wootloops

This thread has been awesome. As someone who just bought a high-ish (100k) mileage 981S with PDK I am really glad there is a potentially inexpensive fix if I ever run into this issue. Once these are available Ill probably buy a couple just in case.

PV997

Passing a long an update on this as we've learned some new info and hit a bit of snag.

Most of our testing and development has been on the Cayman/Boxster PDK as @jjrichar had a spare unit to use as a test bed. This testing has gone pretty well but we realized we also needed to do some tests on a 911 PDK. It's a mirror image of the Cayman/Boxster and we wanted to make sure we weren't missing anything important. The factory distance sensor is the same in both transmissions, but in the 911 it's rotated 180 deg from how it's oriented in the Cayman/Boxster. Fortunately @stjoh had a spare 911 PDK and offered to help. That's where we hit a snag.

Our distance sensor design uses an end-sensing configuration rather than a center-sensing configuration. We did this as the short magnet length (6 mm) would cause ambiguity in a center-sensing configuration as two different locations will output the same duty factor. We know at least one of the aftermarket suppliers uses this arrangement, and we suspect it could be the cause of reported calibration problems.

However, we now realize the end sensing configuration has it's own issues that can be problematic.



Take a look at the picture above and note the shift rod on the right (the distance sensor is removed in this photo). Unlike channels 2 & 3 whose shift rods are offset to the side (the two shown on the left of the photo), the shift rods on channels 1 & 4 are in line with the magnets and spaced about 12 mm from them. Unfortunately the shift rod is made of ferrous material.

On the Cayman/Boxster this isn't an issue at all as the shift rod is on the far side of the magnet (opposite the Hall Effect sensor) so it's impact is negligible. However on the 911 PDK the distance sensor is rotated 180 degrees so the shift rod is between the magnet at the Hall Effect sensor at the far end of travel. Since it's ferrous it significantly affects the sensor readings.

The easy fix for this is to offer two different distance sensors, one for the Boxster/Cayman and another for the 911. On the 911 sensor we would move the Hall Effect sensors to the other side so the shift rod is again on the far side of the magnet. However, this is different from factory which uses the same sensor for both, so I thought I'd throw it out there to see if people have concerns.

One final note is that we know one of the lesser known aftermarket manufacturers uses and end-sensing design like what we are using but they don't have separate units. I seriously question how this unit could work reliably in the 911 based on what we have found.

More to come.

Edit: here's a nice photo @jjrichar made that documents the issue. This is our prototype installed in a Boxster/Cayman PDK. When installed in a 911 the distance sensor is rotated which places the Hall Effect sensor on the other side under the shift rod. The solution would be two different distance sensors so the HE sensor is always distant from the shift rod (on the right hand side in this photo).

tony9189

Update:
1. I replace new T-design distance sensor and calibrate PDK, but the data is the same as before (P gear image) , i think it's not T-design distance sensor problem.

2. I asked a 2011/987/Boxster owner for the data (P gear), as you said it engage 1st gear when P is selected, but unlike mine is the year and Boxster transmission is latest software load.

The car is running fine now, but i'm still worried about the "P GEAR" , maybe i should upgrade the software too?

jjrichar

If you changed the sensor and it's all working fine now then it seems the sensor was intermittently failing.

Regarding the indications in P, I wouldn't be too concerned. If it was incorrectly selecting R rather than 1st in P (or giving an incorrect indication) then there's is no way in my mind that the transmission would be working OK. It does seem to be a software change in later models.

Me personally I'm a big fan of upgrading the software. Car manufacturers learn more after a car has come into service and they will install new software loads in the later cars. I've upgraded the PDK software in both my cars and the improvement to how they drove was immediate. Clearly this is your decision, but from a PDK knowledge perspective, it would be nice to know if a software upgrade changes the gear selection in P.

bzalforno

I have a 2010 base carrera PDK. With 70k miles. I would buy the aluminum part and even $200-30- would be cheap insurance if my pdk failed.
so count me in for at least one, possibly 2 units.
thanks
bob

rodsdream

Holy Smokes. I am super impressed!!!!! I just love the fact that you are able and willing to do this. Count me in on buying one of these. The best of health to you....
Rod

PV997

Thanks rodsdream. I just sent the reprogrammed sensor back to jjrichar in Australia and hopefully we'll be able to close out the design for Boxster/Caymans. We're still struggling with the 911 a bit due to the different transmission geometry (that I discuss in comment #182) but we think we have a solution for it. Unfortunately it needs a PWB update so it will be another month at least before we have updated parts to test.

Prairiedawg

I think I read in this thread or possibly the other that there's no difference between the OEM distance sensors with the 911 and the Boxster/Cayman's. If so, what does the manufacturer do to avoid the issues you are having using the same sensor in different transmission's? Is it just a programming thing with the TCU?

Just curious.

PV997

That is a great question. The factory sensor uses a long inductive coil that spans the entire range of motion of the magnet, so no matter where the magnet is positioned it's close to the sensor. This dramatically reduces the effect of surrounding metal and is a good design from a performance perspective. The problem with this approach is that it takes lots of discreet components to implement (over a dozen for each of the four sensors in the unit). More components means more likely to fail since component failure rates compound.

We (and at least two of the aftermarket manufacturers) are using a single integrated Hall Effect sensor per channel. Unlike the inductive coil it's small so as the magnet moves it can be quite distant from the sensor. One way to fix this would be to use a longer magnet so part of the magnet is always over the sensor. However we can't have people replace their magnets as part of this as it's just not practical.

So the best solution we have so far is to have separate units for the Boxster/Cayman and the 911. This allows us to still use a simple Hall Effect sensor (for good reliability) but to optimize the geometry depending on which car it goes in (thus assuring reliable performance by distancing surrounding metal from the sensor).

tony9189

Update:
Thank you for your suggestion. I have upgraded the TCU software(version from T079 to T082) and the issue with P Gear has been resolved!

jjrichar

That seems to indicate to me that the selection of reverse in Park is a deliberate software programming and has since been upgraded. Great to hear the issues have been resolved. Happy driving.

PV997

Passing along an update and to bump the thread as we should have new data soon.

The updated Cayman/Boxster sensor is in Australia and jjrichar should have it within a day or two. Main changes were reprogramming the sensor to account for influences from the surrounding ferrous metal in the PDK.

For entertainment value and to pass the time until we have new data I thought I'd post the US Postal Service tracking for the package as it has to be seen to be believed.

San Pedro, CA (my local post office) -> Los Angeles - > Denver -> Toronto, Canada (WTF?) -> stuck in Canadian customs for a week - > Chicago -> San Francisco -> Sydney, Australia

Next time I'll try FedEx.

PV997

A brief update to let folks know that this project continues but we are still working out some bugs.

We are still having some issues with reliable calibration on the distance sensor, and we won't release this until we are completely confident in it's operation. We know the factory distance sensor does not take multiple calibration attempts to finally work properly, and a replacement sensor shouldn't either. The calibration issues appear to be caused by variability introduced by adjacent ferrous metal (the shift rods), interference from the adjacent channel magnets, and just plain old manufacturing tolerances. The TCU has a narrow range that it considers a valid input from the sensor, and this variability has led to calibration faults.

We think we have a good handle on this and have resolved it through improved sensor position and sensor angle designed to null out interference. We also now have a very slick bench tuning setup that jjrichar made that ensures very repeatable magnet position during programming. Unfortunately it's been a slow process as I'm in California doing the fabrication and programming, and jjrichar is in New South Wales with the bench PDK and in car PDK demo. We've been shipping the prototypes back and forth across the Pacific and it's been painfully slow.

Prototype #3 is ready to go and I'll be shipping it to jjrichar this week. Third time is a charm we hope.

SpectreH

Wow….impressive work that can benefit us all one day. As an engineer, I’ve always appreciated Porsche’s engineering and I found this very interesting.

Wayne Smith

As an engineer designing industrial controls for 45 years I've found this journey to be extremely impressive!!!