It is sad MEM vanished, a week after he pmmed me how sad it is when people vanish 
The best method also takes effort and cost into consideration IMHO. Half the price 10% less good = better 
I spent $400 on batteries for a flashlight, does it sound like I settle for 10% less good?
I think a spherical reflector could be designed to capture all of the spill light for most all reflectors with typical width-to-height ratios.
Below are some rough to-scale sketches of what I am thinking.
The top reflector is a C8 reflector with 38mm active diameter, 32mm height, and ~12mm dead hole diameter. The bottom is a convoy L2 reflector, which is a bit more relatively deep, with 54mm active diameter, 50mm height, and 16mm dead hole diameter.
Letâs constrain the diameter of the spherical reflector section to be the same as the dead hole, so no light from the reflector is blocked. Then the optimal position for the spherical reflector would be as in the sketches, to just allow the LED light to reach the edges of the reflector, but capture all of the would-be spill light. Then the radius of curvature of the spherical reflector is chosen to be the same as the distance separating the LED and spherical reflector, roughly 12 and 18mm in these examples. (Iâm approximating the LED as a point in these sketches for simplicity, but I donât think a finite size LED would completely change the picture.)
I think this makes sense. Have I got something wrong? :sunglasses:
EasyB youâre correct, this is basically the reverse of a wavien collar+lens, however there are several reasons why this would be worse.
1) the main optic diameter will remain large, since the spherical reflector takes light from the middle rather than the sides.
Something I think is very important is not just getting high lux, but getting high lux per diameter of optic.
Anyone can take a big lens or reflector and put it in front of an LED to get big numbers, but the impressive thing is getting BIGGER numbers with the same diameter, or similar numbers with a smaller diameter.
2) The collar the way you drew it would only collect very little light, resulting in an intensity increase much less than 2.2x like the wavien collar does.
In order for it to be equivalent to the wavien collar, your spherical reflector would need to be collecting 120 degrees, and your main reflector would need to be much flatter and collect 30 degrees on either side.
Again, this would result in a large diameter of main optic.
3) You can see in the drawing that the parts of the reflector near the bottom are still very close to the LED.
This will be contributing to a big corona and less-sharp spot projection.
With a lens, all the surface where the light hits is farther away, giving a tighter beam and smaller spot.
What difference does a âsharpâ spot make? More candela is always more candela. And the collar is doing exactly what you say - improve candela for a given optic size.
@easyb
That looks good. I guess I had an error in my thinking.
Just so you guys know - I am not actually interested in this for this light. I actually want the little bit of dark spill light it produces.
Yes it would increase the intensity, but my point is that you can get a bigger performance boost by using a different type of main optic when using a collar 
Another problem (I made the same drawing a few years ago :party: ) is that in order to prevent blocking the reflected light from the reflector, the mirror must be small and close to the led, so with short focal length, resulting in a larger reflected image than you want (unfortunately the led die is not infinitely small so the projection is an image, not a point), I assume larger than the die you want to project it back on.
Edit: no, it must be just as large since the source distance is the same as the projection distance. I guess??
With a spherical reflector over an LED the image is always the same size as the LED just inverted.
For wavien collars it doesnât seem to matter much which size you use (if the LED is small). The luminance gain is almost the same (the small one is more difficult to focus though).
I finally remembered to calculate this. It was easier than I thought, all you need is Pythagorasâs theorem for triangles since I already now the diameter of the reflector and the maximum focal length (94mm).
So the depth of the reflector is 73.2mm (distance from lens to focal point, straight down).
The real depth all the way to the bottom I donât know yet, maybe an additional 15mm.
Excellent work! :student:
Congratulations! I had not seen it, thanks for sharing
Thanks!
Hi Gaston,
Where is your wavien build? The Driver is making a list of furthest throwing lights in a world and your is also probably in top 20.
Hi luminarium iaculator,
Is an old thread! here it is:
It would be good to do a ranking
OMG the Detail and work for this project. OâŚMâŚGâŚ
And I can only think of one name for it.
UBERThrower
So, who is going to put a Black Flat into his BLF GT to try to beat this light?
sma tested the Osram Q8WP here (a source in China has surfaced) and unfortunately two of his only reached 190-206cd/mm^2 (same as the old XP-G2 R5/S2). It seems they are of the lowest possible Bin and thus very inefficient.
There is a different Osram LED though which has a higher luminance than the Osram Black Flat, the Osram Synios P2720 DMLN31.SG. This test showed that it can produce up to 297.8cd/mm^2, which is 20% more than the Black Flat. It's downside is that it only produces 200 lumens and the small Die creates a tiny spot (0.225mm^2). I think it's too small and dim for this light though. The Black Flat itself is rather dim...
Where did he get the Q8WP from??
All the distributors show 0 stockâŚ
Be more creative ;).
https://www.yoycart.com/Product/547738518217/
Yoycart is a Taobao agent who has an English website showing which products he offers. This makes it much easier compared to buying on Taobao directly (only Chinese).
Huh, interesting.
How do we know that itâs not an old Q9WP being sold as Q8WP?
If no reputable distributor like mouser or digikey have it, why would this site have it?
https://octopart.com/search?q=UW+Q8WP
EDIT- I just found it available here actually, must have recently become available