Reconstructing Constant Velocity (CV) Joints

[Jim Sims]

Pleased to read you're feeling better Grady.

The Schnorr washers have a specified hardness of Rockwell C39 to C45; this is basically the same as the 12.9 fastener strength class hardness of C38-C44.

Likely the Schnorr washer would work better with 10.9 strength class screws (C33-C39 hardness) as they would "bite" into the underside of the head through a combination of permanent plastic and temporary elastic deformation. The strength of 10.9's would only produce about 83% of the clamp load that 12.9's could produce but that still seems like enough torque capability in the joint (see below). Perhaps there was a concern about twisting off the 10.9's in the field or maybe Porsche just wanted a larger design factor (safety factor) to offset joint settling due to the joint gasket.

The Schnorr's still grip 12.9 screws but mostly through elastic deformation which rebounds and leaves little in the way of a permanant mark. As you surmised what marks were made were likely smeared over by the wiping action of two surfaces of equal hardness moving over each other. A result of using the Schnorr's against the hard 12.9 screws are that the washer's teeth are severly blunted and are not suitable for reuse. Hence the instruction (from Porsche and repeated many times on this board by you and others) to only use the washers once.

I recall doing a quick calculation and the 6-bolt circle 911 CV joint pattern using 8mm 12.9 screws properly tightened was good for over 1000 ft-lbs of torque before it "might slip" and start to load the CV bolts in shear and bending. There may be cases where the dynamic loading of a 911 axle could reach this level; perhaps severe wheel hop in first gear or "jumping" the car.

However, I believe almost all the CV joint failures experienced are simply due to the bolts becoming loose; not by backing out and falling out totally but just loosening enough to eliminate the CV joint clamping load. Once the clamping load is gone the bolts are loaded in shear and bending and they rapidly fail by fatigue under these conditions. Then they break in rapid succession sometimes leading to the loose, "flailing" half shaft disaster.

Produce proper bolt tension through the correct torque and keep the bolt tension by using the Schnorr lockwashers or properly applied lockwire in drilled heads (I've seen it installed backwards). Note that lockwire is not a substitute for proper torquing; a lockwired but semi-tight bolt will still fail early due to increased fatigue loading.

I believe you have interpreted the Schnorr equipped joint torque-tension (clamp load) diagram correctly. If one retightens after "settling" with a Schnorr washer in the grip then much of the torque will be consumed in friction between the highly loaded (still has 80% of the original clamp load) interface between the bolt head and the Schnorr washer; I'm not sure much additional tension will be added with the second tightening. This calls into question retightening the joints after a short period (~100 miles)of use - does it do any good especially if one is just further blunting the Schnorr teeth more by moving them under the 80% clamp load?

Enuf, dinner's ready.

[island911]

Originally posted by Wil Ferch
I really can't add anything meaningful to this excellent thread...but the goodness of Schnorr washers cannot be over-stated.
. . .

oh no . . the goodness of Schnorr washers HAS been thoroughly over-stated, here.

Seriously; do any of you even know WHY these bolts come loose?

Me thinks that understanding WHY there exists a problem (for some) is a better start to a proper solution.

[Jim Sims]

Why? Well Junker's theory is a start:

http://www.boltscience.com/pages/vibloose.htm


Then it can become much more complicated:

http://www.google.com/search?as_q=+loosening+theory&num=10&hl=en&ie=ISO-8859-1&btnG=Google+Search&as_epq=transverse+vibration&as_oq=&as_eq=&lr=&as_ft=i&as_filetype=&as_qdr=all&as_occt=any&as_dt=i&as_sitesearch=&as_rights=&safe=images

[island911]

Well there's a start. . ... but do you think that resonance is an issue here?

[randywebb]

I have a question - a back off and look at this another way question: many cars use CV joints nowdays. What sort of fasteners & locking washers are they using? Seems they would have the advantage of another 20 years of study after the ones we are talking about.

[island911]

Well that is a good way to approach the problem. However, new cars become more disposable. As such the design for servicing slides. . . and sliding is where the problem lays.

not following?

Notice that the CV has a thick gasket inset around the center to keep grease off the flange contact surface. Furthermore, on the later Porsches, an additional "cover" was added. Ya gotta ask yourself why would Porsche add this part ? . . . to add weight? . . just to be nice and help keep grit out when the half shafts are laying around a dirty shop? n'No; the covers keep the CV grease from migrating to the contact surface (flange/ CV). . . and to keep that cork gasket in check during re-assem.

Jim talked around this earlier .. .that is; flange/CV contact surfaces being allowed to slide back/forth, rel to each other, will likely start 'walking' the bolts out. (grease is great at helping things slide .. . last I checked)

OE installation they know to put them together clean. . . .they even develop covers to help facilitate that end.

The DIY'er here is told "Schnorr! . . .ya gotta have Schnorr's!!!!!!!!!!!!!!!!!!!!! " key-riest!

[Grady Clay]

Randy,

That GE “Roughness” comparison gauge is calibrated in microns (0.001 mm). Its range is from 4 microns to 2000 microns in 24 steps. It is “feeley” gauge in that you take a probe (your finger for instance) and feel the surface to be measured and then feel the various rough surfaces on the gauge to find a match. When you get to the smooth end, a scribe works well on metal surfaces.

The surface under the head of the screw was off the scale on the smooth end so I estimated.

This particular gauge is pre-WWII and may date to the 19th century but it is digital (you use your finger, HeHe.)

I have regular Mitutoyo micrometers that are calibrated to 0.001 mm.


OK, back to the issue at hand.

As an anecdotal observation, I have never seen a ’69-’73 CV joint come loose except when the screws were left loose. The spiral pin is an interference fit in both the CV joint and the flange. This would prevent any slip even if the clamping force was momentarily insufficient.
The other side of that argument is the engines weren’t as powerful, the tires didn’t have as much grip and the weight was less.


Thinking about other automobiles reminded me that a 911 is unusual in that the CV joints always run at an angle starting in ’69. When the wheel base was lengthened from 2211 mm to 2268 mm, the differential stayed in the same place but the rear hubs were moved rearward by 56 mm (~2.25”). I wonder what effect this could have?

I think the real culprit is the smaller diameter of the contact ring between the CV joint and the flange (100 mm vs. 108 mm) and the clamping force provided by six 8x1.25 mm screws torqued to 33 ft-lbs compared to six 10x1.50 mm screws torqued to 62 ft-lbs. The net result is less clamping force on a smaller radius.

The 100 mm 923 CV joint was just fine with a 2-liter 912E. I think it is marginal in some circumstances with a 3.0 and larger. I think that the improved tire grip and heaver weight are also contributing factors.

Another telling aspect is when Porsche returned to the 108 mm CV joints for all 911s in ’85, the reason they gave was to “…standardize” the joints, not the obvious increase in strength. It seems as if they didn’t want to bring up the subject because it might be asked “Weren’t the joints strong enough before?”


Jim,

I recall doing a quick calculation and the 6-bolt circle 911 CV joint pattern using 8mm 12.9 screws properly tightened was good for over 1000 ft-lbs of torque before it "might slip" and start to load the CV bolts in shear and bending. There may be cases where the dynamic loading of a 911 axle could reach this level; perhaps severe wheel hop in first gear or "jumping" the car.

A 3.2 Carrera is specified to produce 262 Nm of torque at 4000 RPM. First gear ratio is 11:35. The R&P ratio is 8:31.

(262 Nm) (35/11) (31/8) = 3230 Nm at the differential.

Say each axle takes half.

3230 Nm / 2 = 1615 Nm

Since 1.356 Nm = 1 ft-lbs

(1615 Nm) (1 ft-lbs/1.356 Nm) = 1191 ft-lbs.

On one hand I’m neglecting tire slip and transmission losses but on the other this isn’t considering dynamic loading.
Is this calculation correct? If so, “Houston, we have a problem.“


Jim, can you run the calculation for the 108 mm CV joints with six M10 (pitch 1.50 mm/thread) screws?


Correct me if I’m wrong but it seems most of the CV joint attachment failures have the bolts completely backed out. It is unusual to hear of the threaded end to shear off. It seems that usually four of the six screws can be used to get home. Randy Cecale (Pelican rcecale) is a good example.


Island911, when Porsche started using the end cap on the CV joints they did away with the gasket. Earlier in this thread we went through this.

Many of the failures have the end cap (rcecale) so grease on the contact surfaces can’t be the issue.

I agree that there has been too much focus on Schnorr washers and new screws. However those are something that is easy for an owner to control.

..do you think that resonance is an issue here?

With everything going thump and not ding, I don’t think so.


Best,
Grady

[Jim Sims]

Approximate 911 CV joint torque capability calculations (I'll keep them in SI units to avoid units conversions until the end); some of this is from memory as my references are at work.

100 mm CV joint with M8 x1.25 strength class 12.9; I recall the actual bolt circle was something like 85 mm in diameter or a 0.0425 m working radius for a bolt to resist the torque.

M8x1.25 12.9: .2% offset yield load is 40,000 newtons
ISO sets the "proof strength" as 88% of this. It is common tightening practice to only use 90% of the proof strength to provide some margin against twisting off the screw due to torsional and tensile stresses being combined.

The coefficient of static friction of lubricated steel on steel is about 0.16 and dry steel on steel is about 0.8.

Therefore per bolt and with greasy flanges: 40,000 N x .88 x .9 x .0425 m x .16 = 215 Nm; so six bolts would provide 1290 Nm or about 950 ft-lbs. If the flanges are clean and dry this increases to 6450 Nm or 4760 ft-lbs.

Repeating for the 108 mm CV joint (I assumed a 90mm bolt circle since I don't have one on hand to measure) and M10 x 1.5 12.9 bolts (yield load 64,000 newtons):

Therefore per bolt and with greasy flanges: 64,000 N x .88 x .9 x .045 m x .16 = 365 Nm; so six bolts would provide 2190 Nm or about 1615 ft-lbs. If the flanges are clean and dry this increases to 10,950 Nm or 8,080 ft-lbs. If I punched the correct calculator buttons.

I have made no reductions for cyclic loads or stress concentrations in the above calculations.

There was a big gain in torque capability in going to the 10 mm screws at a larger bolt circle (as in the 108 mm CV joints) AND in keeping grease off the flanges with the metal end covers.

Cheers, Jim

[Wil Ferch]

I understand the change over from flat seal ( with no end cap) to end cap ( w/o flat seal). I also understand the change to 108 mm CV diameter from 100...and the use of friction welded axles on the wheel-end CV's from the previous bolted.

However, I'm not quite clear as to when the Schnorr and/or moon washers were used in the context of these changes.

Example...my '85 Carrera is an "early" build date MY 85 where the old style arrangement was used. I confirmed I have 8mm recessed hex cap screws used at both ends. However, no Schnorr washers or moon washers are in-place. The original cosmoline dirt-covering can be seen on the assembly and I don't think these were ever apart previously...so I don't think washers were forgotten in a subsequent re-assembly.

Anyone else with this set up ??

PS---see also the Parts and Technical Reference catalogue, 911 models 1974-89. Dec 1995 edition, p.118. The interesting thing is that an "old" style 108 mm CV joint is shown ( Turbo / Turbo Look up to '84), that uses a flat seal. This requires the CV joint flange to have a ridge for the flat seal just like the smaller 100mm brother. "All" models ( Carrera / Turbo / Turbo Look) in late '85 supposedly use a 108 mm CV flange that has no ridge. This provides more contact area ( for clamping) than either the old 108mm or ( naturally) the 100 mm joint. No grease can contact the CV bolt circle with this ridge-less design..unlike the old 108/100 which uses the flat seal which subtends the bolt circle as part of the recess that is built-in. I think this is significant in our discussions..... no? ( that there were two styles of 108mm CV's, ridge and ridgeless, etc.)

- Wil

[Grady Clay]

Randy,

Good one.
“…the stuff the metrologists up the river at hp use for their chips & etc.”
What is the statistical diameter of an electron you can’t measure? Aah, fun with quantum physics.


That wasn’t a correction for you merely a clarification of what I was lamely trying to say.

One of my fun volunteer activities is I teach a 9th grade science research class to some of the brightest of the brightest. It keeps my brain functioning. Part of their normal science class is understanding measurement. We tend to stay away from digital and get the students to understand resolution, repeatability, precision and accuracy with analog instruments.

You can buy instruments that are far better than necessary and too many loose sight of what happens at the limits of metrology.

For instance, I built a set of linear measuring instruments calibrated in sillimeters. A sillimeter is roughly 1 mm ± 2 mm. Some of the marks are 0.5 mm wide. Some of the “rulers” are mis-marked where they go from 21 to 23 without a 22, HeHe.

We have a wonderfully terrible bunch of digital scales. The display changes by 0.4 g increments. They progressively read; 22.4, 22.8, 32.2, 23.6, etc. When you “zero” them they randomly change to progressively read; 22.6, 23.0, 23.4, etc. The students have to figure out how to calibrate such wonderful junk so they can report the density of various blocks of wood. Of course I made the blocks with no two side measurements the same, some are hollow and others are filled with lead shot in plaster (won’t rattle). From the outside they all look the same.

There is nothing more satisfying than watching smart 14-year olds figure this out.


The science research part is sorta a class within.
The students have to pick a project. Of course they all want to do thermonuclear research or find a cure for aids. The hardest part is to direct them to 3-5th grade level science project. Then they get to apply undergraduate level process to a seemingly easy experiment.

The sequence is:
Decide on a project.
Write a review paper with absolutely everything cited.
Write a detailed plan as if someone else was to do the work.
Do the experimental work the appropriate number of times.
Apply all the normal techniques; double blind, statistical, etc.
Combine the review paper and the experimental work into a paper and abstract.
Present their findings in various science fairs and as a stand-up presentation to the Colorado-Wyoming Junior Academy of Science in front of peers, grad students and professors. Heady stuff at age 14.
Have their abstract published in the IEEE Jr. Academy Proceedings. Who of us were published at 14?

Some of these kids graduate HS in three years. One former student is now VP of Google in charge of this big library project and more. I probably got 80 cards this Christmas from former students who are succeeding in life.


I think part of the fun of this Forum is the degree of intellect everyone brings. When I slip, please hold my feet to the fire.

Best,
Grady

[Grady Clay]

Wil,

You bring up a good point. The change from the flange that
uses a seal and consequently smaller clamping area to the
flange used with the end cap and more clamping area reduces
the pressure with the same clamping force. Is this an
improvement or reduction in preventing the joint from
slipping?

Here are the three flanges:

"
"
© Dr. -Ing. h.c. F. Porsche A.G.

#1 is CV joint flange 911 Turbo and 911 Carrera Turbo
Look (M491) 108 mm diameter, up to ’84 with gasket

#2 is CV joint flange 911 Carrera, 911 Turbo and 911
Carrera Turbo look (M491) 108 mm diameter, since ’85
with end cap (no gasket).

#3 is CV joint flange 911 Carrera, 100 mm diameter
up to ’84 with gasket.

Another important aspect of the change to the end cap
(#2) from the gasket (#1 & #3) is the clamping is now
completely around the axis of the bolt (red area in #2).
This probable prevents deformation of the flange because
one side was formerly unsupported. (I don’t think the
gasket can be considered support.)

Eliminating the gasket prevents the possibility that part
of the gasket gets between the steel contact surfaces.
I can see that situation happening and causing all sorts
of problems.

What else do you guys see going on here?

Best,
Grady

[Grady Clay]

Wil,

There is a third 108 mm CV joint that has the gasket outside the perimeter of the bolts. This was used from ’69 to ’75? This is the 108 mm CV joint with four M10 bolts and two spiral pins. The relief for the gasket was on the joint and not the flange.

Yesterday I tried (again) to write down all the joints, flange and axle combinations. There is still too much to research to try and set it down in print.

The issue of support plates (moon plates, washer plates) for the 100 mm axle I haven’t resolved. Clearly most didn’t have those parts. My (sometimes failing) memory is that a few did. We need to find that part number.

Best,
Grady

[randywebb]

"metrologists up the river at hp * * *"
- I used to date a metrologist up the [Willamette River in Corvallis]. She worked at hp. So, it's a better one than you knew!

I'm glad to hear you are teaching the HS class...
"far better than necessary"
- This is sometimes called the "error of misplaced precision." It [along with some other issues] destroyed the research carreer of the woman I post-doc'd for.

Back to CV's...

One thing that would be useful is to figure out what (if anything other than clean threads, new washers & repeat checking of torque) might make the existing CV joints on our cars more reliable....

Also, somebody ought to do a summary for the non-experts at the end of this series of posts.

[Grady Clay]

Jim,

WOW,

That is a dramatic change with having the contact surfaces clean and free of grease – next to impossible without the cover.


The bolt circle on the 108 mm CV joint measures 94.2 mm. I actually measured against the pins but the pins and screws appear to all be in the same diameter circle. The one I’m measuring is a ’69-’71 with four M10 screws and two spiral pins. This is what I use on the 914-6 SCCA GT2.

Best,
Grady

[Wil Ferch]

Grady:
In an earlier response from you on this thread...you mention that the axle flange ( for the version you had in your hand) had about six threads in its thickness ...and adding the combo of Schnorr washer and moonplate would add about 3.5 mm. With the x 1.5 pitch of an M10, this would result in 3 threads engaged instead of about 6 ( assuming it gets normally engaged to about "flush" on the backside). You go on to say that you prefer 1.5-turns past the flange backside. I presume you mean you prefer this amount of "past flush on the backside" from a purely mechanical fastening viewpoint...for any connection.

How much past "flush" ( on the backside) can we go before this would get us into problems as to contacting internal transmission pieces? This would be important to know if we add Schnorr/moonplates ( @ 3.5mm combined thickness) ...and try to avoid special machining operations on the "next longer" size of bolt to use...that you correctly say comes in increments of 5mm additional lengths, typically.

-Wil

PS (- yeah...when all is said and done.....we need a Cliffs note version of this...)

[Jim Sims]

Possible "moon plate" (washer plate) part numbers:

For CV joints (108 mm?) with 10 mm screws: 911.332.191.00

For 8 mm screws the only plate I could find was from the 914 and I believe Grady posted an image of it earlier in this thread; this 914 part number is: 901.332.191.00.

Perhaps Grady could check his example plate and see if it matches a "100 mm" CV joint. This might be the part to add to a 100 mm CV joint for a moon plate.

Wil, Isn't there a lot of room on the back side of the transmission CV flanges? It appears the bolt could protrude quite a bit before causing problems. This is not the case on the wheel side on the earlier cars where protruding bolt ends cause problems.

Jim

[Grady Clay]

Wil,

As I understand it (and Jim can correct me) flush out the back side isn’t sufficient. As I understand, never ever should the fastener not fully penetrate the flange threads. Penetration of 1-2 threads (with 1.5 ideal) is necessary for proper stress on the threads (male and female). I think every Porsche follows this. I recall that as a MIL-SPEC at some point. I don’t think too much penetration has an adverse effect on the threads but I don’t know. The problem with too much penetration (aside from hitting the case) is the exposed threads get dirty. When the fastener is unscrewed, the dirt is dragged into the threads in the flange and can damage the internal threads in the flange.

With too long screws, the ends absolutely should not contact the transmission casting under any circumstance.

I’ll call our best local metric hardware supplier tomorrow and ask about lengths. My recollection is the screws are commercially available in these sizes in 5 mm increments of length. Porsche lists the screws to exact 1 mm length increments.


I called a local friend who races a 914 and pays close attention to these issues. He uses the same CV joints that I do (’69-’71) but chose to run the unthreaded bolt shaft all the way through the flange and use a flat washer and steel lock nut on the back side of the flange. His engine is about 550 ft-lbs torque and we figure his CVs see about 1700 ft-lbs at torque peak in his 90 mph 1st gear. (This is a 6 liter alu SBC with a huge Hewland.) He is going to send me the specs for his (SAE) hardware for comparison. His plain ‘ol 108 mm 911 CVs, 914-6 only axle shaft, 911 stub axle and hub have worked perfectly for 30 years with annual maintenance. Oh BTW he turns lap records and is competitive with well prepared Cup cars. He just races on a nickel budget and is all home built – with a little help from his friends. It isn’t pretty but at 71 he goes fast.


Jim, this brings up another issue. Normally the fastener has its threaded portion extending into the CV joint, well past the joint interface and flange threads. If someone shortens a fastener where the transition from un-threaded shank to threads is at the joint junction, does that increase the likelihood of bolt failure at the first thread?

Best,
Grady

[Jim Sims]

For use with a nut or through threaded hole the 1.5 threads exposed is a good design rule and and I too recall it to be a mil spec. That being said, there are exceptions depending on the relative strength of the materials of the male and female threads and of course in a blind, female tapped hole it is not usually feasible to have the male threads extend beyond female threads.
In the CV joint application being discussed given the relative strengths of the flange and the screw I would believe it is important to have the thread exposure to better balance the load capabilities of the male and female threads.

Actually if it can be avoided, NOT having threads or the thread run-out in the joint shear plane is preferred but in compact mechanical hardware configurations this is often not possible. The main problem I see with shortening a bolt in these CV joints is not having enough thread and driving the thread to shank run-out into the female threads before clamping load is established.

Thinking about solutions to this problem of 911 CV joints I would propose two approaches:

For new flanges I would tap them using the female "Spiralock" thread form and continue to use standard 12.9 strength class screws. No separate locking elements such Schnorr washers would be required.

For retrofit with existing flanges I would modify the 12.9 strength class bolts using the Longlok "Dynathread II" process which can be applied to all but the very strongest bolts (won't work with the multi-phase alloys such as MP35N - can't make the forming tooling strong enough). Again no separate locking elements such as Schnorr washers would be required.

Links to information on these two "all metal" prevailing torque thread form solutions:

http://www.spiralock.com/

http://www.longlok.com/products_dynathred2.htm

Cheers, Jim

[AvonGil]

In all honesty, the Schnorr washers are a simple and cheap means of insurance but they don't solve the problem. Even if the bolt head was soft or even serrated, impact, vibration and fluctuating loads can still free the bolts.

An honest mechanical lock is the best way to solve this issue. Safety wire is one solution. Another could be tabs that you bend up to prevent a hex bolt from turning. You will find these two methods used everywhere in machinery where it matters, such as airplanes.

Before replacing the boots on my car I tried finding pre-drilled metric bolts, but I could not find any at a reasonable cost. So instead of doing it the correct way I just threw in the washers.

If anyone knows of a source for metric drilled head bolts, please let us know. I would replace and safety wire them in a hear beat.

[newbe911]

Some of you know a lot more than I do about this but I see the main problem being the lack of bearing area under the head of the bolt. The bole is high strength steel and the head diameter is designed to be inserted into like material. The boot flange is soft pressed steel so regardless of what you do the the soft material will cold flow out from under the bolt head. The only way to stop this is increase the bearing area (with plates, washers, etc.) or use some sort of mechanical restrant such as safety wire, tabs, etc.,. I am speaking of the 8mm bolts because thats whats on my 82. With the new type pressed steel boot flanges there is no room between the O.D. of the bolt head and the raised portion of the flange --- a very poor design. The boots I had to replace were thick aluminum with room to put "high collar washers" or something else to get the proper bearing area.