3 - All of the Aliens tested had an effective cam angle greater than the angle of the other cams. This means that before deformation of the cam lobe even begins, the cam is closer to the point of reaching the critical angle of 35 degrees. The higher the starting angle, the less metal needs to be deformed to reach the critical point. Which again, means less force required.

Even if the method of measuring the affect of the misdrilled axles on effective cam angle is imprecise, there's little doubt that in almost all of these samples, it resulted in a larger starting cam angle. Perhaps we don't have enough precision to predict the variance from cam to cam, without laying the lobes flat on a scanner bed, but on average, they were pretty consistently off in the direction of too high a cam angle.

Dammit Cracklover, you're making me want to go do more testing!

We have no evidence that cam angle in this sample is predictive of failure load. I have conducted a regression of failure load on cam angle, adjusted for strength rating, and the p-value for cam angle is 0.81. Cam angle is completely insignificant in the sample. Its effect has either been obliterated by measurement error, or swamped out by the harness issue, or a combination of both. I suspect that it is has mainly been masked by the hardness issue. Regressing failure load on hardness, again adjusting for strength rating, the effect of hardness on failure load is in the expected direction (harder is better). The effect is not statistically significant, but the p-value is 0.12, low enough to suggest that the effect may be real, but that we need a bigger sample size to detect it.

Jay

As we agreed before, I think the margin for error is too big compared to the range of values to see a statistically significant difference in pullout based on measured cam angle.

But you missed my point, (or I didn't make it clearly enough) which is that 1 - Based on the geometry of the Alien cam lobes, the angle starts off higher than that of other cams, and 1(a) even if the margin for error is too great to see a nice even progression, that doesn't change the fact that overall, the cams have an effective cam angle that's even higher than they would if the lobes were drilled properly.

So it's reasonable to assume that this plays a significant factor in explaining why the *overall population* of Aliens pull out before something else on the cams break.

Perhaps if Aric scans in the lobes, we can get a more precise measurement of axle hole to work with. Then we could see if that gets rid of enough slop in the numbers to improve the statistical correlation.

So here's a question for you all... I've got Healyje's 4 Hybrids plus the possibility of a couple regular ones from the 90's headed my way in the very near future. I just ordered 2 pair of granite inserts for my fixture, which should be ready next week, so do we test the additional samples the same as the others so that we have a larger data set or test them on the granite?

EDIT- BTW, I'll be resurfacing the fixture to the UIAA max roughness either way.

Wouldn't you expect that a cam that started with an angle of 34 degrees, all other things being equal, would take much less force to pull out than one starting at 14 degrees?

Ok Cracklover, have at it. The scans of all the lobes are available here: Link.

I can send a half dozen older Metolius and a #1 Master cam that was stuck in a crack for a year, assaulted by various parties, and finally funked out of the slot.

I would think so, but the relation between cam angle and failure load, for a given cam size, may not be linear. There may be a sort of threshold distance between the cam angle and the critical angle beyond which the effect of cam angle is negligible.

Like I said in a previous post, if the tests are done in a granite jig, where the cam angles will be much closer to the critical angle, then we might be more likely observe an effect of cam angle.

Regardless, I think that your hypothesis—that Aliens pull out of the test apparatus, whereas other brands of cam break, is mainly due to the actual cam angle of the Aliens being greater—is likely to be false.

However, the best measure of cam size in the dataset is rated strength, an imprecise surrogate for cam size. Hence, the effect (if any) of cam angle is confounded with the effect of cam size in my regression analysis.

I never said anything like that!

No, the best measure of cam size is cam size. There were four grey Aliens and five yellow tested. So at least you have two small datasets controlled for size.

Funny you should mention it, as I was just going through my pile of gear to see what's available to be sacrificed and it's easy enough to measure the spread on the fixture to calculate the radius.

Yup, as suggested in one of these threads I'll put the pieces in the fixture horizontally for the pic so that we get good shots of exactly where the lobes are making contact.

EDIT- That is, if I'm understanding what you're asking correctly...

Ok Cracklover, have at it. The scans of all the lobes are available here: Link

A couple notes on the scans:
1. The side scanned is the side with less apparent deformation, and in cases where the axle is bent the deformation on the other side could be significant.
2. The reason for this that I needed to clean the edge up with a dremel so that it would sit on the hardness tester properly.
3. The numbers in the pics correspond to the numbers in the hardness tests (the lobes are labeled on the other side).
4. Sample 22 has no numbers as I couldn't do a hardness test on it.
5. Buckled and cracked lobes are notated.
6. Some of the slightly bent axles may not be apparent as the bend wasn't enough to catch on the plate of the scanner and rotated all the way downward.