No announcement yet.

Viscous Coupler Testing (Rebuilt unit vs. 'Good')

  • Filter
  • Time
  • Show
Clear All
new posts

    Viscous Coupler Testing (Rebuilt unit vs. 'Good')

    First off, shout out to TehRaydarlover for paving the way with a how-to on viscous couplers

    Secondly, if anyone has a previously brand new unused viscous coupler on the shelf and is willing to rent it for me to test, please DM me! I would like to have data from the epitome of a factory OE coupler. This will help define what a good coupler should look like.

    I ran across Ray's post when I first bought my car and thought one day I'd like to try this. I figured maybe I could rebuild some for other people along the way and help save some cars from scrap. I rallycross my iX regularly and feared that I might one day need to replace it gave me an excuse to play with my machinery

    I decided I wanted the ability to prove that the rebuilt couplers had the performance desired. I ended up building myself a test stand for the lathe that would allow me to measure the output torque at various speeds using my lathe. I calculated the HP needed based on a Subaru center diff (4 kgf-m @ 100 RPM) and figured I should get the lathe up to speed and quickly stop the coupler. The setup below is what I came up with. It uses an Arduino Uno to monitor the speed of the lathe, the viscous coupler and measure the torque output of the viscous coupler. #Nerd

    Keep in mind my reference to 'good' coupler is one I got that did not have a proven jack test or anything else (I hope to get a second test sample to weigh this against). The bad coupler was very clearly bad. The silicone fluid was on the outside and the case looked like it had been heated cherry red then cooled down.

    Click image for larger version

Name:	PXL_20210209_002037798.jpg
Views:	400
Size:	73.1 KB
ID:	9973357

    Next I ran a 'good' coupler through a bunch of speeds to see what it did (Each color is a different speed). I ran the lathe up to each speed and applied the brake for about 3 seconds. This showed the peak torque as well as the dynamic torque. Unfortunately, the peak torques are hard to capture consistently on a viscous coupler. In order to really capture and characterize each speed, I would need to let the coupler completely cool and settle before rerunning. I figured this is a good starting point and I will improve my data collection techniques as time progresses.
    Click image for larger version

Name:	Multi-Speed.png
Views:	325
Size:	100.9 KB
ID:	9973358

    Here is an example of how to read one of the curves above (140 RPM in this case). The peak torque occurs about 0.373 seconds after a speed differential occurs from front/rear axle. If that 140 RPM speed differential continues the torque will plateau at around 93 ft-lbs at 2.6 seconds.
    Click image for larger version

Name:	One_SpeedExample.png
Views:	335
Size:	63.1 KB
ID:	9973360

    Next I took the peak torques of the 'good' coupler and a very bad coupler. I'm not sure how Subaru measures their center diff viscous couplers; It might be at peak torque or dynamic torque. If they use peak torque the 'good' coupler I have is around 16.9 kgf-m @ 100 RPM. If they use dynamic torque it is closer to 10 kgf-m @ 100 RPM. Either way that is quite a bit stiffer than a stock Subaru which is 4 kgf-m @ 100 RPM. Click image for larger version

Name:	Good vs Bad.png
Views:	325
Size:	42.9 KB
ID:	9973361

    Next I sliced open the bad coupler to see what was on the inside
    Click image for larger version

Name:	IMG_1132.jpg
Views:	335
Size:	68.7 KB
ID:	9973363

    After a bunch of months of the coupler sitting around I finally machined the DOM ring and had a friend TIG weld it on.
    Click image for larger version

Name:	PXL_20210202_025044557.jpg
Views:	326
Size:	58.8 KB
ID:	9973364

    A bit more work and I installed a 3/8" NPT plug (I'd recommend 1/4" NPT to anyone trying at home).
    Click image for larger version

Name:	PXL_20210203_130840894.jpg
Views:	324
Size:	56.5 KB
ID:	9973365


      Now at this point I learned some hard lessons. Pay close attention to Ray's post and what he did. I should have machine my cover's OD down more and made the new ring thicker. But it was good enough to test the fluid to see if the 80,000 cSt actually was the right stuff. I cleaned up all the parts and started to lay in the silicone. This stuff is 10x thicker than joke.
      Click image for larger version

Name:	PXL_20210204_013936274.jpg
Views:	330
Size:	68.2 KB
ID:	9973369
      Click image for larger version

Name:	PXL_20210204_013941326.jpg
Views:	325
Size:	65.7 KB
ID:	9973370


        I ran my first comparison at 120 RPM just to see how it did....something wasn't right though. What was odd was my first test (not pictured) had so much torque I couldn't actually stop it. After thinking about it I assumed I had trapped some air bubbles in it. So I painfully slowly bled the air bubbles out over a few hours/days. The results were not promising... but it had to be air somewhere!
        Click image for larger version  Name:	Hand_Bleed.png Views:	0 Size:	60.5 KB ID:	9973375

        After researching online more I ran across this treasure trove of information.

        I did not realize that so much air could be dissolved and not even visible in the silicone fluid. (See second video

        So I got a vacuum pump and degassed the silicone as well as topped off the fluid and got much better performance.
        Click image for larger version  Name:	Vac_Bleed.png Views:	0 Size:	66.7 KB ID:	9973376
        Last edited by AWD_E30; 02-09-2021, 05:07 AM. Reason: Formatting


          I need to perform additional tests yet on the rebuilt viscous coupler and the known 'good' one. Especially to learn the characteristics of when the 'hump' occurs which self protects the unit from overheating. I wanted to post up the full data as soon as I finished a rebuilt unit to show some of the lessons learned and results. My first thought is that a stock unit uses a fluid lower than 80,000 cSt but I want more data to prove this. If there any other requests or feedback let me know.


            Nice little test rig you built there Sent from my SM-G960U using Tapatalk
            88 E30M3 X2
            89 325IX
            92 R100GS/PD


              I know a VC that needs rebuilding. Very cool stuff here, keep it up!


                Thanks guys! Glad everyone is enjoying the setup and results.


                  Great work AWD_E30! Hope I never need your services but nice to know they are available.

                  Not sure if this is the appropriate place to ask these questions, but is there a way to tell if the VC is starting to fail by how the car performs in the snow? Do the VC 's fail gradually over time with decreased levels of torque transfer? Can an ix pass the jack test but still have a VC that is starting to fail?

                  89 325ix coupe Diamondschwartz
                  1994 Mazda Miata


                    Thanks E30andy! It's hard for me to say definitively on gradual failure over time but I believe it is possible (sorry for the rambling below!). Based on the very extensive testing/history done by Andreas on VW VCs ( it appears there are two main failures.

                    The first is what most of us are aware of; The seals get overheated/worn out and the silicone within the coupler get mixed with transfer case fluid and/or air. This causes a drop in the torque transfer capabilities and would likely cause a fail of the jack test after some time. I think this might be gradual at first as more air/foreign fluid gets mixed into the coupler. How quickly the air/foreign fluid ingresses into the coupler will decide how quick the degradation occurs.

                    If you look at the first graph in my 4th post, the gray line (Rebuilt+HandBleed) vs. the medium blue line (Rebuilt+VacBleed) is almost all difference in air in the system. What you'll notice is that the peak torque is very similar between the two runs but the dynamic torque is very different. What does this mean? If you start your car after sitting overnight to do a jack test, it's possible that first bit of static torque is enough to get the car moving on the jack. Now, let's say you drive the car a bunch and then try the same jack test. I could see the same coupler failing the jack test after the fluid gets mixed up and the air bubbles scattered. I have some test data that shows massive differences in torques based on how many runs I make on the coupler. The first run has the highest torque and the it drops down with each subsequent run. Maybe I will make a bunch of subsequent runs to show how much the torque changes between runs from sitting overnight to let's say 10 runs. I'm guessing with a bit of math we could estimate the torque required to pass the jack test...(and here I go off into Nerdland again :D lol)

                    The second I was not aware of but is present in VW VCs which is silicone hardening due to silicone mixing with the air. This actually causes the torque transfer to increase and the hump mode to activate sooner. Quite the opposite of what I would normally think of 'wearing' out. I don't know if our viscous couplers suffer the same fate but perhaps over time I will rebuild a few units and find this case.

                    Keep up the comments and questions. The more you guys ask the more I want to collect data and catalog my finds.


                      Nice rig for testing and great data. Do you know what causes the difference in torque between the ix and Subaru VCs? Plate geometry? Fluid? Fill? I'm surprised that the ix locks up so much tighter.
                      AWD > RWD


                        Thanks Kershaw! I think all of the above is why Subaru VCs are 'softer' than BMW VCs. The general theory is the fluid viscosity and the plate count are the biggest factors (if memory serves me correctly)... but air volume and plate geometry also impact results. I have a fellow rallycross guy who might have an extra Subaru VC...might see if I can set it up to test and maybe see what the insides look like


                          Check out the big brain on this guy - awesome!!!


                            After reading TehRaydarlover thread on rebuilding the VC, I also thought about cutting one open at work (I've been meaning to do this for a year of 2 now) but never found the time to get started.
                            I wouldn't even have tried to measure the VC after rebuilding so very nice and thorough job! very nice, props to you sir!

                            1990 325iX Touring - November 2018 R3V Car Of The Month

                            1980 Volkswagen Golf mk1 1.1
                            1974 BMW 2002 Touring



                              This is awesome. Building a setup to measure things, and measuring + documenting before and after the fix (which even itself is quite challenging) like this is something not many enthusiasts are capable of. Sharing it is the cherry on top.

                              Would be nice to know the graphs of a brand new factory unit, how close it is.
                              1986 325iX
                              2006 530xd