Testing a Cree XHP50.2 J4 3A led

I have an Acebeam K60, if I install this led and according to the graph I would lose around 600 lumens but gain more throw? Doesn’t sound like a bad trade. :heart_eyes:

It’s a disclaimer actually, because it’s not made by you this way, not based on your measurements but by image manipulation done by me.

of course.

You wrote the red line represents an abused LED (resulting in higher Vf and lower output, but on a proper MCPCB) so i wouldn’t call that ideal circumstances either.

I love the charting, the information it imparts, the graph showing all the pertinent aspects give a great idea of what to expect…

However, I have an XHP-70 in a Convoy L6 making 7245 lumens at 11.97A, far above the plot and respectably higher than the XHP-50.2 shows. Not sure then how to interpret the graph.

While I do have 4 of these 5000K 50.2’s on the way, the main plan is to build 3 of them into a triple. I don’t know quite yet what I’ll do with the 4th, perhaps put it in another L6 and build the same ramping driver as the other for a direct comparison?

The problem with this is that it would not work with all hosts, only those that have a thick enough shelf to allow a screw to be used to mount the bottom side mcpcb among other issues. If you were desinging a light from scratch there are lots of ways to power the xhp50.2 but in a retrofit setup we are much more limited and need a more universal option.

A good preforming buck driver is the best idea but the best we have at this point is the mtnmax with ~5A.

Your xhp70 is a P2 bin where the one djozz tested was an N2 bin, that explains the bulk of the difference in readings.

Also his readings are “hot readings” and yours are turn on / 30 second readings. I notice big jumps in output from a cold start vs a steady state test, easily explaining a large part of the remaining difference.

The final ~5-10% would simply be sphere calibration differences.

Does raise the question though, if one should do a graph based on ‘hot readings’.

…is the post with the corrected XHP70 led.
@Jerommel: it used the same led as the first, imperfect, test, and it does state that the led had suffered somewhat, but the few % that that mattered can hardly be called less than ideal.

@Texas Ace: thanks for the explanation, but if you add that all up you still do not get to Dale’s numbers comfortably: in the above linked post you can find a cold start test as well, at 12A/30seconds that led did 6024 djozz-lumen, the difference between P2 bin and N2 bin is a factor 1.15, so that makes 6933. In my experience OTF of a common ok quality flashlight (without super fancy reflector and lens) you loose about 18, so that makes 5685. That makes Dale’s reading 27 higher than mine.

But that does not really startle me anymore btw, the few attempts (texas pyro, sma, Zweibrüder integrating sphere and Texas Ace) that I made to compare different people’s calibrations, in most cases showed way larger differences than I expected and wished for. I was within 1% of sma, 9% over the LedLenser sphere, can’t remember tp but it was between 20 and 30% different, and between TA and me it was 5% or 20% depending on which light source :confounded: )

Yes, that’s the post i linked and took the graph from.

Perhaps you use a fuzzy incorrect definition of “ideal”. :wink:
In the red graph the Vf shouldn’t be higher than in the purple graph (at least, not below 4.5 Amperes or so, should it?), and the output in the red graph should never be lower than in the purple graph.
This way the result is that the green XHP50.2 graph looks better (compared to the red graph) than it should.
(doesn’t it?)

Either way, i did it just to illustrate things, but with keeping the circumstances of the XHP70 test in mind.
Maybe i should have written: the state the XHP70 was in, was not ideal in the red graph.

@djozz - Thanks for the thread! Can the XHP50.2 be a direct replacement for XM-L emitters? (I'm thinking yes due to the exact 5x5mm base but not sure about the connections)

Also, any tips on perfectly centering the emitter during an upgrade? I saw your home made reflow thread and am potentially interested in making one. After so many years, decided to finally try a little modding... :P

Ok, so at any rate the new 50.2 seriously outperforms it’s older brother the 50, might as well go apples to apples instead of comparing the 50’s to the 70’s.

Yes, I did go to the highest available bin on my L6’s XHP-70, I also put it on a 26mm MaxToch copper mcpcb. No matter how you look at it, it’s a bright mamajama and get’s the job done for sure! :slight_smile:

What determines the max output voltage of a buck driver? i am ondering if it is possible to use a 2s buck driver that is designed for 3v output with a 6v output. I get changing the current sense resistor alters the current output but what limits or sets the voltage.

I would have thought that the buck driver would allow the voltage or rise as long as the current is under the limit but this is clearly not the case. I hooked up an mtg2 to a buck driver like those that can be found in any of the commercially available tube lights, I think this one was an olight s30 driver, and it did not light up. If I remember right it was just a faint glow visible over the emitter. So what would make a driver like this be able to output a higher voltage? Is there voltage limiting circuitry and therefore not possible?

I also have a couple nitecore ec4s hear and I heard mention that the buck driver in this light will light up a 6v led. Pretty sure I heard that somewhere?

So,what’s the skinny :question:

Every buck driver is different in how it works and the limits. sometimes the IC will fry with too much voltage, others the caps and others it simply doesn’t know what to do with higher voltage.

Some will work higher and others will not. almost all will have more ripple if used outside the design specs.

The real issue is not voltage but current. Getting a high voltage buck is simple, getting a buck that can handle more then 3-4A of current in a reasonable form factor is the hard part.

Jippie!

I’m an unlucky noob when it comes to finding or making high current drivers for 6V leds so I put the test led from the OP in a 2x18650 thrower with a smooth 45mm diameter reflector, with no driver at all (used the cleared stock driver board just as a contact plate), un-bypassed springs on both sides of the batteries (tail spring is chosen very long to tame the current a bit).

On VTC4 batteries at ~4.05V I got 12A (too much), with 30Q’s at 4.05V I got 10A (still a bit much). With the 30Q’s I measure 3900 lumen at start, 3050 lumen at 30 seconds, and because the light is pretty lightweight built, at that point the head temperature near the pill was about 60 degrees.

Like other new Cree leds the (very nice!) hotspot is slightly on the blueish side (but by far not as bad as the XP-G3) with a slightly yellowish corona. But who cares with this output, it is great to see so much big throw coming out of a 3000+ lumen flashlight, With some luck it is more than 200kcd :smiley:

edit next morning: did the measurement and it is just 70kcd. :person_facepalming:

It is late now but I will make some pictures tomorrow plus a beamshot or two. And measure throw.

I am actually surprised that the current is that low with the 30Q, although be careful your springs don’t melt. I find that long 24AWG wires are a better option for the resistance and then small 26awg bypasses on the springs.

Cool beans Jozz! Ordered some 50.2 J4-1A’s to compare against a Convoy L2 with a old XHP50 J4 1A that’s shaved and pulling 8.57 amps. at turn on. I plan on putting it in my spare L2 and add resistance to around the same current maybe? I can’t believe what I just said! :person_facepalming: Can only measure throw, but at least it will be apples to apples, and see what it looks like out of a smooth reflector! :+1:

[following on from this post]

I was thinking about having the bolt head hold the LED MCPCB down against the shelf, and the nut hold the FET MCPCB up against the other side of the shelf. The hole in the shelf wouldn’t need to be threaded, so the shelf wouldn’t need to be thick enough for that.

Admittedly, there are other issues, as you say: at least one extra wire taking up even more space in the driver cavity, for example, and there’s a risk that the bolt head would experience clearance issues with the reflector.

Just chucking ideas around, basically :slight_smile:

It is a reasonable idea for a purpose built light that would work, the only issue I see is universal retrofit abilities.

You could actually make something like that with an oshpark printed pcb fairly easy. While not ideal it should be good enough to vastly improve the heat dissipation so it can handle almost any single LED on the market right now. Particularly if you stay away from the 50% range where heat dissipation is highest.

My thinking with a MCPCB mounted FET is that it should fit almost any light fairly easily with just 1 extra wire needed. It might have to use a lfpak33 fet but that should be ok. You could use parallel FET’s but that would get more complex.

Wouldn’t it be less than ideal putting both high heat sources, led and fet, in the same location? Wouldn’t be better to keep the fet in the driver cavity and add a small smd heatsink potted to the fet and the closest body contact point?

As to my earlier question regarding buck driver voltage, I suppose the only way to safely test would be while watching a scope and thermal image simultaneously…

Edit: or the “squeeze it in your fingers til it burns” approach might work. This is how I did some testing on the t25c driver. Actually, I might just put that driver back on the bench for a voltage out test. . :smiley:

I like that OshPark PCB idea :beer:

I see what you’re getting at with the universal retrofit now, though, especially if you were switching to different FET package sizes to make it / them fit. Once you had a supplier for the new LED+FET MCPCBs, you could buy them to use in any size-compatible light, and no worries about space in the driver cavity either.

Having both heat sources in the same place while not ideal is not going to make much if any difference compared to mounting the FET to the bottom side of the shelf. The heat still has to travel through the shelf in order to make it to the body either way. The mcpcb is FAR better at dissipating heat though so it should be able to easily handle the heat from the FET and keep it from overheating.

Now you can have a custom aluminum pcb made for the fet separately but it would add a fair amount of cost.

This option is far from ideal anyways and the best idea is to put the effort into a good buck driver. Also moving to 21700 cells would be a very wise idea as it would give more room for a larger driver.

The opamp driver is simply an easy way to drive high currents if you can heat sink it well enough.