More updates:
This is what the shelf looks like under the MCPCB. It’s covered in thermal goo:
The driver comes out with a bit of force from the inside, and it still works afterward when held in place only by pressure. Here’s what it looks like:
To get the driver out:
- Unscrew and remove the bezel.
- Pop the optics out. Hopefully it isn’t glued to the MCPCB, but if it is, this step could be tricky.
- Unsolder the red and black wires.
- Pull the MCPCB out.
- Carefully push on the driver with a plastic rod like a spudger.
To get the driver back in, you may need to file down the edges of the driver a tiny bit. I did, because the glue around the edge made it slightly too wide.
The driver is 22.3 mm in diameter.
The shelf under the MCPCB is 4.5mm thick. :+1:
I think it may be giving me the finger, to protest being separated from its home.
Good news about the anodizing. I accidentally scraped part of my light with a bastard file while filing the edge of the driver, and after wiping the driver dust off I can’t even see where I scraped it. So the anodizing seems pretty tough.
Also, Narsil works on the D4 without any code changes. I flashed NarsilM.hex from the repository, and it just works. Moon appears to be significantly lower, too — about the same as my ZL SC52, or 0.1 lm. I haven’t actually measured it yet, but it looks pretty close.
However, runtime on moon isn’t dramatically longer. The driver seems to use about 4.5mA even without the LEDs on, so the absolute longest it could possibly run is about 28 days on a 3000 mAh cell.
I did a very low-tech runtime test to see what the curve looks like and make sure LVP works. To speed things up, I used an 18350 cell and did not recharge it first. I measured the output at 285 lm in my light box, then moved the light over to my runtime test setup. Note, the battery was not full; I had been using it since the previous evening and I think it was at about 3.9V or 4.0V. It didn’t really matter for this purpose. I also bumped the table a bit during the test, which made some small bumps in the graph.
What this shows is:
- LVP works. It goes directly to moon in one step though, instead of doing a more progressive step-down.
- Output on this mode, which was about 350mA on the 7135 chip and 350mA on the FET, has a gradual decline as expected.
- I was too impatient to wait for the light to turn off completely, so I ended the test when I noticed it had stepped down to the lowest level.
My “test setup” was pretty ad-hoc. It looked like this: