There is I believe another factor to the problem. There was a change made by both SSC and Cree to the
Lambertian dome. I took some interpolated data from both Cree and SSC's pdf files that describe their emitters. SSC lists this data in graphs called, "Typical Dome Type Radiation pattern". Cree calls it "Typical Spatial Distribution" for the XR-E and "Typical Spatial Radiation Pattern" for the MC-E. The frame of reference here is how far off axis you have to go to get 80% of the axial illumination for each type of emitter. Something totally arbitrary to help show the change from single to quad die emitters. These are visual guesstimates based on the graphs.
Cree XR-E= 27°
Cree MC-E= 36°
SSC P4= 38°
SSC P7= 47°
So the net result of the change to the dome is more of the LED's light is being directed towards the sides. That's why scaling the reflector to match the die wouldn't be enough to make an MC-E behave like an XR-E. If LED's had been designed for flashlights instead of just an accidental "this also works situation", they would have flat covers instead of hemispherical. Regardless of the fact that would mean more light would be spill instead of hitting the reflector, the resulting beam would be much cleaner. It would be perfect for TIR or Aspheric optics.
In 2007 a law signed into effect allowed the DOE to sponsor a $10M competition for the
L-Prize. Some of the requirements for the 60-Watt Incandescent Replacement are greater than 900 lumen output and a less than 10% variation in mean luminous intensity from 0° to 150°. They still need to be placed in the familiar opalescent dome with standard screw in base for the DOE contest. I believe this contest is at least in part responsible for the existence of quad emitters in the first place. So that's good for us flashaholics. The downside is the new dome that makes them inherently flood type devices.
StefanFS had a fine post about
MC-E MagLites with an unusual regulator. The 9th picture is a side by side photo of an MC-E and a P7 MagLite. One thing I noticed about this shot was the MC-E quad die actually looks larger than the P7. That's even though it's the smaller LED. The Lambertian dome among other things behave like a little magnifier. That means nothing to their $10M light bulb contest. But it makes the die a larger, more difficult target for a flashlight reflector. Hypothetically, if the quad die was half the apparent size after removing the Lambertian dome, then you would get a lux measurement 4 times higher. I have no idea what the actual magnifying effect of the dome is.
There is one other problem caused by the larger Lambertian domes. That's the dreaded "donut" effect. Everyone has seen P7 pictures like this.
Notice how the bonding wires on the far side of the LED are much higher than on the near side. You can turn it a certain way to make them reach all the way to the top of the dome. The bonding wires on both sides are in fact the same height. It's nothing but Lambertian distortion.
If LED's were infinitely small and the reflector a perfect parabola, all rays leaving the reflector to go off in the distance would be parallel to each other. So much for theory. I did a simple experiment with my P7 MagLite. I made a pinhole camera from a paper plate. I used a brad about 1 mm in diameter to punch a hole in the paper plate. I folded the paper plate to form a right angle at a point that put the pinhole in the center of the MagLite's bezel with both resting on a table. I was then free to walk across a semi dark room and stick Post-Its on the wall where the images were formed. I slid the plate across the center of the reflector and marked four positions. The entrance to the reflector's beam, the beginning of the extinction zone where you reach the reflector's dead spot in the center, exiting the extinction zone and exiting the far side of the reflector. There are 2 images visible as you move the pinhole camera in front of the Mag's reflector. There is a moving slightly fuzzy image of the quad die which in the absense of the plate you would normally call the spill. Ignore that one. The more sharply focused one that is relatively stationary while moving the plate is the image formed by the reflector. This is the one of interest here. I picked the center of the quad array as my image point.
I hope it was clear which 4 positions of the paper plate I used to mark the corresponding projected image on the wall with the Post-Its. It's a little difficult to describe with just words. Now here's the interesting part. You would expect the order of the Post-Its on the wall to be 1-2-3-4 from left to right just as I slid the paper plate in front of the Mag's bezel in 4 positions from left to right. But that's not what happened. The left to right Post-It order was
1-3-2-4! Positions 2 and 3, which corresponded exactly with the position of the donut when the plate was removed, were
reversed. The Lambertian dome distortion
is the P7's donut. Positions 1 and 4 are the 2 edges of the hotspot. This experiment is worth doing just to see the wildly distorted patterns the LED projects on the wall.
For any experimenters that have read this far...
I had thought one time about buying a hand microtome to shave down the silicon dome on a P7 say 50 or 100 microns at a time. SurplusShed has one for about $25. Most of the ones I found with a Google search were $50 or more. You could make one from a carriage bolt and 2 nuts epoxied in a tube. Drill and tap a small hole in the side to have a piece of plastic put tension on the carriage bolt. The problem would be the stage. A piece of glass with a polished beveled hole in the center is the ideal thing to prevent dulling the blade. But where to get something like that? It would take 30 or 60 passes to remove the 3 mm of material useless in a flashlight. A small piece of wire placed on the cut surface of the silicon directly over the bonding wires while moving your head from side to side and viewing through a high powered loop should be a good test when you get close. The longer cathode wires are closer to the surface so check those. With a brand new razor blade lubricated with something like a 1 to 10 dilution of dish washing detergent (oil might attack the silicon) and water, it might work without destroying the bonding wires. The problem is that silicon is very flexible and the bonding wires are not. Any undue pressure on the silicon will tear the bonding wire right off the emitter. There is a whole science to bonding wires, very touchy stuff.
Another and perhaps better idea I had was those old spy movies where the bad guy smuggles diamonds into the country by putting them in oil with the same index of refraction as the diamonds. Likewise if you could find silicon glue the same or close to the index as the emitter was encased in, pouring more of the same on the top would make the Lambertian dome go away. It would be even better if you could do it while the emitter was mounted in the reflector. That would solve the problem of ghost images by light coming from the side of the newly created silicon cylinder. That might be good for an Aspheric though. Unfortunately a change in the LED's virtual position would also cause a change in the position the reflector needs to be for proper focus. And you wouldn't know exactly where that was until after you put the glue in the reflector. Closer I think, but how much?
The RTV one part glue you buy at the hardware store, the stuff that smells like vinegar, wouldn't work. It's not transparent enough and the acetic acid would probably attack the emitter. Silicon glue is ever so slightly porous. The same reason Cree and SSC say to bake your LED's at 80°C for a day if they're not fresh from the reel before soldering. It turns out there are actually several types of silicon glue. I found this on
ThomasNet but was unable to find a retail source. You probably have to buy it in 5 gallon drums for several hundred dollars, but it does exist.
Silicon remover won't work for several reasons. I don't know if the compound even works on 2 part silicon glue. I guessing they use something similar to the Dow optical glue. The emitter needs the silicon both to protect the delicate bonding wires and to protect the die substrate from moisture damage caused by condensation when the LED is turned off. Cree had a section about this in their Reliability pdf for the XR-E. It would probably attack the phosphor coating even if it did work.
If anybody has experimented with any of these ideas or others, I would love to know what success if any you've had. I have ideas, but not a lot of money. Wow... This is some serious babel. Time to go.
