White LED lumen testing

[jtr1962]

Your testing on an actual part is still hitting CREE's claims even with the die not at 25C, very impressive.

Very nice, your P4 did even better than my P3:

My power LED test jig uses a ~7"x3"x2" U-channel of aluminum as a heatsink. As a result, slug temperature is probably fairly close to room temperature. Also, the room was at about 65°F when I tested. Given that, the die may well have been not much over 25°C at 350 mA. Using 8°C/W for the junction to solder point interface, and allowing perhaps 2°C/W for the thermal grease interface to my heat sink, I'm calculating a junction temperature of perhaps 30°C given a heat sink temperature of ~65°F. This means the output is only derated by 1 or 2% relative to a junction temperature of 25°C.

I found it interesting that our efficiency numbers are close because Vf on your sample was less than on mine. I wonder if the higher binned XR-Es will have output which scales with Vf? If so, then they won't be substantially more efficient than the regular binned ones.

Another interesting thing here is that according to the light meter benchmark testing my meter reads about 10% low for an LED-type spectrum. Granted that the error will be different for different types of LEDs, but this could mean that my test sample was putting out as much as 8.5 lumens more than my test results show. Or perhaps the error in my meter just about cancels any errors resulting from my test methodology. I'm really surprised that all my tests of power LEDs have given the results they were supposed to. A SSC came out at almost exactly the typical 52 lumens it was supposed to. A Q-bin Luxeon was also practically dead center of where it should have been.

 

[NewBie]

I sure would be handy if someone had a certified lab for testing lumens.

Since high power LEDs, that are not abused, have such long lifetimes, I've often considered building up a very well regulated (light feedback mechanism), grossly over heatsinked, and sending it off to a certified lab for testing. I've also considered purchasing a certified reference bulb, but as you mention, the spectrums are different.

 

[Doug S]

jtr: This is one of the true great threads on the CPF. I'm sure many here appreciate your hard work.
I've read your first thread carefully but still have a question about your test setup.
Is any special or at least consistant provision made to remove heat from the cathode [negative] lead of the LED tested? The cathode lead is the primary path for heat extraction from a white 5MM LED. If you are not making any special provision to heatsink the cathode lead, a very interesting comparision would be a comparison with and without the cathode lead heat sinked to a few square inches of copper.

 

[jtr1962]

I added four samples (LCK 40,000mcd, Jeled 50,000 mcd, LCK 35,000 mcd 2-die 5mm, Jeled 15,000 mcd superflux) sent to me by CPF member TMorita to the list in the first post and updated the graphs accordingly. You might need to refresh your browser to see the updated graphs. The relevant spreadsheets were added to the .zip file linked to in the first post of this thread.

 

[jtr1962]

jtr: This is one of the true great threads on the CPF. I'm sure many here appreciate your hard work.
I've read your first thread carefully but still have a question about your test setup.
Is any special or at least consistant provision made to remove heat from the cathode [negative] lead of the LED tested? The cathode lead is the primary path for heat extraction from a white 5MM LED. If you are not making any special provision to heatsink the cathode lead, a very interesting comparision would be a comparison with and without the cathode lead heat sinked to a few square inches of copper.

Thanks for the praise, Doug.

To answer your question, I haven't made any provision to remove heat from the LED because I'm interested in how the LED performs in a typical application where it might have poor thermal coupling. I'll try your suggestion out of curiosity on a few LEDs just to see if it makes much difference. I would guess offhand that it wouldn't because LED output doesn't vary as dramatically with temperature for whites as it does for reds or especially ambers. I've noted that when I test at higher currents the output will drop somewhat as the LED heats up but the drop doesn't seem to be more than maybe 10% at most, usually under 5%. Heatsinking 5mm LEDs with heavy copper planes is usually done for lifetime purposes rather than to increase output.

A couple of quick calculations might shed some light as to how much there is to be gained. A 5mm LED in air has a thermal impedance in the 300°C/W area (junction to ambient). Since the socket I mount the LED in helps a little let's say it's closer to 200°C/W. I usually test at 80 mA maximum since that's around where output levels off. At a Vf of around 3.7V (typical at 80 mA) power dissipation would be about 0.3 watts. Junction temperature rise would therefore be roughly 0.3x200 or 60°C. Looking at some output versus temperature charts and we end up derating the output by around 15%. Even if a heatsink cut the junction temperature rise by two-thirds we would be looking at only a 10% increase in output at 80 mA (less at lower currents).

 

[3rd_shift]

I just finally found this.

However, if anyone wants to send me one without the lens I'll be happy to test it.

Can you email me your shipping address?
I can send over one without and one with the lense, both with the same binning to show the difference.

 

[NewBie]

It would be *much* more accurate to measure one, then pull it's lens off. Otherwise, you don't really know for sure.

 

[3rd_shift]

That could be tried with the one with the lense still on it.
After measuring, then an attemp can be made to carefully remove the lense and ring.
Then re-measure it.

My offer is still good.

 

[jtr1962]

That could be tried with the one with the lense still on it.
After measuring, then an attemp can be made to carefully remove the lense and ring.
Then re-measure it.

My offer is still good.

I sent my shipping address yesterday to the email in your signature. I look forward to doing the testing.

 

[3rd_shift]

Got the email.
Weather permitting, I'll get them out tomorrow.
These are the ones I got from Nitroz a couple months back.
I forgot the binning already.
But I think I can look it up.
PS;
Go ahead and keep the leds when and if you are done with them. :santa:

One more thing.
Be very careful with the domeless one.
It's kinda fragile.
I just killed another one trying to take it's dome off.
I ended up successfully removing and testing a working one from a maglite mod I did last month.
Unpack it carefully in case the gel stuck to the plastic baggie it is in.

 

[jtr1962]

One more thing.
Be very careful with the domeless one.
It's kinda fragile.
I just killed another one trying to take it's dome off.
I ended up successfully removing and testing a working one from a maglite mod I did last month.
Unpack it carefully in case the gel stuck to the plastic baggie it is in.

Thanks for the warning. I'll try not to kill it before I have a chance to test it. :huh:

 

[jtr1962]

If you are not making any special provision to heatsink the cathode lead, a very interesting comparision would be a comparison with and without the cathode lead heat sinked to a few square inches of copper.

I ran a few tests of both the old and new Jeleds with and without a heat sink. The heat sink consisted of a ~0.75" square of copper soldered on the cathode lead as close as possible to the epoxy body. Here is the relative output for the older Jeleds (red is without heat sink, blue is with heat sink):

Here is the same thing for the newer Jeleds (again red is without heat sink, blue is with heat sink):

Based on my tests of these a few days ago they already do substantially better at higher currents than the older Jeleds. There is some improvement with a heat sink but not as much as the older style.

Here is the percentage improvement versus current for both LEDs (older Jeled is in blue, newer one in red):

My tests are telling me that the way the LED is made is probably way more important than any external heatsinking. The newer Jeled probably uses a larger die with more contact area with the cathode "cup". It may even be attached with more thermally conductive material. The heat sink still helps, but only to the tune of about 10% at 80 mA.

 

[Doug S]

I ran a few tests of both the old and new Jeleds with and without a heat sink. The heat sink consisted of a ~0.75" square of copper soldered on the cathode lead as close as possible to the epoxy body.

My tests are telling me that the way the LED is made is probably way more important than any external heatsinking. The newer Jeled probably uses a larger die with more contact area with the cathode "cup". It may even be attached with more thermally conductive material. The heat sink still helps, but only to the tune of about 10% at 80 mA.

Thanks for doing this. The differences are in line with what I had expected. I agree with your general analysis. Heatsinking becomes a consideration for those who are inclined to drive the **** out of 5mm leds. While 10% may not be much, it is better than a poke in the eye. 10% is about the span of one cree flux bin and folks have been known to pay a premium for a single bin step. With the 5mm leds, sometimes it is possible to get that 10% for free with a little attention to the thermal paths. BTW, back in the dark ages before Nichia introduced white LEDs, it was not uncommon for LED datasheets to specify a thermal resistance from the junction to a point 3mm down the cathode lead. I remember that for some of the Hewlett Packard products this value was given as 220 C/W. The practice of providing this thermal resistance seems to be uncommon with the current 5mm white products.

 

[Pinter]

The older Jeled 50k leds were in drawer for a long time, got humidity from the air. Just wondering if their re-measured brightness were the same as in your original tests or were affected somehow?

 

[jtr1962]

The older Jeled 50k leds were in drawer for a long time, got humidity from the air. Just wondering if their re-measured brightness were the same as in your original tests or were affected somehow?

Actually they were still in the antistatic bag they came in which in turn was in the shipping envelope and then in a small box with my other white LEDs. It wouldn't matter if they had lost brightness anyway since my heat sink tests were solely to measure relative increases in brightness of the same LED by adding a heat sink.

I don't think humidity would affect the LED all that much anyhow since the epoxy case is designed to protect against that. I have a BestHongKong UWLC white LED sitting in my tester all the time it's not in use. It's therefore exposed to the ambient humidity all the time. It measured around 14 lux at one meter when new. It still does whenever I check it. It's over 18 months old.

 

[3rd_shift]

My apologies for the delay, I finally mailed the leds to you today.
The icy weather here has been a nightmare all week long.
Enjoy.

 

[NewBie]

Thanks for doing this. The differences are in line with what I had expected. I agree with your general analysis. Heatsinking becomes a consideration for those who are inclined to drive the **** out of 5mm leds. While 10% may not be much, it is better than a poke in the eye. 10% is about the span of one cree flux bin and folks have been known to pay a premium for a single bin step. With the 5mm leds, sometimes it is possible to get that 10% for free with a little attention to the thermal paths. BTW, back in the dark ages before Nichia introduced white LEDs, it was not uncommon for LED datasheets to specify a thermal resistance from the junction to a point 3mm down the cathode lead. I remember that for some of the Hewlett Packard products this value was given as 220 C/W. The practice of providing this thermal resistance seems to be uncommon with the current 5mm white products.

I specifically asked Nichia this about a year ago, on their various leaded parts, and they replied with 200-300 C/W, depending on the part.

It would be interesting to remove the thermal resistance from the base of the package to the cup (grind away the clear epoxy). I know I see considerable differences between these two points when I heatsink the lead. About a month ago, I wound up on a website somewhere, where they had not necked down the lead to the normal size, but brought it out full width- can't recall who it was though. It was also a very high current part for a 5mm LED...

 

[Doug S]

I specifically asked Nichia this about a year ago, on their various leaded parts, and they replied with 200-300 C/W, depending on the part.

It would be interesting to remove the thermal resistance from the base of the package to the cup (grind away the clear epoxy). I know I see considerable differences between these two points when I heatsink the lead. About a month ago, I wound up on a website somewhere, where they had not necked down the lead to the normal size, but brought it out full width- can't recall who it was though. It was also a very high current part for a 5mm LED...

Since the geometry is simple, this is a straight forward calculation. I just grabbed a random 5MM white LED off of the litter on my workbench. Measuring from where the relatively massive die cup structure necks down to the square lead crossection I get 6mm to the point 3mm outside the package. The lead crossection is 0.55mmX0.55mm. Using the 3.8W/cm-K conductivity of high purity copper would give a thermal resistance of 52C/W for this 6mm distance. This shows that the bulk of the thermal resistance from the junction to 3mm outside the package lies in the die attachment. I feel pretty confident that we can generalize that conclusion to the larger power LEDs. The high die attachment thermal resistance of the 5mm products is a function of their small area. As a reality check, lets assume that all of the Cree XR-E 8C/W thermal resistance is at the die interface. Cree uses a 1mmX1mm die. IIRC, the typical 5mm LED uses a 0.2mmX0.2mm die so it has an area of 25th that of the Cree. At best, we would expect the die interface resistance to thus be 25 times higher or 25X8C/W=200 C/W. Add this to the 52C/W for the lead as I calculated above and you have 252 C/W which is pretty near the middle of the range that was given to you by your Nichia contact.

 

[Pinter]

Interestingly, at 5mm leds heat exits not only through the leads; if you hold the "dome" with your fingers, the Vf will start decreasing slowly. Within 10-20 seconds I saw -0,02 V change in Vf, while the current went up by 0,2mA (simple resistored setup from NiMH batteries).

 

[3rd_shift]

I almost forgot to mention that these are P2 binned Cree XRE's that I sent over.