Modifications to the BISY-FL headlamp assembly

My primary concern with performing this modification related to getting heat out of the LED emitter. Most of the heat from the light bulb is radiated out the front of the lamp (per Andreas Oehler, who should know) while the heat from the Luxeon I LED has to be removed via the thermal slug at the bottom of the emitter. Refer to the Luxeon data sheets for details on thermal considerations.

It seemed clear that if I could epoxy the emitter to the end of an aluminum rod and then get the other end of the rod out to a heat sink, that should do the job since aluminum is an excellent heat conductor. Fortunately, the design of the BISY lamp helps in this initiative: the removable cap on the rear has a conical taper inside which is precisely centered over the socket in the reflector. So all you have to do is drill through the end cap from inside, using the conical taper to guide the drill bit, and you will have the rod precisely aligned with the center of the reflector!

NOTE: If you perform this modification the same way I did, you will not be able to go back to the standard halogen bulb setup without considerable difficulty. It would be less nuisance to just buy a stock new BISY-FL lamp assembly.

Sorry about the poor focus, I haven't quite mastered the close-up function of the digital camera...

I started with a length of 1/4 inch diameter aluminum rod readily available at Home Depot and elsewhere, since its diameter was a reasonable match for the heat-sink slug at the bottom of the Luxeon I emitter body. I cut it to a length which seemed ample to project into the reflector a bit and also project out of the removable rear housing of the BISY lamp. As it turned out, on this first unit I cut the rod about 1/4 inch shorter than appropriate and could barely get the heat-sink plate on. On the other three lamps I made the rod a little longer.

I don't have a metalworking lathe, so I had to improvise in several areas. Here's the first one. To face off the ends of the rod so they'd be square, particularly important for the emitter end, I chucked the rod in the drill presss and brought it down lightly against a mill file. Keep the file moving during this process or you'll end up with a bump in the center of the face.

Next I drilled and tapped a 6-32 hole in the end you can't see. This accepts a screw which holds the heat sink in place. I carefully filed four opposing flats on the emitter end of the rod. These are to provide clearance for the wiring to the emitter. Filing V-grooves would likely have been simpler but I didn't think of it. Three flats would have been mechanically more sensible than four, but the emitter has electrical leads 180 degrees apart so four flats were appropriate and also simpler to get right when filing by hand. The size across the flats for my setup was about 0.210 - 0.215 inches but the size you'll need depends on the outside diameter of the insulation on the feed wire you're using.

Now you need a way to adapt the heat-sink rod so it will center itself in the reflector socket where the Philips HPR64 (3W) or HPR60 (2.4W) halogen bulb normally fits. I was intrigued to discover that a standard PR-3 flashlight bulb has exactly the same flange and base size as the HPR bulbs, so a sacrifice was in order. Here you see the flashlight bulb with the case cut through at the base using a Dremel tool with a grinding disk mounted. These flashlight bulbs are a lot tougher customers than I realized, the material inside is mostly glass and impervious to grinding. So after you cut around the base you carefully mount the bulb in a vise and use a hammer and punch to drive out the innards. This sounds a lot easier than it is, since you can easily bend the housing beyond the point of usefulness. Of course if you could turn a heat-sink rod on a metal lathe from larger-diameter stock and leave a flange of the proper size and location, this sort of improvization wouldn't be necessary.

You eventually end up with a usable housing like the view shown above. This one has been cleaned up by filing and careful use of needle-nose pliers to straighten out the flanges etc.

Now you take the heat-sink rod and your feed wires and a couple other pieces of wire, hold them against the flats of the rod, and install the housing. This makes it sound easy, but the first time it definitely isn't -- and it doesn't get much easier with experience. The challenge is getting the width across the flats correct so that the housing fits snugly (but not too snugly) in place over the supply and spacer wires. This takes some trial and error, and would be avoided with a turned heat-sink rod of proper dimension and having a couple slots for the wiring in the flange.

Here's another fuzzy picture, sorry about that but the best I could get. The rod is held in the vise and a piece of paper towel hides it from view.

The housing pressed onto the end of the heat-sink rod, over the green feed wires and black spacer wires. I threaded the feed wires through lengths of heat-shrink tubing so I could push them in and out after the housing was mounted. The spacer "wires" are actually the insulating jackets of somewhat larger wires with the wire removed so they'll squash down when the housing is pressd on. You want a pretty tight fit. After the housing was in place I trimmed the "spacer wires" closely at both ends of the housing.

The yellow band is a piece of heat-shrinkable tubing. Probably unnecessary but I put it on due to concern about potential for a short circuit between the connection to the emitter leads and the heat-sink rod.

Several steps led to the photo above.

First I trimmed the leads on the emitter by cutting off the wide part and bending the remainder carefully downward so it's close to the emitter housing. Use care, the leads aren't well supported inside the emitter package. It's a good idea to support the leads close to the body while bending them downward.

Next I epoxied the emitter in place using Arctic Silver thermal adhesive, with the heat-sink rod held vertically in the vise. Other thermally conductive epoxy could be substituted. Ensure the leads on the emitter are lined up with the supply wires, and that the emitter is reasonably centered on the rod and pressed down squarely. You need only a small dot of epoxy.

Since the adhesive is electrically conductive and the thermal slug in the base of the emitter is not electrically isolated, the heat sink will have some electrical potential and needs to be electrically isolated. For my application that isn't an issue because I won't use the bicycle frame as an electrical path and the heat sink won't be connected to the power source or touching the supply wires.

The Luxeon I emitter is a LXHL-DW01, a sideways emitting diode with 8000K color temperature (looks decidedly purplish to my eyes, similar to the HID lamps on some cars). I didn't have a choice of brightness bin; the bin code supplied by Future Electronics was QYAJ. The five I bought all came from the same bin so I can't comment about comparative brightness. Remarkably bright however, at least to one not accustomed to the current generation of high-intensity LEDs. And yes, the light does come out sideways, most of it seemingly over a range of 20-30 degrees. Pretty sharp pattern.

After the epoxy set up, I stripped the ends of the feed wires, isolated one of the seven strands in each wire, and trimmed the other six strands back. I then pulled the feed wires back until their insulation was just a bit above the level of the housing flange, bent each of the single strands into a C-shaped strain relief, cut each to length and soldered each to the adjacent lead on the emitter. Use a small iron, flux and a MINIMAL amount of solder. You don't want the solder getting under the LED lead and up against the emitter body, since that will eliminate the strain-relief characteristic of the bent lead. The first unit I made failed shortly after I initially ran it, and I suspect but couldn't prove that it was due to making the supply connections using solder-stiffened #22 wire to very delicate emitter tabs.

Please note that because the housing flange fits inside an aluminized-plastic socket at the base of the reflector, the feed wires have to be kept close to the emitter body to avoid a risk of shorting to the plating on the reflector.

Then I pulled the rod back a little in the bulb housing, so the "crease" in the lens of the LED emitter was somewhat below the level of the filament in the halogen lamp, put the assembly in the BISY lamp, powered it, pointed it at a wall in a darkened room, and gently pushed the rod-and-LED assembly inward until the pattern seemed optimum. The pattern won't have the sharp edges and well-defined brightness stripes it had with the halogen lamp, but it will look similar. See the pictures in the beam-pattern section of this site. As you get the LED out of position (too far in or too far out), the pattern will get larger and fuzzier.

Once the pattern seemed about right (and the position is not terribly critical), I carefully removed the rod-and-LED assembly and using my fingers squeezed in some RTV from the non-emitter end to secure everything in position. Messy job. I also put a little RTV over the supply leads to the LED as insurance against shorting to the inside of the reflector. I didn't have non-corrosive RTV (the type which doesn't smell like vinegar) but non-corrosive is the type which should be used. I'm hoping problems don't show up later due to taking this shortcut.

As for the previous picture, there are several steps to get to this stage.

Simplest first: The heat sink is a piece of 1/8 inch thick aluminum plate about 1.5 x 2 inches overall. Nothing magic about the choice or size: I had the material on hand, the dimensions seemed adequate, and the plate was thick enough that I could bore a 1/4 inch hole about 1/16 inch deep that would fit snugly over the end of the heat-sink rod and transfer heat decently. A little heat-sink compound (I used Arctic Silver Ceramique but others would work OK) provides good thermal coupling from the heat-sink plate to the rod, and the screw secures the plate..

The rear cap for the BISY lamp took considerable work and I may have chosen an unnecessarily difficult approach. I decided I didn't want any of the existing electrical connections and that took extra work. First I unsoldered the lead of the bi-directional zener diode from the spring and removed the spring. Then I pried out the various metal contact straps which could be pried out; the other lead of the zener diode is soldered to one of these. Sorry about the ambiguous description, I did the modification a couple months ago and have misplaced the removed parts so my memory is hazy on the details. The difficult part is removing the connection which is part of the attachment tab. It's held by a swaged-on sleeve and the swaged-over end has to be ground out in order to release the parts. Seems to me this really wasn't essential, I could have clipped off the part which interfered with the heat-sink rod. A Dremel tool with grinding bit is indispensable to grind off the end of the sleeve, but do it in stages so you don't heat the parts so much they melt the plastic. Eventually you can push out the metal sleeve and remove the contact.

I mentioned earlier that the rear housing cap has a conical area inside which serves as an ideal drilling guide. Keeping the cap square with the drill bit, drill a hole appropriately sized for the heat-sink rod. A close fit is desired so the heat-sink rod is securely held. I applied a little silicone grease to provide a degree of waterproofing.

The wires will fit under the base of the spring, in the slot formerly occupied by the bulb-contact straps.

The picture above (finally in focus, hurrah!) shows the rear-cap assembly all put together and ready to snap onto the back of the BISY lamp. When the cap is in place the spring is compressed and the heat-sink rod sticks out the rear far enough to attach the heat-sink plate (but just barely). For the next three units I made the rod about 1/4 inch longer.

I was curious whether the heat sink plate would be adequate so I did a rough-and-ready test to check. Here you don't see the thermistor, which is temporarily electrical-taped to the heat-sink rod just below the flange. The thermistor is connected to an ohmmeter.

Here's the test setup: A chunk of soft foam with a slit cut in it, so the emitter and thermocouple are inside. Not a great test but not bad either. The little circuit board at the top is the TaskLED constant-current regulator. The test was run as you see it here, with the pink foam surface horizontal and in still air at about 70 degrees F. The TaskLED regulator was set to provide 350mA. This amounts to about 1.3W delivered to the LED.

I left the power applied for 15 minutes, by which time it seemed clear from the ohmmeter readings that the temperature was reasonably stable. Then I turned off the power and monitored the temperature for an additional five minutes. Plot courtesy of Excel. Overall temperature rise was about 15 degrees celsius, and that was with no significant convection or airflow over the heat sink. I'll hazard a guess that with free-air convection or some circulation from bicycle movement, the rise would be in the region of 10 degrees Celsius.

When the lamps were fully assembled I found that the heat-sink plates were barely warm to the touch after a few minutes operation at 350mA so I conclude that temperature rise is not a problem.

Here's how two of the converted lamps look when fully assembled. The one on the right is the first one which has the shorter heat-sink post. It almost touches the mounting tab.

It's true that you have to remove the heat-sink plate to remove the screw which secures the rear cap. On the other hand there's no obvious reason why that should be necessary since barring mechanical failure or water damage it's not clear why you'd need to open the rear cap. I put a little silicone grease on the heat-sink post where it goes through the rear cap, to discourage water seepage. But there's only a few thousandths clearance at most so I don't expect a problem.

Please see the section on beam patterns to see how the converted lamps compare with the halogen original version.

Return to Converting a BISY headlamp to LED