Dropping Vf over life of luxeon (5W)

I know we've discussed this before - but after measuring the current on one of my 5W mag mods, I decided to bring it up again.

With 1 ohm in a 6AA battery holder, fresh NiMHs would push a pleasant 1A through the VV0U emitter which I got back in February from the sandwich shoppe. At the time, I was surpised to find out that the Vf of this emitter was significantly lower than the U-bin that it was marked with (it was clearly a T-bin).

Now, with a fresh set of NiMH AAs, over 1.5A is pushed through the emitter /ubbthreads/images/graemlins/eek.gif. Same light, holder, resistor, etc. - just 3 months later.
Using my LM317 based current supply, I plotted the Vf curve. Here's the data for it (then and now):

<font class="small">Code:</font><hr /><pre>
VV0U Vf
I(ma) february now
0.1 4.94 4.89
30 5.67 5.47
130 6.04 5.77
310 6.34 6.00
620 6.64 6.24
1060 6.92 6.47
</pre><hr />

Nearly 0.5V drop at high currents. This emitter is well heatsinked, has never been severly overdriven nor overheated, and has only been run maybe 5 hours since then. I checked a V3T 5W light, and it's Vf is identical to what it was 6 months ago (about the same usage between the two).

I can't remember if we came up with a clear explanation or not for this Vf drop phenomenon - so here's some more data to ponder.

Darin, I think it was me that brought this up before. At that time I had seen this happen with one emitter. I have now personally seen it with two. In both cases I was not specifically looking for this phenomenon, I just happened upon it in the course of other testing. In neither case had the emitter been abused. I have no theories about the underlying cause. I would like to know how common this is. It does tend to be another arguement against using stiff voltage regulation for powering Luxeons. Too bad, as there are occasions that I would like to be able to get away with a voltage only regulation scheme.

Darin, I note from your data that Ri has also decreased.

Yep, Rdyn has decreased from 1 ohm at 700mA to 0.8 ohms at 700mA.

I'd hazard a guess this is related to the LuxV Portable's relatively short lifespan.

Well, in case the less knowledgable folks don't follow this thread yet...

When a LED die (the light source in a luxeon) begins to loose it's thermal transfer, the die heats up, dropping it's forward voltage lower and lower, and the Luxeon gets less efficient (less light output).

Too bad you don't have some light measurements on that sucker, evan8162....

[ QUOTE ]
idleprocess said:
I'd hazard a guess this is related to the LuxV Portable's relatively short lifespan.

[/ QUOTE ]

This was a hazardous guess, /ubbthreads/images/graemlins/tongue.gif The two cases of this phenomenon that I have seen involved 1W Luxeons.

[ QUOTE ]
NewBie said:
Well, in case the less knowledgable folks don't follow this thread yet...

When a LED die (the light source in a luxeon) begins to loose it's thermal transfer, the die heats up, dropping it's forward voltage lower and lower, and the Luxeon gets less efficient (less light output).

Too bad you don't have some light measurements on that sucker, evan8162....

[/ QUOTE ]

I have my doubts about this explaination. If it were due to higher die to slug thermal resistance, it would not also be evident at low drive currents due to the negligible heating. In the case of the two examples I have seen, total runtime was in the minutes for one and a few hours for the other. In both cases, the reduced Vf was evident even at low drive currents.

[ QUOTE ]
NewBie said:
Well, in case the less knowledgable folks don't follow this thread yet...

When a LED die (the light source in a luxeon) begins to loose it's thermal transfer, the die heats up, dropping it's forward voltage lower and lower, and the Luxeon gets less efficient (less light output).

Too bad you don't have some light measurements on that sucker, evan8162....

[/ QUOTE ]

Nope, I don't have light measurements, but I do have another 5W that's roughly equal in brightness, and I can't detect any brightness difference between the two. Besides, these measurements were 1-2 second measurements when the light was cold (this is my standard measuring technique).

Anyways, to account for that Vf drop due to the die heating up (0.4V) at -0.004V/C (according to the datasheets) would require a 100C difference between the two measurements. There isn't even a 100C difference between the die temp with no power, and the die temp at 1A.

I have also observered such a Vf drop after short usage (on a VWOV LED).
One possible hypothesis: A reduced contact resistance (between semiconductor and metal pad) caused by the drive current. Such a behavior could be caused by a contamination or very thin oxide between the semiconductor and contact material.

PeterB is on to something I belive. Remeber (ack, who was it?) that had their cat toast their KL4 in the breat toaster? Remeber how his runtime was improved? That would no doubt be due to a lower Vf, requiring less current to be pulled from the batts for the circuit to boost the voltage...

I'm thinking the same thing's going on somewhere... Heat is re-arranging something and making it work/conduct better...

If you got your emitter hot enough, you could cause the solder balls binding the die to the contact pads to soften and reflow, possibly improving contact resistance somewhat. I don't think this luxeon has ever been exposed to that high of temperatures.

The other thing that may cause it would be extreme thermal cycling (heating up very quickly). But just wild random guesses here.

[ QUOTE ]
NewBie said:
...When a LED die (the light source in a luxeon) begins to loose it's thermal transfer, the die heats up, dropping it's forward voltage lower and lower, and the Luxeon gets less efficient (less light output)...

[/ QUOTE ]

This is confusing to me. I'm not technically knowledgeable but I thought lower vf was desirable because it increased efficiency.

Brightnorm

Brighty,
in general, it is true that lower Vf increase efficiency OF the converter board it was used with (step up, i believe, take more advantage if lower Vf). But what newb mention on the *edit decreasing *end of edit* in the Vf is because the increase in the led die temperature and we all know full well (i hope) that the higher the temperature of the die of the led, the less efficient the led become in converting current to photons.
So if the led get hotter and hotter, the converter works more efficiently in giving the current to the led (because lower Vf) but in the end, the drop in the led efficiency in converting the current to photons is so much bigger than the efficiency gain from the lower Vf that overall, you lose efficiency.

[ QUOTE ]
evan9162 said:
[ QUOTE ]
NewBie said:
Well, in case the less knowledgable folks don't follow this thread yet...

When a LED die (the light source in a luxeon) begins to loose it's thermal transfer, the die heats up, dropping it's forward voltage lower and lower, and the Luxeon gets less efficient (less light output).

Too bad you don't have some light measurements on that sucker, evan8162....

[/ QUOTE ]

Nope, I don't have light measurements, but I do have another 5W that's roughly equal in brightness, and I can't detect any brightness difference between the two. Besides, these measurements were 1-2 second measurements when the light was cold (this is my standard measuring technique).

Anyways, to account for that Vf drop due to the die heating up (0.4V) at -0.004V/C (according to the datasheets) would require a 100C difference between the two measurements. There isn't even a 100C difference between the die temp with no power, and the die temp at 1A.

[/ QUOTE ]

Humm, hello!

You put 1A into the part and you show 6.92V, which equals 7.3352 Watts. Assuming your part is the better LXHL-PW03, at 8 C/W, 7.3352W * 8C/W = 58.7 C that means 58.7 degrees C over slug temperature.

Assuming the part is the other devices LXHL-LW6C or LXHL-FW6C, you have 11 C/W * 7.3352W = 80.7 C over slug temperature. (ah, just found you said emitter...)

So, 58.7 C + rise due to thermal resistance of slug to heatsink, + thermal resistance from heatsink to flashlight body + the thermal resistance from the flashlight body to the air + about 25 C ambient air...

Puts you at 83.7 C + rise due to thermal resistance of slug to heatsink, + thermal resistance from heatsink to flashlight body + the thermal resistance from the flashlight body to the air. I could see some pretty decent temperatures inside the Luxeon.

But, you mentioned:

"Now, with a fresh set of NiMH AAs, over 1.5A is pushed through the emitter" (so the part was abused here)

So, if I take 1.5A, which probably pushes you around 8V Vf or so, and 1.5A, 1.5A * 8V = 12 Watts. 12 Watts * 8 C/W =
96 C over slug temp.

So 96C + 25C = 121 C (still missing the rest of the answer)

121C + rise due to thermal resistance of slug to heatsink, + thermal resistance from heatsink to flashlight body + the thermal resistance from the flashlight body to the air.

This means Danger Will Robinson, Danger! Sounds very possible for bad things to happen.

Okay, so we hit 121C quite easily and most likely *way* over this (remember we still need to add thermal resistance from die to heatsink, thermal resistance through heatsink, thermal resistance from heatsink to flashlight body, plus thermal resistance from flashlight body to ambient air...
(seem to recall the Luxeon class instructor mentioning 143 is definitely reflow temps). Humm... Now -0.004V/C...

0.004 * 121C = 0.484 V drop in Vf.

Hows that you said again???

You've made some bad assumptions.

First, in quick 1-2 second tests, the only thermal resistance that matters is the junction->slug thermal resistnace. Since nothing else has heated up, the slug is held at 25C, so the junction temperature rise is:

8*Vf*If

In the case of the 1A test, the junction temperature rise is:

8*(6.47*1) = 52C
which puts the junction temp at 77C


Second, I measured the Vf at 1.5A to be about 6.6V, not 8V as you assumed. Thus, the junction temperature rise is only:

8*1.5*6.6 = 80C
that puts the junction temp at 105C

Note that the Lumileds datasheets specify the negative temperature coefficient for a rise over 25C, not 0C.

Finally, you completely missed that I measured the same emitter under the exact same conditions at two different times. Nothing was different about the conditions which I measured Vf (temperature, power supply, contact locations, techniques). So the whole Vf changing due to temperature argument is bunk, since the junction temperature should have been about the same between the two measurements.

Temperature doesn't even come into play for the 130mA and 310mA measurements, and those changed just as much as the higher current measurements. At those currents, the junction is just a few degrees above ambient, so Vf dropping due to temperature is minimal, yet the Vf is lower at those currents as well.

If you say so.

Abuse, 1.5A, over 2x the parts rating...

Also, VV0U Vf

I(ma) february now

1060 6.92 6.47

Wow, thats a huge change.

Care to measure it again, now?

So I asked Christian today. The guy at work with the CRADA with Lumileds. He says it's a know to happen thing, not all that repeatable and the cause definitely unknown. Answer the question and you win the prize. And he should know.

Think about what this means to those who favor voltage regulation over current regulation....

Or direct drive.

Doug Owen

Good point Doug.

It's looking like the voltage shift occurs most when the parts are being driven hard.
I have noticed that several of my Luxeon lights have dropped a bit in forward voltage when driven hard at over 700 milliamps on foot patrols around the property.