K2 and K2 TFFC Technical Information Thread

[HarryN]

Hi, I thought it might be nice to have a thread dedicated to technical information on the Philips Lumileds K2 and K2 TFFC package.

My goal in starting this thread is not make this a "K2 vs Others war" thread, just a place to post concentrated information the K2, how it works with various optics, reflectors, drivers, etc.

These devices are pretty time tested, so there are plenty of users and user experiences. This is an easy LED package for DIY hobby use, so even I can solder them up, and there are many optics options.

Just to start off with some basic information, here is a link to the main Philips Lumileds K2 page:

http://www.philipslumileds.com/products/luxeon/luxeonK2

What I didn't realize until today, which makes a concentrated thread like this valuable to me at least, is that the K2 TFFC is only available in cool white, while the "conventional" K2 package is available in cool, neutral, and warm white, along with various colors. Perhaps I am the only one that didn't know that. - (note - this is no longer the case, as edited below)


Also, for those interested, here is a link to their main distributor (Future) who is selling the K2 TFFC, 200 Lumen (rated) device for under $ 10 in qty 1 each, and less than 1/2 of that in reel qty.

http://www.futureelectronics.com/en...hifted|*LXK2/-PWC4/-0200*|1|,Ny:True,Nea:True

Thanks

Harry


____________________________________________________

Update 18 June 2009, per blindasabat:

a)

You ought to add to the first post that K2 TFFC Neutral 180L @1Amp is availble at Future.

b) Future now stocks cool white 220 lumens (but not 240) even though they are not listed at the Lumileds main site. http://www.futureelectronics.com/en...erSearch|lxk2/-pwc4/-0220|1|,Ny:True,Nea:True

c) Future is selling cool white 200 lumens K2 TFFC with two Vf bins (1000 ma test current)
http://www.futureelectronics.com/en...erSearch|lxk2/-pwc4/-0200|1|,Ny:True,Nea:True

d) There was a time in the past when Future would only sell in 1 reel quantity, and this frustrated some end users. Just so everyone knows, you can buy directly from Future in single unit qty, and I have done it. Of course, there is a limit to what you can buy in cut tape as far as binning, espectially for exact color and Vf.

e) CPF member and dealer Photon Fanatic (whom I have had excellent dealings) is selling some K2s and other LEDs in his sales thread here:
http://www.cpfmarketplace.com/mp/showthread.php?t=169435

Let me know if anyone else is selling K2s and I will post it here.

 

[Nos]

btw, look what ive found .... LUXEON K2 LXK2-PWC4-0220 TFFC

300 lumens @ 1,5 A 😡
and they offer a 3° 60mm optic for the K2 too :green:


http://www.leds.de/p709/High-Power_LEDs/LUXEON_K2/LUXEON_K2_LXK2-PWC4-0220_weiss_300lm_TFFC.html

 

[HarryN]

Those might in fact be reasonable numbers.

If you look at the K2 TFFC data sheet for the -0200 version, the 200 lumens is the tested rating at 1000 ma. The same data sheet shows that the -0200 is rated at a "typical" 275 lumens for 1500ma.

It is not so uncommon for specialty buyers to come up with special bins of LEDs. I managed to get some X bin Lux Vs (with some real effort) when the "typical" was more like a V bin (about 1/2 the output).

The K2 packages are single die based and modeled after the original Lux I and Lux III package, so there are many years of optics out there.

The advantage of single, 1 x 1 mm die packages in general is that they are much easier to focus. The obvious disadvantage is that it is easier to get more light from a larger area 4 die or single large area die - the downside being the engineering needed to achieve a tight beam is "challenging". 😗

 

[SemiMan]

the downside being the engineering needed to achieve a tight beam is "challenging". 😗

Not challenging, actually impossible. Direct correlation between source size and reflector size to achieve the same angle of beam. Etendue.

 

[ledstein]

Future is actually the global exclusive distributor for Luxeon (Lumileds). Every other company or online shop that offers luxeon buys them from future.

Still dont get your hopes up. Future is not build to deal with small customers because they have "bigger fish" to catch. They are very slow even with companies that buy big time.

About datasheets and highpower leds in general. Someone should make a website for making compares and tests with all of them. Untill than the only place you can do this is on a online shop: http://www.ledrise.com/index.php?cat=c2_High-Power-LEDs.html

 

[HarryN]

Hey someone on here explained how I could calculate my junction temp based on voltage drop for one of the cree leds I was experimenting with.

Can someone tell me how to do this for one of the k2 leds? I am experimenting with some of the k2's (not the new ones with ttfe or whatever they have now) Im trying to figure out the same thing but im unsure what figures to use. Im pushing them to 1000ma.

I looked on the phillips website but am unsure if im reading this stuff right.

Under the section which said electrical characteristics at 1000ma it says mv/C and C being 2 degrees.

So my voltage drop was .480 I assume this means 480 / 2 = 240 degrees C. However this doesnt make any sence. I am using a larger heat sink than my crees and the actual led doesnt seem as hot


so am I doing something wrong?? I just dont think it is running that hot

Hi Axkiker, I hope you don't mind if I answer your question in this thread. This is an excellent question for a technical content thread.

I am not sure about using the mv/C aspect to measure temperature of the junction, but there might be an easier alternative method to estimate the junction temperature.

Lumileds quotes the thermal conductivity of the package as 5.5 C / watt. If you meaure the Vf at 1000ma, then drive the LED at the 1000ma, this will provide V x I = power. (watts) Assuming that ALL of the power is converted to heat, then:

Assuming a Vf of 3.5 V x 1000 ma = 3.5 watts.

3.5 watts x 5.5 C /watt = about 20 C pad to junction difference.

This is not really much power to deal with, and it shows why PL is comfortable to use this same K2 package for 1500ma (about 5 watts). The K2 package is much better at heat dissipation then the older Lux III / Lux V package which were also driven at 5 watt levels sometimes.

 

[Axkiker]

Hi Axkiker, I hope you don't mind if I answer your question in this thread. This is an excellent question for a technical content thread.

I am not sure about using the mv/C aspect to measure temperature of the junction, but there might be an easier alternative method to estimate the junction temperature.

Lumileds quotes the thermal conductivity of the package as 5.5 C / watt. If you meaure the Vf at 1000ma, then drive the LED at the 1000ma, this will provide V x I = power. (watts) Assuming that ALL of the power is converted to heat, then:

Assuming a Vf of 3.5 V x 1000 ma = 3.5 watts.

3.5 watts x 5.5 C /watt = about 20 C pad to junction difference.

This is not really much power to deal with, and it shows why PL is comfortable to use this same K2 package for 1500ma (about 5 watts). The K2 package is much better at heat dissipation then the older Lux III / Lux V package which were also driven at 5 watt levels sometimes.

Okay so im kinda lost which is nothing new.....

Say you did the math and came up with a 20 C pad to junction difference. Would you then take that temp and add it to the ambient air temp. Typically 21 C at my home. which would make junction temp 41 C

Or am I totally off?

 

[HarryN]

Okay so im kinda lost which is nothing new.....

Say you did the math and came up with a 20 C pad to junction difference. Would you then take that temp and add it to the ambient air temp. Typically 21 C at my home. which would make junction temp 41 C

Or am I totally off?

You are getting closer. There are actually several thermal barriers between the die and the air.

- Junction to thermal pad = approx 20 C

- Path through the thermal adhesive +
- Path to the walls of the light +
- Path between the walls of the light and the air =

Commonly another 30 - 40 C.

So 21 C + 20 C + 30-40 C = approx 80-90 C at the die.

This is true no matter the brand of LED.

Lumileds has some very handy guides for approximating these results and some examples worked out in their Application Briefs linked here:

http://www.philipslumileds.com/solutions/portablelighting

Just click on Application briefs, and download AB 05 (Lux III) and AB 23 (Lux V). They are about 5 - 6 years old, but still valid in general, its just that the K2 package is 5.5C/watt vs the Lux V was 8 C / watt.

You don't need to understand it all to follow the pre-worked out examples and the test results.

This is more or less similar to any other power electronic device. As any electronic device gets hot, its performance will often slightly degrade and the reliability will decrease.

It's kind of funny with regard to reliability and the old Lux V. Lumileds took a lot of flack for having a 500 hr rating on that LED package (intended use - flashlights). I added up one time how many times I would have to charge up and completely run down my Lux V light to have it degrade to 70% output - it will never happen.

 

[blasterman]

just a place to post concentrated information the K2

What's the CRI of a warm-white K2/TFFC? Can't seem to find the answer anywhere.

 

[maxspeeds]

If I remember correctly, K2 TFFC has a typical CRI of 70. Not sure on their warm bins :tinfoil:

 

[blasterman]

Yeah...I know the cool tint has a CRI 70, but I can't find a reference for their WW. *Usually* these are higher, but I want to see it spec'd first before I try them for fixed lighting.

A low CRI just means they're pumping specific spectra to meet benchmarks, not light quality.

 

[HarryN]

What's the CRI of a warm-white K2/TFFC? Can't seem to find the answer anywhere.

Hi Blasterman, I didn't know the answer so I took the time to look at their web site for you.

Here is a link to the web page for the K2 and K2 with TFFC:

http://www.philipslumileds.com/products/luxeon/luxeonK2

and here is a link to the K2 data sheet:

http://www.philipslumileds.com/pdfs/DS51.pdf

They did rather bury the answer, but on page 8 of the data sheet, note 5, it lists the CRI for the K2 as:

Cool white = 70
Neutral white = 75
Warm white = 80

AFAIK, there the K2 TFFC is only available in Cool White right now. If a CRI of 80 is sufficient, then maybe the WW K2 would work for you.

If you are looking for an even higher CRI with "nearly identical incan like output, there is a product specifically for this application, but currently only in a Luxeon I format.

The datasheet is here:

http://www.philipslumileds.com/pdfs/DS23.pdf

and a spectral output graph is on page 9.

It is not high output at all and based on the older Lux I package, but if your goal is to mimick an incan, this is your ideal part.

IMHO, if your goal is to have:
- LED based general lighting
- high efficiency
- high color quality (not just high CRI, which can be tricked)

You are always better off to use color mixing than to buy a white LED, regardless of brand.

There are a lot of people that like to use an RGB based color mixing, which is pretty good at fooling your eyes, but I am still a big believer in using at least 9 colors in general lighting, starting at around 450nm (or lower if you can get it) and on up to around 670nm (very deep red).

Just drive them in series at 750 - 1000ma and it will beat most white LEDs in almost every measurement.

A low CRI just means they're pumping specific spectra to meet benchmarks, not light quality.

Often this is true. The data sheets for most power LEDs show the spectral graphs so you can get a better picture of this. Most white LEDs are deficient in red and red / orange, so you can get a lot of benefit from a cool white in series with a red and red/orange. The highest output red and red/orange that I know of is their Lux III line - massive die in those packages.


Its funny, I didn't even have to look hard to find this info. Maybe I should go work for an LED company. :thinking:

 

[Axkiker]

You are getting closer. There are actually several thermal barriers between the die and the air.

- Junction to thermal pad = approx 20 C

- Path through the thermal adhesive +
- Path to the walls of the light +
- Path between the walls of the light and the air =

Commonly another 30 - 40 C.

So 21 C + 20 C + 30-40 C = approx 80-90 C at the die.

This is true no matter the brand of LED.

Lumileds has some very handy guides for approximating these results and some examples worked out in their Application Briefs linked here:

http://www.philipslumileds.com/solutions/portablelighting

Just click on Application briefs, and download AB 05 (Lux III) and AB 23 (Lux V). They are about 5 - 6 years old, but still valid in general, its just that the K2 package is 5.5C/watt vs the Lux V was 8 C / watt.

You don't need to understand it all to follow the pre-worked out examples and the test results.

This is more or less similar to any other power electronic device. As any electronic device gets hot, its performance will often slightly degrade and the reliability will decrease.

It's kind of funny with regard to reliability and the old Lux V. Lumileds took a lot of flack for having a 500 hr rating on that LED package (intended use - flashlights). I added up one time how many times I would have to charge up and completely run down my Lux V light to have it degrade to 70% output - it will never happen.

hey first off thanks so much for helping me figure out all this crazy stuff. I never would have imagined there was this much to learn.

So tell me is this all sounds correct.

I set my led up and allowed it to run for roughly 30 min to make sure it was up to temp. Then I set the current at 1000ma. I then took a voltage reading and came up with 3.94V

So I take 3.94V x 1000ma = 3.94W

3.94w x 5.5C = 21.67c

21.67c + ambient temp 22c + 40c = 83.67c

So from my understanding 83.67c is well within acceptable range for the k2's

am I on track

thanks

 

[HarryN]

hey first off thanks so much for helping me figure out all this crazy stuff. I never would have imagined there was this much to learn.

So tell me is this all sounds correct.

I set my led up and allowed it to run for roughly 30 min to make sure it was up to temp. Then I set the current at 1000ma. I then took a voltage reading and came up with 3.94V

So I take 3.94V x 1000ma = 3.94W

3.94w x 5.5C = 21.67c

21.67c + ambient temp 22c + 40c = 83.67c

So from my understanding 83.67c is well within acceptable range for the k2's

am I on track

thanks

Yes, this looks correct. Of course, the 40 C delta from the pad to air is an estimate, but yes, you are fine. Pretty much any old flat piece of Al that has 6 x 6 inches will do this for you.

 

[Axkiker]

Yes, this looks correct. Of course, the 40 C delta from the pad to air is an estimate, but yes, you are fine. Pretty much any old flat piece of Al that has 6 x 6 inches will do this for you.

Okay well then maybe I am off again because im not using a heat sink nearly that big.... I didnt realize the 40c was based upon a heat sink that large.

Im just trying to figure out a way that I can experiment with heat sinks and get an estimate as to the junction temp. This way I can determine the layout and design of my fixture.

 

[HarryN]

My number was just conservative.

If you are looking for just a way to do some proto work safely (as in, not kill the K2), then you can go with less.

Example - I have a "test bed" that is made from a 2 x 4 with a piece of threaded rod sticking out. The rod sticks out about 4 inches and the untheaded center of the rod is about the dia of the thermal pad. So - how well does it work ? The steel ones will not keep up with a Lux V but an Al one will.

The threading is handy - sort of like fins.

It sounds like you are attempting to understand and compare the performance of various thermal adhesives. As you note, there are a variety of them out there. I spent a lot of time on this in the past, so if you don't object, I will throw out my "experiences". (all at least 2 year old info using Lux III and Lux Vs)

a) Adhesion vs flexible
- The adhesives that were the "strongest bond specs" were "inflexible"
- At approx 1000 psi, the bonds were rated as flexible, at 2-3K psi, they were rated as inflexible.
- I only tested materials with a manufacturere's spec bond rating, and AA did not provide this on any of their products at that time. Not sure about now.
- I did some unscientific testing to attempt to debond these from mechanical shock, and flexible always won for adhesion.

b) Preclean or not
- Precleaning the thermal pad with pharaceutical grade iso propanol also always won for adhesion.

c) Thermal Transfer
- It is very hard to really double check the thermal adhesive specs (at home), easier in a real lab
- Curing while under pressure helps keep the bond line thin.
- I think you can just solder these K2's to the pad, so that is easier.

Testing method
- Once again, unscientific, only 2 - 3 samples of each material
- I made up an example board, heat spreader, etc and did the full mounting
- Test 1 involved throwing the parts up into the air as high as I could and letting them fall on the concrete driveway multiple times until failures occurred. (parts physically fell off) :naughty:
- Test 2 involved throwing the assemblies against a stone wall as hard as I could from about 10 ft until failures occurred. :naughty:
- Test 3 involved a hammer and chisel impacting sideways at the bond line. (only used on bonds that survived the first 2 tests. The hammer won, but did show very easily the value of the pre-clean)


Based on my limited experience, I would use a 2 part epoxy with about 1000 psi bond strength and listed as a "flexible" bond, precleaned with decent rubbing alcohol. These were brand name thermal epoxies recommended on the Lumileds data sheets at the time.

The good ones were VERY hard to debond. :thumbsup:

 

[Axkiker]

My number was just conservative.

If you are looking for just a way to do some proto work safely (as in, not kill the K2), then you can go with less.

Example - I have a "test bed" that is made from a 2 x 4 with a piece of threaded rod sticking out. The rod sticks out about 4 inches and the untheaded center of the rod is about the dia of the thermal pad. So - how well does it work ? The steel ones will not keep up with a Lux V but an Al one will.

The threading is handy - sort of like fins.

It sounds like you are attempting to understand and compare the performance of various thermal adhesives. As you note, there are a variety of them out there. I spent a lot of time on this in the past, so if you don't object, I will throw out my "experiences". (all at least 2 year old info using Lux III and Lux Vs)

a) Adhesion vs flexible
- The adhesives that were the "strongest bond specs" were "inflexible"
- At approx 1000 psi, the bonds were rated as flexible, at 2-3K psi, they were rated as inflexible.
- I only tested materials with a manufacturere's spec bond rating, and AA did not provide this on any of their products at that time. Not sure about now.
- I did some unscientific testing to attempt to debond these from mechanical shock, and flexible always won for adhesion.

b) Preclean or not
- Precleaning the thermal pad with pharaceutical grade iso propanol also always won for adhesion.

c) Thermal Transfer
- It is very hard to really double check the thermal adhesive specs (at home), easier in a real lab
- Curing while under pressure helps keep the bond line thin.
- I think you can just solder these K2's to the pad, so that is easier.

Testing method
- Once again, unscientific, only 2 - 3 samples of each material
- I made up an example board, heat spreader, etc and did the full mounting
- Test 1 involved throwing the parts up into the air as high as I could and letting them fall on the concrete driveway multiple times until failures occurred. (parts physically fell off) :naughty:
- Test 2 involved throwing the assemblies against a stone wall as hard as I could from about 10 ft until failures occurred. :naughty:
- Test 3 involved a hammer and chisel impacting sideways at the bond line. (only used on bonds that survived the first 2 tests. The hammer won, but did show very easily the value of the pre-clean)


Based on my limited experience, I would use a 2 part epoxy with about 1000 psi bond strength and listed as a "flexible" bond, precleaned with decent rubbing alcohol. These were brand name thermal epoxies recommended on the Lumileds data sheets at the time.

The good ones were VERY hard to debond. :thumbsup:

Actually I am experimenting with different types of heat sink materials. Not thermal adheasives. I have a material which is an epoxy base that is meant to dissipate heat once hardened. You can vary the mixture and additives to create more or less heat dissipation properties. So im trying to create the mix which works best for my situation. However until I have a way to measure junction temp I dont know if im dissipating enough heat.

am i making sence

 

[blasterman]

They did rather bury the answer, but on page 8 of the data sheet, note 5, it lists the CRI for the K2 as:

Cool white = 70
Neutral white = 75
Warm white = 80

Thank you - appreciated. The Cree Warm Whites have a CRI of 80, so I have a frame of reference now.

80 is fine for general fixed lighting needs, which is why the Cree is popular, so if I experiment with the K2/TFFC color quality needn't be an issue because it seems the K2/TFFC is on par. Those who think CRI isn't a big deal should put some older Luxeons (CRI 70) in their living room and note how the color is awful.

 

[ledstein]

TFFC is getting started very well now(after 18 month of uncertain information), large quantities are available. Neutral white (LUXEON K2 LXK2-PWN4-0140, white, 195lm) is available too but only in low brightness. For usual white even very high brightness of 0220 are available in large quantities and some small quantities of 0240 have been "seen around" 😛
The main advantage of K2 LEDs in general is the extreme heat resistance. These LEDs can be used at junction temperature of >150°C at which other LEDs are already literally melting.
Especially if using narrow focused optics the K2 can be way better then the MCE, because the light output is much more equal thanks to only one chip.
In terms of CRI the TFFC and usual K2 don't seem to be very much different and are the usual average type.

 

[Axkiker]

So I think im still lost here. Since im dealing with a heat sink material which has never really been used as a heat sink is there a way for me to calculate junction temp. Either by voltage drop or whatever way you all can fill me in on...