Rebel red-orange not living up to spec

[Oznog]

I see someone else has discussed tests on the Rebel 100lm whites.
I have gotten some of these Rebels myself, but am having a difficult time believing their spec. My story's different enough that I don't want to threadjack him.

First a few notes. One breakthrough with the Rebels is that it tolerates up to 175C junction temp and is rated for full life at 135C. Second, if you can, solder it, this gives a very low thermal resistance with no danger of glue leaking onto the electrical contacts. I used like 1.5 sq in of PCB of unknown foil thickness and by watching the delta-V under constant current as it warmed up vs the spec sheet voltage-temp coefficient shows the die temp was still acceptable at 700mA.

Here's my problem. I have a Lamina BL2000 Red here, rated for 270 lumens(!!) at 2.1 amps mounted on an aluminum slug. I have 3x red-orange Rebel LXML-PH01-0050 rated for 100 lumens at 700mA mounted on about 4.5 sq in of PCB.

So the book says the Rebels should be slightly brighter. There is not a significant color difference that I see, but the Rebel is notable dimmer at all viewing angles when I drive these side-by-side. This is troubling because I know the eye generally views intensities as the square of the candela value falling on it so the lumen different is significant.

I have no equipment to measure this, but I can go back and change the drive current on the Lamina. I have a dual power supply and flicked them on at the same time so I can visually compare the apparent output in the initial moment before the dies heat up and begin reducing the light output. From what I can see, 3x of the 100lm Rebels @ 700mA each only compare to the 270lm BL2000 when the BL2000 is being driven at about 1 amp, approx 50% of the current where the BL2000 produces 270lm! So assuming flux is proportional to current in the BL2000 and the BL2000 is operating at spec, the Rebel is only putting out 45% of what it's supposed to. And this is 3 devices in series and they all appear similar in intensity, no single defective one or anything.

I checked to be sure the packaging said this is the red-orange 0050's. It does.

I kinda doubt the Lamina BL2000 I'm using as a reference is performing twice as well as its spec indicates. Both specify lumens at 25C pad temp and I think I handled the temp issue anyways by only comparing in the moment after it's turned on. By Rebel spec, the pad would have to be some 100C for its output to degrade this much!

What's up here? Why can't I get the expected output?

Anybody in Austin, Tx got the equipment to measure these?

 

[evan9162]

How do you have the rebels electrically configured? Are they running in series or in parallel? Have you measured the current to each to make sure you're actually delivering 700mA to each, and not 700mA divided equally among the 3?

Can you drive each Rebel individually with 3V from 2 AA batteries to make sure there's not something wierd with your setup, and compare that to driving it from your power supply? They should be driven at around 350mA by doing that. If they are brighter from the batteries than from your power supply, then there's something wrong with your electrical configuration - perhaps they are running in parallel rather in series, or something else is happening with the power supply to change how much current is flowing.

You should measure the current and Vf of each rebel to find out what is going on.

 

[Oznog]

OK I appreciate the reply with the newbie probs, but I do know the difference between series and parallel. They're in series. I set the current limit on the power supply at 700mA, there's a digital meter on the power supply, and a 2.2 ohm series resistor to check the current too.

Vf was unusually poorly matched between them, IIRC I got 3.4V, 3.4V, and 3.6V but there's no continuity at all between pads that would bypass an LED. Ohmmeter put on the pads in a direction that would reverse bias the emitter shows infinity. Either the Vf is just poorly matched in the devices or there's a huge difference in the pad's temperature. However, I don't see a perceptible intensity difference from emitter to emitter (they're about 1/2" apart) so I don't believe the temp is all that different.

I did play with some freeze spray around it, I wonder if I'm underestimating the temp involved. They can be brighter than the BL2000 at full power if I cool down the board first. I think I need to check delta-Vf as it warms up to get an exact picture of the die temp. I don't think so since the brightness is normally better when first turned on but still not competitive with the BL2000 flux, but data always trumps theories.

To be honest this isn't the configuration planned, I have a supplier who's supposed to be able to get me 5 oz PCB stock which would have much greater heat conduction and thus will make a lower die temp. However the supplier hasn't been answering my emails and I'm kinda making do with what I have for now.

 

[Oznog]

I just noticed the thread by Brlux about the hand soldering damage. That sounds like it could be a factor, I did mine with solder paste and a heat gun but that was hardly a controlled process either.

I don't see any output fluctuations here, nor do I see any visible damage on the emitter, but it's something to think about.

 

[Oznog]

OK data:

I see the start voltage after sitting in an 84F room is 3.45V. A few min later I get 3.425V. Spec sheet sez -3mV/deg C. With more time I saw it climb a bit to 3.439V, hard to say why that is.

From this I get 37C as a die temp.

But wait... this feels way hotter than 37C.

I left this running and found the thermocouple for my Fluke. I pulled off the protective plastic holder so I can touch the thin metal wires straight to the sink. I saw... wow, some 150C sink temp. I was double-checking and it stopped working all the sudden. A quick check shows one of the emitters is now burned out (unless it just unsoldered itself).

Huh. Sure looks to me like dV/dTemp is completely useless for calculating die temp! That and my heatsinking strategy is woefully inadequate.

 

[evan9162]

Well, it's not totally useful for calculating die temp, but you have to know how to use it properly to determine if the die is heating up too fast.

The die has a thermal time constant of about 20ms. So that means that you'd have to watch the -dV of the Vf with a scope to see how much it drops in the first 20ms to see how hot the junction is getting right away. The rebel package is extremely tiny and has very little thermal mass. If your heatsinking isn't up to par, then the die temp will rise almost immediately to a very high temperature, and just using a DMM, you won't be able to see the -dV due to the junction temp increasing (it'll happen too fast). The further drop in Vf after that will be as the heat sink its self is heating up.

So if your thermal interface between the rebel package and the heat sink is poor, then the rebel will heat up almost instantly. You won't see the -dV you expect from the temperature increase, because the Vf has dropped almost instantly, and you can't really see the behavior unless you use a scope. The poor performance in this scenario is due to the nature of the red/orange/amber LED materials, you get a drastic drop in light output when things heat up - much more so than with the white/blue/green variety. A 135C junction temp on one of the white/blue/green results in a light output loss of 20-25%; whereas on the red/orange/amber, that same junction temp results in a loss of about 75%.

 

[Oznog]

I tried to view it on a scope, but had problems. Namely when I put it on my Fluke 97, the best I could get was down to 200mV/div before it left the top of the screen even when I had the vertical trace Move down as low as it could go. The resolution wasn't so good enough at 200mV/div to see anything useful. Putting it on AC didn't work well either. The jump to 3.4V as it turned on created the same problem. Then I really couldn't see what was going on anyways because the DC blocking filter was slowly changing the reading, enough to confuse what I was looking at. So I gave up.

Well, I'm pretty sure I have a good thermal connection with the PCB- otherwise the PCB wouldn't be seeing such excessive temps, the die & thermal pad would. I did see that 150C or whatever on the board itself. That's pretty conclusive that the board itself can't dissipate the heat.

This leaves me wondering if my 5 oz board is really going to do as well as I'd like either. See my available space is 2" wide by 2.5" long and one side of the long end must be radiused into half a 2" circle.
4.57 sq in total, it can be 1/4" or so tall.
There is a lot of airflow available.
I need to mount *6* of these devices @ 700mA, a total of around 15W.

The outer surface needs to be fairly smooth, not sharp metal fins anyways.

I did well enough with the BL2000 in this space running 13.6W, but that was on a 1/4" thick slug of solid aluminum of the same size though without a lot of air exposure on the backside. These Rebels will need to be PCB mounted unless I do something clever. If the copper is sufficiently thick to wick the heat across the surface evenly, this should not be too much worse than the BL2000 situation.

I also have enviromental issues, the PCB area is exposed to the air. I thought of gold plating the thermal copper and conformal coating the electrical traces and around the emitter's edge. I did think of mating the copper with aluminum or attaching the Rebel straight to aluminum but of course there is a galvanic corrosion issue putting Al in contact with Cu and the Al won't take normal solder.

 

[Oznog]

Well, maybe they ARE tough little buggers!

I desoldered the bad one out of the array but when I put the multimeter on it in diode-check mode it lit up a bit. It didn't light or have any continuity before, hot or cold, so apparently this just ended up with a bad connection. I sure hope it wasn't due to the solder melting! That would be hard to imagine, the electrical pads being separated from the thermal pad.

I'm still at a loss for how to mount these. I was thinking of attempting to rip off the bottom contacts and strip off the top covering over the pads for a top attachment, but that was when I thought the device was already dead. I don't wanna tear up a perfectly good one. Really I think this is just asking for environmental contamination (moisture) and all getting into the die anyways.

Mind is churning, but I don't have an answer. The idea of aluminum solder keeps coming up but those products are touchy and I'm not even sure what arrangement would do what I needed even if they were as easy as soldering copper.

 

[chevrofreak]

The old LuxIII red LED's were rated at 1400mA. Did they change that for the Rebel?

 

[evan9162]

The R/O/A Lux IIIs are a unique beast - they have a die that's 4x the size of the Lux I. Strangely, for the K2 and rebel, Lumileds stuck with the smaller, Lux I version of the die. Thus the die is rated for a max of 700mA, not 1400mA that the Lux III is.

In this instance, the Lux III part is actaully better than the newer K2 and Rebel parts. Go figure.

 

[chesterqw]

yeah... the lux III R/O looks real cool... especially with a hand held microscope

 

[frenzee]

I have a few BL2000 and BL3000s (sadly both were discontinued) and made a few successful projects out of them. They are bright and tough/forgiving little buggers.

I just ordered a few amber and red/orange rebels and have a project in mind. Hopefully they live up to the mfg's specs as I have high hopes.

BTW, does anyone know how Jarhead isolated the copper strip leads from the heatsink in his test rig?

 

[Oznog]

I have a few BL2000 and BL3000s (sadly both were discontinued) and made a few successful projects out of them. They are bright and tough/forgiving little buggers.

Yeah I had REALLY good results with the BL2000. They can still be had at Mouser. In fact the whole point here is to replace the BL2000 in the same app. Rebel is stronger in the green and white esp since I can put several in the same area. Red, not so sure. The BL2000 is rated for 270 lm.

OK here's something I don't understand. The BL2000 is 270lm in red and 78lm in green, so the red should be many times more powerful. I need similar intensities, so I expected I'd have to turn down the red and sacrifice the reduced performance to maintain a balance.

However, at the rated current (2.1A red, 0.42A green) they were ALREADY balanced. In fact, if anything the green looked a bit "brighter" IMHO even though this is sort of an apples-and-oranges comparison. That's weird! I thought, OK, green is more visible than red, a 5mW green laser kicks *** over a 5mW red laser, but this is lumens, not mW. Doesn't the lumen scale ALREADY compensate for the human eye response so 1 lumen will be perceived as a similar brightness regardless of its color and instead how many mW of light in 1 lumen varies by color?

Have I misinterpreted what a lumen is, am I incapable of judging the relative brightness of these sources, or is something wrong with the red BL2000's output? I am certain the current is correct and it was mounted on a thick slug of aluminum with Arctic Silver. Eventually the slug would heat up, but it took awhile and the red's clearly not >3x brighter at the moment it's turned on and still cold.

I don't know why Lamina didn't further their flat emitters and even canceled the existing line. These things were AWESOME! Flat, thermal backing is easy to work, top-mount wiring is easy to work. Very low thermal impedance is easy. A real no-brainer to get right.

 

[frenzee]

I don't have any green Laminas to compare it to, but the red BL2000 I do have is crazy bright at spec. The red BL3000 is blinding at 1/2 spec. Both do get hot pretty fast and are not quite as efficient as say Lux IIIs, but the packaing is very convenient.

BTW, make sure you isolate these things. The ceramic backplate is NOT isolated and is NOT a dielectric. If I recall correctly, it is connected to the cathode. Also these laminas don't have any ESD protection, so be careful handling them.