[jjrichar]

A question to the masses about sensor design. See picture at bottom.

There are grooves along the side of the aluminium casing that are aligned with three of the sensors (red arrows). The shift rod magnets are on this aluminium side for those three sensors. There's another groove that doesn't seem to align with anything (green arrow). The magnet for shift rod 4 is on the other epoxy fill side, and there isn't a groove here as it's just flat epoxy.

PV and I have been discussing offline what the purpose of them might be, and we can't seem to find a reason they are there. Anyone got any ideas? It will just make the CAD drawings easier if not required. CAD drawings should hopefully be reasonably simple. Nicely rounded edges around the end connecting holes aren't required as there is loads of spare space here. It's just the center connecting hole that needs to be the nice shape as one of the shift rods gets to about 2mm away from it.

I'm also interested in some feedback on the material of construction. The OEM and other aftermarket sensors are made of aluminium. It seems this is going to be the most expensive part of a sensor. We have the thought of 3D printing it from glass or carbon reinforced nylon. This is the stuff lots of parts in the car are made of. Transmission pan, intake manifold, and lots of sensors are examples. My guess is the transmission speed sensor is made of it.

We ask the question why the OEM sensor is in an aluminium casing. We think it might be because it needed to be a potted sensor, and to try and pot effectively into an extruded plastic casing wouldn't have been possible. The epoxy wouldn't be secure enough in this environment. The beauty of 3D printing is that those problems wouldn't occur. It would be easy to 3D print a casing that has loads of loops/jagged edges etc that when potted will hold the epoxy very effectively. You could even 3D print an inside overhang in the casing edge so it was physically impossible for the potting to come out. The temperature capability of these materials is the equivalent of the sensors, as they have better heat resistance than standard nylon.

3D printing of this material doesn't use a particularly expensive/bespoke machine compared to aluminium. There would the need to reinforce the screw holes. Maybe a stainless tube. It doesn't need to be particularly strong, as the screw is M5 and will take 8 Nm. We are thinking a sleeve could be just pressed into it like they do for the transmission pan or any other sensor made of it.

Can anyone see a good reason for not pursuing this? Strength, resistance to using a plastic sensor, anything else? Ideas on which, carbon or glass would be the most appropriate? Carbon = stiffer and stronger. Glass = better impact resistance.

Some additional info for reference:

Both of the aftermarket sensors have the grooves. The XemodeX website has pictures of their sensor internals, and the sensors themselves are right up one end well away from the grooves. I don't know if they just put them there because the OEM did.

Magnet distance from sensor side is about 1.5mm on the aluminium side, and about 2.5mm on the epoxy side. The grooves are 0.25mm deep.

When I removed the distance sensor the other day from the project car, there was a load of ferrous sludge on the magnets. This is normal, but there was so much it was touching the sensor. The sludge looked like it was being pushed to the sides of the magnet as the gap reduced to nothing and more sludge gathered over time. This car only has about 60,000 km. I would imagine that if you opened up a transmission that had a lot of miles you would have a really good accumulation of sludge on each magnet, with most of it being pushed to the side and the gap between magnet and sensor remaining filled.

Of interest the transmission I pulled apart had only about 15,000 km, and the level of magnet sludge was minimal compared to the car. The 15k transmission is from a Cayman GTS with a bunch of track time, the 62k transmission is a base Boxster with just street usage. I thought there might be initial wearing in of the gears that would accumulate quickly and then settle down to nothing as time progressed. This seems to indicate otherwise, with progressive wear seeming to occur. Very small sample size to come to any sort of firm conclusion though.

 

[schwinn]

My guess is that the grooves may be there from an older design requirement? I think someone said these were used on many different vehicles, and potentially different transmissions as well? They could also be left there just to provide visual alignment, though that's an expensive way to do it, I suppose.

As for the plastic printed housing... I would be concerned with thermal cycling causing the plastic to crack over time? Remember, this part likely sees a lot of thermal cycling over time... while plastic will hold up to the temps, I question if it will hold up over time as well? But for initial testing, I see no reason not to run with plastic for now... easier to work with. Again, OEM was probably metal if this was an older design, since 3D printing wasn't as popular in the past (and certainly not in the automotive world).

I don't know enough about 3D printed materials to make a suggestion between the options you have, so I'll leave that to others. Don't forget about considering chemical resistance as well - I know ATF is typically pretty aggressive, chemically speaking... I don't know what the PDK fluid would be like, but just remember to consider that when choosing the plastic.

 

[deilenberger]

I'd also research temperature-induced expansion and contraction. How much is allowable? How does it compare to an aluminum part of the same design?

 

[jjrichar]

Snip from here for 30% glass filled nylon (30% is pretty typical from what I've seen).
Thermal expansion and strength highlighted.

Link to document
https://www.theplasticshop.co.uk/pl...s/glass_filled_nylon_technical_data_sheet.pdf

Snip of 6061 aluminium below. This is what the XemodeX sensor is made from.
Thermal expansion highlighted.
Link here:
6061 Aluminum: Get to Know its Properties and Uses - Gabrian

But then if I go here, it tells me they are very similar:
Thermal Expansion - Linear Expansion Coefficients

Epoxy from this site has an expansion co-efficient of 45-65.

If I use the information that gives the greatest expansion of the GFN, it's about 2.5 that of the 6061 Al, which would equate to an expansion of 0.003 mm (in height) and 0.01 mm (in length) for the sensor that is 50 x 180 mm. The relative expansion compared to the contact materials is going to be a bit more than half of this. I'm not sure if this is a lot or a little, but certainly isn't going to move the sensors enough to have any significant effect on sensor output.

If the screw holes had a sleeve like they do with most items that are made of GFN, I would think the problem of expansion along the height (where the screws are) would go away. It would then be the screw expanding equally with the stainless sleeve. Steel thermal expansion is about 15, so would expand less than either the Al or GFN.

My only real concern (assuming the expansion of GFN is about 60) is upon significant cooling. Think cold temps of Canada that I've never experienced. In this case the GFN will contract a lot around the stainless sleeve. I see either stainless or brass sleeves embedded into GFN in a lot of places on the car where there is this level of thermal range and it doesn't seem to be a problem.

Maybe the sleeve could be a 'light fit' into the holes, so this contraction wouldn't be an issue. Like with most sleeves into GFN, the screw end is wider than the sleeve (screw head is 12mm wide). So the GFN body can't slide off if it ever came loose from the sleeve.

More than happy to provide more photos and dimensions if required.

Anyone with experience in the field please don't be afraid to chime in.

 

[jjrichar]

I think ZF have already done the R&D for us. Image below of the inside of the pan.

Filter is Nylon 30% glass fibre.
Pan is Nylon 35% glass fibre.

I knocked out one of the aluminium sleeves. Comes out with little effort with a punch and hammer. I could push it back in by hand. It's clearly a relatively loose fit and the wide screw head over the black plastic is what is keeping it in place. The sleeve is just there to torque against and ensure the screw can't come out.

The pan is designed to lower the transmission, so I'm convinced of the material's strength.

 

[ledbette]

@jjrichar You'll never know how much this community appreciates the contributions you have made to demystifying maintenance on the 981. I've been following and contributing to this forum since 2016 and people like you and the many others in this forum have really been a blessing to the DIY-inclined 981 owners. Thanks for all your efforts!

 

[jjrichar]

@jjrichar You'll never know how much this community appreciates the contributions you have made to demystifying maintenance on the 981. I've been following and contributing to this forum since 2016 and people like you and the many others in this forum have really been a blessing to the DIY-inclined 981 owners. Thanks for all your efforts!

Thanks for the kind words.
I've loved every moment of it, and it's nice to hear it has helped a few people along the way.

 

[jjrichar]

Did a bit of an experiment when I removed the sensor on the project car the other day. On install I used the old gasket without any sealant. The gasket is the metal variety that has a raised ridge along it that compresses on install and creates the seal. This is why they say to replace, because this ridge gets permanently pressed down.

When I looked at the seal the ridge was still a little proud. The metal is 0.25 mm thick, and the ridge seemed to be between 0.05 - 0.1 mm proud still. So I fitted with no sealant to see if it would seal mainly because I didn't want to have to clean up RTV when I pulled it again when PV starts sending me sensors to test.

It seems to seal fine. No leaks the next day.

Next experiment was to remove the 5 lower screws along the bottom of the flange to see if there was any leaks. None. Note directly above this is where the pool of gear oil sits, and the seal seemed to be working fine even with these 5 screws removed.

Clearly this is an experiment over the very short term, but to me it shows that if the gasket doesn't get bent on removal and everything is spotlessly clean on install, then you probably don't have to seal with RTV.

 

[jjrichar]

A bit of an update on progress.

Muddtt has done some super work doing the 3D design work. See image below.

PV997 and Muddtt are now talking about the cavity design for the fitting of electronics.

 

[PV997]

Update on the low-cost distance sensor as I got the PWB assembled and installed in a dummy aluminum cavity for testing. The Hall Effect sensors are extremely easy to program and amazing versatile (and they only cost $3 each). Initial measurements look encouraging though I definitely need to make some modifications on Rev 2 of the PWB to move the sensor location to improve the sensitivity. No problems though as it's easily able to consistently measure 20 mm of magnet travel with good isolation between channels.

Here's a picture below showing the breadboard (not a prototype but a functional breadboard, the cheesy mock-up is for testing only). The housing is mocked up using 1/16" 6061 aluminum sheet metal held together with aluminum tape.

The capacitors will be replaced by high temperature SMT chip caps on the final design and the harness is a temporary attachment while we find a good connector candidate. You can see one HE sensor on the top of the PWB on the left (for the shift rod magnet that goes behind the distance sensor) while the other three are on the bottom side of the PWB. The PWB attachment is temporarily held together with nuts and bolts while we find a good solution. Interestingly there seems to be no effect on the HE sensors even using ferrous attachment hardware. Have not tried potting yet but we've found a good high temperature candidate.

Electrically this design is about as simple as they come and so far there are no surprises.

Thanks to jjrichar for his extremely detailed measurements on magnet location and travel distance to help get this dialed it. Also thanks as he's getting the first housing fabricated for fit and alignment checks.

(x-posted from the rennlsit thread)

 

[PV997]

Had to post this to show Muddtt's amazing work. He printed this today using ABS (what he had handy) as a demo for fit checks, dimensional tolerances, and testing purposes. As jjrichar posts above we'd probably use glass-infused nylon if we went with a non-metallic housing but this amazing work by Muddtt is a very nice proof of concept.

The housing design isn't final yet but it's probably at about 95%. Looking forward to testing my updated PWB in it!

 

[jjrichar]

Here's a video Muddtt made of the printing. Cracking job.

 

[jjrichar]

Photos below of Muddtt's design that I had printed in a low quality spec. It doesn't have a sensor cavity as it was just to test for fitting in the transmission.

It fits very nicely with good clearance to other components. One of the holes needs to move by about 0.5mm to be perfect (my fault as I sent him the dimension specs) but other than this I'm very impressed with how it fits. The OEM sensor wants to rock easily on the transmission mounting points prior to fitting of the screws. The printed housing sat there solidly prior to any screws being installed. I'm impressed by the accuracy of these 3D printed housings even at a low quality spec. It's printed in a clear plastic so I can see more easily the clearance to other components etc. The final product won't look like this.

The alignment pins that are part of the OEM housing we've used grub screws that screw directly into a small hole on the side that have been 3D printed. This worked a treat and fit like a glove.

PV is currently investigating much higher spec 3D printing filaments. There is a large variety of filaments available to choose from. At the moment he has his eye on a particular carbon reinforced high temp nylon. We are considering printing them ourselves due to the cost being very low doing it this way.

I've inserted 10mm OD aluminium sleeves into each of the holes. These take all the screw clamping force like pretty much any other glass reinforced nylon component on the car. I could only locally source a sleeve with 1.2mm wall thickness. End game I think we will use one with a 1.6mm wall thickness for better strength. This will give a hole size of 6.8mm. The OEM sensor holes are 7mm.

PV has been sent the housing that Muddtt printed (photos above), and is now waiting for a few components to be delivered prior to getting it all installed and potted. Then it will be on the way down here for installation and testing. Looking forward to see how it all goes. Working with PV and Muddtt has been a treat. A couple of extremely professional and capable individuals whose combined engineering backgrounds are impressive to watch. I have much confidence, and as you might expect I'm really looking forward to getting the sensor installed in the car.

 

[jjrichar]

Update to the project:

All is progressing well with much happening behind the scenes to make the project a reality.

PV has produced the first prototype and this on the way down under so I can test in the project car. I'm expecting it to be delivered in the next week or so. I'm really looking forward to getting it in the car and see how it performs. From his testing it replicates the OEM sensor output. Image below of the sensor.

Since making this sensor, there have been a few tweaks made to the design that will improve individual sensor location, sensor package fitting and also how it fits into the transmission itself. Whilst the first design fit quite happily, there wasn't sufficient clearance between the housing and the transmission rear casing. Combined with the unknown of how the sensor fits to a 911, where the orientation is reversed, we have decided to pretty much replicate the OEM sensor shape and have it CNC milled from aluminium. PV has received a bunch of quotes on CNC housings and it seems it's possible to get them made in relatively small numbers at around the US$60 mark. The housing is by far the largest material cost of the sensor.

The beauty of 3D printing is that we can test using a 3D printing design that has enlarged holes for fitting of aluminium sleeves, and get this all bedded down prior to simply changing the hole design for CNC milling and fitting to any transmission. One-off CNC milled housings are very expensive, and to have the option of doing all the testing using a 3D printed housing is a big cost saver.

Muddtt has done a super job of creating the CAD models for both housing varieties. Images of these below.

We've also have a bloke over on Rennlist who goes by the name jcfx11 offer his services. He has an enthusiast CNC machine at home and has offered a lot of good information relating to CNC design so we can reduce costs, and is also willing to mill a housing for testing when required.

PV started a new thread over on Rennlist that has a lot more information about the design if anyone is interested . Link here:
A $100 3D-printed PDK distance sensor? - Rennlist - Porsche Discussion Forums

Prototype 1 with OEM sensor behind.
Note: the molex connector in the harness is so we can swap prototype sensors easily. These connections would normally be soldered.

Version for 3D printing

Version for CNC machining

 

[schwinn]

I agree with the decision to use 3D printing for prototyping, and then a machined part for final "production". Don't get me wrong, there are applications for 3D printed parts in production... I work with many industrial machines that are setup that way. But for something tucked away like this and in a non-atmospheric location, I'd certainly stick with the OEM material if possible.

One thing you may want to look at, is to be sure to use the "right" potting material to prevent thermal expansion concerns. You're not going through drastic temperature changes, but it's good to check, just in case. I don't expect a problem, but better to look and be sure. Also, might want to see if the OEM part has any special ribbing or anything to help with the potting. I don't expect it does, but might not hurt to look there either.

Lastly, for the potting, are you planning to use vacuum to get all the air bubbles out? I would recommend thinking about that, as it ensures a good seal throughout the part.

Don't let me get in the way... you are all doing a great job! Kudos upon kudos!

 

[jjrichar]

I agree with the decision to use 3D printing for prototyping, and then a machined part for final "production". Don't get me wrong, there are applications for 3D printed parts in production... I work with many industrial machines that are setup that way. But for something tucked away like this and in a non-atmospheric location, I'd certainly stick with the OEM material if possible.

One thing you may want to look at, is to be sure to use the "right" potting material to prevent thermal expansion concerns. You're not going through drastic temperature changes, but it's good to check, just in case. I don't expect a problem, but better to look and be sure. Also, might want to see if the OEM part has any special ribbing or anything to help with the potting. I don't expect it does, but might not hurt to look there either.

Lastly, for the potting, are you planning to use vacuum to get all the air bubbles out? I would recommend thinking about that, as it ensures a good seal throughout the part.

Don't let me get in the way... you are all doing a great job! Kudos upon kudos!

Thanks for all of this. Always appreciate any input.

Fortunately PV has been able to find the same potting as used on the XemodeX sensor. It's listed on the XemodeX website. It has really good specs for this type of application.

I've done a bunch of work with epoxy in the past (I use to make carbon fibre sailboard fins many years ago) and have forwarded what help I can to PV to eliminate the bubble problem.

 

[deilenberger]

I'm moving this thread to the 981 DIY/tech forum since this is the ultimate DIY project... hang on, moving now..

 

[jjrichar]

Update:
I received the sensor from PV in the mail yesterday afternoon.

Spent today testing and installing. The results, nearly a 100% pass, I'd call it 90%. There was much educated speculation on our part about lots of things and it was really nice to find it was as predicted. The only part that didn't work exactly right was the sensitivity of the sensor. It's a little too sensitive, giving a greater distance than is desirable when a gear is selected.

I did a cal and it passed first time no issues. I then put into rolling test mode and ran through all the forward gears without issue. It doesn't like reverse though. I suspect the distance the TCU is being sent is too much and it says no. I can hear it doing something, but the distance moved stays at zero, making me think it's trying to select, gets the long indication and then puts it back to a neutral position.

I've taken some video footage and also learned a few more things about making the distance sensor replacement easier. When I find some time I'll get it out.

To be honest, for our first prototype we are super happy with the results. The sensitivity was something we knew was going to be a long shot to get right on the first go.

I take my hat off to PV. We have a few fellas in the group working on the project, but PV is the real smarts and driving force behind making this a reality. Cracking job my friend.

 

[jjrichar]

Another update for those following this.

PV and I have gathered a load of data from the testing I did. Whilst it worked in many respects, the response was very close to the limits the TCU was expecting so we need to tone these down a bit.

What was really interesting was the interference we were getting from adjacent shift rod magnets. The distance seen was jumping by up to 2mm as the adjacent gears were be pre-selected. Some interfered a lot, some none. After some thinking about it we worked out it was the angle of the sensor. The HE sensor is sensitive to the magnetic flux that is perpendicular to the sensor face. This response rolls off at the sine of the angle in flux incidence. Have a look at the image below of the flux lines and the possible orientations of the sensor face bottom right. Note this works in 3 dimensions, not the 2 of the picture.

Blue is straight up. Our sensor had a tilt of 20 deg so it was facing directly at the magnet. Red is tilted back. As you can see through the range of the shift rod movement if tilted back the magnetic flux is close to 90 deg to the sensor face. PV has done a bunch of testing and he says the best angle tilt is 70 deg.

This has a bunch of benefits. It decreases the interference from the adjacent magnets by about 2/3 because we can turn down the gains to get the response we need. It also removes the magnet height above sensor sensitivity we were seeing.

PV is busily making the next sensor for testing with much enthusiasm. To say we have learned a lot over the last week is a bit of an understatement. Even when I I re-installed the OEM sensor and was too lazy to clear the codes prior to doing a new cal, it all failed terribly with loads of errors I had to work through. Finally got it sorted (after many anxious moments where I thought I'd killed the TCU or something else) and learned a whole lot in the process. It's amazing how much you learn from mistakes. The sensor only just worked, and because it was marginal we had to work through these issues and work out why. The amount of information we seem to have collected as a result of this is really satisfying.

Whilst PV is making the next sensor, I'm in the process of doing some stress testing on the first prototype to see what's required to break it.

Everything in this prototype should have been good to 175 deg C. Put it in my oven at this temp for an hour and the glass reinforced plastic melted and deformed terribly. The epoxy potting cracked a bit but held it's shape. We are using a higher spec potting for the final product, and we are going to CNC aluminium housings. The wiring insulation melted away a bit as well. I suspect my oven was a little hotter than the approx 175 deg C I set on the dial. Image of sensor afterwards below.

Happily the sensors survived no problems giving exactly the same response.

I'm videoing the entire testing process and when done I'll post it. Sensor currently in my freezer. Plan to follow that with some shock and vibration testing.

 

[schwinn]

Excellent progress... great info! I'm glad you did the oven test... and found that it wasn't going to hold up as well as the specs say. Better to see that before it's in a car!

FYI, not that this is necessarily a solution for you, but when we wanted to "quench" some of the stray magnetic fields (back in my semiconductor engineering days) we would put a u-shaped iron on the magnet, so that the field had a "path" to follow. Basically, if you put a metal cap on the disc-shaped magnet in your picture, it might help direct the stray field and keep it from going too wide.

I think the angled sensor is the best idea for now, or moving them a little further away if you can. Or maybe specifying a "weaker" hall sensor component.

In any case, you're all on the right track for sure! Just wanted to give a few other ideas/options.

EDIT: Here's a picture of what using metal to "short circuit" the field looks like:

https://www.researchgate.net/figure/Effect-of-a-metal-layer-on-an-alternating-magnetic-field-On-the-left-the-magnetic-field_fig1_221907309

The article may be related, but I thought the picture was perfect to show what I was trying to put into words.