Lumileds Luxeon V, test of a 4000K 70CRI emitter

After reading this I would like to repeat my question:
“Wouldn’t this LED be a good choice for FET driven 3x AA lights (e. g. Convoy X3, powered by Eneloop cells)?”

I am thinking of a bright flashlight with NiMh cells for my son.

I guess that’s not so good, as the low Vf leads into pretty high current for NiMh on DD drivers
for a regulated driver this LED is good

Would be great for something like the Sofirn SF11 or Lumintop SD4A which are 4xAA lights with a buck driver. Should give nice long runtime.

According to HKJs’ review Eneloop 1900 mAh delivers a current of 5A at 1.1 V, which (with 3 cells in series = 3.3V ) meets djozzs’ numbers pretty well (5A at 3.25 V for 1800 lumens).

Yeah but you loose quite a lot voltage and the efficiency goes down

5A is a lot from AA NiMh

Where do I lose voltage? 3 cells in series sum up to about 3.3V at 5 amps, where the LED is pretty efficient. Eneloops should be good for this current, HKJ tested them up to 10 amps with a max. temperature raise of about 20 °C. Even the capacity of the cell doesn’t decrease much at higher currents.

Coincidentally I was able to do a quick check on 3 Eneloops in series because a few years ago I made a 3xAA carrier with phosfor-bronze springs and Omten switch (using Oshpark boards) for a project that I never finished. Pushed against the Luxeon V-modded VG-10 head I checked the output on 3 fully charged Eneloops, 30 seconds after switch-on I measured 1400 OTF lumen, that is about 1600 led-lumen, which should be about 4.2 amps (measuring the amps directly is not as quick as this was).

Of course upon draining the batteries the output goes down considerably (though not dramatically because the voltage curve of the Luxeon V is relatively steep), but this at least shows that a 3xNiMh set-up should work well enough to make a flashlight with very decent output.

Thanks, djozz, that’s close to what I expected. Looking forward to my X3 mod once the LEDs arrive …

Today I tried the Luxeon V 4000K 70CRI in my black Emisar D4 (see the WDYMT thread). The output and beam were very rewarding, but although it still looks alright, the tint is better in the VG10. I can’t say for sure but I expect it is a reflector vs TIR thing, the hotspot from the TIR is more yellow, and the rosy appearance in the VG10 beam is not obvious in the TIR beam.

Also, while playing with both flashlights, it looks like the tint is better (more red in the tint) at high current when the led runs on the FET than at low current when the led runs on the 7135.

So they fit under Carclo optics without dedome. :slight_smile:
It makes it somewhat more likely we’ll see the triple / quad MCPCBs for them eventually.

The black D4 with LuxeonV’s runs pretty efficient even on highest setting. On a freshly charged 30Q I measured 15A for the first minute. A 30Q at 15A has a voltage of about 3.7V in that minute, so the power use is 55.5W. If you take the output at 12 seconds, 4620lm, the flashlight efficiency is 85 lm/W, which is rather good. This is not surprising because it is a quad, each led receives just 3.75A in which region the LuxeonV is still quite efficient (guess what, it is even well within specs :party: ).

The same story goes for XP-L(2)’s btw, and even in somewhat less extend for XP-G2 and 219C.

This implies for me that a triple channel driver in a multi-emitter light like the D4, or the FW3A, is a nice luxury and should certainly be aimed for, but the huge efficiency boost in the middle modes that some folks promise does not quite happen, it will not be more than an unnoticable 10 or so percent.

Lights that have a complete battery available per led (single led single li-ion, or i.e. the Q8) drive the led on highest setting very close to the inefficient max, in that case a triple channel will have some merit.

I remember someone measuring D4 (not sure which emitter) runtime @350 mA and arrived at under 100 lm/W. Actually I think it was eighty something.

I checked the black D4 at 1x7135: I measured 347mA with an output of 149.6 djozz-lumen. The 30Q was at 4.09V resting voltage and loosely interpreting HKJ’s graph I see at that drainage stage and 350mA a battery voltage under load of 4.07V. The power is thus 4.07x0.347=1.41W. That calculates at 106 lumen/W.

That is with the 7135 burning off 1.3V overhead voltage (.45 W). Would a buck driver exist that converts that 100 efficient in output then the efficiency would go up to 156lm/W. Dreams!

(as a side note: my djozz-lumen, which in some comparisons that I did looks like it is ~9% optimistic compared to real lumens still appears less optimistic than the lumen that most people’s on BLF and most well-known flashlight manufacturers use)

Is the tint also better when FET is running with lower duty cycle (e.g. with 1 Ampere) - or only when FET is near 100%?

Very difficult to see because the tint change that I think that I see goes together with a large brightness change and a brightness change alone already makes your eyes perceive a different tint. So yes, I think it is already better with FET at low duty cycle but the effect is not at all obvious.

If it depends on high current it might be another point against triple channel drivers. After using several lights with Nichia 219C 4000K CRI92 I’m pretty spoilt with tints.

I crunched the numbers, here’s what I measured, LuxeonV on a XP-Noctigon compared to LuxeonV on led4powers 4040 board. They were different leds so some difference could be expected regardless on which board they were. But the performance on the lower currents is so identical that I’m inclined to attribute the slight difference at high currents to the boards and not to the difference between the leds. I checked the quality of the reflow after the test by desoldering the led and checking the solder pads for incomplete reflow, everything looked fine.

My conclusion: the 4040-board has the slight edge, but only at very high current. But apart from performance, the 4040 board makes reflows way way easier and robust.

I wanted to be gentle to the led to be able to use it afterwards, so I stopped the test at 11A, but looking at the chart it looks like the output decrease when the current is over the maximum-output current is clearly lower on the 4040 board. This suggests a healthier led in case of abuse when a 4040 board is used.

Thanks for the numbers!
Would like to see this LEDs in my D4 …

Hey I just noticed this, but “a thermal resistance of 0.8 degC/W, which is 3 times as low as any existing high power led” is not really correct, the CFT-90 has a thermal resistance of .45C/W at the junction.
But this is probably because the LED chip itself is like 10mmx10mm or something huge like that :stuck_out_tongue:

Yes and no. :smiley:

0.8°C/W is by far the best value ever for a 4mm^2 LED (XM-L2, XP-L2 etc. => usually around 2.5°C/W).
Luminus has traditionally reached the lowest thermal resistances because they use larger DIEs and a special direct-to-copper mounting technique (notice that they have two sets of bond wires, the current doesn’t got out the bottom of the DIE).

Generally LEDs with large DIEs seem to have lower thermal resistances. I think it makes sense to compare thermal resistances relative to the Die size, so I multiply them (so lower values are still better):
XHP70.2: 0.9°C/W * 14.2mm^2 = 12.78
CFT-90: 0.45°C/W * 9mm^2 = 4.05
XHP50.2: 1.2°C/W * 8.24mm^2 = 9.89
XHP-35 HI: 1.8°C/W * 5.9mm^2 = 10.62
Luxeon V: 0.9°C/W * 4mm^2 = 3.2 (I had to guess here regarding the DIE size)
SST-40: 2.5°C/W * 3.994mm^2 = 9.99
XP-L2: 2.2°C/W * 3.55mm^2 = 7.81
XP-G3: 3°C/W * 2.06mm^2 = 6.18
Osram Black Flat Gen2: 4.3°C/W * 1,122mm^2 = 4.82

So it seems the Luxeon V and the CFT-90 are indeed very good, but the Osram Black Flat, XP-G3 and XP-L2 certainly hold their own.

I’m not sure theses theoretical values tell the whole story though or even have any meaning at all. I am completely ignoring the size of the LED package and thus of the center solder pad. The Luxeon V has a larger solder pad compared to the XP-L2 and the SST-40 is even larger than the Luxeon.