The jury has deliberated and the verdict is in. KD sku S020121 flat out sucks. I potted it today, using Alumina powder and Devcon 2-ton epoxy, and I don't think I or anyone else could have done a better job. Here's the procedure, followed by test results:
Procedure
- cut a sleeve of Aluminum sheet the exact circumference of the interior of the Mag D neck
- with driver already wired very closely to top of Mag switch, seal all areas on top of the switch with kapton tape where epoxy could drip down
- mix equal parts of the epoxy with so much Alumina powder that it couldn't possibly pick up any more
- while rotating the driver on its side, pour the toothpasty concoction into the space between the two boards of the driver and paint its bottom for good measure
- wrap the Aluminum sheet around the top of the Mag switch and seal around the joint with kapton tape
- coat the inside of the Mag neck and the outside of the Al sheet liberally with thermal paste
- force the switch, driver, and Al sheet collar into the neck with the heatsink
- pour in the homemade AAA concoction until the driver is completely submerged
- push the heatsink down the rest of the way until the switch is in position
- turn the switch on and test the + and - in contacts with a DMM for a short before the epoxy cures
- if a short is found, pull the HS up a bit to stop the inductor coil from shorting on the bottom of the HS and tone test again
- wait several hours for the epoxy to cure and test the much improved heatsinking
Test Results
With gummy sponge things provided by KD:
0-1 min, 3.1A draw at the tailcap
1-1.5min draw at tailcap quickly plummets to ~1.1A
Fully potted with homemade AAA:
0-2 min, 3.1A draw at tailcap
2-3 min, slow descent from 3.1A down to 1.8A draw at tailcap
test ended at 3 min. neck was warm to the touch, but not hot
Conclusion
Potting helped somewhat, but still did not keep the driver from throttling back after a very short period of time. Combined with metal to metal contact of the host to the hose clamps to the bicycle handlebars and all components air cooled by the ride, it might have been enough to keep the driver from throttling back to less than direct drive, however...
- Potting was expensive. It cost $8 for 4oz of Alumina powder and $5 for Devcon 2-ton. Granted, that's enough to do this about 5 times.
- It is very likely that I might never get all the components out of the host undamaged. Certainly it won't be easy. It's almost certain that I'll have to dremel cut the potted driver from the switch housing.
- NONE OF THIS SHOULD BE NECESSARY! Even if the results were that it worked perfectly, it's crazy to have to go through this trouble to get the driver to work as it should in the first place.
Not convinced that the potted sleeve was making the best possbile contact with the neck, and also to prove to myself that I could swap the DX 57779 driver back in, I removed it, wrapped the sleeve in foil and did some more testing. The results enough of an improvement for me to keep the driver, but in my estimation, not worth the effort.
Full details posted in the bike light build thread
here.