It’s only possible because of that flat, dead area at the base of the reflector. The size just happens to be about right. Smaller lights won’t have an unused region.
Your wording implies that de-doming creates an effect. That’s why it was assumed that de-doming to make the emitter smaller was no different than using a concave “dome” which really is an effect. De-doming leaves the emitter “un”affected.
That idea isn't new, was once discussed in CPF. It's sort of an inverted Wavien collar. It would be much less effective though, since only ~25% of the emitted light doesn't hit the reflector (i.e. is spill; assuming typical width/height proportions).
This is a little necro, so please anyone perusing take notice and apply judgement, for example that Dr. Jones might not have a fair chance to rebut.
I had to create an account to comment on this nice but ultimately off-the-mark post which while a great explanation of many things somehow ended up seeming to state that lenses cannot change throw.
Dr. Jones you were completely right about almost everything but made one very key mistake.
You were right up through illuminance and throw.
You were still right that the apparent brightness of a surface, that you’re looking at (the important point), is its luminance and that this does not depend on your distance from it, for a lambertian emmitter. (a reasonable model to use for discussion but not to turn people off, this is just something that emits light evenly in all directions — well all directions away from the surface — apparently rough wood does a good job of this when illuminated)
The connection you missed (and went on a different tangent that I’ll get to) is that a lambertian reflector emits like a lambertian emmiter with luminance (and thus apparent brightness) in proportion to the illuminance on it. I don’t want to lose anyone with the fancy lambertian reflector term. All this says, is that any light (per unit area) illuminated on our surface that we want to light up gets reflected back at all angles equally so that luminance of that surface (light per area per solid angle) is just proportional to the illuminance of the surface by the flashlight. This seems and is obvious: the more light we shine on it, the brighter it looks. It doesn’t matter how that illuminance got to the surface, or how it is produced, by a big fat light, or a tiny point light, since it gets reflected back randomly regardless. So you basically had the full argument nailed down. Throw is apparent brightness of our distant surface, which is proportional to illuminance, which depends on luminosity and 1/d^2.
But somehow you switched wrongly to talking about the luminance of the flashlight and not the surface that we want to see. The only thing luminance of the flashlight matters for is determining how bright the flashlight looks if you’re looking at the flashlight. This is why these tiny flashlights can be dangerous. Their total light output is not more than a ceiling light, but the light is emitted from a small point so you don’t want to look at them.
The rest of your points about lenses not changing the luminance of the flahslight look reasonable, but this just tells you how intense the small flashlight spot looks if you look at it.
Absolutely a lens can increase luminous intensity at some distance, and thus can increase illuminance, and thus the luminance of the surface we are lighting. This is flashlight zoom 101. More zoom, more concentrated light, brighter spot on target. You only need your eyes to see this. Lenses do change throw If our surface was a mirror, then the surface would also preserve luminance and the the flashlight luminance would be relevant, obviously, because we’d just be looking at the flashlight in the mirror. But our diffuse reflector throws out solid angle relevance(at the surface, this can be confusing: the solid angle at the flashlight obviously matters) and only cares about flux, thus why illuminance of the surface by the incident beam matters but not luminance of the incident beam.
Now back to doming. In the end it doesn’t seem like you reach a clear conclusion here anyway. Domed will have more total light output. This seems clear. TIR can re-emit but not without some loss. So de-domed can only have more throw if the light is more focused. Maybe the lower critical angle limits the beam angle but not much if the LED surface is rough, and you stated it’s 120 for both domed and de-domed. I suspect there’s got to be a little change. Anyway yes, lensing of the dome CAN be an issue. How it is an issue, is it seems still one of the devils in the details, one you may have thought about more than me, but one that cannot just be dismissed by “lenses don’t matter”. Of course, absolutely, lenses DO matter for throw. So I think we’re still left with “try it and see”.
There is one final very important point about throw:eyes dilate/constrict.
For two flashlights with equal target illuminance, if one also illuminated the whole rest of the outdoors, especially including the ground near you, this will cause much less contrast, more pupil restriction and worse target visibility. I don’t have any great inside source of knowledge to say that, just common sense about eyes and my own experience.
You seem to have missed something DrJones said. I think that the central point is that external optics can’t increase luminance, so the best possibility is for the whole front of the flashlight to appear as bright as the led is. Throw is limited by luminance and head size, and because leds are fairly large, most practical flashlights come close to this limit. De doming, though optical, interacts with the led, especially with the phosphor, and does change luminance.
No worries, I'm not dead.
The luminous intensity has more meaning than just the apparent brightness.
For throw, the LED contributes it's luminance L, the reflector (or aspheric lens etc.) increases it's apparent area A (and decreases the emission angle), that results in the flashlight's luminous intensity I=L*A (in candela) as written in the OP.
On a target in the distance d that gives the illuminance E' = I/d² (in lux).
That multiplied by the reflectivity r gives the target's emittance M' = r' * E' (also in lux; r is often wavelength-dependent (colors), but I'll ignore that here).
Assuming the target is a lambertian reflector, it's luminance (and thus the target's apparent brightness) then is L' = M' / pi .
Alltogether: L' = ( L * A * r' ) / ( pi * d² )
L, A and I are properties of the flashlight; d, E', r', M' and L' also depend on the target. So the most throw related property of the flashlight itself is it's luminous intensity I, which depends on the flashlight's optics size A (which of course depends on the used main lens) and it's luminance L, this is why I focus on that.
Ok, nice answer. I prefer the math anyway. (Did the 2 in 2pi cancel something or it was a typo?) anyway…
I’m answering a little quickly so,… but basically I agree that there is a best you can do with optics from a given source, and for a fixed sized outer lens this has something to do with the luminance of that source. I guess your OP seemed to jump from target and apparent brightness being luminance to… therefore flashlight luminance is all that matters. Obviously you know what you’re doing but the explanation seemed to imply something not true. It was probably just the “brief” (I joke but as you said, you didn’t write the whole book) wording.
However you can do worse than this. That seems like a paradox because luminance is strictly preserved, not simply an upper limit (assuming no looses), but it’s of course not a paradox and I don’t quite have time to work up an explanation for that at the moment (I honestly don’t quite have the thoughts I want worked out let alone words). Maybe you’ll beat me to it. Anyway, it seems pretty clear that an internal lens can make things worse even if not better than ideal. The obvious case is the external “lens” being a big flat glass and the internal lens being a zoomie lens. Obviously the internal lens, without changing the size of the external lens (or even internal one), can be adjusted to make throw worse. So there are still ways that internal optics can de-optimize the throw and thus that improving that, if not initially ideal, can improve throw. No?
ok… I’m still here. quickly this is I think the crux of various parts of my still not well-technically-formed counter. A 1 square nano-meter emitter with 1 lumen lambertian output or a 10 square nano-meter emitter with 1 lumen lambertian output will both focus just as much light on target with the same optics but will have 10 times different luminance. In either case the source is effectively a point for any reasonable size optics and throw distances and the change in luminance is surely completely irrelevant.
Also you can imagine a light with no reflector and no true lens, but just a flat glass. Increasing the area of that glass does not increase throw obviously (because the light from the glass is NOT lambertian even if the light from the LED was).
It’s not enough to just say it depends only on A and L.
One point to notice is the lens itself does not emit in a lambertian way. You cannot get light on target by just taking some value like L and mulitplying by A and dOmega because while dOmega of the light (or the target, although they are different) might be about uniform over different parts of the area, the actual angle changes, and at parts of the lens area the light isn’t headed toward the target at all. As I recall the place where you actually can do that math that way is at an actual image plane, but not at the lens. Isn’t this right? (edit: ok yes for a zoomie for instance you can do this math at the projected image where A_image then is the projected area and the lens size actually determines the incident solid angle, and yes its equivilant in the end since the answer then is either way is F= L*A_image*A_lens/(r^2 ) either way, did I get that right? so the assumption for this to work is that the optics really are focused at the target)
Again I think it all comes down to that yes for optimized optics there are limits based on lens size but you can do worse than those limits.
I'm seriously really pleased to hear that.
Yes, indeed, all that is based on the assumption that the beam is collimated as good as possible, i.e. the light source is in the focus of the reflector or lens (or lens system), and the reflector/lens is ideally shaped (which often is an issue with strong aspherics). Any change from that (i.e. using a flat glass as main lens or defocussing a zoomie) will of course make things worse. It won’t get any better that this ideal though, and flashlights are usually built that way (except some flooders, but we are talking about throw here).
And also I assume ‘reasonable’ light source sizes bigger than the resolution of the used optics; 1 nm and 10 nm is much below the diffraction limit, and then more complicated optics (wave optics) apply. But that’s not needed with LEDs.
No, it’s just 1*pi, sort of the equivalent solid angle in forward direction for a lambertian beam profile; I=F/pi (in forward direction) and L=M/pi.
I hope someone can make this clear.
What should I expect if I use a dedomed XM-L close to an aspheric lens?
I know the aspheric lens in this configuration has excellent spill and no detectable hot spot.
This =is= the effect I want, but…
With no dome, I can actually get closer to the lens… but that too can diminish returns.
Graphics would be appreciated.
In the same position, dedoming will decrease the total output more than with a reflector light. I don’t know whether moving it closer will make more or less difference. In focused throw mode, it will decrease the total output and increase the throw.
Thanks Fritz!
With an aspheric close to the LED (i. e. not in focus, flood), but at the same distance, the beam profile will not change, but the total output will be less.
Comparing the aspheric touching the dome to touching the dedomed die, the latter will have a wider beam, be less intense and have a bit less total output (unless the aspheric is quite small, say, <20 mm, in that case the total output might increase, since the smaller the aspheric, the more light gets lost when it is at dome distance.)
Does the TIR theory still apply when you slice (most of) the dome off?
With MT-G2’s I don’t think it’s as simply as dipping the LED in gasoline and waiting for the dome to come off. We just slice it off, and I think it significantly improves the throw.
Thank you also, DrJones. This is what I suspected but just wanted to hear it from the experts.
ryansoh3: Yes, it should. The remaining stump should be thin, so that the reflected light (even at shallow angles) hits the die again and doesn't escape through the sides.
This may be nearly exact if we assume the dome is a sphere with the led at its center, but looking through the edge of an aspheric at the led in moonlight mode I it appears to be distorted by the dome so that it is visible farther out than it would be if the dome were spherical.
Dr. Jones: Thanks for writing this up. My understanding had been that the increase in intensity from dedoming was a result of the decreased apparent die area, so this is really making me think. If you could answer a couple questions it would really help my understanding and I would appreciate it.
With a domed emitter, the die looks about 2x the actual area. Doesn’t this necessarily mean that the emitter+dome does not follow the cosine law of a lambertian emitter? So, compared to an emitter with a thick flat silicone layer on top (no image magnification, but no luminance increase), the domed emitter would have twice the luminous intensity directly above the emitter, since I=L*A? And since A*W is conserved, W would have to be smaller with the domed emitter.
If an emitter with a thick flat silicone layer on top (so the TIR does not increase the emitter luminance) is a lambertian surface and follows the typical cosine law, I think I am confused how a dedomed emitter can also follow the same cosine law. The intensity directly above the dedomed emitter would be greater than the flat-domed emitter, so if total flux (lumens) is approximately conserved, the intensities must have different angle dependences.
So if the luminance increase of a dedomed emitter can be compared to using a Wavien collar, would a dedomed emitter emit less at shallow angles (even less than the reduced apparent area would suggest) and recycle this light to steeper angles (more directly above emitter)?
Is my understanding correct?
Edit: just found this test of the luminance of emitters with flat domes vs regular dome vs dedomed, which confirms that a flat dome will not have increased luminance over a regular dome.
I suggest that you look here:
It is not updated with the latest LEDs and doesn’t contain XP-G2 either but it is my favourite efficiency and throw comparison. 
Sadly, since koef3 retired from this forum we don’t have as good source of throw information so it’s hard to update the charts.
I have some relative intensity and lumen outout measurements at different slicing height here for the LH351D : https://budgetlightforum.com/t/-/63712