Started December 29th, 2006...on going

(I have edited things as I see fit, to clarify, and for brevity)

Well, I got home less than an hour ago, and just as I'd been hoping for, my first production Seoul P4 LEDs had arrived, in their overnight shipping package from Mouser.

I know these are a hot item for many, so without any delays, lets get some photographs.

Photographing these was a real pain, as the very, very soft gummy bear surface acts like a powerful dust magnet. I'd clean it, get it set up for a photograph, and it would have new dust specks on it, so after three tries, I finally gave up. Unfortunately, I don't have a laminar flow bench to give me filtered air.

These LEDs are rather cloudy, so I had to play a few tricks with lighting, to get an okay image.





Focused up off the die at the top of the phosphor, notice the tint variation across the die area:







I fitted one of these to a SO27XA, which isn't the smoothest reflector (I'll get it in a more perfect reflector later) which hides a number of the tint issues since the surface is rather distorted, but it was something that I could quickly get pictures of. in other words, the reflector is warped a lot. Due to the tint variation of the image, and AWB, the camera tint kept changing slightly between shots:











IMHO, for the white wall flashlight crowd, I apologize, the camera doesn't render tint variations near as well as the human eye, but that is a common shortcoming of digital cameras. The distorted SO27XA did a decent job of mixing up the tint, so the distorted reflector actually helps. I'd [b]*highly*[/b] recommend one of McGizmos textured reflectors. The "tintyness" of the beam is probably second only to the Nichia Jupiter and Nichia 5mm LEDs- but quite a bit better than the 5mm LEDs.

So, where does the tint variation within each beamshot come from? This image should help explain that, it is the bare emitter shinning on a piece of paper:




For fun, I used a high end camera lens to image the LED on a wall, and took a photo of it:




With some of the smooth reflectors, I could see artifacts of the bond wires on the wall, reminds me a little bit of the hotwires. I'll see if I can get a photo of it if I have time later. Again, I'd *highly* recommend utilizing a textured (or orange-peel) reflectors that the hotwire guys came up with to hide the defects and smooth out the tint variation! One test I did, you can pretty much get rid of the tint variation out of the Seoul P4 if you use the rough/orange peel reflector.

Even though this LED uses the same CREE EZ1000 die as the CREE XR-E LED, the way the optics were done, it emits much more of it's light sideways, so you have a very substantial drop in flood brightness in a reflector. But, conversely, with so much more of the light going sideways, more of it is directed forward, so this will be great news for those folks who like burning bright hotspots which throw better. Unfortunately, if tint variations within your beam bother you, you'll need to use a textured reflector to hide them, but will cause the throw to suffer a little bit.

With an aspherical lens, with a much higher proportion of the light going sideways out of the P4, substantially less hits a given sized lens, so less will be directed forward, as compared with an LED like the CREE XR-E.

Another drawback of this LED is that the slug is shorted to the positive side, where nearly every flashlight has a negative body. If the heatsink spot isn't isolated with anodizing, you will need to come up with a method of isolating the slug (hurts thermal performance- unless you are clever), or risk blowing up your LED and also the flashlight converter.

Be careful with the dome lens on the Seoul P4, it is very easy to damage the very soft gummy dome.

I think we got *really majorly* spoiled with the Luxeons, which had very little if any tint variation at all within the LED. IMHO, the magic trick for the Seoul P4 is the rough textured reflectors. With *plenty* of extra lumens as compared to the Luxeons, you can easily afford to use a rough reflector to blur the tints, IMHO- unless you need every single lumen possible in a tight spot. Personally, I prefer spots that fade, instead of sharply cutting off.

Anyhow, I've got more testing to go do...

Some preview stuff from later in the page (plenty of other things to dig into if you read on):





The above graph has too much data on it to see clearly, so if you are interested, download the larger version:




----Next day

Sorry, I was gone, went to dinner it was great.


This sample of 1 so far, measures 3.758V @ 1 Amp at the emitter leads.

Lumens looks to be similar, but I will need to do fairly precise comparisons as they are close. This one is consuming a good amount more power than my reference CREE P3 I tested, about 0.378 Watts more, or 10% more.

As far as tint variation comparisons, I'll offer the following:

CREE XR-E:





Luxeon III (yes, it is actually on a curved surface, and the smooth tint output causes a lack of depth perception):





Seoul Semiconductor P4:





Here is another point of reference, which helps to show the glob of phosphor in the recess of the LED.





FYI, I rinsed off the Seoul P4 again, it keeps reaching out and grabbing dust out of the air... : (

If I decide to use many of these, I might think about making my own mini laminar flow bench, along with a +/- ion generator for equalizing charges to help keep the static down. Besides rinsing with water or isopropyl alcohol/water mix, anyone come across any great ideas on removing/preventing the dust collection on the Seoul P4?

My batteries started getting low on the flashlight on the right for the shot below, so the Luxeon III here has a bit less light, otherwise the extreme clarity of the Lux III in comparison to the Seoul P4 would have been quite obvious, but here it is anyhow:




Unfortunately, with this sticky gummy dome, blasts of air don't help much, except stir the dust around the area, and the Seoul P4 yanks more dust out of the air.

Okay, this next section will be of concern to white wall hunters and Ring Wraiths.

Anyhow, I finally got one of the Seoul P4 emitters mounted on a post, so I could put it in a nice perfect and smooth MagLite reflector. I let the camera AWB, since due to the tints that come out at different exposure levels, I was getting some really weird looking photos which made one look decent, but the rest looked like the set white balance was kaddywhompus.


First a series of photos at different exposures:









Above, in the smooth "perfect" MagLite reflector, some of the issues of the P4 start to really stand out, like bond wires getting in the way, tints (which the camera barely picks up), and a very, very distinct ring.


I fiddled a bit with emitter placement and adjusted the focus spot a few mils, and instead of dark lines from the bond wires, you get the shiny distinct reflection of the gold bond wires shown below at 2 o'clock and 4 o'clock:






I wish my digital camera had the latitude to show the tints like the human eye sees them.


Like I said before, I'd *really* recommend rough surface reflectors with the Seoul P4, the orange peel reflectors should do a decent job smoothing out the tint variation in the LED, and hide the shadows/reflections of the bond wires. They should also fix the ring wraith problems, as evidenced by the same problem with the CREE XR-E, several manufactures use this technique to smooth the "defects" out with the rough reflectors for their products. I can't emphasize how much better these Seoul P4 look in the bumpy surface reflector, and I'd very highly recommend them for fixing these issues.

There are more than plenty enough extra lumens in the Seoul P4 to spread things out, blend them, cover/hide shadows/reflections, send the Ring Wraiths back to Mordor, and create a nice smooth hotspot transition, and still walk all over a Luxeon III.


I hauled out one of the outstanding bumpy surface reflectors (a prized possession), the rough surface HD45 and got MUCH better results. All the artifacts disappeared (rings, lines, etc), and the tint as well as the rings really blended better. All that is really left is a cold white hotspot which blends to a warm (yellowish) corona that gets warmer for a ways into the flood, then a cold white flood.

It reminds me of why bumpy/rough/textured reflectors are prized possessions! (really handy when you need them)

While fiddling around with the various reflectors I have, I noticed that different parabolic shapes cause different tint effect blend patterns, and some of them cause a "double image" flood when used with the Seoul P4.

Anyhow, I got to looking at the side emitted light color profile a little closer, pay attention to the banding and tints, and then look from one side of the photo to the other and notice how the tint changes (close-up):





Well, I wasn't that satisfied with the P4 tint photos, so I decided to re-take a bracketed series of them. I found if I turned it down to 250mA (instead of 700mA), it wasn't as tinty looking in the photos for some reason. Maybe the sheer brightness range in the photos was messing up the camera or something. Anyhow, these are closer in some ways to what I see:






If one looks carefully at around a 45 degree angle, you will notice the Seoul P4 has an extra, extra milky layer of silicone above the phosphor layer that is down at the die level.



Also, if you watch the die at various angles and rotate the Seoul P4 LED, the lens on it causes different magnifications at different angles, it is somewhat more aspherical shaped than a half circle of the Luxeon, if you look back earlier.


I'll re-emphasize my suggestion I made earlier on the rough reflector surfaces helping with the various defects/reflections/shadows/rings/etc, and the rough reflector surfaces also help with smoothing out the varying tint, so one just ends up with smooth tint donuts/bands- where the more heavy ruddy/bumpy/heavier orange-peel reflectors help out with the tint rings/donuts/bands more than just a slight orange peel. IMHO, the more aggressive finish also blends out the hotspot better, but it is a personal preference for me (the intensity blending, with a minor cost in throw).

I took another photo, with the camera focused on the paper surface, up close:





Side note- from different emitters off the same strip, I am seeing different items for tint variation in the profile and such, in some ways some are better, others are worse. But it is more a trade-off this for that.




Currently, on the production Seoul P4 parts that I've got (Mouser), they are binning by Kelvin temperature, and this one is 6500K. This is more like the old LumiLEDs method that they used with the LXHL-BW01 parts when they had the blop/glob phosphor. LumiLEDs went to great efforts in the early days to sell everyone on this technique in their technical presentations and training seminars. Thank goodness common sense won out, and they went to the color binning method based on the eye response to color that they use now (MacAdam ellipse). Under the old system, they take the Planckian BBL radiator line, and draw lines perpendicular to it, and create a binning structure from that. One of the big drawbacks of this system is that within one bin, you could get anything from an X1 to a X0, to a WA bin, which takes you from green to cool white to pinkish/purplish white, if anyone remembers the old days. So, under the old LumiLEDs system, this would fall in Bin Code 4, if anyone is interested.

Example:



Unfortunately, that chart there doesn't have the tints on it, so use this as a reference:





Here is a side profile of the Seoul P4 beam, with the McGizmo HD45 reflector which has a bumpy reflector surface:






Did a quick Vf run, since folks were asking me about Vf on the Seoul P4 (graphic replaced with later measurements of production Seoul P4 parts from a variety of sources) :




One thing I noticed, is how the overall tint on these changes a lot with current, but how much is for later testing. For now, above 1 Amp, whatever phosphor mix they are using really starts shifting hard, up at 2 Amps, it gets a pretty scary shade of blue (much like the overdriven 5mm do- a.k.a. angry blue). Coming back down, I noticed the tint shifts considerably over the current range as you move it around, but I'll need to some calibrated measurements to get more specific. Extremely low and very aggressive thermal resistance solutions may help this issue, I'll have to look into that further. (later samples, some had a lower amount of shift, and some have the same amount of shift I saw here. See note towards the end on color shift over the first 1,000 hours, as mentioned in the datasheet on page 4)




I've decided to do a set of side profiles in the smooth MagLite D cell reflector.

As you look at the photos at different exposures, you should be able to see the various tints in different areas that are not captured by a single photo. One item to take note of, for example, is in picture 3, you'll notice the cooler white center hotspot on the right hand side at the very end of the paper, then you will see the warm surround, then the very cool white flood area. Also, some of various the ring artifacts can be seen at different exposures more than others. The technique of various exposure levels is called bracketing, and is used to help make up for the short comings of digital cameras.

(keep in mind, when looking at very, very slight tinting, from monitor to monitor, some of the tints will look a bit more greenish, warmish, purplish, bluish, etc, depending on the monitor settings and the quality of the monitor)
















I finally dug out the manual and figured out how to lock the movie camera exposure setting and color correction settings.

Color tint shift with current adjustment
I specifically noticed during my forward voltage testing of the Seoul P4, that the color shifts dramatically with current levels. Two things can cause this, one being the die color output, and the other is the phosphor (some YAG phosphors will drop as much as 60% lower output as the die heats the phosphor)(towards the end of the thread, there is a plot of multiple spectral output at varying currents, the blue light out of the die is not shifting appreciably).

An example of what happens to one type of YAG phosphor is shown here, with the red line: http://ledsmagazine.com/press/14132


It is unknown to me as to what is actually causing this in the Seoul P4 LED.

As I mentioned, I locked the exposure and white balance in the camera for this movie I took to demonstrate the color shift. The current runs from 400mA to 1.1 A in the video.

The movie is located here:
http://www.molalla.net/~leeper/seoulp4t.wmv

I'd estimate the part is moving from a very strong X1 bin (greenish-yellow) to way out beyond the end of the YA bin. At 400mA it is greenish-yellow and at 1100 mA it is blue. Pay attention to the corona color around the hotspot, since the camera is oversaturated in the hotspot and doesn't represent color that well in the movie.


Thoughts-

The P4 LED appears to tint shift a lot more than other LEDs with drive current level changes, and hopefully it is just due to heating effects on the phosphor.


Early thoughts-
I'd consider aggressively pursing very good thermal transfer techniques in an attempt to help mitigate the tint shift in the Seoul P4.

My plans for checking this out further-
Getting a thick copper plate and directly soldering the Seoul P4 to it. Then to thermal paste this plate to a large CPU heatsink with a strong fan. Then re-test to see if it actually will help or not.(it helped)


The Seoul P4 has a rather low thermal resistance 6.9 C/W vs. 8 C/W for the CREE XR-E, and they use the same CREE EZ1000 die in them. The LumiLEDs K2 has a 9 C/W. I don't notice tint shifting in the K2 or the XR-E that is anything like what the Seoul P4 is demonstrating. (or even the Luxeon III with a 13 C/W for that matter)

With the Seoul P4's very low thermal resistance being in the same range of other newly introduced LEDs one would figure the shifts should be similar. The shift in the blue wavelength (affects interaction with the phosphor), and also the shift due to YAG phosphor heating, should be less, unless they are using a different phosphor blend than LumiLEDs or CREE is utilizing (which is quite possible).very likely, since the blue emission isn't shifting appreciably)

I hope to explore this further by directly soldering the part down to get rid of the thermal resistance of thermal epoxy and aggressively heatsinking the thermal energy out of the Seoul P4 die.

The movie doesn't work for some folks, so I snapped a snapshot from two points in the video:






I fiddled around a bit more with my movie camera, and used the white balance to set a neutral perfect white at 700mA. I then varied the current up to 1100mA and down to 400mA. Unfortunately, my video camera has very poor latitude, and does not do a good job of capturing the lighter and darker areas but you would be able to see how the various tints in the beam of the Seoul P4 shift at different current levels if you watch carefully.

Links to the videos:
http://www.molalla.net/~leeper/sp4twb.wmv

Zoomed in on the hotspot:
http://www.molalla.net/~leeper/sp4zoom.wmv

I added a high resolution version of the zoomed in hotspot, which shows things much better, worth the download, imho (3MB):
http://www.molalla.net/~leeper/sp4higrs.wmv




I ran a second Seoul P4 through the same drive levels, it doesn't get nearly as blue, but still shifts around.

The more and more I fiddle with these parts, the more I realize the consistency is not like the Luxeons, especially from one batch to another.(I've seen parts now from seven different batches, and they certainly are not all the same...)

Personally from what I see so far, if I were buying the Seoul P4 in a flashlight, I think myself, I'd really want to try out a batch of them, and choose the pick of the litter, unless it was a 10 dollar flashlight (and then, I'd still be tempted).

So, from what I'm seeing out of my first batch... For some, it is possible that they may get a premo choice part, and they will feel they are as good as a Luxeon (or better, since it is a lot brighter than the Luxeon). Others may get not so awesome of a part, and they won't be all that happy about it, especially those that have some concerns over tint, and tint variation. Right now, about the only idea I can think of, is to ask specifically for the full ranking of the parts that comes off the reel, Flux, Tint, and Forward Voltage for the flashlight they are put in- or to ask what tint bin the flashlight is in at its rated current- like SYN @ 1100mA (roughly like LumiLEDs YA) and then what bin it falls in at it's dim current, like SUM @ 50mA (LumiLEDs UM bin).
if you haven't noticed yet, LumiLEDs has again subdivided their bins, in response to customer's needs. This is a VERY good thing!



For those that haven't seen it yet, due to tint requirements of customers, LumiLEDs has moved to tighter bins, back in May of 2006, so the chart above may be a surprise.


I understand that Seoul Semiconductor offers sub-bins, however, I've never been able to buy a single bin. The ones I've gotten so far are by Kelvin temp (6500K), which leaves you pretty wide open. Possibly if one was able to purchase a single bin, the variation from part to part would be less, and I think if I was buying an expensive light with the Seoul P4, I'd really like the bin stated.(I was easily able to buy parts in a sub-bin, since entire reels of parts come in this fashion)

I'm not sure at what current Seoul bins the P4 at but when I get them directly from Seoul, on the package, you will see USW0H, where U = flux 91-118 (LumiLEDs 87.4 to 113.6), SWO basically equals LumiLEDs WO color bin, and H is the same as LumiLEDs Vf H bin, but 0.03 Volts lower.

I'm currently unsure at what current Seoul bins these at, but since the datasheet has it's Electro-Optical characteristics stated at 350mA, possibly it is done there. They also show the Relative Light Output vs. Junction temperature @ 350mA. Unfortunately, at other currents, white LEDs only generally track, and it isn't possible to specifically nail down exactly what the LED is going to do at other currents.(pay attention to the note on page 4 of the datasheet, about the tint shifts a lot in only 1,000 hours, this affects it's lumen output, fyi. See details towards the end of this page.)

If you are soldering the slug down, if the part has not been in a sealed dry atmosphere, and they have been outside the sealed desiccant package more than 1 week, you will want to bake the parts for 10-12 hours at 60C before soldering the slug down. When washed, they recommend Isopropyl Alcohol. They recommend that if you are storing them for more than three months after shipping from SSC, that you store them in a sealed container with dry Nitrogen. Long term exposure to sunlight or occasional UV will cause lens discoloration. Avoid leaving fingerprints on the silicone resin lens. And they do not recommend covering the silicone resin lens with another resin.

Possibly, with higher dollar lights, we could get the makers to provide a tint variation plot with current or something. It would not need to be exact, only what the light typically does. IMHO, this would be helpful for these parts.


I worked on another video with a different Seoul P4 part to try and show you what I am talking about.


Overlaid bins:






One person was doing a comparison...

The XR-E does NOT come in a STAR format from CREE, just the emitter by itself. As such, if you get them from authorized distributors, they will come in sealed bags marked with the exact tint bin and the intensity bin. Beyond that, you will need to trust the resellers, and the companies they are using to mount the parts, the materials used, and the process involved.


In response to a question from a CPF member:

Don-

I am not sure if you have noticed, but CPF'ers already have a long history of returning Surefire flashlights for tint issues, and it has been mentioned quite a number of times. Surefire even replaces them for tints, and that too is documented here on CPF.

The advantage that modders have, is they can select the parts used for their mods, and maybe this is the approach that should be used with the P4, depending on the current you are driving them at (since they shift so badly).

I dunno, it is just an idea on how to deal with the Seoul P4 parts- do whatever you want, I'm sorry if you take my ideas personally. We have a long history here on cpf, on custom lights and mods, of folks stating or buyers wanting to know the brightness and tint bins, as it is important to some folks.

You yourself have often mentioned you prefer the X1 bins for over-emphasizing green plants and such in the outdoors, myself, that is not what I am usually looking for, but actually what I prefer not to see- since it is everywhere, and I don't go out hunting pine or fir trees. To each his own, I'm definitely not saying you are wrong, as only you know what you prefer best and everyone is different.

Another idea, if I didn't mention it, is just to utilize the Seoul P4 in single level lights, selecting the bin needed, that get you to the color tint you want to be at, say at 917mA, 825mA, 660mA or whatever level you are going to drive the LED at.

I'm working on finishing up a video from another Seoul P4 right now, off the same reel strip, you be the judge of the differences... I'll get it up shortly.




In response to another question from the CPF member:

Don-

I am not telling anyone they have to do anything at all. I'm just offering up ideas on ways of dealing with short comings.

I am not motivated to make profit by making flashlights, selling flashlights, nor am I associated with any company that makes flashlights, or invested in such. For me, this is a hobby, and I intend upon it remaining that.

If I were to start making and selling flashlights, then it would be a job, not a hobby, and imho, much less enjoyable. Personally, I actually enjoy testing things, and finding out the results, as well as designing things, and I find manufacturing things to become a chore after a while. At the point where I'd begin making flashlights, my opinions would also probably become highly biased on what I was trying to sell at the moment, and I'd seriously start having to look at myself in the mirror a lot, a most definite catch-22 situation.

I never asked you to take any of my ideas as absolutely correct, I am just tossing out things for folks to consider.

Lets take your preference for X1 bins. Yes, they are yellow-green when compared to white. So what. Some folks on CPF prefer XO, or WO bins. Other folks really prefer YO and VO bins. To each his own, it is not my concern at all. To just say a light has tint, and to say it has no importance, except yes it has tint, is a tad odd, imho. You have had a long history of preference for the X1 bin. So, even you yourself are thus saying tint is important. Furthermore, you yourself have stressed how much more color rendering the X1 tint bin offers, and a few others support that opinion. That's fine, nothing wrong with that, but tint is quite important to some. It is in my personal opinion a dis-service to say that tint has no importance or that there should be no consideration for the same.

Anyhow, I'd recommend that we take your personal preferences and such to PM or email, and get back on the topic.




Getting back on track...

Here are the results from a second LED over the same range. Both of these Seoul P4 LEDs are actually from the same reel strip, in fact, side by side in the strip.

First LED here:
http://www.molalla.net/~leeper/sp4higrs.wmv

Second LED here:
http://www.molalla.net/~leeper/sp4t2nd.wmv




From another CPF'er:

[ Racer]And the tint debate is alive and well... I don't see an end in sight to the tint debate.

As a consumer, it would be great to be able to purchase only the best of the best as far as bin codes go.

But then, as a manufacturer, what are they supposed to do ? Purchase 200 units of a certain bin, hand select the 50 units that meet the criteria for use ? What of the other 150 units that don't make the spec ? Do they charge you the cost of 4 LEDs to make up for the three that get scrapped ?[END QUOTE]


Some people do purchase the best of the best bin tints, and we have folks making a very hansom profit here on CPF by selling them by tint, output, and even forward voltage. That is what threads that say UXOH and XX1S are about.(so, there is definitely a market for this!)

I have no idea what folks will do, IMHO, the Seoul P4 has really opened up the same old issue we had many years ago with the Luxeons when they first came out. Craig from LEDMuseum (Craig) even coined the term pee-green or whatever it was. What I see is new with the Seoul P4, is that I am seeing extra tint shifting at various drive levels. This could be a big issue in low output dual stage drive lights, maybe one could get a bin that it is just bluish when you want punch, and yellow when you just need a tad of light (splitting the difference). Unfortunately, when you drive white LEDs hard, they generally shift a little blue, and when driving them lightly, they shift somewhat yellow. The Seoul P4 seems to do more of this than other 3W LEDs I've worked with lately.

Maybe folks could apply the method that Henry did with the ARC4 and HDS lights, to the Seoul P4, where you pull a few drive techniques to hold the tint more constant. Not just a PWM or CC thing alone, but a combination of both. I've mentioned this before... Definitely something to consider.


Interesting idea, selling by tint, I imagine that could work, that is kind of what HDS did with their EDC60, EDC60GT, EDC60XR, and EDC60XRGT, which seemed to work for them just fine. It would also allow the maker to get extra profit off the premium parts.

Example here:
http://www.hdssystems.com/EdcUltimate.html



I finished a video comparison testing for now, until I get one soldered directly down to a plate of copper and mount it on a CPU heatsink. Here is a third LED for comparison against the first two:


First LED here:
http://www.molalla.net/~leeper/sp4higrs.wmv

Second LED here:
http://www.molalla.net/~leeper/sp4t2nd.wmv

Third LED here:
http://www.molalla.net/~leeper/sp4t3rd.wmv



I took a trip to Fry's and bought an overkill heatsink.

I then soldered a fourth Seoul P4 to the same 0.165" sheet of Copper that I have some CREE XR-E LEDs soldered on.

I then wired the LEDs in series to assure that each received exactly the same current, and placed a meter in series with them.

Before we get into further testing, I just want to say that not every one of these shifts the same, and you can see the results in the videos from previous posts. Some start tinting blue at a low current and some start tinting blue at a higher current, it is almost like there is a knee.

The previous tests had the Seoul P4 thermal epoxied down to a large solid aluminum rod, definitely more thermal sinking and lower thermal resistance path down the length than you'd see in a typical setup. Soldering and heatsinking is an attempt to see if aggressively lowering the thermal resistance to hold the die temperature extremely low would reduce the color shift.

Enough with the words, here are the pictures, with a meter showing the current going through the two LEDs (wired in series):

























I have a video of the same test that I am working on right now, I will post it shortly.



The video of the test is done.

Unfortunately, the camera over-saturates towards the end, and doesn't adjust it's brightness down, so you end up with a hot white spot in the center of the blue surround.


Here it is:

http://www.molalla.net/~leeper/sp4txre.wmv



Another CPF question:

[Gryloc]Hello Newbie. I was watching this thread for a while and I finally felt like posting. Again, nice work with the testing. Is it appropriate to be so harsh and judgmental over an early production LED? I was just wondering. Do early production parts ever change much from the more mature products that may be available like 6 months later? I was thinking about trying one when more available for some projects, but I hope I am not going to be disappointed.

Normally, when designers design things in, and run them through their own tests, through qualification testing, HALT/HASS, and in many industries, certification. As such, you expect the parts to have a certain amount of consistency and not to be changing, as a simple change to something like a primer that is used for silicone adhesion, can cause issues when going back through testing. These tests take months or even years to do, so you don't want things to be changing, or in a constant state of flux. This is where some designers get in a bit of hot water, where they will design and qualify things with prototype samples, then they find out (even though the manufacturer of the device said...) that something has changed, or often discover in themselves that the parts actually changed in their product production or the field, or that some attribute changed, and is causing failures now.

This is the whole reason for production parts, as the manufacturer has now dialed things in to a point that the part will be stable for a number of years, and will not change. You design using the stated specifications, and run qualification tests to assure the parts will work and hold up in the field. So, you expect/need a part to be the same for a period of time.

Unfortunately, my past experiences so far with Seoul Semiconductor is that they are always fiddling with their production parts, much like you were getting evaluation samples as things are dialed in, which makes them more difficult to use than LEDs from Nichia, LumiLEDs, OSRAM, Toyoda-Gosei, or CREE.

One of the indicators of how well a manufacturer has a part dialed in is the consistency from part to part. I am not seeing that much consistency, imho.

Critical? These are production parts and are what they are. I've not formed an opinion on these parts yet, but it is forming. I'm just checking them out.


Another CPF question:

[Gryloc] That color shifting in your videos are pretty extreme, like the first one (on you very latest set of videos). I am usually pretty cool over various tints, as long as they are not an ugly color, but the super blue that they turn at higher currents are pretty nasty for me (if the white balance on your camera is correct).

Well, the parts are much more consistent at 350mA as far as the tint from part to part. Above that, some parts start to shift drastically at 500mA and some go clear up to 1000mA before they shift much. There is also a lot of color shift over the current range.(Since this time, I've seen some parts that have much less shift, and some with more)

If you go back and look at the Vf curve, you will notice there is some tilt to it, as compared to the CREE XR-E, where they both use the same CREE EZ1000 die. As an LED die heats up, it's Vf drops, typically something like 2.0 mV per degree C. This may be what is going on, is the die is getting hot, and the thermal resistance is not what is stated on the datasheet. Another item one can look at is the change in wavelength with temperature of the blue die, which I hope to get to.

On the fourth production part which I soldered down (instead of thermal epoxy like the other three), just like the other three, it would shift right when it got up to a certain current level the first time(some cases only 500mA), not something that happened after some use.


Another good CPF question:

[Gryloc]
What are they doing wrong with the phosphors? You mentioned using a different type, but something seems wrong about the extreme shifts. It scares me because it reminds me of accidentally overpowering the old Luxeon I and it turns that angry blue like it will fail at any moment.

Wrong? They may just be using a different mix for the YAG than others. Phosphors is a big area of patent activity, as well as how it is mixed in it's binders, and how it is applied. All this affects how well it works. One can even trade off high initial brightness, for long term sustainability of that high lm/W.

I was hoping that using aggressive thermal techniques would help this part, it does, but not as much as I would have expected.

It could also be that the thermal resistance to the die is not what is stated in the datasheet. YAG phosphors will typically start dropping in output around 60C die temperature, and continue to drop as the temperature rises. This is the material that makes the broad band yellow (as well as the rest of the spectrum), and with the blue from the LED, what we interpret as white. Some YAG mixes are better than others. For an example of how a typical YAG phosphor performs, look here:
http://ledsmagazine.com/press/14132


I don't design LEDs for a living, I just use them in my designs, and one needs to look at things carefully, or risk jumping in the cauldron. As far as what is actually going on with the Seoul P4 LEDs, one would have to talk with their designers.


Another good CPF question:

[Gryloc] The pictures of the P4 emitter is discouraging, too. Thought the actual quality does not seem too bad, the way it was designed is odd. People are against the "gummy" domes. If it is like the K2 domes, I would fully accept them. How close are the "gummy" domes of the P4 to the K2?


There is quite a bit of difference, and not so much, depending on your point of view. I have not had the production parts long enough to do any sort of actual abuse or life testing.

[Gryloc] What could they have done better with the internals of the P4 above the die? I see how the soft gel diffuses and disperses the light in a lousy way. Why isn't it clear like Luxeon and CREE?

No idea, this is another area where the Seoul parts have had a good amount of variance. Depending on how you treat the light source, this could help or hinder, depending on your goals.

[Gryloc] Its way too milky, as though they didn't allow the silicone with the yellow phosphor to cure before they gooped on the filler gel. Also, why does the bond wires have to bee so long that they arch up and get in the way?

As to the cause, there are a number of things that could make this happen. It could be the supplier, the ingredients the supplier used, the production process that Seoul is utilizing, interaction with materials/steps in the production process, and a number of other things. No real idea as to the root cause.

[Gryloc] I bet with that long, narrow bond wire, there is some good resistance. Maybe not. Finally, is the conformal coating of the die (where the phosphor is perfectly square and even like the shape of the die) patented by Lumileds? Why did they use that puddle of goop-like phosphor and not like how even CREE did it?


Keep in mind, there are some optical effects that magnify things, however, they are long enough and tall enough to cast shadows or be in the image of the die in a reflector. Yes, LumiLEDs has a patent on the way they do the phosphor application, and at least two other ways of accomplishing the same thing.

[Gryloc] Oh, one more thing. Newbie, when are you going to take measurements of the flux of these P4s? I would like to see how they compare to the XR-E, even if its a similar die. That would be cool. I suppose that even if the LED is horribly blue, and it is super bright, I might have to try some of them. Alright, good luck and thanks... -Tony

If I get access, I'll do some lumen testing, but I am not promising anything.

Meanwhile, there is some variance from part to part, on the production devices I have. Some batches may be better, some may be worse. And since Seoul has a history of constantly fiddling with their parts/process, hopefully we will see the parts evolve or improve. A risk one takes is that unseen or unknown items may cause issues later in the field. Another risk is that you may get in another batch that are not suitable for your product. Luckily, for modders and small time flashlight companies is that it isn't that big of a deal, you can just use something else, or choose and pick from what is available.

It will be very interesting to get parts in from other supply points, and see if they show the same characteristics, or if the Seoul parts just typically have the same types of variance that they have had in the past. It will also be interesting to see how they do over time with their constant evolution on their parts.




[CPF-wquiles]Those Zalmans are in fact awesome for CPU's up to 100 Watts or so, if I recall correctly. I use two of these in my dual Opteron setup motherboard. I am running two 852's at 2.6GHz and these Zalmans keep both CPU's at a very cool 20C or so. These HS's will even support my move to the 285 (dual core) Opterons sometime next year :rock: Will

The one I was using here is the CNPS9700 LED, which is one of their largest versions. The smaller version has a thermal resistance of 0.16-0.12 °C/W, the large one is unspecified.

It supports:
-Intel's Core2 Duo, Pentium D, Pentium 4- All speeds
-AMD Athlon 64 FX, Athlon 64 X2- All speeds
-AMD Opteron and Dual Core Opteron- All speeds

Found here:
http://www.zalmanusa.com/

The small one was reviewed here:
http://www.tomshardware.com/2006/04/25/six_coolers_for_amd_processors/page21.html

The differences between the small one and the large one's performance is shown here: http://www.tomshardware.com/2006/10/16/zalman_cnps9700_led_big-ups_cpu_cooling/page2.html


Lets get back to testing and results. I used a fifth part for this testing, and found it by far shifts tint the least amount out of all five tried so far. This one will take three times the current that my worst on does, before shifting blue that much. Two people tell me they are seeing better results, maybe the batch that these came from wasn't up to snuff, I have no idea, but I do have more of them coming from multiple sources. When I get a chance, I will post some comparison results from the best and worst out of these five.

Relative measured intensity shift with temperature, supplied from a constant current source of 750mA, and monitored during the test to assure it held less than 0.1 mA movement off of 750mA, or 0.013% change, temperatures were allowed to dwell for 15 minutes to stabilize, and were held within 2 degrees C at each point. At ambient temperatures, the Vf was 3.76 and with a current of 750mA, for a power consumption of 2.82 Watts. With it's stated 6.7 C/W, this puts the die temperature under 18.9 degrees above the slug temperature (some of the power goes out as light, especially with the more efficient CREE EZ1000 die that Seoul uses in their P4). Even at a 95C slug temperature, this put the die temperature at 113.9 C, well under the absolute maximum specification of 145 degrees C. The 2.82 Watts is also under the 3.8W absolute maximum power dissipation rating:



I also ran an intensity vs. current curve. Conditions are on the graph, and when I say forced air, I really do mean it...




If you take a look at the curve, you will notice the brightness drops around 1200-1300mA (beyond the maximum current specifications). The part starts shifting blue around here, which may be hurting the brightness measurement. I was very happy to see at least one of the five showed significantly less shift than the other parts I've tested so far, and it occurred at a much higher current than the worst one, which started shifting much the same, but down at the 500mA-700mA range.

I did discover something very interesting, but we will need to wait for another day for those charts...hint- It has to do with phosphor and temperature. Oh, and the blue wavelength pretty much stayed put.




Well, I got good news and bad news.

Tonight, I got to looking at my Seoul P4 that shifts blue very rapidly above 500mA, a little bit closer.

I took a set of crossed polarizers (black), and took a look at the die under magnification on my LED that shifts the most. Just above 500mA I saw this very clearly:




For me it was one of those Eureka moments. I've never seen anything like this in an LED before. It only took a short moment to realize what I thought was going on. I know that YAG phosphors (which produce the broad band yellow to mix with the blue of the LED to make white) drop in output with temperature.

So what to do next.
I decided to sacrifice this LED I paid too much for. First off, I decided to take the dome off. It turns out, unlike other LEDs, this is not a rubbery silicone dome filled with a jelly consistency gel, but the whole dome area is made from the same silicone that the outside is made from. The silicone was bonded well to everything, but a little liquid release agent, and it came loose from the slug. However, it was still bonded to the die, quite well. In the process of trying to tugging and pulling the dome loose from the die (where it was still very well stuck to), the die just popped off the slug, clean as a whistle!


What I found on the back side of the die was this:



What you are seeing here, is the shinny area that was bonded to the slug, and an area that was not bonded, and has oxidized and started turning black.


Here is a closer shot:




And a shot of the die:




Anyhow, for myself, this explains very clearly why I am seeing the variation in the strip of production Seoul P4 LEDs I purchased, especially this one.

Will all parts vary like this? Who knows. In my past experience of working with high power LEDs over the past 5 years or so, I've always seen that Seoul does have variation in their parts, that I was un-accustomed to seeing in other brands that I've worked with- which include LumiLEDs, Nichia, OSRAM, CREE, and Toyoda-Gosei.

This does not mean that one should expect this from all the Seoul P4 parts, it is possible that a person could get a reel of parts that don't show this issue.

I am not absolutely certain, at what current Seoul bins their LEDs. Most of the specifications, and the binning looks to be done at 350mA. Below 500mA, I could not see any trace of this issue, whatsoever, it was uniformly the same color over the die area. Close to 500mA, I could see just an ever so slight shade of pastel baby blue, and right above 500mA, it started shifting hard blue very quickly. As I approached 700mA the whole die was pretty much blue, with the corner where there was no thermal transfer showing a distinct hard blue.

Anyhow, that was my adventure this evening.


Side note- With the die wires not being isolated from the outside by an inner gel fill, I noticed the bond wires actually do move when you press on the outside. It does not take much force to sheer the bond wires from the top of the die, and I'd *highly* suggest that every precaution is taken to avoid distorting the dome.




[CPF-Nereus]Very interesting info, thanks Newbie! Not very good quality control... Btw, in Soeul datasheet they say that the led is rated at 350 mA. But you have an idea that it is not the case? -N

"The P4 emits 240 lm at 1 A of light and features the industry’s highest luminous efficacy (100 lm/W @ 350 mA at the maximum) with only a single die, making it a brighter and more cost-effective light source compared to conventional 70 lm/W fluorescent and 15 lm/W incandescent lighting options."
http://seoulsemicon.co.kr/_homepage/home_eng/product/product.asp?topCODE=1&midCODE=25


If you look at the datasheet, they used 350mA for testing typical performance. The Absolute Maximum Ratings are listed after that.

I don't see 1A listed as in a maximum pulse condition. But they don't list clearly typical current levels to run their parts at.

I know McGizmo and Dat2zip have a number of their flashlight designs where they are driving it up to 917mA. Could be a language barrier, I guess I should get a hold of Seoul Semiconductor to clarify this point.


In case anyone is curious, this part is not the same part as the one I took up to 1.5A. The maximum this LED has been driven at is 1000mA (1A), and I've only taken it there for a matter of seconds. This part started shifting blue from the get go @ 500mA, and that was a sign that something was caddywhompus, so I didn't try driving it hard, or driving it for very long at currents above 700mA.



[CPF-yaesumofo]Newbie, could you please make a diagram which shows what part of the emitter is supplied by CREE? would this come from CREE like this or does this happen after SS get the "part ind install it into their package. No, it would be flat and very shinny on the backside from CREE.

CREE only supplies the little blue die part shown here:(without the phosphor on it)
http://seoulsemicon.co.kr/_homepage/home_eng/product/product.asp?topCODE=1&midCODE=25

[CPF-yaesumofo] Is there a possibility that this effect was accelerated by the high current tests? Can you do the same to a "virgin" emitter?

No, as I mentioned, this part showed the issue the very first time I applied power, as I slowly turned the current up, and it showed up at about 500mA. It has not gone over 1A. The part you are thinking about is another one of the same part.


[CPF-yaesumofo] There are just too many variables here to draw any kind of realistic conclusion. This could be a total aberration, a single bad unit or whole bad reel. Could it have been caused by high current testing? Unfortunately this "issue" raises far more questions than it answers. Such is life.

With some manufacturers they don't have the tight controls of other manufacturers. I often test parts before I will consider utilizing them in a design, even from industry leaders. If you don't, you risk being burned. When I get a production part that has issues like this, where a manufacturer could have worked more on dialing their process in, or if they can't, or hasn't put in tests to prevent process problem parts from getting out the door, it is concerning for me. If it was just one of five, that would be one thing. As it stands, I see significant variations in the production parts I purchased, and have seen things that do concern me. IMHO, if I was to consider trying to use this part in a product for a production run, I'd definitely have to do a lot more testing, and I'd definitely need to obtain parts from multiple batches, and I'd want to make sure the supplier wasn't cherry picking parts for me. (Note-Since this time, I've had a chance to test a number of Seoul P4 emitters, and I've seen a variety of what I'd consider issues, which are covered further down on this page).


[CPF-yaesumofo] I can't imagine a part like this having a realistic service life. Shirley large corporations who plan to use these parts by the 10's or 100's of 1000's are testing these emitters in the same way that you are. before committing to spending millions of dollars to implement these emitters into new LED fixtures in cars and boats and elsewhere.

It is hard to say much, without more testing. If a bad batch did get out the door, what is to prevent that from happening, and why did the QA department not catch it, and what is wrong with the manufacturing process that was put in place, or is it a design issue from the get go?

With some suppliers you have lots of variations, it is the nature of the beast. Luckily, not too many people count on their flashlights, and most flashlights don't see much use, but some people use their flashlights like crazy, so experiences will be different.


[CPF-yaesumofo] As this Nation makes the switch to solid state lighting we will see a lot of companies spending lots of R&D $$ in order to design super efficient methods of lighting our homes while maintaining lumen output reducing heat increasing efficiency all while producing an acceptable "new" type of light for us to see at night by. There will be companies which do not make it. Lumileds will not be one of them. will CREE? what about SS? I do not know. I do see a lot of dust on the horizon obscuring a clear view of the future. the images of the die you show here adds to this obscured view. I am sure most agree.. I look forward to what the future holds. I almost can't wait for a practical solution to replacing the compact fluorescent globes in m home with LED based systems. Wouldn't you hate buying 20 globes for your house only to find out that a MUCH better technology has just hit the market?

There is a city in California that purchased LED traffic lights for the power savings and for the reliability and reduced maintenance. Unfortunately, they went with the lowest bidder. To hit the cost target, the LED traffic lights manufacturer used LEDs from a Far East (non-Japan) company. Within the next two years, every single one of the lights failed, with many of them going dim in as little as six months, and they had much higher maintenance than they had with incandescent bulbs. Unfortunately for the city, they did not put anything in the contract to hold the LED Traffic Light maker accountable. The city had to eat the whole thing. However, there are several other cities that went with quality LED Traffic Lights, and the power savings, reliability, and reduce maintenance held true and they have had an outstanding experience with them.

[CPF-yaesumofo] I believe that it is important to keep in view the much bigger picture. the picture which shows us custom and semi custom flashlight nuts as the absolute bottom of the LED food chain. We are not the reason they are spending billions on developing these new solid state light light sources.
They aren't making these for us little flashlight users in mind!!!! WE are very picky. Will this translate down the road to better lighting for peoples homes? Maybe. I think we are a small annoyance to these companies. Did I drive this post off topic? sorry if I did.

Right, but due to demands, companies like LumiLEDs have gone from CCT Kelvin binning, then under customer demand on to binning like XO/WO, that everyone knows, and under even more pressure and demand, many of those bins have now been sub-divided even further!

If you haven't taken a look at the new binning structure that came out the middle of last year you should. They sub-divided all the bins around white much further yet. Its like wow, nice (page 5):
LumiLEDs new binning datasheet


[CPF-yaesumofo] NEWBIE can you tear apart a few more of those emitters? we need a large sample set. I am sorry I can not afford to contribute much to the Newbie R&D fund but I could send something if needed. BTW I have not been able to get my hands on any of these emitters (outside of a flashlight). Not for lack of trying. I would like to have a few for space needle builds and for some of my own crude (by comparison ) testing. Have fun!!! yaesumofo
These are super simple for even you to purchase. TTI, a large distributor, owns Mouser, who is a catalog house, which is very similar to DigiKey. They have been available there for at least ?two weeks? You could have them in your hot hands tomorrow if you wanted.

Direct link:
Direct Seoul Purchase Link

Anyhow, I did discover a few more things about these LEDs that I will go into later.

[CPF-EngrPaul]I notice over in the group buy for this emitter, quite a few people are backing out. Which leaves me wondering, is the "sky really falling?" I see one part with a poor solder joint, and almost an entire thread of evaluation around that one defective component. (Am I right?) NewBie, could you put this all into perspective?

Sky falling? Looked up at the moon and stars, and they are still up there, so no.

I do wonder why of the five LEDs I've starts shifting color at a different current level. Right now, it may have to do with the method that Seoul Semiconductor is bonding the die to the slug, the materials used, the process, the design, process variation, poor batch to batch control, or whatever- who knows? Possibly they discovered it and corrected it already, but didn't post a public advisory on their website, nor sent out through the distributors? Maybe they just didn't catch it because testing wasn't good enough? It is all idle speculation. McGizmo assured me that the parts are extremely consistent and stellar performers from his experience, and that is why I went out and purchased production parts. Now I've been hearing there is the typical variation, and personally, with the ones I purchased, I've not had the same experience.

Will others have a great experience or an okay experience, a mediocre experience, or a bleh experience? I do not have a clue, only time will tell. I most certainly hope that things end up more like what McGizmo talked about, where the tint is consistent, doesn't change with drive current, and are extremely robust.

I definitely was quite amazed by the level of abuse the CREE EZ1000 die, that is deep in core of the Seoul P4, was able to take, without heatsinking under a whole portion of the die. At least this holds much promise.

Don't forget, the LumiLEDs Luxeon V when it first came out. It was out at for a while, and folks were reporting it was going south in a matter of a couple of days, and I saw the same type of issue myself. It was denied that there was any issue, and it was the fault of the customer causing the problem. This went on for some time, until LumiLEDs fessed up to the problem. Then you couldn't get them for awhile, while they were looking into things. Some changes were made, which definitely improved the lifetime, and the specification was revised to a 500 hour lifetime. Since then, a more specific specification has come out, which gives you a guide on how long they are expected to last, depending on how well they are heatsink, and the drive levels.

To this day you still find the following statement on LumiLEDs White Luxeon V page:

Features
Superior lumen maintenance
Luxeon V Portable LEDs lasts longer than any incandescent bulb

The datasheet was revised back in 2004, and you will find this on page 9, even though they just revised it in August 2006:

Heat sink temperature: 85°C
Current: 600 mA
Average Lumen Maintenance After 500 Hours of Operation 65%

"LUXEON V Portable is designed primarily for portable lighting and other applications requiring operating lifetimes of 500 hours and less. While the device will operate past 500 hours, its lumen maintenance cannot yet be characterized. Longer life versions of white LUXEON V will become available in the future."
LuxV datasheet

The whole point here is even the market leaders make an oops on occasion, and we could go into the K2 for more examples...


Consider things and comments in this one post here, all pure speculation, and no more.

Okay, I need to get the other camera out to get a better picture, but here is what I see:



At the moment, the best I can tell right now, this looks like many of the low silver content thermal epoxies that I have used and have a number of samples of. This technique has been used with very low power 5mm LEDs, due to their low heat levels. It is possible it could be some odd solder, but from what I can see, it looks like globules held together in a matrix (I guess it could possibly be solder paste that wasn't fully heated up to the melting point). I'll know a little more when I get a chance to look at things more tomorrow night.

Thermal epoxies have a much higher thermal resistance than solder and are rarely as durable as solder. You will find that LumiLEDs, OSRAM, CREE, and others actually solder their die. I see that CREE put a 0.00012" AuSn (gold tin) layer on the backside of the die.



I'll need to get in at a much higher magnification level to tell more, and poke and prod it- for the moment, consider things and comments in this one post here, all pure speculation, and no more.

BTW, EngrPaul regarding your earlier comments- I found nothing at all between what appears to be a delaminated die area and what appears to be "die bonding material" on the slug.

I've also seen a couple of cases of leakage current, one was from an ESD diode, and another was from the thermal epoxy making a connection on the side of the die between the internal die layers.



[CPF-EngrPaul]
NewBie,
Thanks for (re?)stating the total number of parts in your evaluation.
Let me restate the failure I'm observing. On the picture above, marked "Thermal Epoxy?", the bottom right corner of the joint was never soldered/bonded correctly, metal to metal. The chip was never seated planar against it's mating face, instead a corner was lifted during this attach. This caused a gap at the high side. Some of that liquid seeped into the air gap during process stages after this soldering/bonding.

This area as I said, is where the thermal transfer was not occurring.

If you read, there was no material found under that area.


[CPF-EngrPaul]
Another point: Seoul's T bin is significantly lower than their U bin. Aren't T-bins are components that "didn't quite" meet the output target due to flaws in materials and workmanship, but are good enough for certain applications that don't need flawless performance? If we put them through rigorous performance tests, we're only going to find out more details about those flaws. It doesn't necessarily mean that the part doesn't meet specifications. I look forward to some U-bin evaluation.
P.S. I'm not trying to stick up for Seoul, just continue to justify my $ order for more than a dozen U-bins :)


IMHO, that is really pulling things out of .... ...

Oh, did I say that? I meant stretching things beyond the breaking point.

The Seoul T bins are the ones that are easy to get your hands on, their normal parts. The U bins are the rare ones, and for many, they have been waiting to get their hands on them. It is expected that by February, they should be more plentiful.

There is a significant difference? Humm. I remember checking out some premium priced Luxeon U bins, and comparing them to a number of T bins, in an integrating sphere, and the difference between them was easily within the measurement error of the equipment. I took over another older T bin set I had purchased, and the difference was 5%, which made me feel a little better about the purchase, but not by much.

I don't know if I'd really put that much emphasis on the difference between them, unless I had a set from the same Vf, Tint, but the U and the T lumen bins, and one was seeing a very clear difference. This would be more prone to better performing die, and if you look at the die datasheet, you will see there are five bins for brightness, and twelve bins for wavelength in each brightness range measured @ 350mA on page 3 and 4:
Link for CREE die used in the Seoul P4


The wavelength also comes into play, as most YAG phosphors have a narrow wavelength range for their peak conversion of blue light to broad band yellow.

This very tight wavelength binning helps, in that it makes it much easier to hit the target of the phosphor. Each bin wavelength range is only 2.5nm wide!

With five die brightness bins, starting out at 200mW and going to 380mW, one at least gets to start out knowing the amount of light output the particular binned die makes, fairly precisely to boot.


[CPF-65535]That is far from a CREE so the only cree in their is the PCB well I won't be buying a Seoul chip thanks Newbie.

I don't know if I'd put it that way, but so far, I'm not too happy with my purchase from this batch. I'm really hoping that I'll get some better batches in the future.

I am now much more leery about the parts, after looking at the construction and the materials it looks like they used, but I need to look much closer at the die bonding method they used.


[CPF-EngrPaul]Going by the pictures only, the metal square is the solder joint. It should be fully shiny through the square, with the exception of some air pockets. Instead, it's dark in the lower left corner where solder didn't join, and liquid (phosphor, whatever) wicked in later. If a cross-section at 45 degrees (a cutting line from NW to SE) was done, it should reveal there is solid material on the surface of the metal at that corner of the solder joint. This is why it pulled apart so readily. Good, complete solder joints should not come apart like that.

Sorry, it is definitely not solder.

I just got home, and I'll get some pictures taken and up shortly.



Die bonding material.

I apologize about some of the dust ahead of time. We are looking at a die that is actually a tad smaller than 1mm by 1mm. When the dome is on, magnification takes place and makes the die look much bigger than it actually is.

As such, dust starts looking pretty big at these magnifications, and I didn't dust my lenses before hand. I noticed it part way through my photos, and cleaned things off.

If you look flat at the surface, or normal the material actually looks grey, and if you look closely, you can see a few shiny spots, those are flat areas that were on contact with the die. When you are perpendicular like this, it appears black, grey, and white (flat spots):




However, if you look at any angle it has a tan appearance. If you look carefully, you'll see a spot where the thermal epoxy did not break loose of the slug, and instead broke loose of the side of the die, and left a bump there at 1 o'clock. The thermal epoxy that was used sticks well to the glassy sides of the die, but it didn't stick as well to the slug as it did there. However, on the solder bottom of the die, it stuck better to the slug than the die, and that is most what we are looking at:




An angled shot of the material up closer:




Looking again from angle, closer yet, so you can see the side view, instead of where the occasional flat surface touches the die, like you see from the top.




Looking even closer yet:




Here is a top view, perpendicular to the surface, where you can see the occasional surface that touches the die, and the valleys in between. The white blurry areas that you see, if you focus up and down, you can actually see the slug surface below in the porous matrix:




Here I went over to the edge to make things very obvious. I am focused on the slug surface in the top half of the picture, and you can see the slug surface between the clumps of goo in the bottom half:




Same shot from another angle:




And closer yet:



Here is a shot where I focused on one of the pieces of epoxy that didn't stick to the glassy sides of the die:




And the dome + bondwires + die, just as it came off as one assembly, the RTV + RTV+phosphor, actually stuck better to the die than the thermal epoxy did!




Finally, lets go back and look at the die shot, taking notice where the thermal epoxy stuck to the sides of the glassy die better than the AuSn bottom of the LED die:




Anyhow, enjoy! It was much more fascinating in person than it is in photos.



Seoul does not use the SiC ESD diode sub assembly, like you find in the CREE XR-E LEDs.

The failure is in the epoxy used by Seoul to direct bond the die to the slug, and the porous open epoxy matrix delaminated from the CREE die at the thermal and rear electrical interface, but yet stuck to the glassy sides of the die better than even the substrate for 3 out of 4 sides.

--> Disqualifier: My own personal feelings, not anything I'm stating as fact or whatever, below

IMHO, especially after looking at the construction methods, personally I don't think I will personally be using these, and if I did, I personally will not put them in any of my personal expensive flashlights. Maybe something like a Lumapower or a Nuwai Q3. For myself personally, I'd definitely be very tempted to run them at low powers (less than 350mA), and would not personally use them in anything I personally would want to last or personally count on. I personally would be very tempted to get as much heat as possible out of the LED with heatsinking I'd personally consider quite robust.

With the failures I personally had with parts I personally purchased, and the variations I personally experienced between the parts, it reinforces my personal own opinion I mentioned above.


[CPF-McGizmo]Hi Newbie, How are dice typically bonded to the slugs? If epoxy was used here, was it also electrically conductive epoxy and hence the anode connectivity? Would one or could one measure resistance between the anode lead tab and the slug to test connectivity of die to slug?

The dice are typically soldered to an ESD sub-carrier assy, like with the Luxeons and the CREE. The CREE SiC ESD diode is soldered to the substrate. I know this as fact, as I have actually unsoldered a die from an XR-E and re-soldered to a heat pipe.

I've also unsoldered the Luxeon die, shown here:




From the Luxeon ESD diode shown here:



The electrical resistance of even a few points of silver is going to be rather low, and if you look carefully at the Seoul P4, you will notice there is a bond wire that is opposite the three negative bond wires. The electrical resistance of this gold wire might mask anything that could be measured with a sensitive milliohm or micro-ohm meter.

What might work, is to watch the Vf shift from when the die is cold to when it gets hot. Hotter die will rise more than cooler die.

How one would assure long term reliability from some sort of test one could do might be a little tough. Here I'd be inclined to feel that you'd have to build 10,000 or so of them up, and run a test for a few years. The problem with this, is Seoul Semiconductor is constantly fiddling with their processes, so that would only tell you if that one batch held up over time. And you'd want to thermally cycle them at the same time, or at least turn them off until room temp, then back on, until the temperature stabilized.

One could also run the LEDs across several different current levels and capture the shift in tint, but since tint shift can be caused by other things, it isn't a definite thing. But, you might be able to weed out the ones that have obvious issues from the get go. The only reason this might work, is I'm seeing almost a break point where to color takes off blue, and that should be obvious.


[CPF-McGizmo] This part was obvious from the get go to you as being suspect. Do you think other less obvious LED's will have premature failure due to an epoxy bond? Will this failure be due to thermal build up?

Yes, I've worked with Seoul's LEDs before, and have some experiences with their previous parts. Whether the other LEDs that looked better will have failures over time is pretty hard to say, but it is definitely something to suspect. One could do a few hundred or few thousand thermal cycles, or even a little bit of thermal shock (tough to know at what point you are exceeding specifications on weak items like the epoxy), and not exceeding the slower thermal pre-heat ramp they specify for soldering. This would be a better question for a materials scientist, imho.

This part here did not experience thermal build up, shock, or whatever. It was simply thermal epoxied, with more than adequate heatsinking. The very first time it was turned on, it showed a rapid blue shift at around 500mA. I've got other parts that shift above this, and one that holds on clear up to 1500mA (which is exceeding it's rating).

Now, if you are saying thermal build up in a product and causing pre-mature failure, that would be something I'd be thinking about, and addressing any way I could, as that should help lower the stresses that will be present inside the part.


[CPF-McGizmo]
Is epoxy not acceptable as an adhesive? I have had no trouble lifting LED's from stars or even from anodized sinks that have been epoxied as the mechanical bond to some of the smooth surfaces is just not that great. There appears to be delamination in this bad part. Was the delamination a result of some trama to the assembly after assembly or do you think it could have been a bond flaw at assembly time?

Not being a materials scientist, it is hard to say for sure. Surface prep is often of utmost importance when using many epoxies/RTV/silicones. What ever they are doing for their silicone bonding process is pretty decent, imho. When I've seen failures with bonds, on something that should handle the forces according to the datasheet on the substrates that were bonded together, it is often poor surface prep, substandard epoxy was supplied, or the designer forgot to consider something like the CTEs of the two materials joined, or in the case of two materials with differing CTEs- the designer didn't specify a minimum bond layer thickness of epoxy, the production process has poor control, or they are not doing it according to the specification. No idea if it is adequate or not, but I am seeing variations in these production parts I purchased from Mouser- and one that could cause what I am seeing is delamination, but the others could be due to something else. I've got one that has blue tint around it's four edges when driven at 700mA, and the one that runs up at 1500mA without excessive shift doesn't show this.

It would be hard to cause trama, it is recessed, but I do not know their process, so anything here would be pure speculation.


[CPF-McGizmo] On this particular LED, what has its history been with your testing? Could it have suffered some thermal abuse prior to sufficient sinking? Was it soldered to a heat sink and if so, could the process of soldering it have had ill effect?

It was thermal epoxied, so no thermal shock or anything. It had great heatsinking from the get go. It showed the blue shift the first time I slowly ramped up the current, right around 500mA. The tape carrier packaging showed no signs of being crushed, bent, or other issues that might indicate something happened after they were packaged at Seoul.


[CPF-McGizmo] In fairness to potential and unwitting users of these LED's should Seoul be contacted and told to cease production of the P4's until they get a better assembly program in place?
Humm- It really isn't my place to say what they should do. They may want to screen portions of production runs, and life screen, if they are not already. The screening may need to be more rigorous. I've seen a video of a Luxeon lookalike part that the only thing that was done was to test them at the binning current, and that is it, the binning was the only screening. Hard to say, but one might want to talk to the designer of this part himself, and not thru the sales/marketing force, and get his take on things, not some toned down marketing speak.

Better yet, if others find this issue in parts they have, contact their engineering, send them back to Seoul, and let them evaluate what is going on, since only they know what is actually used, the process, the intent, and such.

IMHO, the customers using these parts might want to consider doing some testing of their own (in any case), especially on something that is such cutting edge technology and new processes/materials used.


[CPF-McGizmo] In assuming that your personal opinion on these LED's is not a case of over reaction, what do you suppose should be done?

Be cognizant of any failures or weird behavior, even odd tint shifting, and if one finds them, to be dynamically proactive. Most especially for the consumer of the end product. If I was the consumer, I'd really want to consider the warranty offered on the light, and the past reputation for warranty handling of the company.

I've got more parts inbound from various sources, it will be very interesting to see what happens with those parts. I'm really hoping for the best, these Seoul parts only require heatsink shimming and reflector grinding to retrofit into existing Luxeon flashlights, so they could be useful for a lot of CPF'ers.

Right now, I'm not going to go out and buy a bunch to modify my own lights, but I may wait, and see what the long term experience is on these parts for others.

As I've been writing this, I've been considering making a circuit that would turn the parts on for a period of time, then turn them off for a period of time, to see how this Seoul P4 construction method holds up. One of the questions, is do I heatsink the part excessively, or do I heatsink it such that the die hits 125C, which is 20C lower than it's maximum specification. Or do both...which is where I am leaning....

It is possible, that somehow, I beat all the odds, and ended up with bum parts. It is really hard to say, without buying a few hundred parts from each of multiple runs, and doing a whole test program...



[CPF-chimo]Newbie, great job on the detective work, pics and write-up.
Seoul Semi states a fairly low thermal resistance (6.9 degC/W). Do you think they achieved that while using thermal adhesive instead of soldering the die by omitting the ESD diode?

Paul

No, not really. CREE uses a SiC ESD diode, SiC is one of the areas that CREE is also a leader in, and has been for awhile.

"SiC is an excellent thermal conductor. Heat will flow more readily through SiC than other semiconductor materials. In fact, at room temperature, SiC has a higher thermal conductivity than any metal. This property enables SiC devices to operate at extremely high power levels and still dissipate the large amounts of excess heat generated." CREE SiC properties webpage

CREE actually makes very durable rectifiers/diodes from SiC some that have sub 1 C/W thermal resistance.
CREE SiC High Power Diodes



[CPF-McGizmo]Newbie, Thanks for the point by point responses.
After Luxeon and CREE solder their dice to a ESD device, is the ESD device then soldered to the slug? I assume the answer is no in the case of the XR-E due to the isolation of slug to die.

I know for certain, that the CREE ESD device is soldered to the "slug" or the bottom ceramic that has a thin copper layer on it. The isolation comes from the ceramic layer in the "slug" or base.


[CPF-McGizmo] Is it possible that the bonding material used by Seoul in this assembly is up to its required task providing a proper bond is accomplished?

I figure you'd need to get long term testing data across multiple production runs and process tweaks the Seoul constantly does. IMHO, long term thermal cycling would be very important.

[CPF-McGizmo] Without speculating on the materials used and their ultimate viability and acceptability in this application, there are a few things you observed which certainly point to a part that should be considered a reject. The obvious color shift with variation in Vf and the lack of uniformity of tint in viewing the die itself. This led you to taking the LED apart and your discovery of an incomplete and questionable bond layer.

Had I only binned it @ 350mA, I doubt I would have seen any issues at all. I've got five individual different parts, and they are not the same, but at least two of them are similar, and one is a real performer and not one of the two.


[CPF-McGizmo] I believe there is speculation that this poor bond layer would put the die in a position of excessive heat and inadequate thermal relief when operated at specified drive levels. This would lead to poor lumen maintenance as well as color out of spec, presumably. Would this also likely be evident in flux output being below expectation?

Flux output is affected by a great many items, such as the root die, the phosphor layer, it's thickness and application, the electrical bonds, the bond wires, and such. If you had a part that was specified as a U bin, but it wasn't up to par, then I'd really be scratching my head over it.


[CPF-McGizmo] I have no idea how these LED's are made or what type of QC is present in the various plants. From the outside and with some experience with being sampled numerous times by Seoul, I will throw out some observations and comments of speculation more than anything else. It has seemed to me that Seoul have been very fast at taking a new die and packaging it for market evaluation/ acceptance. In the past, I have received a number of LED's which were pretty good but not up to market level; close but no cigar. It seemed that Seoul also recognized this and back they went and out came the next iteration.

IMHO, they have a bit more variation than a number of competitors, and constant process tweaking, so it is hard to say what is final.

[CPF-McGizmo] I received a sample from them a few months ago that looked like a Nichia ceramic square box (jupiter style) but with the gull wing lead frames like the Luxeon. This LED had the flux of the XR-E and I believe it was host to the EZ1000 chip. The LED had a large phosphor dome for lack of better description and it may have been viable for fixed lighting but the light image or object was way to large for consideration of collimation. It seemed that Seoul now had the flux to compete but not the color consistency or image size. About a month later, the P4 shows up.
So I too have been given the impression that Seoul has been under an accelerated chain of trial and error and numerous new designs in an attempt to get to market level for acceptance. Tint variations aside, I believe the P4 has now brought Seoul to the forefront and in some applications to the lead. In their rush to get there, have they brought a flawed design with them or have they not dotted their I's and crossed their T's with the intent of first checking the market acceptance and viability of this latest itteration?

Hard to say, without long term testing of a large sampling of multiple production runs, IMHO.


[CPF-McGizmo] If this P4 assembly is sound with proper compliance to procedure in assembly and manufacture, that is one thing and likely something they can fine tune if they can rest on the design for a while. They can go back and dot the I's and cross the T's I would think.
Since I haven't experienced first hand a "dud" like this one here but I have experienced numerous exceptional performers which do not have tint shift as a function of Vf change, I am real impressed with the P4. The LED does perform admirably! I do not know if the design is sound when manufacturing is in compliance nor what to expect if manufacturing is not in compliance.

That is a real can of worms, and hard to say. I could very easily believe you got a real awesome batch of parts, but it is hard to say if every batch will be like this.


[CPF-McGizmo] The Luxeon V was found to have problems and it seems the design itself was weak as the part never got past "portable" applications standards.

Yes, I talked about that earlier in the thread. (Comment- If you look at the datasheet, on page 4, there is a note stating a rather substantial color shift in 1,000 hours, I've seen the shift in 500 hours as well as over a 10% loss of output).


[CPF-McGizmo] Ultimately in the case of the P4, where will it end up being? Are the good parts not only good on first impression but built for some time to come? Given the impressive efficacy of the P4 and its claimed low thermal resistance, it would seem that it should and could be a part that produces light and adequately dispels its excess in heat. A 90 lumen/ watt part is not saddled with a lot of thermal energy it needs to dissipate. Even this obviously bunk part reviewed here was not a case of instant death or catestrophic failure. We don't know for certain that this blue boy wouldn't have limped on with its blue output for a good number of hours, do we? If the die in this part did not exceed its specified opperating temperature as a result of incomplete bond it may well have held on, yes?

It wasn't the die that was the issue, so obviously, yes, if had been bonded properly, it might worked like one expected. Keep in mind, I'm looking at a variety of performers here, with different characteristics in the same batch.

Once you get delamination started, typically it progresses with time, and especially with each change in temperature.


[CPF-McGizmo] A poor bond is a poor bond and no two ways about it. If the bond were proper and complete, would it be considered appropriate? It is really a shame that this "dud" wasn't tested for flux. I wonder if the part would be capable of high flux with its flawed assembly in the first place. Can you get a U bin flux part with a voided die bond, I wonder?

One would have to do many months to years of testing, across multiple batches, with a large number of samples to prove the technique is viable. Obviously, the part was tested for flux at the factory, being a 6500K W42180 T bin.


[CPF-McGizmo] We have seen a bad apple here but what does that tell us about the good apples?
This LED had an obvious birth defect and one that was seen immediately. Do we know the ramifications of this defect and can we assume a life expectancy for it as a result? Because it had an obvious defect, does this imply that others will have the same defect but not be obvious?

One would have to run the testing I mentioned before, especially if you want to know the long term reliability.

Some companies get deep into the materials, the material suppliers, and the processes of the parts they purchase for their designs, requiring notification of any changes, and specifically what was done. These sorts of changes often lead to problems after the company that is using the parts had ran their system and those parts thru qualification. I've seen it happen too many times to count on all my appendages.

Had I had one part that was an anomaly, and four others that were close to the same, I'd definitely be a little less concerned.

As it stands, IMHO, I'd personally be *quite* anxious to do a lot of qualification testing on the Seoul P4, getting with the manufacturer to assure they will not change anything whatsoever over my expected production period, and get it in writing and signed by the president. That way, I'd be qualifying the parts I plan on using for production.

However, with some companies, you will find they keep sending you samples to re-qualify every 3 to 6 months (after the part is in production), so it makes it hard to do any sort of reliability testing, without a lot of real in depth analysis by some high zoot, highly specialized folks, with lots of experience in this sort of thing.

Anyhow, that is typical of life on the bleeding edge of things...



[CPF-McGizmo]Thanks Newbie.
Isaac, Yes, I understand. Do we know how the die/ ESD device is bonded to the ceramic and how the ceramic is bonded to the sink? The point I should be more specific on, I guess, is is it practice to use epoxy in any of these "laminate" constructs? On dome XR-E's that I have sheard apart, I see a green material which I assume is some type of bonding adhesive, whether it's an epoxy or not, I have no idea.

This is from the other thread:




The square hole in the thermal pad copper:




The electrical via on the electrical connection pad ends:





The green is over the copper, and between the electrical pads and the isolated thermal patch on the backside of the LED. The green is often called solder mask, and is put over copper to prevent oxidization, and between the traces to prevent tree growth.


Here is a great photo by Anglepoise:




The XR-E ESD diode is directly soldered to the copper, much like this XR7090 shown here:




The copper is not epoxied to the ceramic, I've looked at that before. It looks like some direct thermal bond, or growth right on the substrate.


[CPF-McGizmo] In Newbie's defect here, I am convinced that the bond layer was flawed but that doesn't tell me that the bonding material is inappropriate nor does it tell me that the design where such a bond is used is flawed. Without knowing the properties of the adhesive used, can we fault it because it was not applied properly in this case?

That is the million dollar question. Infant failures in adhesives are also often signs of later long term failures.


[CPF-McGizmo] Now I am not inclined to take apart a perfectly functioning P4 to see if it in fact has similar voids in its bond. How many parts could I end up taking apart and see no such voids and yet they could still be there in some small percentage?!?! Maybe? :thinking: :shrug:
This part here that is obviously flawed was also obviously suspect in its behavior when Newbie fired it up. I have had obviously flawed Luxeons and set them aside. Does the obvious in this case imply a concern for the not so obvious?
Seoul in the past has tweaked their parts, IMHO, to get a contender to the starting line. Now that they do have a contender, IMHO, is it possible that further tweaking will be done if and only if it is known to be an improvement? We have to wonder if they are aware of these defects like Newbie has found and we have to wonder if they are inclined to strive towards a part that can and will be dependable. Is the P4, in the majority at least dependable now? They brought us the part before the specs it seems to me. Has maximum drive current been updated from the TBD status that I have seen?

Since links for you are broken like you mentioned above I'll give a link for everyone else, and a page from the November datasheet for you:
SEoul P4 datasheet




[CPF-McGizmo] I agree that these parts are at the bleeding edge. What remains to be seen is how much blood will be let and will the edge smooth out with traffic to the point it no longer cuts?
Ironically, the hours I have spent on this thread and topic could have been spent at the bench where I am modifying heads and reflectors to host the P4 as well as building them. :green: I guess I am out here on the thin ice and I might as well get in some play time while it holds? :)

Hopefully your ventures into new territory will result in some real winners for you. I just got a package from dat2zip with another source of production arts for the Seoul P4 (thanks a million!). I need to run down and get some more 0.165" copper sheet metal for thermal epoxying these to, so I can get some testing done.



I just got production Seoul P4 U bin parts from dat2zip for testing, this one shown here is #1.

I briefly ran the current up to 1 Amp, and unlike the other production batch I got, it is *very* much improved to say the least! I know Seoul has a lot of variation from past experience, but I would have never believed they could be so night and day from one batch to another. But I will let the testing speak for itself, which I'm getting ready to start.


The test rig:




Collecting data takes awhile, to write down, and then type into Excel.

Sorry for the two hour delay.

I took one of my Seoul P4 T bins and also ran the data.


First, here is the Vf comparison chart:




Here is the intensity graph vs. current for three of them:




in reply to CPF-3rd Shift

It appears there is not that much difference between the T and U bins of the ones I've tested here, you are correct- the numbers for the ones that were tested, @ 1A is a difference of 8.65%, from the T bin to the top U bin.

However, if you were to set the light up for a specific brightness output, it could gain you 9% more runtime, and 9% less heat. If you set them up for the same current, I highly doubt anyone could actually see the 9% difference in brightness.

Here is a side view shot on a wall, of the Seoul P4 U bin #2 from Dat2zip, about 2 feet from wall:





[CPF-ViReN]Hello NewBie, it's an excellent setup, so in test currents ranging from 0 - 1A the die temprature is kept at 25 degree celcius? (like they mention in datasheet) ....
Is your wall having gradient color or the LED has that tint variation? or there is yellow at edges and Blue in center... (like old generatin 5mm LED's ?)

No, the die isn't at 25C, but the slug is pretty close.

That is the LED, lol, not my wall changing tint!

Go back and look again, I just updated the post with pictures on a piece of paper about 1" or so away.

Here is a bracketed set of shots from a bumpy reflector, of the U bin I got from Dat2zip, this is #2:










The tint shift on this U bin Seoul P4 from Dat2zip is *vastly* improved over different current levels, watch the movie of #2:

http://www.molalla.net/~leeper/2sp4ud2z.wmv


A shot of the U bin die @ 1A:




Sorry, wish I had.
Here are some side profiles of the U bin Seoul P4 LED #2 in a bumpy reflector (used to hide defects), and bracketed:















[CPF-chris_m]Which is exactly what you'd expect given expected practice of making them all the same way and then picking off the parts which make the higher bin. Suggests the consistency of production is pretty good (at least for that limited sample). Presumably U bins are all towards the bottom end of the bin (though still possibly 95lm at 350mA), and T bins towards the top, in which case the T bins are pretty good value for a ~90lm at 350mA part.

For brightness consistency, yes, in other areas, not really, which has been shown earlier. I'm at least very glad to see these two don't move much in color from 400mA to 1000mA, like several of the others obviously did.

For tint in the different areas of the beam, I'm not seeing a major improvement like I was also really hoping for- for the white wall warrior type. I believe some of this could be mitigated by reflector shape, depth, and textured or jeweled reflectors.

The differences in the new graphs I made for the Vf and the curve shapes is interesting, as shown in the earlier graphs, and I am still wondering what is going on there. Normally I just see offsets with the CREE, but the shape is the same. It could be resistance, and variation in thermal transfer, especially for the shape- but I'm speculating there. From the limited sample, it is hard to say.

I've got more parts from other supply points coming still, it will be very interesting to see how it all pans out.


[CPF-McGizmo]It seems we are moving away from the scope of the problem as may be indicated by one bad apple. No worries.

It was one really bad apple, two questionable apples, one decent apple, and one stellar apple. Both of dat2zip's apples were fine.

I'm hoping this sort of thing doesn't become a batch to batch situation, so far, it is looking like it, future samples will tell us more.


[CPF-McGizmo] Like I have stated before, I don't think ghe XR-E or the Seoul P4 can win a white wall contest against a Luxeon.

Personally, the Seoul P4 reminds me a lot more of the older Nichia products, like their Jupiter, as far as tint vs. angle.(or the old glop phosphor LXHL-BW01 Luxeon from LumiLEDs



At this point the sheer output of CREE's EZ1000 used in the Seoul P4 and the CREE XR-E is a really big deal, IMHO. With LumiLEDs having patented the three best phosphor application methods (so far) for uniform tint, it is unfortunate that other manufacturers are cut off. Combining it with the rough textured reflectors that the hotwire crowd created to get rid of filament images, however, does go a long way towards helping to blend contrasting tints.


[CPF-McGizmo] To relegate the Seoul P4 to cheap or disposable lights is a blunt denial of what its potential can provide, IMHO. It seems this thread is headed towards the white walls and white halls. No problem, but I suggest the P4 is better suited for the real world and outdoors.

Running LEDs over current, to look at tint shift, as well as white walls (since we very obviously have a major white wall crowd presence here on CPF- e.g. Fenix P1D-CE/Lumapower/Huntlight thrashing), not looking at this aspect and not disclosing it would be doing those folks a big disservice. It is a very important aspect for this crowd.

Out of doors comparisons are better done with flashlight running at the same input power and also with the same reflectors. Often these lights are each customized to take advantage of different aspects to achieve a certain goal. IMHO, due to this, these sorts of comparisons are best left at the flashlight level, especially since additional factors come into play, such as the system thermal solution and such.

On a side note:
I did find some white wall shots I did on the CREE XR-E utilizing the same camera, for comparison, over in this thread, about post 57 (and there are now plenty of other XR-E beamshots from various lights on CPF now):
http://candlepowerforums.com/vb/showthread.php?t=138503&page=2&pp=40



[CPF-McGizmo] I set up my integrating sphere and computer to measure a set of 27ST heads I just assembled as I wanted to see if there were any real and measurable discrepencies from what my eyes had told me weren't there. All of these (18 of them) ranged from the runt at 135 lumens to a couple that just cleared 150 lumens. Most were in the low to mid 140's. The 27 heads seem to measure within a few percentage points of [i]real[/i] lumens.

[CPF-McGizmo] HotBeam,
Those lumen measures were at approximately 917 mA of current. They are also net measures out the front of the light that has a deep reflector that isn't hubble telescope quality in reflective finish and through 2 mm of sapphire window that does not have unobtanium A/R coating on both sides and edge. 30% transmission loss maybe?

30% loss for the reflector and 20% for the sapphire? So 50% loss total?

So, are you saying your U bin Seoul P4 are making +280 lumens? Oh wait, you need to further de-rate for heat, so you'd need to make +380 lumens to get 140 lumens out of the flashlight. And some of them need to make more than that!

Scratches head....do I look that gullible?!? Or did I misunderstand you?


Because I certainly have not been able measure any differences between similarly binned Seoul and CREE parts...


If we take conservative numbers:

You are running at 917mA.
917mA/1000mA = 0.917
So, 240 lumens * 0.917 = 220.08 lumens @ 917mA

25% loss for the reflector:
220.8 * .75 = 165.06 lumens @ 917mA

20% loss for die temperature rise:
165.06 * 0.80 = 132.048 lumens @ 917mA

14% caused by the sapphire lens:
132.048 * 0.86 = 113.56128 lumens @ 917mA

So, we are looking at a more realistic number of 114 lumens. Even jtr1962's testing show the lower binned CREE XR-E P4 bin running neck and neck with a top binned Seoul P4 U bin.



The high losses of sapphire/mineral glass really showed up clearly in the Gransee's public testing of the HDS light, if you recall that. I think the measured impact was 14% or so there, with the special sapphire window you installed in that light. How about 16% loss caused by the sapphire/mineral glass? (see chart below)






[CPF-Hellbore]So is this better than the Osram Golden Dragon or Nichia Jupiter?
Better is a relative term. IMHO, each LED has it's own advantages and trade-offs. (IMHO, there is not one LED to rule them all- at the moment.)


Here is the link to the reflector loss testing:

Reflector loss testing


jtr1962 did some lumen testing of the various LEDs, I've overlaid the two graphs below. Looks to be an 8% difference, and who knows if either one is at the top or bottom of the bin:


White LED testing


Keep in mind with LEDs, less of the light is hitting the reflector than with an incandescent bulb, so you have less losses. I don't have time to look, but the 35% losses (65% bulb to flashlight output) was measured with UCL, which is much lower loss than sapphire/mineral glass.


[CPF-Kiessling]I'd be quite interested about some solid data for LED lights with our current reflectors and windows. This would help us tremendously to "guesstimate" the real luminous flux of a device. bernie
P.S.: hey NewBie ... it's more like 35% loss ... :p :D

LOL, you are right, that is what I get for thinking right after the alarm clock goes off.

I believe the reflector losses drop to 25% with a luxeon output distribution (which would be lower for the CREE XR-E, where it would be similar for the Luxeon and the Seoul), according to my simulations.

Toss in 14%-25% additional losses for sapphire (mineral glass)(depending on the grade), or use the actual tested 14% loss from the HDS lights in the integrating sphere.

Say 39%-45% overall losses with sapphire lens plus reflector losses for the Seoul and Luxeon? Then tack on the additional 20% thermal losses @ 917mA.

IMHO, that would still put Don's numbers at way too high a number for the Seoul P4 to achieve, especially once you account for thermal losses.


My chart numbers are not scaled to lumens or lux.

I don't have access to equipment for measuring lumens very often, and have to pull favors. I've been considering purchasing my own NIST traceable reference standard so that I could get in the ballpark at home. I've also considered getting a diffraction grating and sweeping thru the spectrum, recording angle and intensity once reference angles are established from neon, argon, etc. Or, possibly wiring up a linear 2048 cell CCD array. Somewhere around here I have a 256 cell linear array, but if you are going to put out the effort...

The CREE EZ1000 used in the Seoul P4 is a very nice die. It is not pushed at it's maximum, and the thermal resistance in the CREE XR-E and the Seoul P4 are both lower than even the Luxeon K2. On the Seoul datasheet, they specify flux was done at 25 degrees C ambient, and give no other further details about their test conditions- and when LEDs heat up, their flux drops, so if they used an emitter mounted on a MCPCB with a 1" square heatsink, you'd obviously get lumen drop as things heated up. When I tested the Seoul P4, I was quite aggressive with holding the slug temperature as close to ambient as possible. This was not available for other tests. So consider those factors. See picture below:




I just measured the temperature of the slug of the P4 after it has stabilized for many hours, it is at 23.9C @ 1.000A drive. The inlet air temperature to the air cooled heatsink is 23.1C. As a point of reference...


[CPF-Gryloc]
I also wanted ...
Sorry if I got way off topic from the original thread. I always have almost too many questions. Anyway, Thanks!
-Tony

For some folks, punching lumens down range is important, others like more balance between beam and flood, others want a perfect looking beam, some want longer runtimes. Be it the Seoul P4 or the CREE XR-E, they both use the same CREE XR-E die, just implemented differently. You'll get a good amount more lumens out the front, or consume less power doing it, and even run cooler, than options that were not available in the Summer of 2006.

Anyhow, more Seoul P4 parts showed up from yet another source today, so I better get to testing before the night is over.


[CPF-ViReN]From what I have been reading If i understand correctly for SP4 (summarizing)
1) The "Tint Shifting (whole LED turns Blue)" (is a function of Current) will occur only after you cross the limit (perhaps 400 mA) and that Tint shifting is also partly a function of heat. better the heat sinking, you have greater chance of reducing it at least partly.
2) where as "Tint Variation" (on parts die), is majority a function of LED Construction & resultant of poor die-bonding (explained in detail in above posts)

So Basically, as long as you are within current limits, have better heat sinking & that you have a LOP Reflector, you are fine.... the SP4 will "Out Throw" any flux bin of luxeon at any current level at any time

Viren, it all depends-it appears, on the batch, and in the Seoul P4 emitters I got from Mouser, I get four shifters (each at different currents), and one really great one. One of the four had an issue with it, which is covered earlier. I got two Seoul P4 emitters from dat2zip that don't shift much, I took a video of them earlier. I don't feel it would be fair to say they all do X, as it could be better or worse. Example, since they all do not shift blue at 400mA, that would be an unfair statement. Some do it at 500mA, some at 700mA, some at 1000mA, and I've got one that hangs on pretty darn good, even at 1.5A. I'm up to my seventh batch now, and there definitely is a variety of what these Seoul LEDs do.

McGizmo said he has a whole batch of flashlights that don't shift at all.

IMHO, it is pretty hard to lump or stereotype something that has variance.


For a relative size of things, I placed a dime on the old vice in front of the heatpipe based heatsink:




I almost forgot this graph which could be useful for some. I also took my least shifter (my very best Seoul P4), and ran a set of measurements on it.

The blue line is the shift over current, with slug held pretty darn close to ambient. The pink line is with the current held to exactly 750mA, and the slug temperature moved from 25C to 95C. This should help to understand what happens to the shift with current and while a flashlight heats up.





I've only got a few emitters tested from these next batches I have. 3 to be exact.

The one I called Seoul P4-T Lot1 #1 pulled 50mA and made no light, had a lower Vf, and right around 53mA it started to make light, and the Vf read 2.3V, about 0.5V lower than typical, and it's output was lower than typical.

I figure the ESD diode in it is leaky, and no, I did zap it. In my experience, if it was the LED die, it would not have had a typical LED die curve, just offset.

The funky one is in light (bright) green squares.


Updated charts:







[CPF-Opto-King]So, what you are "telling us" is that the T and U bin are almost the same when driven at 1-400mA and that you only "need" the U bin if you are going to drive the LEDs at a high mA?
Also, if I'm reading your chart correct it seems that the LEDs you have tested are above 118lm @ 350mA.

Hell no! You be a dreamin' bigtime!

118lm @ 350mA would be a pipe dream on these. For that, you would need at least a V bin, maybe more once you account for die heating.

The chart is NOT in lumens or lux @ 1m. Just a relative comparison.



[in response to DougS of CPF] Could you elaborate on the typo? Are you talking ( ) or funky and Funky?

The tint shift direction does vary a little bit, depending from which bin you start off at. I wouldn't take this as a rule, the tint shift direction, but as a stereotype. I've ran some Philips LumiLEDs Luxeon III XO bins that follow or parallel the line closely also, but not every one of them shifts identically. The left and right are generally similar, with some variation of the parallelism from one LED to another, from what I have seen in the past. Other color bins have more departure from parallelism, but I have not ran a big enough sample. Sorry guys, I'm not going to run a bunch of these for color, it takes many hours to do...

Here is the XRE chart below, and I'll see what I can dig up on the Luxeon tonight, if I can remember.





[CPF-Doug S]Sure. Text describes #1 as the funky one but graph labels #2 as the funky one.

The funky one is in light (bright) green.
In the first graph (Vf) it is labeled as:
Seoul P4-T lot1 #2 (funky)
In the second graph Relative Brightness it is labeled as:
Seoul P4T Lot1 #2 Funky

That is the second real odd ball part I've found in 10 so far (leaky=funky), and also the two more 700mA tint shifters, but at higher current levels the one that started feeling blue around 500mA.



I finished up three more Seoul P4 emitters, nothing odd found (good thing!-thank goodness)

All the Seoul P4-T Lot1 parts are from the same Intensity, same color bin, and the same single Vf bin. One of the nice things about buying reels, they label exactly what is on the reel, on the reel.

The series I am starting on now, are all Seoul P4-U, and are from the same Intensity, same color bin, and the same single Vf bin. In fact, they are from the same Vf, and color bin as the Lot1 parts.


The updated graphs:







I've updated the Vf chart with the results of the rest of the Seoul P4-U bin. The Vf bin for lot1 and lot2 is 3.25V - 3.50V and Seoul P4 binning is done @ 350mA.




I've expanded the Vf plot a bit:




And the Relative Brightness chart:




One of these last four U bins was below the T bins. Otherwise, you can see a slight split up at 1000mA, between the T bins and U bins. Generally, the U bins have a higher Vf than the T bins.




I have some more K2 parts and Luxeon III parts that I think I might run a couple of curves on tonight, just for additional comparison.

I would be interesting to blast it with one of those de-ionizing compressed air guns usually used for static neutralization, to see if it would reduce the miracle dust magnet personality.




[CPF-nightrider]Great work Newbie. I've been following this thread from the start... lots of nice photos and discussions.

Any chance of getting an XR-E and a Luxeon III thrown in the Current vs Intensity chart?
Thanks.


Sure, I got home early tonight, so I'll hop on it right after I finish eating dinner.




In response to CPF-Yaesumofo, The Seoul P4, and the way it actually pulls dust out of the air is really very uncanny. I've never seen anything like it. Whatever they did, they could make a product out of whatever they used and market it to housewives. Once it is covered, they'd simply rinse it off and re-use it.

But, you can clean it, and get it into a sealed light and you should be fine.


Back to testing...and we have updated graphs...

I tested a number of Luxeons and tossed them on the graphs below:




Take note of where the Seoul P4 Vf fall in reference to the K and H bin Luxeons, above.






The light output comparison above is very interesting, especially when you compare the various bin Luxeons to the Seoul P4 which uses the CREE die.




I've made another graph to put things in better perspective.

Current does not tell everything about an LED.

To the converter and battery, what really counts is the power it consumes, which is known as Watts. Watts is the product (multiplication) of Voltage times Current. W = V * C (a.k.a. P = V * I)


So, I've taken the Power consumed and plotted it against the light output on the vertical scale below:



There are enough data points there, that it is getting hard to make things out, so I've produced a larger version of the graph that you can use, download it here:
Big version of graph


On the larger graph, you will notice that there are some kinks in the lines, which were due to the resolution of the readings on the early parts. I tore apart the power supply and added dual 10 turn potentiometers on the current adjustment, which allows for much finer adjustments, so the later measurements are much smoother, as well as utilized different meters for better resolution.

I have the Seoul P4-T that is still mounted to the heatsink, which allows me to verify the setup is correct, before the start of each set of measurements, it is my reference to ensure consistency.

Both lot1 T and lot2 U are all from the same Vf and tint bins. Lot1 T is specified as 70-90 lumens, and lot2 U is specified as 90 to 110 lumens, both at 350mA.

I hope to get to the CREE tonight, if time permits. I should say CREE XR-E, since the Seoul P4 also uses the CREE die...




I tested a second CREE, which is #2 on the overkill heatsink.

CREE #1 is already mounted on a 2" by 3" by 0.165" piece of copper sheet metal only.

The device used for "brightness" testing was an Extech mounted to a white pail, such that the light from the LED would not have a direct path to the sensor. LED was inserted in a hole in the jug.

Meter here:
Extech 407026


--->>> Caveats <<<:
-Like all light meters, it is not a spectroradiometer, and will have some error depending on the spectral output of the LED. Expect any errors due to spectral differences between LEDs to cast doubt on accuracy of measurements.

-CREE #2 was not mounted on the overkill heatsink, and it's thermal solution is different than the rest of the LEDs under measurement.

-CREE LEDs were both directly soldered to copper, NOT a MCPCB.

-The CREE LEDs were both from the "low output" P3 bin (73.9 - 80.6lm), [u]NOT the Q3 bin (93.9-100.4 lm) which would be in the Seoul P4 U bin range (91.0-118.5 lm).[/u]
-CREE #1's bin appears to be higher than what the vendor that supplied said it was. Possible mix-up at the vendor, or a top performing part out of the bin.

For the testing to be fair, I'd need to obtain parts from the same lumen bins from Seoul and CREE. This comparison here is a bit biased, as I do not have any Q3 bins from CREE yet. Some folks have been running around saying the Seoul P4 is brighter than the CREE XR-E, which is dicey at best to say, since we are not comparing the same lumen bins. Of course a part from a higher lumen bin will be brighter!!! It may be that companies are paying premium for the top end CREE bins, like Surefire, and thus the lack of availability- *conjecture on my part*.(Arrow NAC says that the CREE XR-E P4 bins are available to readily purchase now...)



The new graphs:



The BIG version:






In response to a CPF'er

I fully understand having to grind down the reflector for the Seoul P4, and also having to shim it up ~0.30" to fit into existing solutions. Non-perfect reflectors sure help with a variety of things, and also the rough surface or jeweled ones.

Here is what I got out of a "perfect" MagLite D cell reflector with the Seoul P4:




There are a number of different pictures of the Seoul P4 and various reflectors earlier in the thread, along with some videos of the tint shifting.


I had five of the 13 I tested for this that had various issues. 3 of them would shift extremely blue at 1 Amp. One definitely had a die delamination problem. And I had the leaky one, that drew 50mA and made no light (normally they make light under 1 mA). Above that, it made light, and for the most part, acted like any other Seoul P4, just hampered a bit. Oh, and there was the Seoul P4 U bin that put out less light than a Seoul P4 T bin.


The Seoul P4 is most definitely quite a different beast than the Luxeon III, really smoking the current LumiLEDs parts for sheer light output. It isn't a perfect part, so the white wall warriors will want to keep in mind:
Do you want lots of light, or a virgin white wall beamshot..,

I'm very glad to hear you really like your Seoul P4!

It is a new part, and I imagine it will continue to evolve and improve as they dial things in.




[CPF-yaesumofo]Newbie, I wonder if the Seoul has some sort of charge on it. If it is positively charged (static) and the the dust has a negative charge on it (or the other way around) maybe one needs to work in a static free with non charged air (filtered)flow around the part maybe that would solve the problem. Maybe just applying some sort of anti static charge to it would work.
Remember in the days of Lp's they used to sel