JamesHoward referenced a well documented procedure for increasing the preload on the pinion which ostensibly was used to insure that the bearings were also not being overloaded in the process. I found several very arbitrary issues with the procedure described by the author and wrote to ask for some reference to his chosen increments of turns (1/8 inch of a turn or 15 degrees of rotation per increment of measurement) of resistance of the pinion to rotation. Perhaps he'll answer or perhaps not. [Addendum Apr 4: he answered -- see his response at the bottom of this message. He confirmed my suspicians in his own way]. I must however point out that the procedure is applying in increase in loads on the bearings... and the magnitude of load increase is unknown and can't be calculated or approximated analytically because the proportion of total compression being applied to the crush sleeve and that being applied to the bearings is unknown.
To figure out how much load is being applied to the bearings (and hence to prevent overloading them), each instance of this procedure on each automobile's rear-end must know the bearing surface area's of the pinion shoulder on the crush sleeve, the crush sleeve's elastic properties (modulus of elasticity), the pitch of the pinion threads (on which the nut is being rotated), the bearing area of the rollers on their races, and the sundry other compressive loads being absorbed... to name just the obvious variables.
In fact, shouldn't the bearings actually have no load on them... i.e. roughly 0 load rather than being compressively loaded? I make this assertion because wheel bearings for example never have a compressive load on them in their steady state condition (i.e. they actually are supposed to have some "end-play" when being adjusted... which means no compressive load of bearings against their races. The bearing compressive loads only occur with the weight of the car on the spindles and in turns (lateral forces when turning corners addto the compression on the wheel bearings... though not that much).
I also make this assertion because I've worked in an industry where bearing's on spindles must not only have extremely low "wobble" (measured in nanometers ===> billionths of meters), must induce nearly no vibration (measured also in nanometers), and last for 7-10 years of continuous (always on) use.... as well as frequent on/off utilization (3-4 times / hour for life time)... which applies great loads during the torque required to get to some rotational speed... oh, did I mention the rotational speeds are measured in double digit thousands of rpm's.? A bearing "pre-load" is indeed applied during asm, but the ability to define how much pre-load requires development experiments with many measures of "wobble" and many other parameters(vibration, transient vibrations,etc) on hundreds of units in relation to the reliability of the bearings to funtion for 7-10 years without degradation in performance attributes.... certainly not wearing to the point of failure. In mfg'ing the issue of how to reliably measure the set pre-load is a non-trivial pursuit with frequent tooling calibration required.
Now, I admit that automobile bearings are not nearly as sensitive in function, but that they have a relationship of load to life of reliable function before bearing failure. If you've even had a front wheel bearing go out on you while driving you'll now what I'm referring to ---oh, and the trailers or F150's or other light duty load hauling vehicles you see stranded on the side of the freeway because the wheel bearing failed is not a rare event... they failed due to either insufficient lubricant, which is usuually due to overheating, which is most often due to bearing's being over-loaded..... either because they were improperly adjusted when assembled to the spindles, or the weight on the vehicle was beyond the load bearing capacity of the bearings.
I'm not saying the procedure described didn't or won't work on the vehicle to which the procedure was applied, at the condition of the bearings at the time it was applied, but that the procedure leaves everything to be desired about the function to which it's applied... which is to insure the pinion bearings have the required load --- not to much, not too little, and most probably actually no load ---- and the procedure has no means of insuring that too much load isn't being applied... no matter what the increment of Nut turns are between measurements of torque (by levers and weights or with a torque wrench) unless the prescribed mfg'er's torque limts (tolerance) is being used.
Author's response to my query to him about the procedure's inherent lack of bearing load risks and issues:
This procedure is intended as a last resort. [bold highlight is mine.]
Example. The noise is driving you crazy, and before a diff rebuild you are going to try something radical.
I did not buy an inch pound torque wrench because this procedure is such a risk. As you noted there is no defining point where the pre-load moves from the crush sleeve and directly to the bearing.
Since I don't believe that simply reproducing the original pre-load will guarantee the elimination of the whine, the idea was to limit the excessive movement of the pinion gear which should result in a reduction of the diff noise.
Longtooth
67 250SL US #113-043-10-002163
'02 SL500 Sport
