than one notch" movement of
>>> the balls on the outside race.  That means the balls don't fall into
>>> both sets of "notches" at the same time.  This assumes there is wear
>>> on both races, of course - if you replace just one or the other race
>>> the notches should be the same through the rotation of the fork (now
>>> assuming the headset is installed truly perpendicular to the head
>>> tube).

>> Oh no.  With 20 bearing balls, (typically Campagnolo) dimples form at
>> uniform angular spacing around both races although not uniformly deep.
>> Therefore dimples will match with every 18 deg rotation of one race
>> with respect to the other and the bearing balls moving at half the
>> rate of the moving race, will fall on a dimple pair every second time
>> the dimples are aligned.

> I agree that there will be 20 dimples of uniform angular spacing on
> each of the races in the example above.

> When you turn the fork, the inner race "drives" the ball bearing in
> a vernier motion.  As the race rotates under the ball bearing, the
> ball bearing will eventually drop into the next dimple on the inner
> race.

> In doing so, the ball bearing will have "rolled" on the surface of
> the inner race a distance equal to 1/20th the circumference of the
> inner race's surface.

> But, during the same time, the ball bearing can roll only the same
> amount of distance (linear, not angular) along the surface of the
> outer race.  Since the outer race's circumference is larger, the
> ball will not have yet reached the next dimple when it falls into
> the first dimple on the inner race.

Excellent.  And because the balls could only drop in at the second
coincidence of the inner and outer race on a linear bearing, the
offset of the ball to that position would be far enough off that it
would not fall in.  However, (assuming 20 balls) they would fall into
a dimple every 36 degrees on the rotating race and a slightly
different angle on the stationary race, the reference race being

This makes the perceptibility of indexing to less than half of what it
is straight ahead primarily from differing angular velocities and
secondarily by the diminishing depth of dimples away from the straight
ahead position.

From looking this up in bearing analysis texts, the ratio of motion
along the inner and outer races, or period at which the ball will roll
into a dimple is:

(pd+bd)/(pd-bd)  pd : pitch diameter of the ball (center)
                 bd : ball diameter

This is the basic effect but angular contact only changes the effective
ball diameter and puts spin into the ball on an axis about the contacts
with the races.

I seems evident that unless 'pd' is an integer multiple of 'bd', that
passage of a ball over dimples in the inner and outer race will not
coincide, at least not in one revolution (of one race with respect to
the other).

I think this ought to be included in the FAQ item on the subject.
Thanks for the correction.

Jobst Brandt      <>

From: (Jobst Brandt)
Subject: Re: Indexed Steering: a movie
Date: 21 Jan 2000 21:06:42 GMT

Damon Rinard writes:

> While the wheel and stem were off my wife's Merlin I noticed she has
> an indexed headset. Because of the discussion about 36 degrees being
> (or not being) the next alignment spot, I checked. It definitely has
> a secondary alignment spot near 36 degrees (by eye). I'll count the
> balls when I take them out.

> I made a short 5 second MPG movie of the fork snapping into
> position.  See (136KB).
> In the movie I can hear the brake's barrel adjuster rattle as the
> fork drops into each position.

I see, you took out the clearance before making the test.  That is a
great demonstration but it probably doesn't conflict with the analysis
that the balls don't exactly reach the coincidence of the dimples at
36 and 72 degrees.  The difference in ball pitch diameter and ball
diameter is so large that the (pd+bd)/(pd-bd) difference places the
ball just beyond the edge of the dimple and it could fall in, there
being some overlap of the dimples in the inner and outer race at the
ball position at 36 degrees (with 20 bearing balls).

I believe that with larger bearing balls with respect to the pitch
circle, overlap would probably diminish, or at least it would be a
closely spaced double detent, of half the depth.  The one at 72
degrees should have a greater spacing for these paired detents.  Can
you detect a double detent?

Jobst Brandt      <>

From: (Jobst Brandt)
Subject: Re: Ceramic balls in headset
Date: 15 Apr 2000 00:24:17 GMT

Ben Davies writes:

> I've never really bought the "fretting" explanation either.  If the
> molecules really "migrate" at points of contact, then they would bo
> this on the shelf, wouldn't they?

No they wouldn't.  I don't know what you mean by migrate but there is
no fretting (motion) just sitting on the shelf.  However, spare head
bearings for lathes, stored in a shelf in the machine have been ruined
from fretting.  This occurrence is cited in an engineering text as a
typical example.  I recall in the 1950's that GM experienced damaged
differential and wheel bearings on RR shipped autos whose rocking
motion was to tightly restricted.  Fretting damage to bearings is a
well understood phenomenon in industries other than bicycles.

> Bearings tend to wear differently, based on the materials that they
> are made from, and the type of loads that they carry,(assuming
> lubrication is a constant).  If you can "fret" a brand new headset,
> by running your bicycle into a solid obstruction, then the cause
> seems pretty clear.

That's a big if that will not occur.  You can pound on the top of a
stem bolt or under the fork crown with a hammer to no avail.  You
cannot damage a head bearing by such of impact.  This experiment has
been unwittingly done often.

> It isn't worth inviting argument about causes however, when there
> are so many variations in headset design and construction materials,
> operating in so many variations of conditions.  What has come to be
> known as "fretting", may actually be many different types of bearing
> wear, each generated under different circumstances.  There are
> simply too many variables for a single conclusion.

There are few variations and the result is that they all fret more or
less and develop dimples with use.  Just the fact that these dimples
develop over as much as 10 to 30 thousand miles proved that it is not
impact.  If it were impact, one hard bump, or for that matter one
Paris-Roubaix on the cobbles would produce them, but they don't.  if
the impact is below the yield stress, repetition has no effect on the
bearing other than fretting damage (the metal to metal contact that
causes micro-welds and tear-outs).

I think you should consult an engineering text on fretting damage
rather than conjecture about this phenomenon, putting forth erroneous

Jobst Brandt      <>

From: (Jobst Brandt)
Subject: Re: Ceramic balls in headset
Date: 17 Apr 2000 21:05:36 GMT

Ben Davies writes:

>>>> You can pound on the top of a stem bolt or under the fork crown
>>>> with a hammer to no avail.  You cannot damage a head bearing by
>>>> such of impact.

>>> Yes, I've accidentally struck a bearing, and the result was
>>> indentations in the race and rollers.

>> That would be a first.  How did you do this?  Considering that
>> hammering on the stem and stem bolt is an old custom in bike shops
>> and that it continues to be done because there is no damage to the
>> head bearing, you must have either done something drastically
>> different or your head bearing was already dimpled before you hit
>> it.

> Ok, first of all, this happened to a SKF type caged roller bearing
> on a motorcycle, not on a bicycle headset or bottom bracket. The
> races were clearly peened by what became elliptical rollers in the
> process. The amount of force required to create this type of damage
> to a new bearing was not negligible, but not enormous either. The
> impact was lateral (90 deg. from axis).  I expect you could
> duplicate this type of damage in a bicycle headset merely by
> overtightening it.

Not so.  I think you should place a bearing ball on an anvil and
strike it with a hammer to see that your scenario is the result of
great imagination.  Bearing balls and rollers can be cracked but not
plastically deformed into elliptical rollers.  Besides, what has this
got to do with bicycle head bearings of which we speak here?

> My position is, and has from the beginning been; that the common
> damage to bicycle headsets, is probably due more to poor
> maintenance, harsh operating environment, lateral impacts, or
> improper installation and adjustment, than the longterm effects of
> fretting.

That may be your contention and it is consistent with the folklore of
bicycling, the mechanics being just that and not scientists who
analyzed the phenomenon.  You seem to ignore that this is well
understood in other industries that employ more highly skilled
technical people, the machinery being far more expensive than bicycles.

> I don't know of many people, not even racers, who can boast the kind
> of longterm use that would result in fretting damage, unless one of
> the other causal factors is also present.

I don't know what you think this statement does other than say by the
back door that fretting is not a cause of bearing damage.  Head
bearings have been damaged in 5-10000 miles of use.  As I pointed out,
this can impossibly be caused by impact because if it were then it
could be caused in a single ride.  Steel does not get soft with time
and therefore, there is no time dependence of such loading.  Either it
exceeds the elastic limit or it doesn't.

In contrast, fretting is a progressive erosion of the bearing contact
that eats away at the interface between ball and races whenever there
is vibratory motion, such as occurs when the fork vibrates fore and
aft from surface roughness of even a paved road.  Inspection of the
bearing reveals a milky finish in the indentations that is not found
on other parts of the races.

> I was never trying to suggest that fretting is not a real phenomenon
> in other operating environments.  Though I am still not convinced
> that it is common in bicycle headsets that are properly maintained.

But you do directly state that it is not the mechanism by which
steering bearings fail, something that flies in the face of decades of
engineering work to prevent exactly this failure.

>> As I have mentioned, before the advent of the square taper crank
>> assembly, bicycles had steel cotters that were driven in by
>> hammering on the end of the crank assembly.  Even with a support
>> under it, something that was routinely omitted by experienced
>> mechanics, no dimples were caused in BB cups.  This is even more
>> telling because this is a load on a single 1/4" ball rather than a
>> 20 ball complement in a head bearing.

> This is why I don't let other people work on my machinery.  You
> cannot convince me that hammering on a part will leave no mark, not
> even if the part is carbon steel.

I don't want to.  I want you to try it and convince yourself.  That's
why I take the trouble of describing methods by which you can
determine this for yourself.  I don't expect you to take this on faith
or because "I have assembled more bearings than the next guy" or have
a PhD in tribology.

> I can think of almost NO circustances where you might want to hit
> something with a hammer, if you plan to re-use the component.  They
> don't make cottered cranks anymore, wedge fit stems are losing
> marketshare, and this may be a very good reason why.

I think you ought to toss out your hammers then.  Besides, these
methods required a degree of skill that was not reliable and other
assembly methods are far more repeatable.  You are assuming a link of
cause and effect that is not supported by good evidence.  The wedge
retention of handlebar stems is the cause for many seized stems
because they are held only at the bottom end and pump water in at the
top, regardless of prior lubrication.  The lubricant emulsion is a
corrosion accelerator that serves to firmly lock the stem in place.
This is the major advantage of clamp on stems.  My road stem has an
expander at the bottom and a clamp on the extended steer tube to
prevent that.

>>>> There are few variations and the result is that they all fret
>>>> more or less and develop dimples with use.  Just the fact that
>>>> these dimples develop over as much as 10 to 30 thousand miles
>>>> proved that it is not impact.  If it were impact, one hard bump,
>>>> or for that matter one Paris-Roubaix on the cobbles would produce
>>>> them, but they don't.  if the impact is below the yield stress,
>>>> repetition has no effect on the bearing other than fretting
>>>> damage (the metal to metal contact that causes micro-welds and
>>>> tear-outs).

>>> I offered no hypothesis at all on this matter.

>> Well then you are lying.  I think you were better off leaving your
>> presentation as tentative rather than a statement of fact.

> Now that's a little strong...

Your statements are a bit strong for the conjecture they are.  Maybe
you don't recognize that to persist something you are not sure of as
fact is not truthful.

> If you look at the beginning of the thread, you will note that I
> replied to someone's post that had stated that he felt the fretting
> conclusion is not correct.

You can't do a hand-off here.  I am talking about what you have
posted, not what someone else opined.

> You will also note that my reply was, (quoted from original post):
> "What has come to be known as "fretting", may actually be many
> different types of bearing wear, each generated under different
> circumstances."

You have since stated emphatically that it is not fretting but rather

> I was very careful to make no presentation of conclusion, apart from
> anyone's ability to produce the same type of bearing damage by force
> of impact. I did not quantify the force required, but I was thinking
> of forces strong enough to damage wheels or forks (common speeds).
> I am not back pedaling here... Please read my original post.

So let's not beat around the bush.  Do your presentations on the
subject conclude that this is an impact phenomenon as has been put
forth in bicycling up until now or not?  And if not, what causes the
dimples in bicycle head bearings?

>> Maybe you can expand on the "numerous conditions" that cause
>> dimples in bicycle head bearings to extricate yourself from this
>> appearance of repeating myth and lore.  Also, how does this vary
>> with the brand name on the ball bearing?

> I have already listed other types of known bearing wear in my
> previous post.

There you go again saying that these are "known bearing wear" when in
fact they are not know to cause head bearing dimples.  They are
believed to be such by misinformed people.  Besides, these included
mostly problems that do not apply to the majority of head bearing
failure, ie no lubrication, dirt in the bearing, water intrusion, etc.
Perfectly maintained bearings suffer from fretting damage.

> I've seen cheap head bearings that were "streaked", and some that
> were speckled.  I have seen balls removed from headsets, that were
> clearly flaked.  Bearings will differ in both quality of materials,
> and manufacturing methods.  Even you, cannot refute that not all
> products are created equal.  Nor can you argue that identical
> products will perform the same way in separate and different
> environments.

Let's restrict this to Shimano or Campagnolo and toss out all the red
herrings and ill maintained ones.  These bearings suffer from dimpling

>> The "impact indentations" residing casually at the end of your list
>> of causes is the root of this discussion and you try to keep it
>> alive by sweeping it in as one of many causes.

> Ok, then lets assume for a moment, that the fretting conclusion is
> correct in most circumstances.

> Does it explain why the top bearing is more likely than the bottom,
> to develop this type of damage?

It doesn't and it doesn't.  It is neither true nor does it explain
anything about that.  It is however true, that the upper bearing
suffers, but not more so, than the lower one.  On the other hand, how
do you explain the dents in the upper bearing considering that the
lower bearing receives the major impact, the fork being primarily
axially loaded under impact loads.  That is the reason for the rake

> Does it explain how the typical single dimple, or less common pair
> of dimples, are only likely to appear at the front and rear of the
> race, and not over the entire surface?

There is no single dimple but rather a series in the front and rear
quadrant that fade out toward the sides where vibrational movement
approaches zero.

> Is there something revolutionary about the idea that smacking
> something real hard will leave a dent?

I already replied to that contention.  I put my hand on the ground,
you place the front tire on it and pound on it with a hammer.  The
impact load is trivial.  Meanwhile the front axle supports these same
loads on two balls and suffers no damage at all.  Wheel bearings fail
by spalling and never show any dimpling.

> I'm going to make a quantified statement here, instead of bowing to
> the master...

Oh oh.  Now you are using the old, "well if you know everything..."
routine.  Don't bother.  The ploy has been over used here by others who
attacked these subject with folklore and less than adequate analyses.

> If only the top race is damaged at North and South, and the damage
> is at the working end of a two foot lever (fork), that is exerting
> forces along this plane, then the result might be fretting, but not
> the cause.

Your hypothetical model is faulty to begin with.  Both top and bottom
bearings suffer from this failure and it is not a two foot lever that
causes the forcing function.  I think you should go back and analyze
how you can get enough force to indent a bearing.  New bicycles that
crash into a car or wall, for that matter, still have perfectly good
head bearings.  Brand new components have been recovered from new
wrecked bicycles.  This quite aside from the experiment where you put
a complete head bearing on an anvil and give it a whack with a common
1/4 lb ball peen hammer without damaging effect.  Try it, you may be
surprised.  On the other hand you may not because you should know this

> Fretting is defined as wear caused by axial movement, and where this
> does occur, most of what I have seen in bicycle headsets looks like
> wear due to tangetial movement.

That is an incorrect definition and I have no idea where you got it.
fretting damage in ball bearings is causes by micro-motions that cause
partial rolling of the ball so that a portion of the contact patch is
never unloaded.  That is, no new lubricant can enter the contact area
and the sliding that must take place in rolling Hertzian contact
causes welding.  This effect occurs more frequently in splines of auto
axles where elastic deformation causes sliding in the interface.

> The only assertion that I have made, is that not all headset bearing
> failure is attributable to fretting.  I don't think this can be
> called a wild speculation.

And I will counter by that ALL dimpling of bicycle head bearings is
caused by fretting, there being no other mechanism that can cause it.

Jobst Brandt      <>

From: (Jobst Brandt)
Subject: Re: Campy headset - how tight?
Date: 28 Apr 2000 21:16:07 GMT

Mike Jacoubowsky writes:

> #1: Your headset should *not* be coming loose during a ride!
> Something's not right here...either your top cone and locknut aren't
> tightened up against each other, *or* your locknut is bottoming out
> on the top of the fork instead of tightening against the washer
> below it.

You didn't say what the remedy is although this is the most common
cause of loosening head bearings.  Maybe you don't like self promotion
but what the bicycle needs is a thicker lock washer or a shorter steer
tube, the first is easier to come by and leaves open the use of a
taller bearing at a later date in the event this one wears out.  Mike
has such washers and you can double up on them as well.

> #2: Proper adjustment of a headset is something very few ever
> accomplish.  Too bad, it's not that tough.  The problem is that it's
> *very* easy to overtighten a headset without realizing it, and an
> over tightened headset does mean nasty things to your headset parts,
> just like too loose a headset does.  In other words, you have to get
> it just right.  How is this done?

Yes the bounce test you mention is one end and the swing test is the
other.  The bounce test assumes you don't have a loose crank or front
wheel bearing, because both of these can give the same head bearing
type chatter.  Check them first.  The swing test is more important
because ball and roller bearings work best with light preload or in
the case of BB spindles, moderate preload that can still be turned by
a fingertip grasp.

The swing test is done by hoisting the bicycle on the shoulder (road
bicycle) with the front end tilted down just enough to make the front
wheel want to swing straight ahead.  The steering should swing side to
side slowly when pushed off to one side.  A little drag should be
noticeable.  Of course this is also a precise way to check if any
indexing dimples are developing, a much finer test than bounce.

> Properly adjusted, even the cheapest low-life headset in the world
> will seemingly last forever.  Improperly adjusted, you might get
> half a season before suffering from a notched, or "self-centering"
> headset that requires replacement.

Wellllll... I wouldn't go that far, but proper adjustment with a
slight preload helps prevent fretting damage.  A loose or free bearing
will allow vibratory fretting motion even on smooth asphalt.  I think
that is why indexing accelerates once it begins, because the bearing
cannot be adjusted with with preload or it will jam in the dimples.

Jobst Brandt      <>

Index Home About Blog