Another frozen luxeon!

[evan9162]

With one difference...This time, I didn't hold back...Instead, I instituted a full-on tourture test of a 1W White HD, bin P2K.

I used a similar setup to the 1W R-O supercooling experiment:

The same 35W peltier, dipped in ice water for heat dissipation.

The rig to the left is my LM338-based voltage regulator. This will crank out over 5A, and is adjustable from 1.2V to 10.5V. I have 2 computer power supplies, one for 12V to the peltier, one for 12V to the adjustable "luxeon frying" power supply.

The luxeon is stuck to the cold plate of the peltier using the "sticky" properties of normal silicone heat paste. This is a bare emitter, so the junction-cold plate thermal resistance is about 15C/W. The cold plate is kept at -20C.

This is run through a 1.2 Ohm, (and later, 0.3 Ohm to enter "fry" mode) resistor. I need 0.3 Ohm since 1.2 Ohms only allowed me to crank it up to 2.4A. The luxeon power supply seems to puke out after about 2.4A.

The data/pictures I have are for 400mA, 800mA, 1.6A, and 2.4A. First, projecting on the cieling at those currents through an NX05 optic:

Notice that even at 2.4A, there is still an increase in light output. Also notice something else: at the lower currents, this white had a slightly bluish tint. At extreme current (2.4A), it didn't get angry blue, it got ANGRY GREEN! /ubbthreads/images/graemlins/eek.gif

And the data:
<font class="small">Code:</font><hr /><pre>
Current Vf P Tj (est)
400mA 3.34V 1.34W 0C
800mA 3.59V 2.87W 23C
1.6A 4.04V 6.46W 77C
2.4A 4.52V 10.8W 142C <-- WOW! (Angry Green color)
3A FAIL!!!!!
</pre><hr />


I examined the luxeon after the failure. I noticed a bubble around the negative bond wire. Closer inspection reveals discoloration and what appears to be soot-like deposits on the inner surface of the bubble.

Observe:

It appears the bond wire fused open at this point.

One more luxeon down in the name of science! /ubbthreads/images/graemlins/tongue.gif

 

[kj]

Great! /ubbthreads/images/graemlins/thumbsup.gif

 

[shiftd]

nice /ubbthreads/images/graemlins/thumbsup.gif

 

[Mark_Larson]

You really need a better mounting method.

Also, use superchilled ice+salt mixture.

Also, a simple fan or stirrer in the mixture will drastically lower thermal resistance.

 

[Azecos]

Correct me if I'm wrong, but it looks like the die is still intact?

 

[jtr1962]

You should be able to get a lot cooler using a liquid heat sink and a larger wattage module. Using a home-made copper liquid heat sink with tap water cooling I can obtain temperatures under -35°C even in the summer, and under -45°C in the winter. In the absence of a liquid heat sink, a larger air-cooled heat sink will work wonders. I can get under -25°C using forced air cooling in a 70°F room. A couple of points to remember are:

1)Use stainless steel screws to clamp the module between the cold plate and heat sink. Stainless steel conducts heat less than regular steel.

2)Have all surfaces machined to a flatness of 0.001" to prevent losses at the thermal grease interface.

3)Have a clamping pressure of about 300 psi to ensure good thermal contact between the module and heat sink.

4)Use fiber insulating washers between the screws and the cold plate to reduce thermal shorting,

5)Ramp the module current in small increments(5 to 10% of Imax), and measure the cold plate temperature after it stabilizes(usually takes about 5-10 minutes). Eventually, you'll reach a point where further increases in current result in higher, not lower, temperatures. Back off a bit from that point and use that as your running current.

Number 5 is particularly important. Many people using peltiers make the mistake of driving them at Imax or some set voltage(usually 12V). If you're using a small heat sink you will reach the point of diminishing returns long before than, and will actually have higher temperatures, not lower ones, by increasing current further(i.e. the delta T may be higher at higher currents but the heat sink temperature rises faster than delta T past a point). Also, driving a Peltier at any more than about 90 to 95% of Imax is pointless even with a liquid heat sink. You just don't get any colder at that point. I cringe when I hear about people "overdriving" Peltiers. All they're doing is making heat.

BTW, interesting results but seeing the size of the bond wires on the LS 1 watt I would be afraid to go much past about 500 or 600 mA. I'm amazed it lasted up until 2.4A. I'm thinking of doing something similar with one of my Q3 LSs. I plan to cool it and see how much difference in light output there is at a given drive level. However, I'm not planning on overdriving as I plan to use the LEDs in projects. Any suggestions for a crude light meter? I'm mainly interested in relative, not absolute, light levels here.

 

[WildRice]

hey, since you are calling it dead, can I have it to disect one if these things. /ubbthreads/images/graemlins/grinser2.gif always wanted to see what held them together.
jeff

 

[Doug S]

Darin, I love this stuff. Your data is the best that I've seem for bracketing the actual fuse current of the bond wires. While I look forward to seeing for myself, some here on CPF have suggested that the bond wires on the Lux IIIs are actually smaller than those of the 1W. If so, user beware.

 

[evan9162]

Mark, Jtr:

I wasn't really after good supercooling, just "good enough" for this experiment. The point of this was to push the luxeon to its limits, but not be thermally limited. I wanted it to break because of extreme current.

The die is, in fact, completely in tact. If one were to pop off the dome, I'll bet it would still light up if current could be applied to the remaining part of the negative bond wire.

This "instant luxeon death" current of 3A probably should be derated, since the cooling would have prolonged the life of the bond wires by cooling them too. However, I suspect that there is only so much that cooling a conductor of that size will do to prevent it from fusing. In addition, it's entirely possibly that I had an extremely tough luxeon (I often gripe about others using a sample size of 1 to declare some fact, and here I'm nearly doing the same /ubbthreads/images/graemlins/tongue.gif)


So far, the most suprising thing was that at extreme currents, this particular emitter turned angry GREEN, not blue like others have so often observed in their extreme overdrives. I really don't know what to make of that.

 

[Xrunner]

/ubbthreads/images/graemlins/thumbsup.gif Great work. Do you happen to know what the rated operating temperature of the luxeon is supposed to be?
Thanks

-Mike

 

[Doug S]

[ QUOTE ]
evan9162 said:
So far, the most suprising thing was that at extreme currents, this particular emitter turned angry GREEN, not blue like others have so often observed in their extreme overdrives. I really don't know what to make of that.




[/ QUOTE ]

I have long been doubtful about the belief that overdrive universally yields bluer output. I believe that this is individual LED specific. While I suspect that your result is specific to the sample, there is another factor worth considering. Both drive current and Tj have independent affect on the operating point in the Chromaticity diagram. Your test shifts the relative effects of the two compared to the usual overdrive scenario. BTW, the independent affects of drive current and temperature are best illustrated by the latest *Nichia* datasheets. The physics should be the same for the Luxeons.

 

[IsaacHayes]

Maybe since it was so well cooled the extra bright blue light didn't overwelm the phosphur as much and as a result was still putting out a lot of yellow to mix green. Maybe overdriven non-supercooled ones the phosphur tends to degrade and not emit as much yellow light due to the heat and thus more blue comes out and looks just blusih?? /ubbthreads/images/graemlins/icon3.gif

 

[evan9162]

Issac,

That's a possibility. When I get home tonight, I may have to browse Lumileds presentations to see if they mention a shift in spectral output from their phosphors due to temperature.

However, the 1.6A test had a Tj of about 77C, which is a pretty normal operating point. I would suspect that the phosphor was also at or near this temperature as well. So at this current level, I shouldn't have seen any real color shift caused by cold temperatures.

Doug,

I would suspect the 40-50C difference in operating point that I created did shift things around a bit. At 800mA and 1.6A, the junction (and likely phosphor) temperature are at "normal" levels, although the drive current isn't. The phosphor is affected by input light from the die, and (possibly) temperature. The die's output light is likely affected by temperature and current.

Could it be possible that at higher current levels, there is some sort of (non-permanent) breakdown in the electrical properties of the junction, allowing electrons to jump across a shorter band gap? Maybe some kind of "saturation" happening? This might explain an increased output in longer wavelength light...

DS48.pdf (Luxeon DCC display backlight) showed a graph of blue output vs. temperature. The blue luminous flux actually increased with temperature, due to the temperature-induced shift to lower wavelength light, and the eye's increased sensitivity at those wavelengths.

 

[evan9162]

[ QUOTE ]
5)Ramp the module current in small increments(5 to 10% of Imax), and measure the cold plate temperature after it stabilizes(usually takes about 5-10 minutes). Eventually, you'll reach a point where further increases in current result in higher, not lower, temperatures. Back off a bit from that point and use that as your running current.

Number 5 is particularly important. Many people using peltiers make the mistake of driving them at Imax or some set voltage(usually 12V). If you're using a small heat sink you will reach the point of diminishing returns long before than, and will actually have higher temperatures, not lower ones, by increasing current further(i.e. the delta T may be higher at higher currents but the heat sink temperature rises faster than delta T past a point). Also, driving a Peltier at any more than about 90 to 95% of Imax is pointless even with a liquid heat sink. You just don't get any colder at that point. I cringe when I hear about people "overdriving" Peltiers. All they're doing is making heat.


[/ QUOTE ]

Woah /ubbthreads/images/graemlins/eek.gif

I totally didn't get what you were saying first time I read this. Good idea! I'm going to give this a shot tonight. Thanks for posting this.

I've tried insulating, and even pre-freezing the whole apparatus, and the coldest I've ever gotten is -32C. But the constant for every attempt was that I was driving the peltier with 12V, and just leaving it at that. Now I'll try incrementing the power to the peltier and see what I get.

Thanks /ubbthreads/images/graemlins/thumbsup.gif

 

[Doug S]

[ QUOTE ]
evan9162 said:
Doug,

Could it be possible that at higher current levels, there is some sort of (non-permanent) breakdown in the electrical properties of the junction, allowing electrons to jump across a shorter band gap? Maybe some kind of "saturation" happening? This might explain an increased output in longer wavelength light...



[/ QUOTE ]
I don't know. Interesting idea.

[ QUOTE ]
evan9162 said:
DS48.pdf (Luxeon DCC display backlight) showed a graph of blue output vs. temperature. The blue luminous flux actually increased with temperature, due to the temperature-induced shift to lower wavelength light, and the eye's increased sensitivity at those wavelengths. {Doug's comment: I believe you mean *longer* wavelength or *lower* frequency}



[/ QUOTE ]
This would make sense as an explaination for the increased "blueness" at higher temps. Another consideration is that the phosphor has an optimum pump wavelength, above or below which it absorbs less of the pump photons. If the blue die peak is shorter than optimum, increasing temps should enhance absorbtion [greener output], if blue die is optimum or longer wavelength then increased temp should result in decreased absorbtion [bluer output]. Sorry, I don't have a picture in my head of what the phosphor efficiency vs wavelength curve actually looks like for the Luxeon phosphor so the above is just theoretical speculation only.

 

[evan9162]

Doug,

I looked at the Lumileds presentations. One thing that they showed was a broad excitation spectrum of blue for their yellow phosphor. One thing that I did notice was that at extreme temperatures, the peaks of the output spectrum of the phosphor did not shift. However, relative intensity of the spectral peaks of output from the phosphor did seem to change -- that is, the green spectrum seemed to maintain output more than the other spectrums.

 

[Entropy]

I don't think the phosphor will significantly change its color with temperature.

BUT the phosphor probably changes its efficiency and linearity properties with temperature. So while the phosphor's output might not change in color, it WILL change in intensity with respect to the amount of excitation.

I believe the reason Luxeons normally become bluish when overdriven is because the phosphor's response is not completely linear (thus an increase in blue excitation light does not necessarily increase yellow transmitted light) and I'm pretty sure the phosphor becomes less efficient with decreased temperature.