Betatronics^®

Comments on Automotive Hypoid Differential Axle Assembly.

This is copywritten material and may be copied only if the entire web page including this paragraph and the copyright notice at the end are copied. The material may not be broken into pieces. This applies whether in computer file format or printed on paper or some other media. Some of the material may differ from what others in the field may believe is a satisfactory procedure or description. Where photographs are referenced that are on a different web page these should be included in the copy. We do not want to be quoted out of context.

What is a web page? It is this entire section you selected by picking AXLE ASSEMBLY, and are currently viewing and requires scrolling. This requires a number of printed pages.

BACKGROUND.

We have been involved in electronic gaging of differential axle assemblies for shimming and preloading and other functions since 1971. Primarily we have designed and built the electronics, and created the system concept, and have been indirectly (sometimes directly) involved in the mechanical equipment. Our equipment has built many 10s of millions of axles.

Some of these other functions are: backlash measurement, flange or yoke runout measurement, gear ratio measurement to a resolution of about 0.000,3 in ratio, force check, transfer case testing, and drag torque measurement to predict case preload or verify pinion preload adjustment, and case preload by the null balance method. On a stable set of pinion bearings drag torque repeatability is on the order of 0.2 #-in over a moderate time (many minutes at 30 rpm). The gage alone is better than 0.1 #-in over many hours. But each machine cycle the gage is autozeroed so any drift is eliminated.

E-mail us with information on your needs at info@beta-a2.com. Please, no attachments as these will be ignored.

Over these years the electronics has resolved .000,1" or .002 mm. This is 1/10 or 1/20 of the shim increments used. Generally the electronics over the same time span from 1971 has had long term stability on the same order of magnitude as the resolution. The electronic resolution could have been higher but in this application was not required because of the shim increment size. However, we have made considerable improvement in the mechanical part of the system. In 1971 the variation in a repeat cycle measurement on the same part for pinion position was somewhat above .001". Today with all gaging machine mechanical components in good condition the peak to peak variation is typically within +/-.000,15". For a case shim in 1971 variation was generally .002" to .003" and sometimes worse. Today on optimum equipment with good parts we may achieve +/-.000,2".

SOME DEFINITIONS.

"open cup" --- this is a term I have used for many years to refer to a condition where one or more cups or cones or other components are not fully seated. The causes are many, but usually the end result is a loss of bearing preload after a period of time.

"stationary process" --- One where the statistics of a process (for example the mean value) are invariant from one sample group to another. In the past I had used the term "ergodic", but mathematically its definition has changed so that today ergodic should not be used to refer to a stationary process. Ref. Dr. Theodore G. Birdsall, EE Uof Mich. Birdsall also created the name N. P. Psytar.

"composite runout" --- This is the TIR (total indicator runout) measured in the plane perpendicular to the pinion axis of rotation passing thru the nominal center of the universal joint relative to the differential carrier. Note that TIR is twice the radial offset of the universal center from the axis of rotation. One wants the composite runout measured because this is what puts the center of mass of the drive shaft off of its axis of rotation. Also note that simply dynamically balancing the differential yoke or flange can not correct the system unbalance resulting from runout. System means differential, universal, and drive shaft. If you want to balance the system then the drive shaft and universal that will be assembled with the differential must used during balancing. If disassembled for shipping, then these components must be marked so that correct reassembly is possible. If instead you use a master universal and driveshaft for balancing, and in final assembly use an arbitrary shaft and universal, then the results in the vehicle may not be as good as would have resulted with balancing the actual parts. In axle photos see yoke, flanges, and runout distribution plot.

"a priori" --- from what is before. Ref. Funk & Wagnalls.

"a posteriori" --- from what comes after. Ref. Funk & Wagnalls.

"torque" --- the rotary force in a mechanism. Ref Funk & Wagnalls. The English units of measure are pound-feet, pound-foot, etc. Ref. Physics books, Handbook of Physics and Chemistry", and others. Most torque wrench manufacturers incorrectly label their wrenches as foot-pounds, etc, but are correct in the metric system with Newton-Meters. The unit of measure foot-pounds is related to work or energy and not to torque. In other words ---- for "torque" the unit of force is first and for "work" the unit of distance is first in the compound unit. We have some newspaper writers that use pounds-feet which is a double plural. But that is not as bad as using foot-pounds.

"drag torque" --- the torque to rotate a shaft in bearings resulting from the frictional drag of the bearings and/or seals. I want to further define this as the average torque. Sometimes it is difficult to determine average torque when reading a hand torque wrench. Note peak or minimum torque can vary substantially for a given average torque. In most cases average torque is the easiest estimate of power dissipation and therefore temperature rise. Drag torque is used to estimate axial force preload via coefficient of friction. Generally drag torque is speed sensitive.

REPEAT CYCLE TEST.

A repeat cycle test is very useful to evaluate a machine and troubleshoot problems. When used with real parts it may show interface problems between part and machine.  Data from a machine in good condition can be used to compare with data later to identify problems.  For example if on a pinion shim machine we normally expect on a repeat cycle test to see a worst case variation of +/-.000,15 but actually see +/- .000,3, then we might suspect a bad thrust bearing in the machine. An absolutely solid master or simulation of the part usually produces better repeatability.

The repeat cycle test is run with the same components for each cycle in the test. In one type of test the components are not removed from the machine, but the machine is simply recycled. Some case shim machines can be recycled without change of the gear mesh. Another test uses a single operator and the parts are installed and removed from the machine each cycle. And a third uses several (typically 3) operators to sequentially run the cycles.

If a machine is on an automatic line with pallets, then the test may be run with the same pallet, and also with different pallets. The different pallet test would be done by doing regular production and identifying potentially bad pallets, then later use these in combination with good pallets to run a controlled test with the same part or different parts. There is also the need to pull parts that were on the pallets that are suspect and keep these with the associated pallet because a problem may result from a unique combination of pallet and part.

CASE SHIM.

In general, the applications we have been involved in for case shim have measured position based on the no-backlash position of the gear set. Then it is assumed that the offset from no-backlash to correct backlash is a constant for that gear set part number class. This is generally approximately correct. Using the same gear set and changing the gear mesh point may show a small change in the reading. The mechanical design of some machines does not allow a fixed mesh on a repeat cycle test. The assumption of no-backlash to correct backlash being a constant does introduce a bigger problem between parts, especially between gear set batches, or when bad gears are hidden within a batch.

The obvious solution is to gage at the correct backlash position, but this takes much more cycle time. On balance the no-backlash technique does a very good job in production. Typically repairs for backlash during production are less than 1%. Keep in mind that over a one shift period hundreds to a thousand or more units may be built and different baskets and batches feed into this production. So 1% is very good.

The coefficient of thermal expansion of steel and cast iron is about the same, whereas aluminum expands much faster than steel. Thus, in the design of aluminum carriers the bearing cup bores are generally smaller than a comparable cast iron carrier. This means that the tightening of the cap bolts will have more effect on case preload on an aluminum carrier than a cast iron unit. Thus, most of the development of case preload in an aluminum carrier comes from tightening the cap bolts and the resulting squeezing of the cups.  Note, if the two side bores are not the same diameter, then there is also a shift of the backlash position. In an aluminum carrier this also means that pinion and case preload vary more from a cold start (preload high) to high load (200 degrees or more in carrier) than in a cast iron carrier.

In general, if you build a cast iron carrier assembly and do not stretch the carrier to install the case, then I believe you have a high probability of building low case preload. It is very hard to put a case into a cast iron carrier and obtain correct preload and not use some means to stretch the carrier to ease insertion. Instead of pulling on the manufacturing holes to stretch the carrier one can squeeze the cover plate face with a big special c-clamp. In some aluminum carriers the case will almost fall in with no stretch.

The Dana specifications at the Dana site www.cyber-fish.com/fordeec/light_axle_service.pdf for DIFFERENTIAL BEARING PRELOAD are a good reference, lacking anything else, to check case preload. This might be used on other manufacturer's axles as well, but you need verification from them as to the correct values.

If you build a cast iron carrier unit and do not stretch the carrier for installation, then the Dana values may be a guide as to whether your preload is low.  However, you should get correct values for the case component of drag torque at the pinion from the manufacturer of the axle.  Always keep in mind that you must have an accurate measure of pinion only drag torque (meaning including seal).  When you measure total torque with case, the component from the case is only about 1/4 of the total, so a small error in the pinion only drag torque produces about a 4 times greater error in the case component.

Suppose we use Dana's figures for a 4.10 axle of 6 to 8 #-in increase from the case component (DIFFERENTIAL BEARING PRELOAD chart).  If the error in reading the pinion drag torque is 1 #-in, then you have a 50% error in the allowed range for case preload.

On a second reading of this Dana site I have a question about the PINION BEARING PRELOAD, DIFFERENTIAL BEARING PRELOAD, and EXAMPLE charts.  Why is the new seal drag torque brought into the equation in the example?  It is my opinion that it should be brought up in the PINION BEARING PRELOAD chart. My assumption is that "Torque to rotate pinion" includes seal drag.  Because once pinion and components are assembled one can not separate the seal drag from pinion drag torque.  I also assume that the pinion preload torque specifications include seal drag. 

In production on a Trio Pinion Preload Adjust machine the build and test points are based on the total of the seal and pinion bearing drag torques, same for Cooper machines.  Seal drag is assumed to be a constant based on an average value and this is effectively built into the pinion drag torque specification.  Whether the Dana specifications on pinion drag torque in the PINION BEARING PRELOAD chart include seal drag or not I do not know. 

In any event to measure case preload thru the increase in pinion drag torque when the case is installed one does not care what the seal drag is.  What you want is a stationary statistic of the average drag torque of the pinion and seal.  This pinion plus seal drag torque is what you want to subtract from Total Torque after the case is installed and seated to estimate case preload. This analysis is based on the above assumptions. In the field for service or in production the only measurement you can make is the pinion plus seal drag torque, and truly this is what you want for the case preload measurement.

As noted elsewhere in this discussion a 1463# axial load on a set of pinion bearings produced 25 #-in of drag torque at 30 rpm. Normally case bearings will be in the same ballpark. So if your axle ratio was 4 to 1, then 1463 # axial case preload would produce a component of about 25/4 #-in ( 7.5 #-in ) at the pinion.

PINION PRELOAD ADJUSTMENT:

There are four mechanical pinion assembly structures that may have different means used to adjust pinion preload. These are: (1) No Spacer, (2) Collapsible Spacer, (3) Hard Solid Spacer, and (4) Soft Solid Spacer.

I know of no application of type 4, but we almost ran a machine in this neat mode until product engineering changed the specification on the solid spacer to require that all the different shim thicknesses be hard (the solid spacer is actually a stack of thin shims --- like 2, 3, 5, 10, 20 / 1000") . Whereas, previously the thicker shims in the stack were soft. This method would have had the advantage of shortening the machine cycle time yet allow very good control of drag torque in a solid spacer machine. Otherwise in a solid spacer axle you have no way to control pinion preload except by changing the shims.  The reason the collapsible spacer method is used on so many axles is the efficiency of building pinion preload.

The "no spacer" method is used in some very large axles used in the construction industry. Some of these axles may weigh several tons. The pinion alone may be several hundred pounds. There are many minutes allowed per operation, maybe 7 minutes, and thus slow adjust to correct preload (about same drag torque as an automotive unit) is no problem and thus no spacer is needed. After adjustment the pinion nut is locked in place with a keeper of some type. It should also be noted that pinion only drag torque is different depending upon whether the pinion nut is up or down due to the pinion weight and the difference in the inner and outer bearings and that the drag torque is so low in relation to the weight of the pinion.

The "collapsible spacer" method is the dominate method in terms of the total number of axles produced. In this method a spacer that crushes (collapses) at an approximately constant force for a substantial distance is effectively placed between inner and outer pinion bearing cones. Collapsing force is generally in the range of 25,000 to 30,000 #.

The collapsible spacer adjustment cycle consists of putting the various components together without the nut, then pressing everything together to a fixed point that is guaranteed to not collapse the spacer, or the press force is set low enough to not collapse the spacer and all components are pressed together.

The concept behind the collapsible spacer method is that the initial length of the spacer is great enough that no combination of part tolerances will allow bearing contact. This typically leaves an endplay of about .06" (about 1 nut revolution to bearing contact). Some threads are finer than 1/16" but this is a ball park figure. Bearings must be rotated always during adjustment. When everything is together, the nut is rotated moderately fast until bearing contact (maybe 6 rpm) (in other words 10 seconds to bearing contact), then nut rotation is slowed until correct drag torque is achieved. It only takes maybe .003" axial motion to go from bearing contact to correct drag torque. This is about 20 degrees of nut rotation. After this the part is tested. In a collapsible spacer application you CAN and DO control "pinion preload drag torque", you can and should measure "pinion nut torque", but you CAN NOT control nut torque in this application. The nut torque will be whatever the components produce. Nut torque is a function of coefficient of friction and the force loads. Nut torque in this application is much more variable than previously because of government regulations that have banned cadmium plated nuts. For a drag and nut torque plot vs time for a Trio preload adjustment station goto http://www.beta-a2.com/pa_plot.html .

The "solid spacer" method uses a shim stack to make a solid spacer to fit on the pinion where the collapsible spacer would go. The shoulder on the pinion is higher than for a collapsible so that a moderately thin shim can be used. The parts are gaged to determine the shim thickness. The part is built and drag torque measured. If outside the allowed range the pinion assembly is partially disassembled and the shim is changed. This is done until drag torque falls in the allowed range. A typical range might be 15 to 30 #-in for a small axle. A solid spacer machine CAN NOT control "drag torque", but DOES and MUST control "nut torque". The "drag torque" is only controlled by the operator via the shim selection and rebuild if necessary.

The solid spacer method typically takes most of this allowed "drag torque" range. Whereas we might achieve a +/- 2 #-in range with a collapsible spacer machine and not require the rebuilds.

A Trio machine with Betatronics gaging is in production that runs both solid and collapsible spacer parts. All other Trio machines with Betatronics are strictly collapsible spacer. These machines may have a part to part cycle time of 30 to 60 seconds, but more typically 45 seconds. There are many different factors that determine this time. Some relate to the particular axle plant.  In the past we ran faster cycle times, but now more is done in the cycle and better consistency is achieved. In 1977 I clocked a single station line at 23 seconds part to part.

Most if not all Cooper machines do either solid or collapsible in the same machine.

Final drag torque measurement should be done at a pinion angular velocity of 30 rpm. This number goes back a long time, possibly the 50's or earlier. There is a rotation speed where drag torque is a minimum. This minimum value for pinion bearings may be in the range of 10 to 35 rpm, but could be grossly different. It may not be the same for inner and outer. I would like to test at 20 rpm, but 30 rpm is better from a cycle time perspective and has a lot of history, but this criteria might change as bearing design changes.

The variation of pinion drag torque verses speed is dependent upon viscosity, lubricity, and bearing design. The reason I bring in lubricity is that some "oils" produce very unstable results. For the most part, pure sliding friction should not be speed sensitive unless the coefficient of friction was not constant. Because lubricants have viscosity (implies power loss --- heating) this ultimately means that as speed is increased that torque increases. The shipping oil on tapered roller bearings is a rust preventative and not a lubricant. I have been told that WD40 is a better lubricant and it is not very good, thus this classifies shipping oil as a very poor lubricant. Note that a preloaded roller bearing has a very high pressure at the line of roller contact, and on the roller end.

How are pinion drag torque specifications defined. If you hand build a solid spacer axle it makes sense that the specs are without seal. But if it is a collapsible spacer unit then the specs should be with seal. In a production environment, whether solid or collapsible, the specs should be with seal.

Circa 1987 I measured an assembly that had the following values: 39 #-in @ 3rpm, 22 #-in @ 10 rpm, a minimum of 12 #-in @ 40 rpm, and 15#-in @ 85 rpm.

A totally different set of bearings (different design and different product), circa 1998, was 28 @ 3, 25 @ 10, 27 @ 20, 26 @ 30 TO 85 RPM. A different manufacturer of these 1998 bearings produced 26 @ 3, 18 @ 10, 16 @ 20, minimum 14 @ 30 to 40, and 15 @ 85 rpm.

Generally I am not in favor of a hand torque test for various reasons. But hand torque wrenches are some times the only available device.  One factor is speed of rotation.  Good averaging can not be achieved in a reasonable time.  Also there is more unknown variability at 3 rpm than at 30 rpm.

On a pinion bearing set that I measured at 30 rpm and variable axial force I had 25 #-in of drag torque at 1463 # axial force with no seal. Bearing manufacturers indicate that you can expect a +/- 20% ( total range 40% ) variation in drag torque from their specified values. If I do everything correctly, good instrumentation and constant conditions, then I have not seen this much variation for one vendor for the same part number. However, I have not had product from many different batches. Vendor to vendor may be a different story, especially when one vendor may have a greater variation with speed than another vendor and thus speed becomes a factor in comparing the vendors.

I believe most engineering development is done with hand torque wrench measurement, but all production testing with Betatronics equipment is at approximately 30 rpm whether in preload adjust machines or in separate test gages. I believe Cooper is also approximately 30 rpm.

When various production people use a hand torque wrench there are several problems that show up.

(1) Typically 100 #-in wrenches are used because 50 #-in units do not last long. The problem in any case is that the operators do not constantly check noload reading for zero. They may not even understand how close they need to set zero.

(2) The operator will use the peak indicator and read this value. Wrong --- we need the average value. And they may not even know what average means.

(3) Need to seat bearings first before reading torque. Sometimes seating rotation is too short.

(4) Some operators can not estimate average as they rotate. Note, hand readings are about 3 rpm if you try to read while rotating. The peak memory needle on the torque wrench is of no value because this does not give you the average reading.

(5) Some will rotate, then essentially stop and hold a force on the wrench and read it. This reading could be almost anything below that required to rotate.

Note: if you rotate a fixed displacement preloaded tapered roller bearing set in one direction until the rollers are seated, then reverse direction you will notice a small drop in drag torque until the rollers reorientate in the opposite direction. This is because the rollers run slightly off perpendicular and this causes a slightly higher drag torque. Also forward and reverse torques may be slightly different.

There is more preload adjust information on our pinion preload plot page. If you pick
PIN PRELOAD PLOT page then to return here use your browser BACK button.

CASE PRELOAD MEASUREMENT:

How do you measure actual case preload? Not easily.

The best method is essentially a null balance substitution method, but this typically takes 20 minutes to perform per axle. It also requires removal of the case from the axle.  If you produce 200 axles per hour on a line, then only every 66th axle can be tested.

Another method uses the spring rate of the carrier, or average spring rate. Spring rate equals dF/dX where F is force and X is displacement. In this method the change in displacement between the manufacturing holes is measured for a known desired applied force on the centerline of the case. This provides the spring rate information. If individual axles are not calibrated, then an average spring is determined and assumed valid for all comparable axles. Then when the carrier assembly is built the procedure below is followed. Stress relieved carriers must be used. In an unloaded state, no force on bearing seat surfaces, the change in displacement between manufacturing holes may be 0 to 3/1000" from before stress relief to after. Note: green castings typically have a substantial amount of internal stress. These can be stress relieved in several ways --- put them to pasture for 6 months or so (this takes too long), pound, vibrate, heat soak, and stretch or squeeze. When you do this preload test or build the axles you want a stable carrier. Stretching seems to be a good method.

A special gage is used to measure the distance between manufacturing holes before case insertion and zeroed. The case is installed, caps bolted, and bearings runin. The same gage without change of its zero position is used again to measure the distance between manufacturing holes. The now non zero reading is the change in displacement from before to after case preload. This value in relation to the calibrated value above gives you the case preload. Note: typical mean values may range from .002" to .008" depending upon carrier stiffness and therefore instrumentation error is a problem. However, this is an effective means and can be done as a 100% inspection. This can not be used to determine case preload after tubes are inserted. Both spring rate and shape of carrier change from tube pressing and thus the gage zero reference is lost. It is quite possible for the carrier spring rate to change by almost a factor of 2.

Measuring case preload via "total drag torque" is done by first measuring "pinion drag torque" alone just before insertion of the case assembly into the carrier. Note this of necessity includes the pinion seal drag and that is of no importance. Pinion preload drag torque must be a stationary process. See definition of stationary above.

This is an a priori method. After pattern test the "total drag torque" is measured. The "case drag torque" as viewed at the pinion is "total drag torque" minus "pinion drag torque". If you want this reflected back to the torque that would be seen if the pinion was removed and the case was driven directly, then multiply by the gear ratio and divide by efficiency. Efficiency is a decimal fraction. For example use .95 as divisor for 95% efficiency. And .95 is a good typical value to use.

An a posteriori method could be used but would require removal of the case to measure the pinion only drag torque.

Note: Dana uses the term "PINION BEARING PRELOAD" to refer to the drag torque of the pinion only ( meaning no case ) but including the seal (I believe it includes seal drag). Since they indicate that seal drag is about 3 #-in that means that the actual pinion preload drag torque is 3 #-in less than what the torque wrench measures. I have in the past seen some seals that had about 8 #-in drag, not Dana. I doubt that high seal drag is found today.

Dana uses the term "DIFFERENTIAL BEARING PRELOAD" to refer to the component of total pinion drag torque that results from the case bearing preload. This is not the case drag torque you would see if you removed the pinion and drove the case directly with the torque gage.

See DIFFERENTIAL BEARING PRELOAD table in Dana Service Specifications at www.cyber-fish.com/fordeec/light_axle_service.pdf   This table does not indicate whether or not this specification applies to the Hydra-Lok axle.  Hydra-Lok probably adds about 3 to 5 #-in as seen at the pinion as compared to a standard differential. This results from the Hydra-Lok seal and mechanism. Check with Dana for specifications.

One can see 0 to 5 #-in drop in total drag torque from no tubes to tubes pressed in. Typically 1 to 2 #-in. If the nominal expected increase in pinion drag torque is 7.5 #-in, then this loss from tube pressing is substantial. But nothing like a change of 2500 # down to 800 #. The carrier stiffness on case centerline almost doubles on tube installation. This might lead you to think that case preload would increase, but on axles of our experience it does not. Obviously other changes occur in the carrier when pressing in the tubes. There are old stories that some axles built to about 2500 # axial case bearing preload might drop to 800 # after the tubes were pressed in. I did not run these tests so I have no first hand knowledge. These were run with the null substitution test method about 25 years ago..

Note that available tube pressing force may be 75,000 to 100,000 #, or more. Typical force to press may range from 25,000 # to 50,000 # and is dependent upon the interference fit, surface finish, and lubrication. To remove tubes is not easy. Obviously with these forces you can see why a carrier changes shape from before to after tubes are pressed.

The drag torque method, if done correctly, can produce effective results. There are lots of tricks to getting good results.

Note 1:  If a search engine brings you to this site and you do not know why, then save this page to a file on your computer (under FILE you would use SaveAs) and after saving, then open the file with a word processor and search for the words individually that you used in your search.  For example if the words --- drag torque --- bring up this site and this page of this site, then you would first search for drag or torque and look around those locations.  Pick If_search to return to beginning.