OSRAM CSLNM1.TG & CULNM1.TG 1mm², CSLPM1.TG & CULPM1.TG 2mm²

Collars can only be used with lenses.
The angle of light that will hit the lens will not change whether you use a collar or not.
So essentially the lumens that the LED puts out is irrelevant.

What matters is the lumens that actually hit the lens and come out of the flahslight.
Everything else is lost.
By doubling the intensity of the LED, a collar also doubles the flux output from the lens, doubling the effective lm/W.

If the LED is doing ~50lm/W at 5.5A only 25% of that will be captured by a lens that collects 30 degrees.
That is ~12.5lm/W.
The collar will double the intensity as well as the collected luminous flux, giving ~25lm/W.

Just thinking out aloud here… if you had a small enough collar with a large enough parabolic reflector in a recoil configuration where the focal length the parabolic reflector was long enough to use a collar, would there be a benefit?

This is over-simplified, doesn't it? Typical emitter datasheet spatial distribution graphs reveal the light intensity is higher at the center and decreases gradually with increasing angle offset. This means emitters deliver most of its light out the front. Topmost 60° out of an XP-G3 are above 90% relative intensity, whereas at 120° cone borders we're already below 55%.

You’re basically replacing a lens with a recoil reflector, yes you can do that and get the same 2x benefit with the collar, you just need a recoil reflector that has NA0.5
The main reason to use recoil is to collect almost all the light of an LED into the beam though, and using a longer FL reflector with a collar does the opposite.

30 degrees half angle (60 full angle) is 25% of emitted light from a lambertian distribution.
This already takes into account the higher intensity at smaller angles.
That is why 30 degrees is 25% of light, and 60 degrees is 75%.

Just thinking in terms of practicality too. I understand about having the shorter FL, but then you’d need a very large reflector. If one set the FL so that the diameter of the reflector coincided with the 30 deg. half angle of the emitted light, then the RLT would “catch” the rest and give the most intense beam with the smallest possible reflector. Or not?

I was toying around with the exact opposite of a collar - a spherical reflector with a hole in the middle that encompassed the angles where “spill” would occur. Basically the angles that the light does not hit the reflector. I was actually doing this in an attempt to reduce glare in an outdoor spotlight build. But what if we could take that spill light and reflect it back somewhat onto the LED? My test setup was not the best - I was using an undomed BLF GT70, the spherical reflector was hand-made and polished, and the lens was Lexan with an aperture for adjusting the FL of the spherical reflector. I was able to get the spill to almost nothing, but did get some rings around the hot spot. There’s a bevel on the edge of the reflector I made (a drop-off from a large DIY RLT collar) that possibly caused this. Did not do a lux reading before and after because it was cold, and I was too lazy to set the GT70 on a stand. The hot spot did look a little more intense, but hard to tell definitively.

The other thing I was thinking of… there are quite a few “hybrid” reflectors that have TIR optics - either molded as a whole, or as an assembly. Wouldn’t there be a way to create a side emitter TIR that could work with a reflector, retrofit based on its geometry?

Using a collar with a longer focal length reflector would simply mean you can have half the front area and get the same lux.
That means that the diameter is ~30% less.
That’s not a lot.

If you’re going for highest lux, then you choose a diameter FIRST, then decide if you want short or long FL.
Having a long FL (in order to use a collar) means that the light assembly needs to be much bigger.
Think of the syniosbeam, but instead of being 4” thick it needs to be 12” thick.
That’s almost a 1ft cube, way bigger than having a short FL.

A collar is literally a hemispherical reflector with a hole in it.
I think you mean a concave spherical mirror without a hole in it.

The main reason people don’t do it is because the best case scenario is 1.11x intensity, since typical reflectors like the GT collect between 30 and 90 degrees, so ~75% of the light, which leaves only 25% to be recycled by the collar at 33% recycling efficiency (based on the normal collars which collect 75% of light and turn the remaining 25% into 50%).

So a lot of work for 11% increase is not worth it.

I’ll give that argument to Optiforms :smiley: Which is part of the “practicality” aspect. 12” vs. ~8.5” sounds like a lot to me…

Now I’m lost… You just mentioned earlier that ~50% of the light from a Lambertian emitter comes from the front 60 degrees. That leaves ~120 degrees to collect the other 50. In other words, while the surface of the parabolic reflector is collecting more light, that light is less intense. So, if we have a “thrower” flashlight with a beam divergence of around 4 degrees, then about 44 of the total light the LED puts out is going toward spill. Which also explains my theory for having an aperture in the spherical reflector - since 4 degrees of that “spill” light is actually going into the beam. So at 33% efficiency, that’s a 15% potential increase. For a cheap concave mirror, tube, and lens with a hole in the middle - probably about $3 in parts.

  1. you need to think about volume not just diameter.

2) yeah i made a typo, it’s 25% of lumens that are emitted from a 30 degree half angle cone (16 degrees full angle)
This has nothing to do with beam divergence.
A typical forward-facing reflector collects between 30 and 90 degrees, aka 75% of the light. That leaves only 25% of the light to be recycled.
As opposed to a 0.5NA lens which collects 25% of light, leaving 75% to be recycled.

I think it is possible to design a reflector with a particular shape suited to match a crowning aspheric lens beam converging wise. Of course it should.

:-)

There are many optics that work with LEDs. Some of them don’t work well with a collar. But I didn’t mean to fix this parameter. For example it makes sense to compare LED+TIR to LED+collar+aspheric.

ADDED:
Or one can compare out-of-focus zoomie performance.

For me LED+collar together make an emitter. That emitter has limited emission angle. And it has a power-performance curve. I was interested just in that curve.
Actually I didn’t end up drawing it because the numbers above turned out to be enough for me.

Let me try this again… if the typical forward-facing reflector is capturing between 30 and 90 degrees that’s 66. But If 25 of the lumens are emitted between 0 and 30 degrees, that only leaves at most 25% of the lumens hitting the reflector. In other words, the light hitting the reflector overall is half as intense as the light not hitting the reflector.

So with an RLT collar with a 30 degree half-angle aperture, while it’s recycling 66% of the emitted light, it’s only recycling 50% of the lumens, of which is about 33% efficient. Or in other words, with RLT, about 16% of the total lumens gets reflected back into the LED.

If I remember correctly, someone made a build with a parabolic reflector with an aspheric lens “floating” in the center in an attempt to capture and redirect the “spill” light. There are “hybrid” reflectors that do this on the market, as well as reflector/TIR hybrids. I had also considered (and may still try) a floating “zoomie,” though the rear-facing spherical reflector was easier and faster to make.

Pretty nice focus with my handmade gasket from POM plastic. Centering is slightly off. From around 40cm to the wall I get a hole in the hotspot and from 3m the hotspot looks a little like a flower. Does this mean the focus is good?

Yes, that is a good sign.

I continued this mod by adding some resistance in the same style as I did with my D1S in a post above, by adding a length of nichrome resistance wire. But with the D1 there is not enough room for a length of big wire like I used. I found some nichrome 80 ribbon wire (0.8mm wide, 0.1mm thick) with 4.15 ohms per foot. With its higher resistance and flat shape I figured I could fit some into the tailcap and make electrical contact with the battery tube pressing on it. You can’t solder to it so I needed pressure to make the electrical contact.

I made a PET plastic gasket that mostly covered the ground ring of the PCB, then made cutouts for the ribbon wire so the battery tube makes contact with the ribbon wire only, then the other end of the ribbon wire makes contact with the ground ring. I used two pieces in parallel and it makes around 0.1 Ohms. I used thermal adhesive to keep the wire segments in place. This arrangement made it difficult to measure the current, so I used the gas tank method (charging the cell then discharging in the light for 1min, then charging again and measuring the input mAh) with my reactor 300W charger. I got 6.2A average current over the minute with a charged 30Q. The cell was cool so I would probably get a bit more current with a warm cell.

Edit: with a charged cell (4.12V) I put in my pocket to warm up, I measured 72 kcd.

I would like to buy an s9 18350 with white flat (or s2 with biscotti)

The amount of light collected is not linearly proportional to the angle.
0 to 30 degrees is 25%.
30 to 90 degrees is 75%.

I think I misunderstood your post earlier… :person_facepalming:

Still, the relative luminous intensity of the LED from 30deg. to 0deg. is around 90–100. From 30deg. to 90deg. it ranges from 90% down to <10. So putting a spherical reflector with a 60deg. arc in front of the LED captures the most intense light and reflects it almost straight down onto the LED (so I would hazard to postulate the efficiency goes up), whereas an RLT is capturing a light that’s, I’d say on average 50–60% of relative luminous intensity.

That said, RLT with optics are easier to implement; however getting a Wavien collar is not so easy. I’ve made a few for experimenting however. I think having a downward-facing spherical reflector has some merit still; even though it takes more tuning, it is a relatively easy mod, and in one wished, precision spherical reflectors can be easily obtained from Edmund Optics or ThorLabs… Using the GT70 as an example - even if I could get only an 11% gain - can you think of any other mod I can do that I can gain 11% and still have money left over for gas and food?

According to what I can infer from a typical spatial distribution graph, the integral of the first 30° half angle cone has nearly the same surface (≈14,6 squares) as the remaining 60° half angles (2 × ≈7,5 squares = 15 squares). To me this means the first 30° half angle cone (60° frontal cone) delivers 14,6/29,6 lumens = ≈49,324% of the total output. As I see it right now, of course.

Edit: Had a very quick read at Lambert's cosine law @ Wikipedia, suffice to say there's a lot there which is out of my league but still, the first half 30° angle looks to be ⅓ of the output and the remaining 60°, ⅔ LoL. :-D

Mon, 02/11/2019 - 04:18

I think it’s no accident that many parabolic reflectors for flashlights are designed such that the diameter of the reflector is roughly equal to the distance from reflector edge to the focal point. Because that would describe exactly a 30° half angle…