Flashlight concept #1 - Multi-emitter light

Update 07/08/2020:

Thank you so much for your valued feedback so far! Meanwhile, I tried some new design approach that is easier, hopefully less complex to design as well as simpler to operate UI-wise, beginning as of post #23.


Target output max.

4.200lm white (mixed)

400lm red

Max. output could possibly be higher but efficiency should be more important?
Host design

21700 shape

magnetic rotation ring

mechanical tail switch

integrated USB-C port

stainless steel bezel

Non-AR coated glass

Head section to be designed as a new part (> 38mm)

Center section design adopted from Sofirn SD05 + IF25A

Tail section design adopted from Sofirn SP70

AR coating may "bend" the light's wavelength and cause additional tint shift



(wide beam, 35°, 38mm diameter)

This optic will probably be a good trade off regarding maximum beam distance and the avoidance of tint shift. A more frosted design might be even better tint-wise but it will become very floody.

3x Samsung LH351D 4.000K 90CRI

3x Samsung LH351D 5.000K 90CRI

1x Luminus SST-20 DR (660nm)

The center LED could also be something else instead of red light. I chose red because SST-20 DR is rather easy to source and with up to 3A current more powerful than Cree's XP-E2 red/deep red. The downside of less perceived brightness with 660nm might be eradicated by more current resulting in a much more pleasing red color than standard red (625nm).

A thrower LED like Osram CSLNM1.TG/CSLPM1.TG would probably not work well underneath a wide angle beam optic. It would require to have an optic with different angles for each LED - I have not found something like that. A UV LED would not work either as this wavelength would not pass through plastic material.

Driver Fully regulated buck/boost driver That's wishful thinking, of course. But it will make this concept much more interesting. Driving 6 LEDs efficiently will (hopefully) result in a more useful and longer operation on higher modes than having short bursts of Turbo using FET/DD. However, such a driver will definitely increase the (development/material) costs of this light.
Power 1x 21700 flat top Inner tube diameter should be large enough to use off-standard 21700 cells like Molicel M50-A (more a "22700" cell)
User Interface

Magnetic rotation ring

Tail switch


controls brightness only (stepless ramping from moonlight to high/turbo)


controls ON/OFF (fully press) or mode selection between white and red light (tap to select)

switch design is yet to be defined - most people will probably prefer a reverse clicky over a forward clicky


Hello everyone,

I did some thinking on new flashlight concepts and mostly about those things we seem to miss in some good but not perfect flashlights, e.g. good thermal management, good heat transfer, efficient drivers, convenient user interface and interesting LED configurations. With Fireflies' E07 and ROT66 V2, Mateminco's MT07 and Sofirn's IF25 in mind I wrote down a rough concept. I would appreciate any feedback on this, especially if this is something reasonable and feasible in your opinion.

I like the tint mix version!
Would it be possible to change the preferred tint by giving the preferred LED’s more or less current ?

In this size that copper head makes it really massive, not a light that I would carry with me. The tint-mixing version is nice, I think it is a pity that Sofirn went for their own tint-mixing version instead of Anduril.

Yes, that's the plan. Adjusting the current for WW and NW LEDs will change the mixed tint.

That's a good point. Maybe they™ could use a press-fitted copper pill (like Nightwatch IRA) instead of a solid copper head and/or solder the MCPCB to the pill which should result in better thermal transfer. I wish LEDIL offered a partially wide beam frosted and partially clear narrow spot optic. This would give us a whole bunch of new ideas like 3 emitters for throw, 3 emitters for flood and a wide beam frosted auxiliary light in the middle (it does not necessarily need to be deep red ;-) ).

For me an EDC light would have to have a head width of 35mm max. It can be a tiny bit more but depends on the head design. I think 7 LH531Ds might be too floody. Four would be better IMO.
Here’s some things I would like to see

Decent throw 25+ kcd for high cri version. More would be nice but the LH351Ds are domed so I don’t think it’s possible.

Tailstanding ability

Strike bezel so it’s harder to leave on by accident

Three selectable colour aux LEDs if there’s space (red, purple, orange)

Yes, two of them WW, one CW. That would compensate for the lesser efficiency of WW LEDs.

Lux, when to expect this true threedimensional (brightness, tint, beam) flashlight exactly? :smiling_imp:

Almost forgot: Add this CCD element to autoadjust WB :sunglasses: .

I just added the "USB-C port in thread section" feature. How could I've forgotten that? :-D

Solid copper head is good for sustained performance but bad for turbo. It conducts heat away so well that it keeps the LED - and (what matters more) the LED surroundings much cooler than other materials would. Since they are much cooler they fail to store much heat.
For this reason solid copper lights tend to overheat faster than aluminum ones despite the fact that they weigh 3 times as much.

And thermal transfer improvement is marginal anyway.

I really like 7-up optics because they make good use of the frontal area. Almost like 1-up…but unlike 1-up they are way flatter.

I never use red lights, so for me the extra red LED would be a waste. Especially on a thrower. But if you like it - why not.

Tint mixing - I’m a big fan of it

Having a variant with 3535 LEDs and another with a mix of 3030 and 3535 would require 2 different MCPCBs. I’m not sure if that’s a big deal though.

I think White Flat 1 mm² would make a better thrower choice. For ~45 mm head 12 mm² (let alone 14 mm²) is quite a lot. Very short turbo and actually quite floody beam. 7x KW CULNM1.TG (aka Boost HL) would truly rock.

As usually with dual switch lights - I believe it’s worthwhile to give user a choice whether they want the tail clicky or just a flat cap.

A Powerbank capability is a boon for any rechargable light with reasonable battery size

A boost driver capable of driving 12 mm² White Flats? Not easy. Not compact. Not cheap. But doable.

1. you are talking quite nonsense, just because Copper conducts 3 times better than Aluminum it will be completely even tending to be even cooler, why?

Just check a Cree XPL here at 3A
let’s assume LED runs 120°C on aluminum and 100°C on copper

- the amount of power with a regulated driver gets a bit higher as the LED voltage increases 0.04V*3A=0.12W more

  • but also the photon conversion is more efficient almost 50 lumens more at 350lm=1W 49lm÷350lm/W= 0.14W

On an unregulated DD driver your observations are also not true, why?
With a lower temperature the LED voltage is higher so the light draws less current

copper lights may feel hotter as they heat up more evenly through the whole body

4. I think tint mixing is not so much-needed on a light
Also why is 351D 2700K so much less efficient?

5. I am sure such a MCPCB with both emitters can be made
I can see if that would also allow an aux board probably the positions of the main LED wires soldered is critical here to get a symetrical pattern

6. agreed

So…roughly 0.14-0.12 = 0.02W or 0.2% more, right?

Yes, more than 0.2% with FET.

Now the amount of heat stored is the integral of (temperature hike) * (volumetric specific heat) over volume.
Assuming constant surface temperature of 60C (that’s the moment when the light needs to step down) and ambient temperature of 25C.
Temperature difference between surface and near the led is inversely proportional to thermal conductivity.
With 100C for copper we would have ~160C for 6061 alu, regardless of head size shape etc. Yeah, that 100C is likely too large. I think 80C copper and 130C alu would be more like it, though I am not sure about it really.
This gives temperature hike over ambient of ~55C for copper and ~105C for alu. Adjusted for difference in volumetric specific heat (copper is better), with these parameters alu would store ~33% more heat near the LED. Now, near the surface it would be the opposite, both have the same temperature, alu stores about 31% less. But then the temperature at the hottest point of the surface is much smaller; heat storage near the LED is more important. At the cooler points of the surface the difference will be larger in favour of copper. But then - the amount of heat stored there is minimal.

In practice - through the whole head. And not much elsewhere. :person_facepalming:

Wrong, 0.02W is not ~0.2% but more like 0.3% because energy consumed != energy dissipated as heat.

You do the math wrong, we already have a thick copper MCPCB which reduced the temperature difference between a copper or aluminum head
Ask any here in forums why we don’t want to use aluminum MCPCBs even if they are DTP, the heat conduction is absolutely critical to keep the LED temperature low

Nobody wants to store the heat near the LED, the cooler it runs the better its photon conversion is, so people asked manufactures for thicker shelves on the BLF lights to move the heat faster away and give more heat buffer

I have done several thermal imaging videos of aluminum heads, the heat from the hottest to the coldest spot is minimal as the conduction of heat is already very good on flashlights with a thick shelf and thick walls

Your 0.2% on DD drivers is also wrong math
look on a Samsung 30Q battery for example and look how much 40mV voltage difference between a hot and cold LED result in current change
I did here the graphs between 5 and 7A and on average I would say about 40mV difference

so your 0.2% I don’t know how you calculate that number results in my calculation for 40% current change without any resistance added
so lets take bypassed springs and switch with 50mOhms and driver and LED leads also 50mOhms
40% change on 16mOhm of the 30Q cells, now lets do the math with all resistances I get 5.5% less current on DD light just by 20°C lower LED temperature

I mean a 100% copper light with copper tube, most people touch a hot rod light only on the tube, the better it conducts heat the more the human heat sink plays a role

So, a press-fitted copper pill and/or soldered MCPCB could probably be more efficient in terms of weight, thermal conductivity and dissipation as well as being more cost-effective?

Using a full copper body would definitely increase both weight and cost dramatically. I was hoping to match the best of three worlds (Titanium for a sturdy bezel, Copper for best thermal transfer, Aluminum for best and fastest heat spreading/dissipation alongside the battery tube with the benefit of less weight).

The idea about tint mixing originates from Sofirn's IF25 and the LT1 lantern. Basically, I would like to enable every customer to choose his/her favorite tint between warmwhite and neutralwhite. We could also go up to 5.700K 90CRI - no big deal. There might be certain situations when people like using neutralwhite (outdoor in summer) but also want to change to warmwhite (outdoor in winter with highly reflective snow/iced surfaces or when inside) with the same flashlight. If tint-mix is not an option we could still mix LH351D 2.700K 90CRI with 5.700K 90CRI to get rid of any greenish tintshift and drive the mixed tint below the BBL resulting in a rosy tint pattern somewhere around 4.000K. Just thinking...

Why is 2.700K so much less efficient? Because most (all?) warmwhite emitters seem to have less efficiency than coldwhite emitters. I do not have accurate numbers for LH351D 2.700K 90CRI but I quickly translated the reduced output by the expected loss in efficiency. We may get a bit more output here but I wanted to keep the numbers conservatively calculated.

If multi-footprint MCPCBs are feasible, how about replacing the inner XP-E2 deep red by an Osram White Flat KW CSLNM1.TG 1mm²? This would give us a flashlight with both flood (6x LH351D) and throw (1x Osram White Flat 1mm²).

Looking on Digikey as a source for LH351Ds
I get 2700K CRI90 310 and 330LM binnings
for 5000K CRI90 390-410 binnings

so 4000lm for 5K
and 3200lm for 2.7K

Looking on Digikey as a source for LH351Ds
I get 2700K CRI90 310 and 330LM binnings
for 5000K CRI90 390-410 binnings

so 4000lm for 5K
and 3200lm for 2.7K

also why not simply compensate the lumens in the firmware to have on any color 3200lm just do the 5000K only on 80% max current?

Just to get it right: Do you mean a balanced output, i.e. the flashlight will always have let's say 3200 lumens output regardless of the tint's CCT?


you could then also make in the firmware an option to blast out Turbo with all LEDs and max current at about 4000K

Okay, I got you. :-) Good idea. Sofirn did not balance the output on the IF25 (yet). Effectively, not only in perception but also in actual numbers it is brighter on coldwhite than on warmwhite. They also did not implement the fancy feature of using all 4 LEDs on Turbo (yet). However, your ideas definitely belong to the wish list of the UI a perfect flashlight should have.

A revised draft concept...


Hmmm… Tint ramp is not a huge thing for me, but then again, I’ve not used a light with tint ramp.

I like the idea of a separate red in the center. Sometimes you need a more gentle and nighttime friendly light. Then again, dropping what I assume is an Osram thrower in the middle defintely adds flexibility. The unicorn flood/throw combination light.

Size — I’m torn on this one. As a pocket light, even the 21700 might be a bit big for my tastes, let alone a 7 LED head. As a holster / shelf light, I like the feel of a 26650 tube, which might be a better size match for a 7 LED head. (reference - I have most recently used an FW3 and EDC18 as my pocket lights, and they are IMHO, on the edge of being to big for pocket EDC)

I was wondering about this the other day, and since I saw this, I figured I would ask. Is there any potential advantage to using the red LED in the tint mix? Most phosphor-based white LEDs tend to lack red, right?

ETA: Or is that already the plan?