PV997

As discussed at length in the PDK repair thread sticky (link) and the comment section, the PDK distance sensor is one of the most failure-prone parts within the transmission. If this part fails Porsche will not replace it and insist you buy a re-manufactured transmission at a minimum cost of $12k. In addition, Porsche will not sell you the part and their subcontractor (ZF) is contractually forbidden from selling it to the public. I can understand Porsche’s unwillingness to replace this part on cars under warranty as it’s probably safer for them to just swap the PDK. However their unwillingness to provide the part to those facing an out-of-warranty failure, who are willing to accept repair risk, really stinks.

Here’s it’s location within the PDK shown by the red arrow. You can see three of the magnet appendages that are used as part of the location sensing system discussed below.




Here's a shot of the distance sensor itself. Note the yellow-brownish component on the left is the speed sensor which also sits in the gear case. It's made of high temperature nylon which got us wondering if the distance sensor housing could be also (discussed below).




Three aftermarket suppliers have stepped into the gap and provide replacement distance sensors. While it’s nice to have this option, these all sell for about $2000 (+/- a couple hundred bucks) which IMO is excessively high compared to development and recurring costs. These suppliers have customers in a desperate situation and have priced their parts based on what they can get away with rather than a fair-market value (again just my opinion). That's capitalism after all, but why haven't prices come down with competition from three suppliers? Jjrichar and I set out to see what we could do to provide an alternative. Note that this a not-for-profit open-source project we are describing and not a business venture.

(Some of this has been covered over in the PDK thread comments but I started a separate thread here to document comments and progress. Also take a look at jjrichar’s thread over on Planet 9)

https://www.planet-9.com/threads/aff...roject.253044/


What it does
The distance sensor is an incredibly simple device than simply measures the magnetic flux of four magnets mounted on the shifts rods. It then converts that flux density to a pulse width modulation signal proportional to the magnet distance, which is read by the TCU. Using this info the TCU knows the position of the shift forks at all times. The videos below from jjrichar shows how this works using mulitmeter capable of reading PWM duty factor on a factory distance sensor.




Our Design
When designed back in the late 2000’s, ZF used four complicated inductive pick-up coils, each coupled with a separate IC that converts the analog signals to a PWM output signal. In addition it required nearly fifty discreet components (resistors and capacitors) to set levels and signal filtering. With modern programmable Hall Effect sensors, this circuit can be vastly simplified to just four active HE sensors and eight filter capacitors.




For our design we chose the Micronas HAL2455 as the Hall Effect sensor due to it’s low cost (< $4 each), high temperature range, and ease of programmability. Micronas supplies a very nice Labview-based GUI that allows chip programming without having to bang out line-level code. Micronas also provides a relatively-inexpensive programming unit ($170) for use in configuring the HE sensors.

HAL2455 datasheet, an example GUI screenshot, and a youtube showing use of a comparable programming GUI:

https://www.mouser.com/catalog/specs...hal%202455.pdf





We configured the HE sensors in an end-sensing layout rather than our original plan of using a center-sensing configuration. The reason for this is technical and has to do with the magnet length in the PDK. The magnets are short (6 mm) and thus would produce a non-monotonic sensor response over the required 20 mm travel range if center sensing. This could be resolved using a multiple sensor array but that would dramatically increase part count and reduce overall reliability. End sensing solves this problem as the response is fully monotonic over the 20 mm range.

You can see this below which shows the magnet flux density for various magnet lengths (5 mm is close to the 6 mm magnets used in the PDK). End sensing provides a far larger sensing range where the response is monotonic (doesn’t reverse direction). The end sensing flux is somewhat non-linear, though that is easily corrected using the programmable linearization feature of the HE sensor.



Programming allows custom settings of signal scaling, offsetting, linearization, clamping levels, and full temperature compensation for both the HE sensor itself and the magnet characteristics. We use all of them. It’s an incredibly versatile component for less than four bucks each.


Printed Wiring Board
We laid out our printed wiring board (PWB) using KiCAD, a free open-source layout tool. We then had OSH Park fabricate our PWBs as they specialize in low-cost, low volume production. The PWB is standard 1/16” thickness using extended temperature FR-4 material (Tg = 175 deg C). Individual PWB cost is about $17 each, lower in high quantities.



Here’s an early PWB breadboard used for testing, the updated PWB shown above is out for fab now.



The black transistor looking thing sticking up is the Hall Effect sensor. Three magnets travel behind the distance sensor and one above it, hence their arrangement on the PWB. Future PWB layouts will likely replace the ceramic capacitors with surface-mount chip caps.


Sensor Housing
All of the aftermarket distance sensor suppliers use a machined aluminum housing to replace the cast aluminum housing used by ZF in the factory distance sensor. We created our own housing 3D model and started shopping around quotes assuming CNC machining using 6061 aluminum. Costs in small quantities (1-3 units) were excessively high due to the non-recurring setup involved. Costs in larger quantities were more reasonable, with 20 units coming in at $153 each and 50 units at $94 each.







In keeping with the low-cost goal of the project, we started looking at other housing options. The PDK speed sensor shares the same environment as the distance sensor and its made of injection-molded plastic using PA46-30GF glass-reinforced nylon (known commercially as Stanyl). This high temperature plastic works absolutely fine in the PDK without any problems. The distance sensor itself carries no mechanical loads but simply needs to maintain it’s shape and not deteriorate from the temperature and chemical environment. Why couldn’t it also use a durable plastic housing?

Injection molding has high start up costs so we experimented with 3D printing the housing, first using ASA (an ABS derivative) on a test article. We knew ABS wouldn’t be suitable as a final material as it softens at too low a temperature. However, it would give us a good indication of print quality and dimensional tolerances. Muddtt from Planet-9 printed the first article and we were blown away by how good it looked and the dimensional accuracy. The time-lapse video below shows this amazing process, it actually took about 2 hours to fab.




Note that the bolt holes are oversized for aluminum bushings that would be required with a nylon housing.

With the concept demonstrated we started looking for a higher temperature material that didn’t require $10,000 industrial equipment to print. The leading candidate is Carbon-X HTN CF filament which is close to Stanyl in its temperature and strength performance. This is really impressive stuff as it consists of high temperature polyphthalamide with 15% carbon fiber. It has a quite high heat deflection temperature of 240 deg C, a melting temp of 265 deg C, with strong tensile and flexural strength (due to it’s carbon fiber reinforcement). Best of all, it’s printing temp of 300 deg C and bed temp of 120 deg C is just within the range of higher-end consumer-grade printing equipment.

Muddtt used the Bambu Lab X1C (~$1200) for the ABS housing seen in the video and it worked great. It’s also capable of printing the HTN CF material.





Potting Compound
Last off was potting and for that we went with MG Chemicals 832HT high temperature potting compound. It has a working temperature range up to 225 deg C with intermittent usage to 250 deg C. We know at least one of the aftermarket suppliers uses this exact same material as they say so right on their website. In addition, the potting compound’s high flexural strength (101 Mpa) and high flexural modulus will add stiffness to a printed housing cavity area. Knowing this material seems to work fine in the PDK despite its maximum rated 225 deg C operational temp also gives us confidence the HTN CF will work fine for the housing.




Putting it all together

Assembly is simple:

• Populate the PWB and install the connector/harness
  
• Half fill the housing with potting compound
  
• Install the PWB and harness using attachment screws
  
• Complete potting fill and allow to cure
  
• Program the HE sensors and lock their settings (so they cannot be changed)
  

Programming can be done at the sensor, PWB, or housing level as the programming functionality remains even when installed in the circuit. In fact, you could even program it after installation in the PDK if you wanted to tweak the settings after some trial runs (provided the sensor wasn’t locked first).

Programming is a little tricky as one of the sensors (#2) is flipped 180 degrees with respect to the other sensors in the original ZF design for some unknown reason. This sensor needs unique programming from the other three as it’s duty factor is reversed from the others (90 – 10% rather than 10 – 90%). It’s no problem though, just something to keep track of.

As with the existing aftermarket sensors you will need to run a PDK calibration using a PIWIS clone after installation to account for sensor variation and positioning variations from one PDK to another. We’ve used a $400 PIWIS2 clone previously on PDK work (that has been discussed on several threads here) and plan to use it again for this work.



Distance Sensor Costs
Depending on housing material and quantities, the recurring cost of the distance sensor can vary greatly. Note these do not include costs of non-recurring items such as the programmer and a 3D printer which we’ll talk about below. Also, some things such as printing filament have minimum buy quantities so I’ve included that under non-recurring start-up costs.

Here’s a comparison :

Sensor using 3D Printed HTN-CF housing (quantity 3)

• Components (HE sensor, capacitors, connector): $40
  
• Printed wiring board: $17
  
• Potting: $20 (~90 ml)
  
• Printed housing using HTN-CF: $15 (filament and electricity costs, not including wear)
  
• Total material cost: $92 (plus 1.5 hours labor to populate PWB, assemble, and program)
  

Sensor using 3D Printed HTN-CF housing (quantity 20)

• Components (HE sensor, capacitors, connector): $31
  
• Printed wiring board: $11
  
• Potting: $9 (~90 ml)
  
• Printed housing using HTN-CF: $15 (filament and electricity costs, not including wear)
  
• Total material cost: $66 (plus 1.5 hours labor to populate PWB, assemble, and program)

Unlike the printed housing a CNC machined housing has a huge cost curve versus quantity as seen below.

Sensor using CNC machined aluminum housing (quantity 3)

• Components (HE sensor, capacitors, connector): $40
  
• Printed wiring board: $17
  
• Potting: $20 (~90 ml)
  
• Machined aluminum housing: $465
  
• Total material cost: $542 (plus 1.5 hours labor to populate PWB, assemble, and program)
  

Sensor using CNC machined aluminum housing (quantity 20)

• Components (HE sensor, capacitors, connector): $31
  
• Printed wiring board: $11
  
• Potting: $9 (~90 ml)
  
• Machined aluminum housing: $154
  
• Total material cost: $205 (plus 1.5 hours labor to populate PWB, assemble, and program)
  

As you can see the aluminum housing cost curve changes exponentially with quantity:




In small quantities the aluminum housing is cost prohibitive but it’s very affordable in quantities of 100 or more (100 units is $68 each). All the aftermarket suppliers use machined aluminum housings, and I don’t know their sales counts, but I have to assume it’s at least a 100 units. So why are their replacement sensors so expensive? Good question, you’ll have to ask them as I can't explain it.


Non-recurring (start-up) costs

I’m not going to get into stuff like paying the janitor that a real business has to deal with, but the non-recurring costs to do this as a hobby.

• Micronas HE Sensor programmer: $170
  
• Bambu Labs X1C 3D printer: $1200 (if printing the housing)
  
• Filament drying kit: $150 (if printing the housing)
  
• Carbon-X HTN CF filament 0.75 kg: $175 (if printing the housing)
  
• Good soldering iron and accessories: $120
  
• Cost of noise-canceling headphones for wife who is annoyed by 3D printer sound: $300 (optional)
  

And that’s really about it. If you can find someone with a good 3D printer maybe you can get them to print it for you for free. Alternatively, there are fab houses that will print the part in a high temperature plastic for around $120 each, and unlike CNC machining the cost curve vs. quantity is minuscule.


Conclusion
We’ll follow up in the comments as to where we are going to go with all of this and results of usage tests. We already have a breadboard unit programmed and it seems to operate just fine. Next step is to install and pot it in a 3D-printed housing and try it in a real PDK.

Our thinking is that we could provide these to people at cost (see above) as an alternative to high cost $2000 aftermarket distance sensors if they need one. Over in the PDK thread comments I’ve noted that it’s time for the distance sensor costs to come down as we now have three suppliers, and despite the competition they are all charging around the same amount. So far those comments have fallen on deaf ears so we hope to provide an alternative. Neither jjrichar or myself have any intention of going into business with this, but view it as a community project for those interested.

We'll also post CAD files and programming info in the comments as we get them ironed-out. The intent is that this is a fully open-source, not-for-profit project. We are also very interested in people’s thoughts regarding the risk (real or perceived) of using a printed HTN CF housing versus aluminum.

groovzilla

Excellent write up and Engineering!
This reminds me of the person who 3-D printed the small hing piece used for the broken 997 Cup Holder. Instead of having to buy the $300-$500 cup holder assembly, they could purchase the $10 hinge part.
I bought a few of them just in case and appreciate the savings.

Good luck on your project!

PV997

Thanks Groov, I always appreciate your insightful (but often contrarian) takes as I think along the same lines. The fact that you didn't point out any obvious flaw in our thinking is a good first step!

hexagone

As someone in product development professionally and as a hobby, hats off to you on this. If there's a low cost, DIY, reliable solution for PDK problems - this takes a $20k problem to a $2000 solution (parts & labor for PDK removal/disassembly), truly a game changer.

And that's on top of the few players who were offering the distance sensor replacement and were bringing the $20k problem down to say, $5k all in.

I've passed on some dirt cheap PDK cars just because of the failure stigma - and knowing the solutions out there now I'd probably not hesitate to buy one.


Fantastic progress.

Wayne Smith

From someone who spent over 40 years designing Cat 4 and 5 industrial machine controls ... you've blown me away!!! Very impressive in so many different ways. Congratulations on a fantastic journey.

rael

For someone who spent zero years designing anything, I'm truly blow away as well. Very impressive.

jbkusa

On behalf of all the fellow PDK owners… thank you for your ambition and determination to do what Porsche should have already done.

Are you adding this to your PDK sticky or creating a new sticky?

Wayne Smith

This. Absolutely this.

Given the prices I've seen in multiple threads on this forum recently, it would appear that we are reaching the point where MTs will be more of a repair cost liability than the PDK.

cwheeler

I 100% support this. Even without owning a PDK yet. I'll buy one (or more?) when this is done just to protect my future. The aluminum sounds like the best option to me.

I love the project and have tried to stay somewhat current on the major thread. Thanks for your work!

Cw

PV997

Thanks! I'm not a mod and there's already a bunch of stickies up there already so we'll just update this thread as we go along which should bump it up. The PDK sticky has nearly 1100 comments and I started discussing this in the comments there but thought it warranted it's own thread. Nice to see shops are actually repairing PDK's now based on that thread and the comments. PCA hosted a livestream with Todd Lamb from ATL Speedworks on PDK repair a few months ago and they used a lot of graphics from that post. Happy this is getting attention and the myth that PDKs can't be repaired is dying.

The problem I see is two-fold:
.

• Porsche's PDK mandatory replacement scheme is fine for warranty work but ridiculous for older cars and threatens resale values
• Aftermarket suppliers wrongly think they are entitled to the Porsche Tax. This isn't just a PDK issue but across the board.

So we'll see where this goes and hope to have one actually in a PDK within a few weeks. Big thanks to @jjrichar who posts here but is also a mod over at Planet 9 as he's done incredible investigative work and got the 3D printing up and running with Muddtt.

Wayne Smith

Cwheeler brings up an idea ... If there are 100 enthusiasts willing to put one of these in their parts bin for future proofing their car (or a future car with PDK), that could act as a crowd source funding project. Even if not needed, there is resell value in having the part with the car. Or sell it on to another enthusiast.

docdrs

I would be in on this. I have been following this closely and am impressed with the resolve and ingenuity of PV, JJ and Muditt. I figure the surest way to never need one is to have one on hand. It sure would be the best insurance. It would be great if we could do this for the HPFP which both MT and PDK share.

hexagone

group buy?

Jabamusic

Fantastic work you Guys. I would certainly be up to having one on the spares shelf.

sandwedge

Amazing post. I guess you had the day off as I can't imagine putting all that together was done during a coffee break. Thanks for the time and the effort!

And I agree with you in not understanding why Porsche won't sell the sensor to out of warranty owners or to anyone else including indys as I understand it. Why not just warranty the sensor in case it fails again with no further obligation on their part? No need to warranty the whole transmission because they sell a part that's needed.

Btw....the picture above that you provided of the sensor is a picture of a $20,000 part. Doesn't look like one but I guess I don't understand modern technology. I say this since my latest out of two sensor related PDK failures came to $20,000 as full replacement was the only option. Anyone who can get a full PDK replacement for $12,000 these days is either getting a bargain worth bragging about or isn't getting the real thing.