It occurred to the lion kingdom that the use of self tappers as load bearing farsteners has been a disaster of biblical proportions. The latest thinking was to use captive 4-40 locknuts & hot glue to farsten the locknuts. The new motor module still managed to have acceptible tolerances with the .8mm nozzle.
The hot glue actually seemed to work. If it got in the way, it managed to be compliant enough for the screw to not push it out. The captive nut walls are still too thick. The nuts aren't going to break PLA. It's going to compress.
The mane unknowns are if the 4-40 bolts are going to stay on as well as the self tappers did & if the motor will find another way to stall when it overheats or if it'll just destroy itself.
After 3 miles with the locknut motor plates, the right motor let out the magic smoke. It manifested itself as a loss of torque despite accurate commutation. There was no evidence of the encoder failing, the locknuts getting loose, a stall, or debris. The motor just stopped providing torque & overheated after successfully ascending part of a hill.
Decided to rewind it over the next 4 hours & see if the problem happened again. Sadly, the 32 AWG got tangled. If 32 AWG gets tangled, the only solution is to cut it off & throw it away. It can't be untangled. Ended up doing 2 of the fields in the last of the 26 AWG & ordered another spool of 26 AWG since there is going to be a lot of magic smoke.
What's probably going to happen is a switch back to self tappers. There was no evidence of the locknuts causing the failure, but they were the only explanation. The best explanation is the rebuild was slightly off, causing the encoder to miss just enough to suck more power but not stall. The overheating would have caused the motor to stall before burning up, with self tappers. The locknuts allowed it to overheat without stalling until it died.
Chinese direct drive motors don't have these problems because their sensors are epoxied into solid steel & their motor mounts are solid steel to suck away any heat.
So 8.7 miles with the latest motor rewinding sucked a whopping 190mAh/mile. Solid core 26 AWG might actually be more efficient than quad 32 AWG. The best explanation lions have is the skin effect. In a high frequency PWM application, skin effect causes more resistance as wire diameter decreases. Laminated quad 32 AWG has 4 times more surface area than solid core 26 AWG, so the quad 32 AWG would have a lot more resistance than solid core 26 AWG. You'd think it would matter more for kilometers of wire & megahertz than a motor winding.
It's become clear that the motor encoder is getting knocked out of alignment & the reason probably isn't loose cables snagging it. Lacking any reason, the idea of farstening thermisters to the motors is gaining traction. In 1985, thermisters might have been available in every corner Radio Shack, but in the future they're a little hard to come by.
Daily repair work turned to a failure of the happy meal controller to charge anymore. It had salt water damage around the battery management chip, so the latest clamshell still isn't repelling water.