Preview of updated K2: testing of "new" K2 sample

[chris_m]

I agree c-c, exactly the same thought occurred to me seeing those Vf figures. Of course the problem is that to have a Vf over 4.2V implies 20% worse efficiency to start with even if the lumens/current is the same, so you'd only need to find a 70% efficient buck/boost.

A quick search found me the TPS63000 which would actually seem to give >90% efficiency through the discharge of a LiIon cell driving a 3.5V LED at 200-1000mA, so it's simply the application of such drivers in flashlights that's the issue than the non-availability (I have a feeling I've just generated myself another DIY project!)

 

[BentHeadTX]

Thanks Evan for the update

Sad to hear that the Rebel 100 did not transfer over to the K2. Strange that the Reb100 and K2 are almost opposite to each other with the low Vf/high output Reb smashing the K2 in all measurements.

With the news of continuing K2 problems, I ordered a Fenix L0D Q4 to replace my keychain mounted FireFlyIII. The FF3 will roar again once I get a K2 240 with lower Vf. 2008?

 

[mofiki]

I fired up one of the K2's I have and ran it for over an hour. It got pretty warm but I didn't have any problems. I was somewhat impressed with the amout of light. It was driven with a Lux-drive 3021-1000 off 4 RCR123 Lithiums. I had to run out and do some errands so I shut it off. When I got back I turned it on and notice everything was running hot within a minute. I did some tests for shorts and found nothing. I turned it on again and poof. It smoked.

 

[saabluster]

I fired up one of the K2's I have and ran it for over an hour. It got pretty warm but I didn't have any problems. I was somewhat impressed with the amout of light. It was driven with a Lux-drive 3021-1000 off 4 RCR123 Lithiums. I had to run out and do some errands so I shut it off. When I got back I turned it on and notice everything was running hot within a minute. I did some tests for shorts and found nothing. I turned it on again and poof. It smoked.

Was it heatsinked in any way?

 

[mofiki]

You must know your stuff because you ask the right question. It was heatsinked to a copper pennie which was JB Welded to a threaded barrel peice that was part of a Brinkmann Legend flashlight which fitted into the head of the light and the reflector was modded to fit perfectly against the led body. It did put out a nice beam with even spread of the hot spot.

BUT; when I noticed it getting very hot on the second run I thought something was wrong so I took it apart and started doing meter checks for a short. The ohms checks found nothing. The checks with the meter set to diode made an audible beep and showed a value on the display via the test leads touching the led solder lead and to the heatsink. When I reversed the test leads no beep. Switch them back and got a beep. I thought something was wrong there so I just pulled the led off the heatsink, which was attached with JB Weld also. I confirmed that I had no shorts and I fired up the led again. That's when it went poof.

So you think because it wasn't heatsinked is why it smoked?

What about that contact I found in the diode test setting of the meter too. I never fully understood what the beep meant. I know a steady tone means I have continuity. Does that mean there is some capacitance there? I just don't think it was normal for the light and the buckpuck to run so hot that it is difficult to touch. It did not get that hot the first time I ran it.

 

[BentHeadTX]

BUT; when I noticed it getting very hot on the second run I thought something was wrong so I took it apart and started doing meter checks for a short. The ohms checks found nothing. The checks with the meter set to diode made an audible beep and showed a value on the display via the test leads touching the led solder lead and to the heatsink. When I reversed the test leads no beep. Switch them back and got a beep. I thought something was wrong there so I just pulled the led off the heatsink, which was attached with JB Weld also. I confirmed that I had no shorts and I fired up the led again. That's when it went poof.

So you think because it wasn't heatsinked is why it smoked?

What about that contact I found in the diode test setting of the meter too. I never fully understood what the beep meant. I know a steady tone means I have continuity. Does that mean there is some capacitance there? I just don't think it was normal for the light and the buckpuck to run so hot that it is difficult to touch. It did not get that hot the first time I ran it.

Mof,
You gave up the life of the K2 for science! Congrats on passing your first mod test of making "magic smoke". Did some pondering on this one.

This is just a wild guess but it would explain what happened. When you first ran the LED it heated up all the connections and JB Weld. Then you allowed it to cool which caused something to crack (total guess, don't throw anything at me) When you cranked it up again, there was conduction which really heated things up. Once you removed the LED from the heat sink, the amount of heat a high voltage K2 generates smoked itself.

I once wanted to test an AMD Athlon 850 CPU without a heat sink to see if it worked. Put my thumb on the die and turned the computer on. It reached the surface of the sun temp in about two seconds so I pulled the power cord. My thumb had an AMD tatoo for a few weeks. LEDs do the same thing to a lesser extent.

Diode checks and you. A diode allows current flow in one direction but stops it when the polarity is reversed. This explains why you get the beep when the positive and negative is correctly connected (beep) and it reads "open" (no conduction) when wired backwards. If it beeps in both directions, you have a shorted diode and if it don't beep in either direction the diode is open.

 

[jtr1962]

You must know your stuff because you ask the right question. It was heatsinked to a copper pennie which was JB Welded to a threaded barrel peice that was part of a Brinkmann Legend flashlight which fitted into the head of the light and the reflector was modded to fit perfectly against the led body. It did put out a nice beam with even spread of the hot spot.

BUT; when I noticed it getting very hot on the second run I thought something was wrong so I took it apart and started doing meter checks for a short. The ohms checks found nothing. The checks with the meter set to diode made an audible beep and showed a value on the display via the test leads touching the led solder lead and to the heatsink. When I reversed the test leads no beep. Switch them back and got a beep. I thought something was wrong there so I just pulled the led off the heatsink, which was attached with JB Weld also. I confirmed that I had no shorts and I fired up the led again. That's when it went poof.

The problem was using JB Weld instead of some sort of thermal epoxy. First, JB Weld has lousy thermal conductivity. Second, it ends up in a pretty thick layer which further hurts thermal conductivity. It often ends up with a few air bubbles when you apply it, further inhibiting thermal conductivity. Third, expansion and contraction from heating/cooling undoubtedly caused some air pockets to form, perhaps partially detaching the heat sink. Reason number three was the ultimate cause of failure. Not surprised this happened. You're dealing with a few watts trying to get out via a few hundreths of a square inch. That's a power density well over 100 watts per square inch. The interface has to be very smooth and thin or you'll encounter problems.

 

[frenzee]

I agree with jtr1962. I gave up on JB Weld because it really doesn't make a good thermal interface compund. For one, it is filled with iron particles (hold a supermagnet over the blue tube or some dried pieces and you'll see what I mean). Iron isn't a good heat conductor to begin with and it makes the compound's electrical conductivity unpredictable. Secondly, when heated to high temperatures, it loses it's adesion properties and crumbles like crayon. If you didn't do good surface prep, it will almost certainly just come off and you won't even notice it until it's too late.

I suggest using AA or AS. AA, with proper surface prep (use fine steel wool or Scotchbrite or better yet use a small steel brush to spread the epoxy on both surfaces), makes a very strong initial bond with both copper and silver. I believe the bottom of the K2's slug is copper with silver coating. And if you let AA cure for a few minutes at around 200F, the bond becomes even stronger and it turns into this ceramic-like substance like the stuff Lumileds uses to bond Luxeons to the star's slug. Strong stuff. If you really want to go all out, mix AA with an equal part of fine (50-100 micron) copper or diamond powder. You can make copper powder yourself by rubbing copper tubing or solid-core electrical wire against a diamond file or knife sharpener, and diamond powder can be bought fairly cheaply online ($5-10 for a 2 gram vial) e.g. here. Of course don't use copper if you need electrical isolation.

If you don't care about electrical isolation, soldering the slug directly to the heat sink is by far better than any epoxy, but you have to use a low temp solder (and flux of course).

 

[mofiki]

I just want to thank you guys for taking the time to help me out. I ordered some AA yesturday and I'm halting any more building until I get it. I guess there are no short cuts and I should pay more attention to the details; i.e. heatsinking.

Heatsinking seems to be somewhat of a science in itself, based on what I've read here. I have no real methods, tools, or enough knowledge to factor proper heatsinking based on actual temps. I have seen some here disregard it by making projects that have such small areas for the heat to escape and state that as long as you keep moving it's ok. I don't want to build projects like that. I've built quite a few halogen lights for biking and sold them to the biking community. I would feel ashamed if I sold junk. Same goes for this new LED technology.

Thankx again everyone.

 

[BentHeadTX]

Thanks jtr for the warning on JB Weld. I have never used it for electronics as Artic Silver thermal epoxy is something I keep in stock for my weirdness.

Now for the K2-240 with lower voltage to roll out so I can mod my existing lights. I am not holding my breath.

 

[Gryloc]

I ordered 4 of the 0180 parts last week, and I never had the chance to test them out. I tested 2 out of the 4 tonight and here are the measurements that I got:

LXK2-PWC4-0180

Current   Voltage
  (mA)      (V)             <Update:>
         #1       #2       #3       #4
  50   3.149    3.089    3.055    2.88
 100   3.28     3.222    3.184    3.011
 150   3.38     3.30     3.26     3.102
 200   3.456    3.37     3.33     3.176
 250   3.52     3.42     3.38     3.236
 300   3.57     3.48     3.44     3.27
 350   3.62     3.52     3.48     3.31
 400   3.66     3.56     3.53     3.35  
 450   3.70     3.60     3.56     3.39 
 500   3.73     3.63     3.60     3.42
 600   3.79     3.70     3.67     3.48
 700   3.84     3.77     3.73     3.53  
 800   3.90     3.82     3.78     3.58  
 900   3.95     3.84     3.84     3.62  
1000   3.99     3.89     3.88     3.66  
1100   4.03     3.93     3.93     3.70  
1200   4.09     3.96     3.97     3.73

In my test rig, I have a Fatman boost regulator powered from a battery pack. The Fatman has an external potentiometer attached so I can change the amount of current going to the emitter. The small sense resistor on the Fatman was used like a shunt resistor where I can measure the current with my volt meter by measuring the voltage drop.

So, I will try to get the voltages of the last two K2-0180 emitters and update this post later.

I am a little saddened by the high Vf of the new K2 emitters, but this isn't the end of the world. I have a certain application for all four, even if it means that a fraction of a watt more will be consumed compared to other emitters to produce the same amount light. If I needed a more efficient emitter, I would have grabbed a few of the older P4 XR-Es. Then again, the beam pattern of the lambertian emitter will work better than that of the XR-E. All that matters is that these emitters work, and they are still better than the old LuxIII and K2. Besides, I really like the tint of the batch of new K2s that I got. They are very white. Since the Rebel 0100 is so hard to find easily, these will do...

-Tony


<Update:
So, after too many days, I got around to testing the other two. Just a reminder, these are fresh from the cut reel packaging. Vf may drop, but maybe for just a select few of the K2 emitters and only a small fraction of a volt. Emitter #3 was used for more than 30 minutes over the last week in the test rig, and I found that the Vf did not drop one bit since the initial testing. Too bad.

I was rather amazed by the low (or maybe normal) Vf of emitter #4. At 1500mA, I measured a Vf of 3.83V (this is the only emitter out of the four so far that I went to 1500mA since I do not want to ruin my only test rig). This Vf is close to the Vf of the first K2 0180 emitter at 700mA!!! Wow! Good thing that these will be used in series in an application where heat and efficiency is not the most important. As I find anything more about these emitters from my testing, I will post somewhere in this thread. I ordered another K2 0180 with another Future order, so I may get some more variety of color and Vf due to the good 'ol "Luxeon Lottery".>

 

[TorchBoy]

I never knew diamond powder was so inexpensive. At (from memory) ten times the thermal conductivity of aluminium, and an electrical insulator, a big solid diamond would be the ideal stuff to make a heatsink from... if only it wasn't so expensive. :sigh:

 

[SemiMan]

I got my K2s fired up and did some testing tonight. I am running strings of 4 in series, 2 neutral and 2 cool. I have not done any detailed testing, but when I run these strings of 4 at 700mA, I got forward voltages of 13.7 for one string and 13.8 for the other. This was right after they were turned on with just about no time on them. I would expect the Vf to go down with temperature and a bit of burn in. 13.7 and 13.8 represents 3.425v and 3.45v respectively per LED. I double checked the current and the voltage and everything looked good. 3.45 at 700mA is pretty good.

Semiman

 

[Curious_character]

So at 700 mA, we have. . .

evan9162: 3.85 - 4.12 v.
Gryloc: 3.77 - 3.84 v.
SemiMan: 3.425 - 3.45 v.

Looks to me like if we wait a few more days, Vf'll be down to a couple of volts.

Seriously, does anybody have a clue about the cause of the apparent wide variation?

c_c

 

[SemiMan]

Technically it is all within the specification of the forward voltage of the product which is quite broad. My measurements are just at the typical. The die is the same as the Rebel and some Rebel-100's I had had a Vf of approximately 3V at 350mA, so I imagine we are seeing process variation and things will get better as they start cranking of manufacturing. I am interested to see what happens with burn in. I may get less voltage drop than those with higher Vf bins.

Semiman

 

[soffiler]

Technically it is all within the specification of the forward voltage of the product which is quite broad. My measurements are just at the typical. The die is the same as the Rebel and some Rebel-100's I had had a Vf of approximately 3V at 350mA, so I imagine we are seeing process variation and things will get better as they start cranking of manufacturing. I am interested to see what happens with burn in. I may get less voltage drop than those with higher Vf bins.

Semiman

Just a thought (and I may be off base here, as I am well behind you folks in my knowledge of both Rebel and K2):

Is it possible that, if both Rebel and K2 are using the same die, that they are binned and Rebel is getting all the lowest-Vf bins? That would mean Rebel has less waste heat to deal with for a given drive current, which is consistent with the small package size. Then the higher-Vf bins have to go somewhere - voila, K2.

 

[SteveDavis]

Curious_character:

Thank you for the insight on Vf in relation to total efficiency. You're certainly correct that a buck or a boost is almost always more efficient than a buck-boost, and also certainly cheaper.

However, I think it's important to point out that this data could be very suspect. The range of Vf's for a die is extremely high. The new K2 is spec'd to have a Vf between 3.03V and 4.71V at 1A.

When I recommend to a flashlight manufacturer a driver circuit, I look at the entire Vf range for this reason.

 

[SemiMan]

Technically they should not be able to do this...put the high Vf bins into the K2 as they would violate their typical Vf number by doing it. We are currently working with a sample size of <20 which may have been only two reels worth. It looks like we all got them about the same time, so it is possible that they were broken off the same reels. Gryloc, Evan did you get them from Future?

The max on the data sheet is pretty high. I wonder if under burn in this would come down? Cree does not list the max at 1A, but based on the max at 350mA and typicals at 350ma and 1A, it is likely in the 4.3-4.4V range. SteveDavis....you distribute Cree, can you find out what the max for Cree would be at 1000mA?

Semiman

Just a thought (and I may be off base here, as I am well behind you folks in my knowledge of both Rebel and K2):

Is it possible that, if both Rebel and K2 are using the same die, that they are binned and Rebel is getting all the lowest-Vf bins? That would mean Rebel has less waste heat to deal with for a given drive current, which is consistent with the small package size. Then the higher-Vf bins have to go somewhere - voila, K2.

 

[evan9162]

Yep, got them from future.

The wierd thing was that, for every other order I've ever made from future, I've had a bin code on both the invoice, and package label on the anti-static bag.

For these, there were no bin codes to be found anywhere.


Anyone else have bin codes on their package labels/invoice for these new K2s?

 

[SteveDavis]

Max Vf for an XR-E at 1A is 4.3V