Disk brakes and quick releases - what you need to know

(Updated Jan 2006 with the first out-of court settlement)

---

Yes, that's right - a UK cyclist has recently settled out of court with a fork manufacturer, in what is I believe the first successful case of its kind. There are a few more details in this thread here - the rider himself is hoping to avoid the limelight but thought that others might like to know.

---

The purpose of these web pages is to explain the danger of using a disk brake with a standard quick-release fastening for a front wheel on a bicycle.

• This page you are now reading covers the main description for failure, and evidence in support. What happens and why..
• This page quotes several experts who have commented on the matter. If you aren't prepared to take my word for it (and why should you?) then have a look at what the independent experts say.
• Here is the press coverage.
• The CPSC letter gets a page of its own
• I've had a problem with my disk brake + QR. What should I do about it?
  
• A chronological look at how this all happened, which will also be kept reasonably up-to-date with the latest news and current status of the problem.
• The manufacturers have hardly responded so far, but I'll put up anything more they say.
• One year on, and the manufacturers deny they have ever heard of this problem.
• Finally, I got hold of a copy of Cannondale's tests. Read and weep (or laugh).
• At last, an admission from Marzocchi.
  

[NB for 'quick release' please read 'conventional axle in slotted dropouts'. A simple nutted axle may be slightly better than a QR in theory, as it can provide a higher clamping force. However I'm not convinced that in practice it is safer, as it will be less obvious if it loosens or slips, and in fact the greater bolt diameter will generate a stronger unscrewing force. The safe alternative is a properly clamped axle combined with fork ends with a blocked exit slot, as shown in the pictures below.]

A brief `Executive Summary'

There are two main aspects to the failure.

• The disk brake generates a massive force largely downwards in the direction of the open fork ends. The friction of a quick release skewer is often not sufficient to stop the axle slipping down in the dropout slot.  This is explained in more detail here,
• The QR is initially restrained by the retention lip on the fork (assuming it is present), however over time the slipping of the quick release leads it to unscrew, which is described here. Once it has unscrewed enough, it can be forced over the retention lip and the rider will crash. 

Below the analysis, we have a section of  anecdotal and experimental support, and I've put the small number of  manufacturers' comments on a separate page.

But let's start off with a little bit of background:

The myth of the infallible quick release.

Most people agree that quick releases are a very safe and secure method of fastening the front wheel to the fork. For example, a quick web-search on the subject of quick releases turns up such links as this discussion here where the noted cycling expert and writer John Forester says

---

"I have never seen an authentic case of a properly locked quick-release device coming loose during cycling. The same device is used on both front and rear axles, and when used on rear wheels the device withstands a force equivalent to a direct pull of about 4 times the weight of the cyclist. That is, about twice the weight of the cyclist on the chain side, which is equivalent in a direct pull of 4 times. If the rear quick-release is not sufficiently tight, the wheel pulls over and the tire jams against the left chainstay. Presumably, the same effect would occur at the front unless the pull was direct, in which case the resistance of both sides must be counted, which is why I say 4 times. In any case, we know that wheels collapse at much less side load applied at the rim than is required to move the axle. The idea that such a direct  pulling force can be developed during cycling is absurd."

His comments seem entirely reasonable and correct - with one important proviso that is not stated. HE'S ONLY TALKING ABOUT BIKES WITH RIM BRAKES (his recent comments on disk brakes are now on the experts' page). Unfortunately, many cyclists and manufacturers do not realise this, and have come to accept this point of view as some sort of universal truth: the quick release is infallible and therefore any failure must be due to operator error. Here's another similar comment:

---

"In twenty five years of avid riding and a few years of working in bike shops, I have *never* seen
a quick release hub fail to retain the wheel except when used improperly."

---

The first basic problem of a disk brake is that the position of the disk calliper means that the frictional force of the brake pads on the disk acts largely to push the wheel downwards, broadly in the direction of the open fork ends. The force that the pads exert on the disk is very large - much much greater than the force from a rim brake (due to the smaller diameter of a disk compared to wheel rim) and it is also at an inappropriate angle. Here's a photo of one of my forks - a Marzocchi Dirt Jumper 1. The geometry is basically standard. This photograph is the 'QR20' version which is a bolt-on arrangement with caps over the open fork ends, but the standard quick release version of all of their forks (and most other manufacturers) has open fork ends aligned parallel to the stanchion, so I use that assumption in my calculation.

The relevant forces and dimensions on the wheel are indicated. We have the braking force B and the ground reaction force R acting at the contact patch. The disk calliper exerts a force D as indicated, tangential to the disc at the point of calliper contact. The radii of the disk and wheel are r1 and r2, and finally the angle of the dropout exit is 'a' in front of vertical. For this fork and calliper, the force D is virtually vertical. If it wasn't, there would be another angle 'b' for the angle the disk force makes behind the vertical.

Now for the simple sums. Let's assume we have a bike + rider weighing 90kg in total, braking hard and decelerating at 0.6g (6m/s^2) with the front brake alone (this is a reasonable estimate for maximal braking). The rearward force is 90 x 0.6 x g = 540N, and the vertical reaction force is 90 x g =900N (all the weight is on the front wheel). Taking moments around the axle, the force D exerted by the disk is given by D = 540 x r2 / r1. Here r2 = 13 x 25.4mm is the radius of the wheel and r1 = 72.5mm is the effective radius of the disk (ie to the centre of the  force at the pads, rather than the outer edge = 10mm less than the full radius of 165mm / 2). So D = 2460N, acting vertically downwards. This acts entirely on the left hand side of the wheel, but the ground reaction force and braking forces are split equally between the two sides. So we are left with resultant forces of 2460 - 900/2 = 2010N vertically down and 540/2 = 270N rearward. The sum of those is equivalent to a single force of 2030N acting downwards at an angle of 8 degrees behind vertical. Resolving this parallel to the dropout opening angle (= 18 degrees ahead of vertical for a head tube angle of 72 degrees) leaves a force of 2030N x cos(18+8) = 1825N out of the dropout on the left hand side!

UPDATE Cannondale's own measurements suggest that the ejection force could be about twice as high as this estimate, as the peak braking torque can in fact be significantly higher than my figure.

Some typical fork end geometries

Marzocchi Dirt Jumper 1 - the disk force is even slightly forwards of true vertical here! The fork opening would be lined up with the `quick  release' lever on the qr version of the fork (this is the 'QR20' with a metal cap bolted over the fork end).	Stratos FR4(T), with the disk force at about 10 degrees behind vertical, perhaps 30 degrees from the stanchion angle. Again, I've got the bolt-through version, but in most QR forks, the fork end opens in line with the stanchion or perhaps fractionally forward of it.
Both of these forks are actually 20mm bolt-through axle versions, but the geometry of stanchion and disc mount is typical of quick release forks too. The brakes are Hope 4 piston models (2 opposing pairs of pistons, the position of which is clear). I believe that most disk brakes would apply forces in much the same direction, although perhaps a 165mm disk would exert a force even more closely aligned with the stanchion than my 185mm disks do. The smaller disc generates a larger force too. Possibly the DJ is slightly worse than most other forks, however changing this angle by 10 or even 20 degrees won't affect the result enough to turn a poor design into a good one.

1825N pull - is that a lot?

The most recent ISO standard (1996) specifies that the quick release should be able to withstand a direct pull of 2300N, symmetrically applied, without slipping. That sounds like a big number, but then you realise that this figure only works out at 1150N per dropout. For that matter, the standard specifies a symmetric pull and no requirements are set for withstanding an asymmetric force. Compare that to the ballpark figure of 1825N on the left dropout alone, and you'll understand why I am worried. The limited testing I've seen (which is available on the web as this pdf file) found that all the skewers tested do meet the ISO standard quite comfortably, but only when carefully installed with the correct torque (various 'rules of thumb' were found to be unreliable). It's obvious to me that the ISO specification was designed with conventional rim brakes in mind. Prior to 1996 the ISO standard was only about 500N force, which is still far more than a rim braked-bike will see in normal use.

It is very easy to find anecdotal evidence that the wheel frequently slips under hard braking, as predicted by the analysis. There are one or two examples of such stories here.

Part 2. The self-extracting quick release skewer.

Up to now, all I've demonstrated is that the skewer can slip in the dropouts, even if correctly used. Although clearly undesirable, this isn't obviously all that dangerous at a first glance. Almost all forks have significant retention lips at the fork tips and it's hard to see how a tight skewer could pull over these even under the force of a disk brake. For starters, it implies the skewer stretching by about 2mm, at which point it would probably snap or show obvious signs of damage. The massive force would either shear or smear off the lips, and although the disk brake generates large forces, they aren't this big.

And we all know that the quick release cannot unscrew, don't we? That belief is every bit as fundamental as the belief that the skewer cannot slip. It now seems clear to me that it is every bit as false, too. From http://www.boltscience.com/pages/vibloose.htm :

---

It is widely believed that vibration causes bolt loosening. By far the most frequent cause of loosening is side sliding of the nut or bolt head relative to the joint, resulting in relative motion occurring in the threads. If this does not occur, then the bolts will not loosen, even if the joint is subjected to severe vibration.

Pre-loaded bolts (or nuts) rotate loose, as soon as relative motion between the male and female threads takes place. This motion cancels the friction grip and originates an off torque which is proportional to the thread pitch and to the preload. The off torque rotates the screw loose, if the friction under the nut or bolt head bearing surface is overcome, by this torque.

There are three common causes of the relative motion occurring in the threads:

1. Bending of parts which results in forces being induced at the friction surface. If slip occurs, the head and threads will slip which can lead to loosening.

2. Differential thermal effects caused as a result of either differences in temperature or differences in clamped materials.

3. Applied forces on the joint can lead to shifting of the joint surfaces leading to bolt loosening.

Work completed during the 1960's in Germany indicated that transversely applied alternating forces generate the most severe conditions for self loosening.

---

But condition 3 is precisely what we have just shown to occur with quick release skewers! The belief that a correctly fastened skewer cannot unscrew, is based on the premise that it cannot slip, and when the latter fails to be true, so too does the former. It's not clear how technically precise it is to label this as 'vibration loosening', since it may be forced more by intermittent slip rather than regular vibration. However, disk brakes do generate a lot of vibration, especially under the hard sustained braking of steep fast descents when these failures tend to occur. In any case, the important point is clear: if the skewer slips, it cannot be regarded as a secure fastener, and will tend to unscrew.

This should not be particularly surprising. After all, many other bolts connected to the disk brake system (ie holding the calliper to the frame, and the rotor to the hub) work loose regularly. It happens on all of our disk-brake equipped bikes, I've seen it happen on others, and loctite is widely recommended as a solution. The disk brake can generate a huge amount of vibration and it's all focussed directly in the hub area. There's no reason why the quick release skewer should be immune from this, it's just a bolted fastener.

Again, stories of quick release skewers loosening in use are common. Some of them could theoretically be due to 'operator error', but I find most of them very convincing. Read for yourself and make up your own mind.

UPDATE Cannondale's own tests show that once the QR is loose, dropout failure may result even if the QR is not actually loose enough to be forced over the lawyer lips. Some of the failures I have heard about definitely involved the rear of the left dropout being sheared off, but in other cases this did not happen. A tight QR apparently shares the load better across the dropout so that failure is less likely (but they only performed a single test of this).

Some experimental and anecdotal evidence. 

Since first presenting this theory at the end of March, I've found a large number of anecdotes that fit the description very well. Often for any given story it is difficult to prove beyond any possible doubt that there was not some user error involved, as although the cyclist is generally confident that they fitted the wheel properly, they cannot swear on oath that this is the case - it's such an everyday and simple event that they do not remember every precise detail. Given the strength of the belief in QR infallibility, they are persuaded by friends that perhaps it was their fault all along. However along with the sheer weight of entirely consistent stories, there are many completely convincing descriptions of the failure, including several real gems of careful observation and description that put the matter beyond reasonable doubt. This tale comes from this thread on singletrackworld.com. I have marginally edited the text to clarify a couple of points that were subsequently explained by the original author.

---

---

NEW Some 'experimental' data has also been generated by bike shop owner Ben Cooper, on this page:

Disc bike: after 3 days, the “bite point” was 80 degrees - the experiment was halted.
V bike: After a week of use, the “bite point” was still 90 degrees."

I understand he has repeated the experiment at least one more time, with the same result: the skewer loosened, when disk brakes were used.

Axle slip has also been measured by Ernst Brust of the "privately-owned, industry-respected testing lab velotech.de", as reported on bikebiz and STW.

There are plenty more personal descriptions of the failure. For example, on the subject of axle slip, we have this comment here:

Of course the pull down on the left hand side is exactly as expected from the positioning of the disk brake. He got it to stay put eventually, but how do you think "wearing a glove to reduce the pain of the skewer in my palm" compares to the manufacturer's recommended installation torque? This sort of story is very common, but I don't see the point of putting up a pile of them.

Skewer unscrewing is also regularly reported, here is a particularly clear description:

---

"Both my G/f and I have had problems with Hopes.

1) They were done up f@cking tight.

2) Every time they've come loose, the lever has been shut, but instead of being next to the fork leg, its pointing straight down to the floor, implying they've "wound" loose. Spin the lever back through 180° to its usual position and its "normal tight" again."

---

Here the evidence of unscrewing is absolutely crystal clear in the rotation of the lever. It's not always so obvious, as when the lever is on the right, the rotation of the nut on the left cannot easily be observed. However loosening of the skewer is quite a frequent occurrence.

The whole thread from which that last comment is taken was a bit of an eye-opener to me. Seems like there are a lot of people with skewers loosening, and wheels dropping out, due to the disk brake. The stories are convincing, consistent and worrying. Of particular note also within that thread is the comment by Chris Juden, Technical Officer of the CTC, the UK's national cyclists' organisation.

---

"I've been corresponding with James about this recently and although his tandem fork is an oddity, I'm convinced he's exposed a real problem here. That's backed up by all the reports coming into this site of ejected disc-braked wheels and loosened fasteners - some obviously rotated."

---

For some more comments from the experts who have looked at this problem, click here.

Possible solutions

Just put the calliper on the front of the fork (RH leg) and be done with it. This will result in a generally upwards force applied to the hub through braking, into the fork ends with no tendency to eject the wheel. Erickson have already adopted this design with their tandems (I'd be grateful for a photo from an owner). As an alternative, any proper restraint on the axle that supports it (not just the QR, but the axle itself) against the braking force, like on 20mm through-axle fork models. Perhaps some retrofitted clamp over the fork ends would work, that could be manufactured and supplied freely to users.

There are two common ideas that I'm not so keen on:

Several ideas have been suggesteded for threadlocking the skewer. This still leaves it under a stress far greater than it is designed for, with a massive pull in the direction of the open fork ends. If it breaks under this force (and I would not be surprised to find that skewer breakage is more common with disk brakes than rim brakes) then the wheel will be ejected. There is precious little room for error here, even if these solutions do actually mean that a genuine operator error or some other breakage is required for failure.

Simply changing the dropout angle (to a much more forwards-opening position) will reduce the tendency to eject the wheel. However note that the unscrewing phenomenon only requires a very small amount of movement, and the alternating up/down forcing will still be present. There may not be a 'magic angle' at which this problem suddenly vanishes, so although this might reduce the risk, it may not completely eliminate the problem.

If you've had a front wheel pull out of your disk-brake equipped bicycle, or even had a skewer work loose, then you should definitely inform the manufacturers and also the CPSC, either through this web page or via email.