heinzboehmer's 2002 Topaz 6MT Coupe

[YoitsTmac]

I love this project so much. Also makes me want to build some thermal read to all my brakes

[heinzboehmer]

Finished up the thermocouples.

First off, got tired of the noise in the readings, so I added four extra parameters that smooth out the last 10 or so readings so that the display doesn't jump around as much. Looks like a moving average to me, but the smoothing is a built in Gauge.S function, so hard to know what's going on under the hood. Also, still gonna log the raw readings just in case.

With that set, I boiled up some more water and recalibrated the sensors. Voltage offsets were small (on the order of tens of mV), but, as discussed previously, this still results in large changes in temp readings. They're not perfect, but good enough for what I need:



Then came running the wires through the firewall. These will not live in the engine bay permanently, so I did my best to make it easy to revert.





And lastly, all four sensors got attached to their chosen probe points. Used double sided thermally conductive tape and kapton tape to fix them in place and ensure accurate readings.

I attached them to the front and rear of the windshield mount, as well as to the passenger side of the firewall plug and the underside of the passenger brace (to roughly approximate the temps seen by the cabin filter housing). Strain relief box got stuffed next to the brake fluid reservoir:







Here's some very preliminary data: https://datazap.me/u/heinzboehmer/e8...ta=19-21-23-25

Cool to see how the engine bay thermocouples heak soak very quickly upon turning the engine off, but the ones behind the firewall stay pretty steady.

This was just a short drive around town at a very reasonable 16 C ambient temp. We'll have to see what temps these things log on track next week.


Also, I somehow managed to find the time and motivation to continue working on the modified cabin air filter housing.

Unfortunately, the first thing I realized is that there is no way the stock filter is going to work. My scan of the engine bay didn't fully capture the edge of the cowl, so I wrongfully assumed that the filter would just barely clear. Test fit crushed my dreams pretty quickly:






Time to find a replacement filter then!

Turns out the E9x M3 uses a set of four smaller filters instead of the big panel filter in the non-Ms. I don't know why, but I find the design so funny. Feels like BMW spent all of their brain power on designing that intake then remembered they should probably do something about the filter that no longer fits (sound familiar?) and just stuffed whatever they could find in there.

Anyway, hopefully they work out?



Well, that's a no. Passenger side is extremely close to fitting, but driver side interferes heavily with the windshield wiper arm linkage.

Did a bit more digging and found 64119237159 from the F25 X3. OE part, activated charcoal, super cheap and can squeeze in no problem:



Area of the faces of both F25 filters comes out to 45% of the area of the E46 filter. BUT, the F25 filters are twice as thick, so there's way more filter material stuffed in there than you would think. I still think they're gonna flow worse than stock, but honestly not too concerned with that.

Time to sit down and actually finish the CAD this time around.

[heinzboehmer]

Thermocouple wiring complete.

Decided on a harness with three "steps":

1. Gauge.S pigtail since getting 3.3V from the board involved soldering to it.
2. Loong harness for each thermocouple.
3. Small "manifold" harness to go between (1) and (2), that feeds 3.3V and GND to the thermocouples and also routes their outputs correctly back to the Gauge.S inputs.

I did make a rookie mistake and only ordered three wire colors. Should have ordered more so that each thermocouple output was easy to identify, but too late. I'll make some labels and attach them to each one before putting them in the car.

Everything seems good electrically, but you can see the weirdness in the ADC quite clearly:



Each thermocouple assembly measures in at 3.2m from tip to connector. Made them long so that they can reach all four brake calipers, in case I ever want to log detailed caliper temps.

Also, slope of the output on the AD8495 works out to be 5mV per deg C, so I'm betting the voltage drop from the long harnesses is messing with the calibration. No big deal, can always boil more water and change the offsets in the Gauge.S config.

Next up is routing everything to the engine bay and I can finally get data on the E86 brace windshield mount.

[ac427]

I have not, but i probably should. I just didn't want to get sucked into another long saga of trying to debug closed source firmware, like what happened with D bus init on the V5 board.

Will reevaluate when all four thermocouples are set up and I've logged some data.

I know what you mean about getting embroiled in that sort of misery. Maybe just let them know and leave the with it.

[heinzboehmer]

Have you asked Sorek about the Gauge.S readings randomness?

I have not, but i probably should. I just didn't want to get sucked into another long saga of trying to debug closed source firmware, like what happened with D bus init on the V5 board.

Will reevaluate when all four thermocouples are set up and I've logged some data.

[ac427]

Small side project motivated by a comment from karter16: https://nam3forum.com/forums/forum/m...983#post322983

I did notice some weirdness with the ADC on Gauge.S. On a multimeter, the analog output from the AD8495 is rock solid and matches the expected temp (tested with boiling water). On Gauge.S, the voltage readings are offset by 27 mV and they seem to jump around. Annoying, but the offset is easy to fix and I guess I'll just have to deal with the noise.

Just waiting for more wire to arrive in the mail and then I can build a quick harness for thermocouple power and data.

Have you asked Sorek about the Gauge.S readings randomness?

[heinzboehmer]

3D printer has been the greatest addition to my capability set of anything I can remember. I regret not doing it years ago.

It's pretty much been continuously running since I finished building it. Every freaking project benefits from it somehow. As your project above :P

Completely agree, I use for most projects as well.

[Obioban]

3D printer has been the greatest addition to my capability set of anything I can remember. I regret not doing it years ago.

It's pretty much been continuously running since I finished building it. Every freaking project benefits from it somehow. As your project above :P

[heinzboehmer]

Rest of the print was waiting for me at home after work. Works great!

[heinzboehmer]

Small side project motivated by a comment from karter16: https://nam3forum.com/forums/forum/m...983#post322983

Basically, I don't have real temperature data for that area in the engine bay. I've only ever measured with an IR temp gun, which is not super accurate, as you need to park the car and get out to measure. Ideally, there would be a set of thermocouples attached to different surfaces that can be logged continuously.

So, that's exactly what I did.

Resisted the urge to make everything custom and went mostly off the shelf for this one. Used the following parts:

• Adafruit AD8495 breakout
• Adafruit Type-K Thermocouple with Stainless Steel Tip
• M2 x 3mm Threaded heat set inserts
• M2 x 4mm Screws

I was worried about breaking the solder joints on the board with everything just hanging out, so I designed a quick enclosure with some strain relief features. I added some mating features to the sides of the enclosures to make them modular. For now, all four thermocouples will be places in roughly the same area, but it would be nice to be able to run each to different parts of the car without having to redesign the enclosure.



Here's all the parts for one enclosure laid out:



And assembled for testing on the bench with Gauge.S:





It works!



I did notice some weirdness with the ADC on Gauge.S. On a multimeter, the analog output from the AD8495 is rock solid and matches the expected temp (tested with boiling water). On Gauge.S, the voltage readings are offset by 27 mV and they seem to jump around. Annoying, but the offset is easy to fix and I guess I'll just have to deal with the noise.

Just waiting for more wire to arrive in the mail and then I can build a quick harness for thermocouple power and data.

[heinzboehmer]

Aaaand documented to the present day: https://nam3forum.com/forums/forum/m...fit#post321060

[heinzboehmer]

I have a DIY router with a steel frame and Mitsubishi servos which handles 6061 pretty well, so I'd be tempted to make a fixture and see what I can do - might be good enough. I'll keep an eye on the thread for if/when you go public with the models.

Oh nice! You should definitely give it a shot then. Would love to see how you approach the machining.

There's a couple surfaces (the sides and the bottom rear of the flat) where flatness is not at all critical. You could leave a good chunk of stock there for rigidity and fixturing, machine all the other features, then hold from the sides and machine the bottom surface flat.


Edit: I have no CNC experience, so this might be dumb, but this is roughly how I would approach the machining on a three axis. Excuse the crudeness of the images.

1. Prop the stock on some angle blocks (blue), then machine the angled surface (square end mill) and pocket (ball end mill). Surface with the hole gets bonded to the chassis and needs to be fairly flat, so prefer to do it this way vs ramping to avoid interpolation artifacts.
   1. 
   2. 
2. Reposition stock + blocks, drill out hole and clean up draft in the front clearance feature.
   1. 
3. Lay stock on its side and machine one of the pockets (ball end) + part of the surface (square). Surface gets bonded to chassis, so same comment about interpolation. You might be able to get the entire surface from this side, depending on the capabilities of your machine.
   1. 
   2. 
4. Lay stock on its other side and machine the other pocket + rest of surface if necessary.
   1. 
5. Lay stock flat and machine/slit large chunk off of the rear. This top surface also needs to be flat for bonding.
   1. 
6. Grab part by the sides in vise (red arrows), support large flat cantilever with something (blue arrow) and machine the bottom surface. Only the front part of this surface (where the three holes are) needs to be flat, so deflection in the rear is not a huge deal. This setup should result in a fairly flat surface though.
   1. 

So yeah, fairly involved process, but totally doable.


Edit 2: While we're at it, here's how I would approach the smaller piece. Actually seems easier than how I imagined it yesterday. Again, excuse the crudeness.

1. Start by drilling the three holes for the mounting hardware.
   1. 
2. Prop on angle blocks and drill holes for studs + bore out clearance for stud heads.
   1. 
   2. 
3. Flip over and prop on another set of angle blocks to machine surface that the E86 braces bolt to.
   1. 
4. Lay flat and let the CNC do what it does best. None of these features are too critical, they're just there for brace and fastener clearance.
   1. 
5. Bolt the part down to the work surface using the three mounting holes and clean up the perimeter.
   1. 

[ethan]

That being said, no "machining-as-a-service" company is gonna bother spending the time to do these on a three axis. They're just gonna throw it at the five axis and charge you for it. Really only worth attempting the three axis if you're doing it yourself (or your buddy or whatever).

I have a DIY router with a steel frame and Mitsubishi servos which handles 6061 pretty well, so I'd be tempted to make a fixture and see what I can do - might be good enough. I'll keep an eye on the thread for if/when you go public with the models.

[heinzboehmer]

This main mounting is a good candidate for a forged carbon part and cost about $400 in materials to make. The small stud plate is more difficult. Some of the details are a challenge. Might be worth modifying the part to make a carbon part. I think it would make sense to keep the small stud plate aluminum especially in a street car where the mounting needs to be 2 pieces.

Can definitely modify the smaller part to make it more composite-friendly, but it would require some work. It's currently optimized for weight and strength in aluminum. Machining complexity be damned lol

Can it all be machined 3 axis?

The part that gets bonded to the chassis, absolutely. You would need some clever fixturing and a couple setup changes, but it can be done.

The other part is maybe still doable on a three axis, but getting the bearing planes flat would be fairly challenging.

That being said, no "machining-as-a-service" company is gonna bother spending the time to do these on a three axis. They're just gonna throw it at the five axis and charge you for it. Really only worth attempting the three axis if you're doing it yourself (or your buddy or whatever).

[ethan]

Can it all be machined 3 axis?