His graph has both the CSLNM1.TG and the CSLPM1.TG.
One is square die and the other is not.
The most recent test he is talking about right now is the CSLPM1.TG which has a rectangular die.

It is too early morning for me! Need a coffee!


Thanks for the test!
That’s a somewhat dissapointing result in terms of brightness! The low Vf is great. It might just be that you tested a low performing sample.
The Q8WP that koef3 tested a while ago got up to 9.6A (although his cooling setup might be slightly better).

Yes, I get consistent lower thermal performance than other led-testers, if it is the reflow and/or the amount of cooling from my copper mount I do not exactly know. I could order a mounted led from l4p (he uses a different solder) and see if that tests different but I was planning a sort of half test-retirement anyway in which I mainly test for my own builds (and post the test for others to get at least an impression), and the results that I get are certainly relevant for myself.

Djozz after seeing your cooling setup I think that that is your problem.
You have too much copper between the LED and the fins. Copper is a great heatsink and can store a good amount of heat but it’s not as fast in conducting it.
I think the temperature is getting too high as it can’t get rid of it quickly enough.

It might be better if you just took a modern cpu cooler (something like the last generations from and or Intel) and mount the led directly on it.

Q8WP and PM1 basically use same chip, so why you would expect some drastic increase in performance? Q8WP is in bigger package which could explain a little bit higher max. current, but you can also see that lumen output barely rises over 8Amps, which means chip itself is close to its limit at 8+Amps anyway.

Also,Q8WP has some side emitted lumens, PM1 emits light only from front side, which should make it better for cd/mm2 performance, even for same lm output vs Q8WP.

You basically have best 2mm2 throw LED which is brighter than any other 2mm2 LED,has lower Vf than any other LED (per mm2), it's factory dedomed,has nice neutral (non-green) tint and you are disappointed?

I don't if it's just me, but lately BLF = Never good enoughTM

Q8WP had the same peak intensity as HWQP. Same generation, same performance in a nearly 2 times larger die.
NM1 is a significant improvement over HWQP, better intensity, lower Vf. I expected PM1 to be improvement over Q8WP as well. I didn’t expect it to have the same intensity as NM1 because it has the same thermal pad size with double die size, but nevertheless I expected that Q8WP would be beaten. Now it doesn’t seem so, though we lack data to know for sure. Due to symmetric package as well as common footprint it is easier to use, but that’s just that. Well, it is also easier to get. :wink:

I think that PM1 is an awesome LED and I would find many uses for it, but nevertheless I expected more from it.

This could definitely be a reason for lower peaks, I don't know dimensions of that piece of copper rod, but I took 25mm diameter and 35mm length, and result is about 0.2C/W of additional thermal resistance, this could be enough to cause lower peak currents and lower lumens, but not by much for smaller LEDs,depends a lot on LED C/W itself,and total power.

I sometimes directly after a test feel the ledmount with my hand and it was never hotter than luke-warm, even when testing the XHP70.2 led. So if keeping the mount under the ledboard right at 20 degC or 40degC explains a much lower thermal performance then yes, my long mount is to blame for lower performance. In that case most flashlight builds will suffer from the same problem. Btw, among metals copper is about the fastest thermal conductor on the planet/universe.

The copper rod is 30mm diameter and about 60mm long (not at home so I can’t exactly measure). I need it to lower the led inside the sphere while still keeping access to the ledwires.

I could save half the length of the copper bar by using 4 times longer ledwires and guide them through the heatsink to the side of the fan. Longer ledwires would influence the measured voltage at higher currents somewhat, or for each led-test I should solder separate wires on the board for voltage measurements.

But my life does not depend on these tests, nor are my flashlight builds different if they were that bit more accurate. so I just may leave my set-up as it is and hope for Koef3 or maukka to test these leds too and go the extra mile in accuracy.

It might sound surprising but as soon as you remove/turn off heat source(LED in this case), temperature of that copper piece drop very fast(if bottom cooler still works). So you can't "feel" by finger how hot is top of copper rod when LED is working.

It's like you are trying to measure capacitor voltage after you charge it and disconnect from power supply; copper rod is thermal capacitor itself,but it also is resistor in parallel with that capacitor because it's very thermally conductive. So you have parallel RC circuit. So when you connect this network to power supply(=heat source analogy), it will charge to for ex. 5V (volts = temp difference analogy) and stay there as long as it's connected to power supply.

But as soon as you remove power supply(=heat source analogy), voltage (=temp. difference analogy) drops rapidly due to low value of parallel resistor (electrical resistance=thermal resistance analogy).

That's why you never feel that top of copper rod is hot.

I’m only somewhat dissapointed ;).
The tint and Vf are certainly a big deal.
The brightness not so much.

I will try to think of a new mount with the ledboard closer to where the heat is removed, but I’m thinking of a new integrating sphere also so that will be a combined design and not likely to happen in the near future, if only because I’m rather low on cash and like to spend what I have on other things too.

Crumbled white flat 2mm seems overheated or cooked for long time and while it was in that ultra hot state of matter some kind of pressure with some metal thingy was applied… Also specks of dirt everywhere(too much flux or solder paste) in fact if any of you guys got burning crater inside emitter(especially dedomed one) I would say that in 80% of case it means that hot speck of dirt got sucked into phosphor.

It is nice to see under skin of such emitter. I like that.

G2 looks like hot de dome failure since pressure was applied on wire which eventually broke.

This one looks good. It has L4P DTP and your testing rig seems better than any flashlight diy test(luxmeter clampmeter) so I don’t see any possible failure reason.

Hmmm… Why should anyone be disappointed with new White Flat 2mm ?

-Better lumen and lux performance than G2 (How much? I guess you guys will find out in your builds)
-Better Tint
-Low Vf
-No need for dedoming
-Probably no need for spring bypasses either?

So what is bad in that? :smiley:

Ok… I know what is bad in that. It is not suitable for mine single cell FET drivers(if they could be called drivers). To bad for me.

But I don’t see any reason not to put such type of emitter in regulated single or multicell mildly driven host?? Beautifull emitter better than any Cree :stuck_out_tongue:

LOL, in this case what you see is not what you get, both leds work perfectly fine. That XP-G2 is not broke, in fact it happily did 235 kcd in my mini-GT. And the mini-GT is to blame for both leds, its centering ring is so tight in the reflector that the twisting action of the bezel when opening/closing it already sheared off two leds and cracked that PM1 led.

OK that also sounds logical :slight_smile: Except that decisive pic of g2 with broken wire which in fact is not broken.

“Wounded and still working” :+1:

Sounds like the 7135 died.

If the switch-led report is correct, the led still failed to light up while the ramp was well into the section that the FET was used (2 blinks after halting the ramp up). In that case there’s more to it than a dying 7135.

Those 7135 chips can handle some low voltage btw, I have made an infrared sk68 once (working voltage around 1.8V) with a 8x7135 driver that functioned fine with the 8 chips all burning off close to 2 Volts (I did take care of a good heat path for the chips).

djozz, I just noticed something about your test. You tested the XP-G2 S4 2B with dome on. De-Domed it only goes to 1100lm according to koef3’s test. You numbers of the new Osram are very close to his of the Q8WP, both LEDs are much better than the XP-G2 de-domed.

6.5A seems to be about the sweet spot for the KW CSLPM1-TG (white2 as vinh calls it). I wouldn’t go above that when driving it with a buck driver inside a light that heats up.