Subaru 5mt: Case Distortion - Fact or Myth?
Background:
Subaru's current symmetrical AWD system, as used in models with the 5 speed manual transmission, has roots in a very old design. The gearbox itself (housing 1st through 4th gears) traces its basic layout back to the 4 speed front wheel drive transmissions found in Subarus from 1970-on. The design has evolved incrementally since then, with an optional 5th gear later added in a separate section at the rear of the transmission. The full-time AWD system that newer Subaru owners may be familiar with, using an additional center diff section behind 5th gear, also eventually found its way into the line-up replacing earlier part-time 4wd models. This system became the mainstay of Subaru marketing throughout the 90s and beyond. Many other small changes, including numerous revisions to the synchros have also occurred over the years. The basic design of the case, however, has changed very little. The only really notable change in case design to the AWD 5mt transmission can be found after the 1998 model year, where Subaru added a substantial amount of aluminum along the top and bottom of the case. The bellhousing was also reinforced and 4 mounting holes were added on both the transmission bellhousing and engine block (bringing the total number of bellhousing bolts to 8, up from the previous 4 bolts).
Does it flex and is it a problem?
Within the Subaru community, the subject of whether or not the 5mt's longitudinally split cast-aluminum case contributes to a seemingly high failure rate has been debated many times. While failures are most likely due to more than one factor, it seems that the idea of case distortion as a contributing factor has largely been cast away as a myth. However, there are some design points that seem relatively weak in comparison to many other production transmissions as far as the case is concerned. Admittedly, the 5 speed case, as mentioned above, was upgraded in 1998 with what appears to be a focus on increasing torsional rigidity (the ability of the case to resist twisting forces). Though signs of compromise abound in this particular revision. This was a matter of changing the casting patterns of the case halves while still keeping the longitudinally split design in place with minimal change to the layout of the internals. These changes were probably made in such a way that production tooling could continue on with minimal modification. Shutting down a production facility to accommodate a major engineering change can be very costly, and most manufacturing companies avoid doing so if at all possible. While the newer 8-bolt case design is without a doubt stronger than the aged 4 bolt design, it still has numerous weak points that become quite apparent under load. Because various bearings, and very importantly, the front ring & pinion are all held together by the case, a weak case can more easily accommodate separation and misalignment of transmission internals under heavy load.
As a note: There are aftermarket gearsets that allow the 5mt to handle quite a bit of abuse even with the weak case... Though it should be understood that these gearsets often utilize much wider gear tooth profiles which generate too much noise to be considered by a major auto manufacturer for use in a mass-produced vehicle. The advantage is that these larger gear tooth profiles have a higher tolerance for gear separation.
Down to the hard data :
I decided to test the case and either cast the case distortion theory forever into the realm of false claims or provide hard data that supports case flex as a contributing factor to the 5mt's failure rate. Admittedly, the test fixture was built on a shoestring budget, but by mounting measurement tools to the transmission itself, relative measurements could be taken without any worry of weaknesses in the test fixture itself showing themselves in my measurements.
How the test was performed - The transmission was mounted to a frame by its axle stubs up front along with the output shaft at the rear of the transmission. The stubs and tail shaft where clamped into the fixture and then tack welded to the clamps in order to ensure a complete lack of shaft rotation. This essentially provides a scenario where the transmission is in a vulnerable position, without any support from an engine block and under the simulated load of an infinitely heavy car (no movement allowed at any of the 3 output shafts). A very large torque wrench was affixed to the input shaft, allowing a measured amount of torque to be applied to the input shaft. Results were indeed interesting. The test fixture is pictured below with the torque wrench mounted and the dial gauge in one of its many positions used throughout testing.
Before revealing actual test data , I want to list limitations of my test procedure:
-The transmission is not bolted to a strong engine block which may affect the overall strength of the case relative to real world driving conditions with the transmission installed.
-Torque is being applied with a torque wrench. While this is generating a mostly axial load, there may be some lateral loads introduced which would not be normally applied by the clutch assembly.
-This simulates an infinitely heavy car. There is NO torque dissipation allowed through wheel spin or energy used to move the "car" forward.
-The dial gauge is mounted to the bellhousing as a reference point. Since flex in the bellhousing itself is unknown, there is a possibility that any flex in the bellhousing could affect test data.
With a dial gauge rooted to a bellhousing bolt hole, the above measurements are relative to this position on the passenger side of the transmission bellhousing. These figures represent measured dimensional change with positive values showing movement of the case toward the dial gauge and negative values moving away. Because my dial gauge only has a precision of .001 inch, I used a > or < symbol to show where the gauge needle was in between marks on the dial gauge. These measurements were taken with the transmission in its center most gear (3rd) at a close to stock 230ft./lbs. of torque being applied to the input shaft. Torsional stress data was focused on major bolt locations, where case expansion would have a minimal effect on these figures. At a heightened 300ft./lbs. continuous, many of the measurements had increased by as much as .001 inch.
Since these measurements were taken in relation to a point on the bellhousing, I can't say for sure that the measurements are accurate against a fixed point away from the transmission. What can be noted is that any flex in the bellhousing should be the same as long as input torque stays the same. So we can trace differences in measurements (increases or decreases from one spot to another) allowing paths of torsional flex to be deduced from this relative data. In the Subaru 5mt, the case is torsionally stable at the front, however, internal stresses cause the case to twist substantially aft of the front input shaft bearing and front pinion bearing positions. The top rear half of the transmission sort of starts to lean toward the driver's side of the car, while the bottom rear half "kicks out" toward the passenger side of the car.
As far as expansion and contraction of the case are concerned, a couple of interesting characteristics could be noted. At 300ft./lbs. of input torque, the force of the gears pushing away from each other actually causes the top and bottom of the case to expand slightly (about .001" respectively) peaking toward the rear of the case, while the sides of the case actually contract... The case seems to narrow down from side to side by around .001"-.002" in some spots (measured with the dial gauge mounted along the top "spine" of the case) when under load. The peak area of expansion in the case structure occurs right above of the driver side axle stub behind the front differential assembly, which was measured at well above .003" even at stock input torque. This is part of the case that surprisingly lacks support. I say "surprisingly" because it is located under or behind the ring gear, where a lot of force is applied by the front differential's pinion gear. Under conditions of severe front wheel hop, a thin section of the case can actually crack between a forward case bolt location and the thicker structure of the casting that extends rearward toward the pinion gear (pictured below)! This is a rare failure, but where there's smoke, there's fire. Aluminum has a relatively low fatigue life, and low-cycle fractures like the one pictured below can be a clear indicator that the material has been flexing fairly violently.
Conclusion:
Why is a few thousandths of an inch such a big deal? Subaru's spec. for recommended gear lash on the forward gears and front ring & pinion only carries a tiny +/- .001" tolerance. So for some dimensions of the case to change by up to 3 times that tolerance at stock torque could be a problem. This is an indicator that the gears inside are potentially moving beyond their engineered clearance range even at an unmodified power level. Consider that a harsh launch or even a hard shift can sink a huge amount of torque into the driveline for a split second, a condition often referred to as shock load, and case distortion figures could increase well beyond the figures I measured at 300ft./lbs. of input torque during testing. The problem is that when the case flexes, it is allowing various bearings inside to change position ever so slightly which can, in turn, accommodate added flex of the shafts/gears which ride on those bearings. When it comes down to it, the gears are fairly small for their application in a reasonably powerful AWD car with a pretty narrow gear center line spacing, and flexibility in the long forward half of the mainshaft could also be connected to changes in gear clearance under load. Flexibility in the case is only one of many weaknesses in the aging 5mt design, but I think this test proves that case distortion is indeed potentially a contributing factor to 5mt failures. It's unlikely that the flexible case is the main factor in most 5mt failures, but I think it's definitely at the top of the list right behind limited gear centerline spacing and flex in the main and driven/pinion shafts.
