No real benefits to overdriving XML

[LMF5000]

Regarding running LEDs at very cold temperatures - there's a very easy way to do it. Very very easy in fact. All you need is a thermoelectric cooler (T.E.C.), also known as a Peltier cooler. If you've never seen one in action, you'll be amazed. I have two of them. It's basically a flat plate with wires sticking out of the edge. You apply power to the wires, and one side of it gets ice cold (I've reached -10 deg. C with mine) whilst the other end gets hot. With current technology, for every watt of electricity you put in, you get a watt of cooling.

In other words, if you put in 10W of electricity, the cold side will suck heat at a rate of 10W. The hot side will pump out heat at a rate of 10W (transferred from the cold side) + 10 W (from the electricity being pumped in) = 20W total to dissipate. So the heatsink must be good for 20W in this case.

Or to apply it to an XM-L with 5A at (only guessing here), 4V, that's 20W. So you'll need 20W to power the LED, 20W to power the Peltier cooler, and your heatsink will be connected to the hot side of the peltier and dissipating 40W. Obviously there's a limit to the temperature difference a Peltier can maintain (about 60 deg. C max, normally - and that's without the LED loading the cold side), so the cooler the heatsink stays, the cooler it can keep the LED. In other words, use as big (or effective) a heatsink as possible. Big CPU coolers like the XigmaTek Dark Knight with fan at full speed (what I use) work best.

Peltier coolers are cheap and plentiful on probably all of these online sites, but here's one that can take a 60W load and costs less than £3 : http://cgi.ebay.co.uk/12V-60W-TEC1-...al_Components_Supplies_ET&hash=item19c3456308

 

[Curt R]

I've been thinking about incorporating a heatpipe setup into a flashlight... Has anyone heard of such a thing?

Shao

We have tried that and the problem is most heat pipes are designed for
vertical use only. You need special internal parts to make them usable
for all orientations. The limiting factor is the interface between the flashlight
head and the air. Air is a very good insulator and using a heat pipe verses
what we have designed for our lights is within a few percentage points of
efficiency. We even made our own very small heat pipes using C12200 tubing.
Just not cost effective.

We call our design DTT, or Direct Thermal Transfer. We solder C14500 directly to
the thermal pad of the Cree XML and the Luminus SST-90 LEDs. These are then
inserted into a second larger C14500 slug with an interference fit. This then goes
deep into the Aluminum head that has fins to increase surface area. There is only
so much that can be done with the limited physical size of a flashlight without
making it too bulky and heavy to use.

Curt

 

[CKOD]

We have tried that and the problem is most heat pipes are designed for
vertical use only. You need special internal parts to make them usable
for all orientations. The limiting factor is the interface between the flashlight
head and the air. Air is a very good insulator and using a heat pipe verses
what we have designed for our lights is within a few percentage points of
efficiency. We even made our own very small heat pipes using C12200 tubing.
Just not cost effective.

We call our design DTT, or Direct Thermal Transfer. We solder C14500 directly to
the thermal pad of the Cree XML and the Luminus SST-90 LEDs. These are then
inserted into a second larger C14500 slug with an interference fit. This then goes
deep into the Aluminum head that has fins to increase surface area. There is only
so much that can be done with the limited physical size of a flashlight without
making it to bulky and heavy to use.

Curt

Everything else is right on, but the bolded portion isnt, at least the "most" part. Almost every heat pipe used in PC applications is not in the vertical positions. CPU coolers for desktops look like they would work with a "vertical only" heat pipes, until you put them on their side in a tower. Laptop heatpipes would all be horizontal, and same with video card coolers. If youre buying heatpipes though, its something to keep in mind.

The only places where heatpipes may help in a LED application are areas where thermal gradients are high(very close to the LED). If you had a heatpipe with sufficient capacity, and soldered an XM-L or other LED directly to the heatpipe before severely overdriving the LED, I suspect there would be some advantage over just being on a copper slug. IVe seen some heatpipes specified as .01 K/W or less, I.E. putting 20W in from an LED would yeild only a .2 C temp gradient from end to end. Now, you have to consider size of the heatpipe and actually fitting these tiny LEDS on, bends reduce their efficiency etc. Even if there was a 2C temp rise, that would still be great. In the long run though, youre absolutely right in that the flashlight -> air junction is the limiting factor.

There is also another variation of the heatpipe. The "vapor chamber" same thing as a heat-pipe, just not in a pipe shape. I think they would be more suited to flashlights in some cases then a heat-pipe would be, but I doubt enough high power flashlights could be sold to justify the cost of getting a run made.

 

[Curt R]

Ideally we would like to have 10 square inches of surface
area for every 1.3 watts of power drive to the LED. When
driving a Cree XML to 3.5 Amps at 3.2 volts the size of the
head would have to have 84 square inches of surface. So
we do with what we can, and the outside surface of the
flashlight head can reach over 170 degrees F. I have seen
many import lights that do not get hot. Taking apart those
3 watt cheapies reveals that the star mounted LED is totally
enclosed in a plastic carrier. Any guess as to how long the
life of the LED is going to be? Optics and thermal engineering
is the key to any good quality high performance light. Just
keep them away from children until educated. These lights
can damage the eye if not used correctly. And they can cause
skin burns. Some of our lights can now equal the output of
three automotive headlamps.

Curt

 

[Walterk]

FYI: I tested my XM-L.
Setup:
XM-L (T6 bin) from DX on copper star bolted with screws and nuts with thermal paste Shin-Etsu to copper slug of DYnatron H185 PC-fanned heatsink run at 7,4V. Driver: Variable driver Der Wichtel.

Without power to fan: 689 Lux
With power to fan: 782 Lux, thats an increase of 13 %.
T6 is supposed to be 280 to 300 Lumen, so thats 270% overdriving.

I got 782 Lumen, but I dont see how that can be managed in a flashlight.
Without a star and another bin and a fortunate lottery might turn the odds though.

It had a very clear point of no-Lumen-increase, and short after that angry blue steps in. Using a welding glass it is clear that while cranking up currents, the phosphor near the bondwires lights up harder and earlier then the other end of the die. Welding glass didn't permit me to see if these differences occur around the angry-blue point. Maybe with a darker glass I could.

Apparently there is a clear and recognizable limit to what current is safe for the Led. I suspect that for all higher currents, the brightness of the die is unevenly spread. That means there can be an edge with higher then expected, and an edge with lower then expected intensity (or was it luminousity). I think I appreciate this flaw, as I guess it accounts for the unexpected results in throw for this led.

 

[LMF5000]

Walterk - interesting theory. I've tested an XM-L stuck to a heatsink designed for an Intel Core i7 CPU (130 Watts) that barely got warm, and using my eyes to judge, I noticed an obvious increase in brightness from 3A to 4A but no significant increase on going from 4A to 5A so the absolute maximum for my T6 LED seemed to be at 4A with air cooling.

But the real purpose of this post of mine is because you mentioned a welding mask. There's actually another way of looking at bright light sources such as LED dies and halogen filaments - using a floppy disk! You basically take an old floppy disk and open the flap (or break it off), then you look through the brown disk itself. One floppy disk blocks enough light for you to be able to comfortably look at the sun without your eyes being dazzled at all. If the LED is too bright for one floppy disk, just grab another disk and put it behind the first one so the light passes through the two of them before reaching your eyes (which would reduce the brightness twice over). Try it and tell me how it goes. I don't have a welding mask so I use floppy disks - it would be nice to hear from someone who has both what the differences are between the two.

 

[MikeAusC]

. . . . Driver: Variable driver Der Wichtel.
Without power to fan: 689 Lux
With power to fan: 782 Lux, thats an increase of 13 %. . . . . .

Walter - how much current was going to the XM-L ?

 

[Walterk]

My main goal for posting is that it would be very interesting to know if its specific for XM-L or that all leds have unevenly distributed illuminance with higher currents. Then overdriving might mean:
270% overdriving for the whole die then (measured from test) 270% of the designed Flux (Lumen).
Average die is 782 Lumen divided by 2x2mm2 surface of the die comes to 195 cd/mm2.
But by my theorie:
Say 25% of the die is brighter , say 235 cd/mm2,
and 75% of the die is slight less, say 185 cd/mm2, together that would still be 782 Lumen for the whole.

But then you would have a large enough spot to focus the optics on, with a higher luminousity then the Cree XRE.

Walter - how much current was going to the XM-L ?

Measured: not with a lab set-up (needed 6 pair of hands I didn't have so didn't measure Lux at same time).
low 1,14A - 3,18V
high 6,2 A - 3,7V

using a floppy disk - it would be nice to hear from someone who has both what the differences are between the two.

Thx, but I will leave that to some one else.
A - I dont have any floppy disks, but thx for reminding me their existence
B - Welding glass (12 or better 13) are designed to protect from harming UV-rays. Floppys are not. Especially important with HID and XSA, and some leds.
C - Welding glass is standardized, so it is easyer to compare results between observations.

FYI I found this link with another test: Testing XM-L, MC-E, and SST-50 emitters up to (and over) 5 amps.
FYI Saabluster tested the XML directly soldered to copper and got to 6 Amp. So there is some insight in the difference between with or without copper star!

 

[jtr1962]

The uneveness you saw might simply be a product of one part of the die getting much hotter than the rest due to uneven thermal impedance. The XM-L has traces to distribute the current evenly over the entire die, so I doubt uneven current distribution is the reason for what you saw. Saabluster's tests confirm this in that he was able to go to 6 amps with better heat sinking. All it might take is an air bubble where the LED is attached to the heat sink to introduce uneven heating. Or maybe you just have a defective XM-L. If you have any spare XM-Ls, try and see if you can observe a similar phenomenon with them.

By the way, I think ALL LEDs will manifest what you saw if you crank the current high enough. Chances are good thermal impedance won't be even across the entire die. However, these tiny differences won't result in large temperature differences until the current is cranked really high, like perhaps 5 to 10 times maximum (assuming the bond wires can take it). To see noticeable differences at not much over twice rated current might indicate the LED is marginally defective (i.e. it's fine if you stay within operating parameters, but is not a candidate for overdriving).

 

[Walterk]

@JTR; Thx a lot for the input.

So uneveness occurs probably for most Leds when it is overdriven high, but this instance it is probably on the early side.
It happened far before it reaches the current that turns it angry blue.
I tested an XRE and SST50 and I couldn't repeat the phenomen.

Still can be typical XM-L behaviour. I have no other XM-L spare so I will test a new batch in time.
I remember Ma_Sha witnessed bad reflow-job on DX with copper star.
(Hope manufacturing standards have improved between now and half a year back. )

Thanks.

 

[saabluster]

The uneveness you saw might simply be a product of one part of the die getting much hotter than the rest due to uneven thermal impedance. The XM-L has traces to distribute the current evenly over the entire die, so I doubt uneven current distribution is the reason for what you saw. Saabluster's tests confirm this in that he was able to go to 6 amps with better heat sinking. All it might take is an air bubble where the LED is attached to the heat sink to introduce uneven heating. Or maybe you just have a defective XM-L. If you have any spare XM-Ls, try and see if you can observe a similar phenomenon with them.

By the way, I think ALL LEDs will manifest what you saw if you crank the current high enough. Chances are good thermal impedance won't be even across the entire die. However, these tiny differences won't result in large temperature differences until the current is cranked really high, like perhaps 5 to 10 times maximum (assuming the bond wires can take it). To see noticeable differences at not much over twice rated current might indicate the LED is marginally defective (i.e. it's fine if you stay within operating parameters, but is not a candidate for overdriving).

The unevenness is not due to heat but current density issues. I have noted this for quite some time. The brightest portion of the die can be found right by the two bond pads. It makes sense as this is where the current inters the die and is at its strongest. Also if you look at the lit die you will notice that it is brighter closer to the current spreaders than farther out. In fact it is quite a huge difference. I can try and quantify it if you would like although there is not much that can be done with this knowledge.

I have wondered for quite some time what kind of performance increase we would see if they made the current spreaders thinner and more numerous. The max drive level should rise a bit as well as a good bump in lumens output. It has been noted that the ez900 die has less of a varience in the highs and the lows seen in measuring them in aspheric setups and I contend this is due in part to the smaller distance between current spreaders.

Based on my tests current spreaders should be no more than about 120 μm apart. Wish I had control over some of Cree's manufacturing.

 

[Walterk]

I have tested another two XM-L.
They all seem to have their highlight at 6.1 Amp where the meter measured 500 Lux.
Without the fan, just passive cooling the peak was 480 at 5.8 Amp.
The die lit up even, well spread over the full square.

To answer my own question from a few postings back:
Is uneveness in the brightness of the die helping the XM-L getting a high throw rating?

My thinking: Lighting up uneven over the surface, although a defective led, it meant all current was going over a smaller part of the die, which handled it well.
Might have to do with internal heat distribution of the ceramic/copper/name-it to the star, after all the led was built to handle heat over the whole die but suffered heat-load only from a small hotspot.

XML leds on copper star from DX seem attractive, but somehow they are not that good performers.
For me, I find overdriving XM-L is not that much adding to output.

 

[bshanahan14rulz]

I have noticed this same effect in an EZ-1000 based XR-E. The space between the current spreaders is slightly darker than the area immediately surrounding the current spreaders, i.e. the centers of the three rectangles are dimmer. On the topic of current spreaders, what happened to the Atlas? (at least, I think that's what I'm thinking of...) Those had the most artistic-looking current spreaders!