Started July 17, 2007...on going, updated again July 19th, various places, such as Vf, die pictures, and tried out another "heatplate". And updated again on the 20th.
Philips LumiLEDs Rebel LED
This is the newest Philips LumiLEDs LED to hit the street. It incorporates die improvements that were developed at various US National Laboratories and also the results of University research. Personally, I consider it our tax dollars at work.
I got home and had a much anticipated package sitting on my desk, which arrived on July 17th, 2007, direct from LumiLEDs sole authorized distributor for North America:
So, what does it look like?
Unfortunately, I do not have any current generation LEDs to compare the Rebel part against. However, I do have the old 3 die wire CREE XR-E from last year- newer parts have an additional die bond wire which helps a lot to drop the LED Vf voltage. So, this is not a comparison against current CREE devices, which have the lower LED Vf, as well as improved lm/W specifications. Recent comparisons I've noted around the web are still comparing the Rebel to last years model of the CREE XR-E (which I will also be doing).
The Philips LumiLEDs Rebel LED is a bugger for folks to mount at home. It's small size, combined with tiny pads on the backside, and the delicate silicon (RTV) that is on the whole top surface combine to make it a tricky proposition. I mounted my Rebel by direct soldering to a 1/8" thick plate of copper for thermal transfer.
Here is my prepped copper and pads for backside electrical connections:
Here is the Rebel mounted:
The Rebel was a bugger to get mounted, but it is direct soldered to the copper plate for superior thermal transfer. When doing this by hand, the Rebel is much more difficult and delicate than working with the XR-E LED. For the testing the LED on the copper plate was then mounted to this heatsink (shame on you evan9162 for idle speculation, instead of asking first, assuming things, you have seen my heatsink before...see additional dome comments towards end- more added):
Below is the Vf graph of the Rebel, and I threw in the old CREE XR-E from last year, and also included the Philips LumiLEDs K2. Of the three, the Rebel has the lowest forward voltage, which will help lower the total wattage dissipated in the device package, thus making it more efficient, and this will be brought out later. Please keep in mind, the LEDs it is being compared with are from last year (one month short from a whole year ago). Note: Added second Rebel, Vf is higher.
The next test compares the Lumen output vs. Current in. Unfortunately, I do not have an LED of the same color bin (VO), so comparisons are really apples and oranges, due to spectral differences between the two bins, which can add considerable errors when utilizing a light meter to perform the test, instead of a Spectroradiometer or other accurate spectral sensor which can duplicate the human eye for sensitivity vs. wavelength. I did not bother with the LumiLEDs K2 for this test, as it is very poor and a waste of time, IMHO. As you can see below, the Rebel performs well against the old CREE XR-E, beating it performance wise. According to the datasheet for the Philips LumiLEDs Rebel, the LXML-PWC1-0100 Bin NVOD should have produced a minimum of 100 lumens @ 350mA if the pad temperature is held at 25C. Likely the pad was closer to 30C, but according to the charts for thermal de-rating, I'd not expected this much of a shortfall, 93 lumens vs it's rated 100 lumens minimum. It should have also produced 180 lumens @ 700mA, but it also fell short here, producing 162 lumens. However, it is still beating the older 3 bond wire CREE XR-E from last year.
Recently on one of the forums, I saw some folks arguing over the mounting technique on an LED. Basically one LED admittedly had a thick layer of epoxy under it, where as the other was carefully preped to make the layer as thin as possible- which happened to be the Rebel. Again, doing this, one is comparing apples and oranges, both should have the same mounting technique, or it places the results in the realm of suspect. Most especially with higher thermal resistance materials like thermal epoxy. If the epoxy is kept thin in one case, it is just as important to keep it thin in the other case, otherwise the test setup is *quite suspect*.
Off the Rebel datasheet, on the bottom of page 4, Notes for Table 1:
"1. Minimum luminous flux or ratiometeric power performance guaranteed withing published operating conditions. Philips LumiLEDs maintains a tolerance of +/-6.5% flux and power measurements."
Also keep in mind, on page 14, is the thermal de-rating chart for flux output. If the device's thermal pad temperature rises by 40C, you will loose 8% output.
Note on the Rebel datasheet that the thermal resistance from the die to case/thermal pad is higher, up at 10C/W, instead of 9C/W for the K2, 6.9 C/W like the Seoul P4, 8 C/W like the CREE XR-E, 3 C/W for the OSRAM OSTAR E3A.
The improved Vf characteristics really start to help the Rebel at Wattages above 1 Watt, as can be seen below. Again, the old 3 bond wire CREE XR-E is showing it's age, and if I get a chance, I'll work on obtaining one of the newer 4 bond wire XR-E devices that started shipping at the beginning of this year, which has a lower Vf. Meanwhile, it is included for comparison purposes.
Here is a plot of the new Rebel against last years XR-E showing the lm/W vs. Current in.
Here I held up a sheet of paper to the side of the LED, to look at tint vs beam angle, and if you look, you'll notice the stuff at the top of the sheet, is more blue in color, which represents the light coming from the top of the LED die. As you come down to the side, down by the LED, if you look at the area where the light is not saturated, you will see that it is a bit yellow-green in tint. Looking at the beam in a reflector, I noted that the hotspot is warmer in tint than the light in the flood area.
Contrary to popular belief, this part does not have a smaller die, it is actually a tad larger. Sometimes people get confused when looking at the die through the dome, as some domes have magnification in the lens, and some do not.
I saw no evidence of the anti-droop technology Philips LumiLEDs talked about back in January, hopefully we will get to see that by the end of the year. I hope they part number the newer devices differently, so one can tell the difference.
All and all, it is a very nice LED, and should help to heat up the competition in the market place. Hopefully the capitalistic market forces will continue to drive these exceedingly rapid improvements in the market place, and pricing pressures along with higher volumes will hopefully will start to drive prices on power LEDs down.
This isn't a device I'd recommend the typical home hobbyist, however, if purchased pre-mounted, it would be *MUCH* easier to work with. Be careful with the silicon coating on the surface of the device, it is quite easy to damage. If you look carefully at the picture at the top of the page, you can see where I accidentally scraped some of it off with my finger nail when removing it from the tape and reel package (look at the lower left corner). Luckily this damage should not affect the Rebel for the sake of this testing here.
One note- is that the focal spot is such that utilizing it with most of the common reflectors on the market will mean that a person will probably have to file out a notch in the bottom of the reflector to allow it to sit at the proper focal spot to get optimum focus. Also of note is that from part to part, the lens or die appears to drift in their placement (or both), and can lead to interesting effects when placed in a smooth reflector. For the picture up top, I used the center cardboard of a paper towel roll to create a form to assure the camera was perpendicular to the LED surface, and centered the LED die in the lens- yes, it is a bit off center, but this one isn't that bad.
(new day)
While testing the LED again tonight, I accidentally slipped, and damaged the silicon RTV dome of the Rebel LED. That basically ended testing on this particular LED, or so I thought.
Being curious, I inserted a small shinny Chromel-Alumel thermocouple, known as a K-type, just below the surface of the dome. At 1 Amp, I saw probe temperatures of 83C. A 7um-14um IR thermal camera recorded 93C dome temperatures. During this dome temperature testing, the copper plate was not mounted on my large heatsink, and the base of the Rebel was @ ~40C. Keep in mind that IR thermometers will read the dome temperature and not the die temperature. See measurement device below:
In regards to above, I used a 42 guage thermocouple to test the temps of a second undamaged dome rebel and placed the probe on the surface. The reason I use such small thermocouples, is that larger thermocouple wires heatsink the heat away, which result in a lower reading. This is especially true when using them on low thermal conductivity materials like the silicon dome on the Rebel. At 1 Amp, I measured 27.5C on the copper plate against the Rebel, and 61.4C on the dome surface. At 2 Amps, I measured 31.7C on the copper plate against the Rebel, and 87.6C on the dome surface. But this un-damaged dome (refer to earlier section where I mentioned slipping and damaging the delicate dome) this one measured differently with the IR camera. I got 41.8C with the IR cam @ 1A and 61.9C @ 2A. It is possible that any optical damage to the dome causes light to reflect back inside, heating things up- which would help to account for the difference between the damaged and un-damaged Rebel domes. As far as the difference between IR and the thermocouple, on the un-undamaged dome, I'm still debating that.
I also tried the test with a large thick aluminum plate, and got slightly higher temps, but the deltas are pretty much the same.
In the picture below, you can see where I accidentally yanked on the kelvin connection wires:
Here is the copper plate mounted with thermal paste to aluminum plate, with large heavy steel block sitting on top:
Here is the copper plate sitting on aluminum plate to show size of "heatplate" or heatsink:
Some folks have claimed the Rebel die is smaller than the Luxeon and also smaller than the CREE die. It is not. It measures 1.00mm with a set of Multitoyo calipers, which is the same size as the rest of the recent die on the market.,
Here we have a -100 Rebel with the dome removed, showing the bare die:
Here we have a -100 Rebel with the dome removed, showing the bare die lit up:
Here we have a -100 Rebel die with the phosphor removed, partially showing the naked die. If you look carefully, you will see the light extraction coating which is quite similar to what CREE started putting on their devices last year (see my CREE XR-E page for details):
Same shot under polarized light:
A better shot of the coating and die:
An even closer shot of the light extraction coating:
I decided to blast the LED with 35V from a large charged capacitor. If you look carefully, you can see various layers of the LED blown/vaporized, and make out several layers in the LED die:
Here is a close-up, showing various missing layers, and note the "trace" and the via are empty troughs:
Focused on the bottom of the trough where the via and trace used to be:
Another shot focused mid way between top and bottom of trough:
And an IR thermal image of the Rebel @ 2A, when copper plate is thermal pasted to that Aluminum plate. Basically we are seeing in radiated heat. If you look carefully, you can see reflections of the room (labeled). Bare Metal, even unpolished, often makes for a nice mirror in the deeper IR wavelengths. If you look carefully, you can see the outline of the solder on the plate, where the thermal pad of the Rebel is directly soldered to the copper plate. Oxides at the edge of the solder flow, carried there by the action of the flux, increase the emissivity at this point. I used a high emissivity coating to write HI! on the photo. As you can see, Kapton also has a fairly high emissivity. Of course, you can also see a water droplet, which has a higher emissivity than bare metals:
Main site
Mirror:
Main site mirror