I did some more work on the hub on Sunday and got it running again
although not on the bike yet
To get the old hall sensors out I found I had to use my heat gun on low because high(1500W) heated the magnets on the rotor quicker than the epoxy holding the sensors and I did not want to cook the magnets or their glue so on low and by continusly turning the rotor I manager to keep the magnets at a safe temperature while softening the epoxy just enough to scrape some off the top and then leaver the old hall sensors out.
The dead one was in 2 pieces due to its encounter with phase volts which dosen't bode well for the 6 FET controller it was attached to at the time. Once all three sensors were leavered out they left nice pockets in the origional epoxy in the correct shape that just needed a small clean up with a screw driver while warm and then the new halls slipped in snug. I used SS413A hall devices.
I checked the halls with clip leads before epoxying them and the PCB down. I then soldered them to the PCB and tested again through the main cable and one was flakey.
Turned out I got some epoxy through the hole in the PCB when I pushed it down before soldering that I needed to scrape before I got a good solder join. I had been suspsious because solder didn't flow right but the lighting was poor and I let it slid knowing I would give it a good inspection and test before reassembly
I have replaced all the hall sensors, it was easier to replace all three because the legs of the good two were in poor condition after removing the PCB and would have complicated reinstalling the PCB. STraight legs make the job of reinstalling the PCB much easier.
This video shows a test run on the bench clamped in a vice before gluing the wires down and sealing the PCB.
The power supply is in series with a Lipo brick to get enough volts because my other supplies were in use. The Cav3 was not calibrated to the shunt of the controller used but the Amp meter on the power supply is the correct current.
[youtube]KFPHGOdcopY[/youtube]
You can clearly see the movement of the hub arround the key under acceleration.
I have since reassembled the rest of the motor and glued the wires down but have not test ridden it yet because I want to be sure the epoxy is well set and not going to let the wires come in contact with the inside of the side cover.
You can also see a slight wobble i the rotor. It dosen't seem to be perfectly round although It works fine.
The key on the clutch side made much better contact with the freewheel than the hub makes with its key. I actually had to encourage the freewheel back into place with a few gental blows of a hammer.
For now I have reinstalled the old clutch because it had only slipped a couple of times on really cold days when attempting to start off with lots of torque (within the default limits of the 6 FET controller though). I will run it till it dies or becomes a pain as I do have a spare I can swap in quickly now I have all the tools I need to tear the motor apart.
Forgot the temperature sensor I used is actually an LM335 the 1% version and it gives acceptable readings with the gain and offset values calculated based on the datasheet. The 100degC limit might be an issue but I don't want to push this motor too hard anyway. One thing to note if swapping an LM335 for the origional LM35 is that the LM335 only needs 2 wires, you can see I left the 3rd unused and will probably cut it off since there is no need to trim the accuracy in hardware. the LM35 needed +V, 0V and output.