Dual Cree Bicycle Headlight

It's also made possible by using the newest LED technology. LEDs have just recently surpassed all other light sources (including flourescent) in terms of efficiency. My light is probably brighter than any mainstream commercially available headlights of similar power (and much higher cost), even commercial LED headlights. Commercial vendors have to lock in one part and use that part for months, whereas you can order the newest brightest LEDs -- and this market is changing rapidly. For comparison, the 5W vistalite, which uses a halogen lightbulb puts out light that would fall between my light's lowest two levels (and my light consumes 0.6W and 1.3W at those two light levels. At it's brightest, I'm probably about 1/3 as bright as a car headlight, and should have a runtime of almost three hours.

I made my case from a hammond 1590G, cut in half (enough left over for another light head!). I hand-carved an aluminum heat sink to fit the case, with holes drilled through top to bottom for cooling, and a center hole for a mounting bolt. This bolt also serves as an additional bolt for closing the box, as it would only have two screws at one end after being cut in half. It will thread through a vistalight mounting bracket on the bottom. The case has matching cooling holes top and bottom.

I'll be gluing the heat sink into the bottom half with thermal epoxy, and sealing the lid on with silicone. I could have (should have) ordered the 1590WG which has a waterproof liner on the lid, and would have reduced the amount of sealing I need to do.

I tried to turn the the drilled thru-holes into a channel, and it didn't go well (I was trying to cut too much at once), so I didn't try it on the other side.

As it turns out, the inside dimension of the case is 1.6 mm shorter than the Ledil optics I'll be using. I've trimmed the optics down just a bit to fit.

There's also a hole for the tiny NKK illuminated switch I'm using. The hole is 2 mm diameter. The recessed area will let me glue in and seal the plastic switch bushing, which is 4mm diameter. The remaining wall thickness here is about 0.6 mm.

You can see this hole in the (blurry) picture above of the case, in the back on the left side.

These Cree leds are much smaller than you think they are from seeing blown up pictures.

These pictures were taken after making the first corner cut. This is necessary because the led has little copper vias at each corner to connect the upper pads to the lower pads, and the Arctic Silver thermal adhesive I'm using is somewhat capacitive. I could probably have skipped it but I'm paranoid. I could also have used Arctic Alumina thermal epoxy but it's only about half as heat-conductive as the Silver. I could also have scraped off the bottom tracks, but the copper vias would still be exposed, and again, I'm paranoid.

Snipping the corners with wire cutters turned out to be quick and easy, and did not shatter the boards. If you cut right at the dots on the corners, you'll probably expose the via without cutting it off, so go just slightly past them.

Only work on these things if you have magnification. I had a head-mounted lupe that I was using. I also used it to take the closeup pictures here, by holding the lupe up to the cell-phone lens.

Click for a larger image.

This is the heat sink, after gluing on the LEDs. I've also formed the wires and put them in place, but they have not been soldered yet. Gluing was done with the thinnest layer of epoxy possible. The ledil optics were held in place over the LEDs, to get the position right, until the epoxy was holding them. Careful not to glue the optics to the heat sink!

The heat sink looks to be a funny shape because of the pincushion introduced by using a cell phone and a lupe to take the picutre. I did get the focus just right on this one, so it looks good enlarged.

I'm using 24-gauge solid copper wire. Actually, I had some radio shack intercom wire with red, green, black, and white joined together in a strip, so I used that. If the insulation looks oddly thick, or transparent on the surface, that's why.

Click for a larger image.

And here after soldering.

Gluing first was a terrifically great idea. I have no idea how I could have held everything in place otherwise. The heatsink made everything stable. I used a pair of vise grips to hold the heat sink, just to make everything that much more stable.

First, I tinned the wire leads, and also the solder pads on the leds. I taped all the wires in place (with painter's blue tape) for soldering. I made sure that the wire leads were perfectly placed to be in contact with the pads for the full length of the pad. I added a tiny bit of flux to everything with a toothpick. Then a couple of seconds with the soldering iron got everything to flow together. I had to redo one joint because the wire had come up on one end when I soldered it.

And here, with a minimal circuit test. I hooked the two leds up to four partially charged NiMH cells, so about 4.8 volts. In real life, this was even more feeble, and you could see a grid of illuminated lines inside the led dome.

Click for a larger image.

Now I've soldered on the bFlex, and also soldered the illuminated pushbutton onto the bFlex. The pushbutton gets it's own board made from a standard radio shack PC board. The black wire going to SWB doubles as the ground for the LED, hence that jumper wire. On the other side of the LED I needed about 30 ohms, but hadn't (correctly) calculated that initially and didn't have the right value on hand, so I used two 62 ohms in parallel.

FYI, the bFlex puts out 2.5V for the status led, expected to drive at 20 mA. The NKK switch optimally wants 1.9V at 20 mA. so R = (2.5-1.9)/0.02 = 30 ohms.

A closeup of the backside of the pushbutton board. Someday I'll learn to solder.

Click for a larger image.

Here the same assembly is in the case, with the pushbutton more or less in place. The bFlex will stay at that odd angle, because it's too tall to stand upright inside the case.

I need something to brace the pushbutton board in place so that when I push against the button it pushes against the heat sink (which will be glued into the case with thermal epoxy). Two layers of corrugated cardboard should do the trick.

Click for a larger image.

This is what the button looks like when it's in it's hole.
You might notice that I forgot to drill the holes for the battery cables, so work stops until I can find the time to do that (I live in an condo building, and can't really drill late at night).

So I have drilled the holes for power now, and hooked up my battery pack. Everything is soldered together and it all works. YAY! I need some pictures of the assembled inside of the light. Maybe I'll get to that someday. The parts are a very tight fit inside, all bunched up wires. Where the

One of my battery holders fits inside a water bottle (the 2x2x2 holder). But since I don't want it to rattle around, I cut a couple of foam spacers to keep it softly in place inside the water bottle. The batteries fit nice and snug.

I don't have a clear picture, but this battery holder happened to have two holes at either end, and the wires fit through perfectly. So for the purposes of strain relief, the wires go through the battery, come out at the terminal end, are tied in a UL knot, and then soldered. I could probably swing this thing around my head with batteries loaded, and not hurt the solder joints at all.
And the bottle is ready. The whole through the foam has just enough clearance that the battery holder slides in and out with only a little resistance, but it's still held comfortably in place.

Here's the final assembled light. The handlebar bracket comes from the Vistalight, which is also pictured here for scale. This is a pretty small light.

The bottom of a vistalite has a couple of raised metal spots which fit inside a plastic piece in the handlebar bracket. This keeps the screw from tightening or loosening when you swivel the headlight. I could add this feature to my light, but so far, it doesn't look like it will be a problem, so I'm not bothering.

You can see the influence of the bikecurrent mailing list, where the advocated for using a pigtail instead of a barrel connector. The connectors are are AMP powerlock, which are Anderson Powerpole knockoffs. I managed to crimp them with the crappy crimper/stripper tool that is everybody's first crimper/stripper. But I also soldered them. Note that once you slide the read and black (or colors of your choosing) AMP powerlock connectors together next to each other, they don't come undone without breaking a little plastic doohickey. This is not a big deal if you simply got the sides wrong, because then you just use the other unbroken side.

Might as well mention here that the other end of the pigtail has a UL knot inside the case before it's soldered, so it should be nicely strain-relieved.

You'll notice on the picture of the underside that the handlebar bracket actually covers a couple of the ventilation holes. The holes are all connected on the inside, so things should still be fine.

I have another battery holder in the works. The battery holder used here is the more typical 2x4 configuration. I found this great Ape Case at Microcenter that's designed to hold four batteries and some flash cards. It's a tight squeeze, but I can get this battery holder inside and have a great battery pack that's easy to open to change the batteries, and even marginally water resistant without a ziplock. It's got a belt loop on it, which I'll use to add a velcro strap to velcro this wherever I need it.

Summary

The bFlex is fabulous. I love it. I have the UIB2 interface, and I've got it configured for 5 light levels to a maximum of 1000 mA. The light levels are all distincly different from one another. And I'll guess that the dimmest one is almost as bright as a 5W Vistalite (I'll actually test this if I can get the Vistalite to work again). Here's the measured voltages for the five different light levels: L1=5.8V, L2=6.3V, L3=6.6V, L4=7V, L5=7.5V.

Assuming 8 2650 mAh batteries (which is what I bought), and 88% efficiency, this is my expected run time at the light levels that the bFlex produces:

mA	V	W	lm	battery A	t (hours)
100	5.8	0.6	80	0.07	38:35
200	6.3	1.3	150	0.15	17:46
350	6.6	2.3	214	0.27	9:41
700	7.0	4.9	340	0.58	4:34
1000	7.5	7.5	440	0.89	2:59

The LED voltage to current for the last three comes straight from the data sheet. The current for the two lower voltages was roughly estimated from a graph in the Cree datasheet. The lumens are also estimated from a graph.

In practice however, eight cells cuts it close on regulation. As the battery pack voltage drops, the bflex no longer has enough cushion to keep the light in regulation. With fresh batteries that measured over 11V, I powered the light at full power. Initially that battery voltage with load was 10.76V. After an hour, this had dropped to 9.86V. I think things will go out of regulation around 1 to 1.5 volts above the load, so around 8.5 to 9 volts. At this point the battery still has lots of juice left in it, just not enough for full power. That's a good argument for a 12-cell pack, which should stay in regulation until the pack is drained.

In theory, I can hook up any battery pack that's over 8.5 volts (9 volts is probably more ideal). I could hook up 16 cells for longer run time. Or 3 Lithium cells should be enough to run it (four would be a safer bet for the same reason that 8 NiMH cells cut it close).

Indoors, this light gets HOT. Hot enough to shrink some heatshrink that I used inside the case to shield the bflex from the aluminum case. Hot enough to burn my fingers if I hold it too long. Outside in the cold, with wind from riding a bike, it isn't a problem. I'm not sure how it will work out in the summer nights. But I'll be using the temperature limit option in the bFlex. If I had it to do over, I'd seriously consider my custom heat sink in a plastic case. Then the entire case wouldn't get so hot. I think I'd add fins to the heat sink top and/or bottom since I wouldn't have the case to distribute the heat.

Parts List

You can do the math, but it comes in somewhere around $180, which is a little cheaper than commercial dual led lights, and this should be brighter than those, and is all-around better (brightness, run time, light levels, low-battery warnings, overheating, charge time). Also, I really pushed the price up with my $40 charger, which is not at all necessary. But be aware that there will be additional cost in tools: soldering iron, solder, flux, wire cutters, wire stripper, crimpers, magnifying glass, saw, drill, drill bits, and more. If you've already got stuff lying around (tools, batteries, charger), and you're frugal, you can probably get this down close to $100. On the other hand, if you need all new tools, you could easily spend more than a new commercial light.

Also a large amount of time. At least twenty hours, realistically fifty or more. Most of your time will go into designing and building the packaging for your light.

How I carved the heatsink

The rockler vise is nice. You can adjust for tighter, harder, more accurate action, or looser, floppier, faster action. Even at it's best adjustment, there's still some slop, and an eighth turn of backlash, but most of the errors I get come from the stand and the tool, not from the vise.

But, as crappy as it is, it lets me do things both more accurately and more quickly than I could without it. Without the vise, I could probably do things fairly accurately, but setup times would be significant, especially things like drilling multiple holes in a row, or machining a smooth edge.

The limited reach and clearance of the dremel makes it hard to mount things and make some cuts, but on the other hand, the dremel workstation lets you orient the dremel in different ways, which helps this problem somewhat.