DiY LED - Cree CXA3070

wirat · Mar 21, 2015

Abiqua said:
I would bet low thermal impedance, like the the rest of the "layer" types....I have looked at a few....even bought some from Digikey in the past, the 3m mounting tape....but all have shitty impedance compared to paste, which sucks, because applying paste manually imho, is one of the variables that needs to be eliminated....

Thanks, exactly my point about why I thought maybe they were not any good since I haven't seen anybody using them here.

Abiqua · Mar 21, 2015

wirat said:
Thanks, exactly my point about why I thought maybe they were not any good since I haven't seen anybody using them here.

I don't think they are "not any good" at least not yet...because even though they do have lower thermal impedance, their thermal conductivity IS superior to pastes.....sometimes by as much as 2x or more....

and my feelings on the thermal impedance rating on Thermal paste, is that is based off the thinnest possible line that is achievable without breakdown and I have huge doubts whether the best impedance [least] can be achieved by manually applying it...just imho...
I wish I had more time to investigate! :joint:

Doer · Mar 21, 2015

Abiqua said:
I would bet low thermal impedance, like the the rest of the "layer" types....I have looked at a few....even bought some from Digikey in the past, the 3m mounting tape....but all have shitty impedance compared to paste, which sucks, because applying paste manually imho, is one of the variables that needs to be eliminated....

Well, what is preferred attach? Must be those three screw holders. At least there aren't 4 screws.

Abiqua · Mar 21, 2015

Doer said:
Well, what is preferred attach? Must be those three screw holders. At least there aren't 4 screws.

More testing is needed imho to flush out that answer...I rely on the datasheets for the time being, there is nothing else....

Doer · Mar 21, 2015

Abiqua said:
I don't think they are "not any good" at least not yet...because even though they do have lower thermal impedance, their thermal conductivity IS superior to pastes.....sometimes by as much as 2x or more....

and my feelings on the thermal impedance rating on Thermal paste, is that is based off the thinnest possible line that is achievable without breakdown and I have huge doubts whether the best impedance [least] can be achieved by manually applying it...just imho...
I wish I had more time to investigate!

For those who need to keep hitting the books like me, here are the basic definitions.
http://www.arlon-thermal.com/thermal_prop/

Just quickly:
Thermal conductivity, the rate of passage of heat / time, thru a material

Thermal resistance, the insulation value of a material

Thermal impedance, material resistance values plus contact resistance values

We add it all up what do we get? Impedance. What do we want? Lower impedance. When do we want it? Always. We don't want a breakdown of any of the materials, but beyond that, we don't want to impede any heat flow, and we want the fastest outflow.

Doer · Mar 21, 2015

Abiqua said:
More testing is needed imho to flush out that answer...I rely on the datasheets for the time being, there is nothing else....

Well, so many were talking about "all those screw holes to tap." They must be talking about something other than COBs.

But, there is something else. There is the Cree XLamp design guide I posted this morning, here. There is where I thought I saw those hold down rings. They said another benefit was it protects the emitter, providing a raised edge.

Abiqua · Mar 21, 2015

Doer said:
For those need to keep hitting the books like me, here are the basic definitions.
http://www.arlon-thermal.com/thermal_prop/

Just quickly:
Thermal conductivity, the rate of passage of heat / time, thru a material

Thermal resistance, the insulation value of a material

Thermal impedance, material resistance values plus contact resistance values

You add it all up what do you get? Impedance. What do you want? Lower impedance. When do we want it? Always. We don't want a breakdown of any of the materials, but beyond that, we don't want to impede any heat flow, and we want the fastest outflow.

Right, you could have the all the Thermal conductivity you wanted, but if thermal impedance is not low in an object, rate of transfer is affected....

Abiqua · Mar 21, 2015

Doer said:
Well, so many were talking about "all those screw hole to tap." They must be talking about something other than COBs.

But, there is something else. There is the Cree XLamp design guide I posted this morning, here. There is where I thought I saw those hold down rings. They said another benefit was it protects the emitter, providing a raised edge.

Can you explain a littler further? For cree you need holders to mount with screws, which is specified DS and recommended...Bridgelux builds it in to the Led case itself....

IF that doesn't answer your question, could you elaborate a little, there are several points I see being made :peace:

SupraSPL · Mar 22, 2015

Hey guys, currently I still use the paste + kapton method and no longer polishing the heatsinks. I will sand the heatsink surface if it is warped. It is a no drill method and if you use the Vero29 with pico ez mate no soldering either. Kapton has a tenacious, long lasting grip on clean aluminum and copper. It is worthwhile to clean the heatsink surface before applying the tape and I would warn against reusing the tape. Its job is to stop the COB from moving laterally or twisting and to to keep gravity from allowing air to get behind the COB. So when you apply the tape you want to stretch it over the corners of the COB to lock it in place.

I have tested the Vero 29 @ 2.47A (96W) and temp droop was 2% with this method. Using a pair of heatsinks, one sanded/polished with PK3 paste and the other stock with stock paste, same exact result 2%. I tried pressing hard on the COB to see if the Vf/junction temp would change and it did not, so I do not believe there is anything to be gained by increasing pressure. In other words, we are here, not a lot of room for improvement:

SupraSPL · Mar 22, 2015

In order to be sure of a good thermal transfer, I waste paste. It goes against the traditional wisdom of using a small amount. I dab the paste across the back of the COB, then press hard on the corners for 60 seconds. The paste slowly makes its way out the sides as the layer gets thinner. This allows air to escape and you can visually verify that every part is covered. Maybe you could waste less if you used a tool to spread the paste on the COB but a 30G tube is good for 50-75 COBs.

Then I clean up the excess with paper towel and alcohol and tape 2 corners.

There is some risk with this method. If the wire is pulling on the COB or the tape loses its grip due to contaminants on the heatsink surface etc, you could end up with a fried egg. I have never had that happen, but Dloomis lost one and someone else mentioned it a few pages back. But if drilling is stopping you from DIY, it is always good to have options. A high conductivity thermal adhesive pad might be worth a shot. It would be interesting to test it on the 85W Vero29 setup and see if it performs the same as the thin layer of paste.

Doer · Mar 22, 2015

Doer said:
Well, what is preferred attach? Must be those three screw holders. At least there aren't 4 screws.

Abiqua said:
Can you explain a littler further? For cree you need holders to mount with screws, which is specified DS and recommended...Bridgelux builds it in to the Led case itself....

IF that doesn't answer your question, could you elaborate a little, there are several points I see being made

Thanks. I am total clear now. I was indeed in a planning mix up about attaching Bridgelux vs Cree. All straight now. Thanks again and again.

I try to pay all this forward. I found the holders last night, but after I saw a certain someone's use of them, I am not sure I like the look. And I think they might be over large for a 2 inch water sink.

I am leaning toward just epoxy for all time and forget it. The sink tube is $50, so, I don't need to worry about recovering a $300 heat sink. And the COBs? They will last a long time at 700ma.

I will probably just get it prototyped with Kapton and Thermal paste.

Doer · Mar 22, 2015

SupraSPL said:
Hey guys, currently I still use the paste + kapton method and no longer polishing the heatsinks. I will sand the heatsink surface if it is warped. It is a no drill method and if you use the Vero29 with pico ez mate no soldering either. Kapton has a tenacious, long lasting grip on clean aluminum and copper. It is worthwhile to clean the heatsink surface before applying the tape and I would warn against reusing the tape. Its job is to stop the COB from moving laterally or twisting and to to keep gravity from allowing air to get behind the COB. So when you apply the tape you want to stretch it over the corners of the COB to lock it in place.

I have tested the Vero 29 @ 2.55A (86W) and temp droop was about 2% with this method. Using a pair of heatsinks, one sanded/polished with PK3 paste and the other stock with stock paste, same exact result 2%. I tried pressing hard on the COB to see if the Vf/junction temp would change and it did not, so I do not believe there is anything to be gained by increasing pressure. In other words, we are here, not a lot of room for improvement:

View attachment 3377669

Extrapolated curve to 85w. I can't go without praising that. Good job!

Doer · Mar 22, 2015

SupraSPL said:
In order to be sure of a good thermal transfer, I waste paste. It goes against the traditional wisdom of using a small amount. I dab the paste across the back of the COB, then press hard on the corners for 60 seconds. The paste slowly makes its way out the sides as the layer gets thinner. This allows air to escape and you can visually verify that every part is covered. Maybe you could waste less if you used a tool to spread the paste on the COB but a 30G tube is good for 50-75 COBs.
View attachment 3377671
Then I clean up the excess with paper towel and alcohol and tape 2 corners.
View attachment 3377672
View attachment 3377674

There is some risk with this method. If the wire is pulling on the COB or the tape loses its grip due to contaminants on the heatsink surface etc, you could end up with a fried egg. I have never had that happen, but Dloomis lost one and someone else mentioned it a few pages back. But if drilling is stopping you from DIY, it is always good to have options. A high conductivity thermal adhesive pad might be worth a shot. It would be interesting to test it on the 85W Vero29 setup and see if it performs the same as the thin layer of paste.

Waste paste. I do that too. I fried a Peliter Junction once. My idea was to scrap it really thin and just fill the voids only. Nope. Not a good idea.

Doer · Mar 22, 2015

They do mention a thermal pad, on that chart that does as well as the grease. I found this one.
http://www.farnell.com/datasheets/99609.pdf
SilPad's, Q-Pad II is a composite of .0015" aluminum foil coated both sides with .0025” thick thermally/electrically conductive Sil-Pad rubber. It is designed for those applications where maximum heat transfer is needed and electrical insulation is not require . Q-Pad II is the ideal thermal interface material to replace messy thermal grease compounds.

Thermal Resistance,°C-in2 /W = 0.10

Wow, at digikey a 19mm x 12mm piece is $0.70!!! I would need 4 per COB. And they don't sell another size. So, that's out for cost alone.

Doer · Mar 22, 2015

I was reading that pre-heat is recommended for soldering the COB pads, I think it is 150F.

Does anyone bother with that?

SupraSPL · Mar 22, 2015

I believe greengenes found some thermal pads that were even higher conductivity than PK3 but they were hella expensive.

I pre tin the COB pads but I don't preheat the pads. I use 63/37 lead solder + flux gel, so it is very easy to keep low soldering temps. It is possible to damage the solder pads if the wire is pulling on the pad as it cools. That is more likely to be a problem if using large stiff wires. Despite that I have never seen a COB disabled by a dislodged solder pad.

Abiqua · Mar 22, 2015

Doer said:
They do mention a thermal pad, on that chart that does as well as the grease. I found this one.
http://www.farnell.com/datasheets/99609.pdf
SilPad's, Q-Pad II is a composite of .0015" aluminum foil coated both sides with .0025” thick thermally/electrically conductive Sil-Pad rubber. It is designed for those applications where maximum heat transfer is needed and electrical insulation is not require . Q-Pad II is the ideal thermal interface material to replace messy thermal grease compounds.

Thermal Resistance,°C-in2 /W = 0.10

Wow, at digikey a 19mm x 12mm piece is $0.70!!! I would need 4 per COB. And they don't sell another size. So, that's out for cost alone.

Yes, but what is the thermal impedance? That is the kicker....if we were truly after the lowest impedance, we would go away from the paste imho.....too many inconsistencies...I know SupraSPL has great data, but more paste does not equal less thermal impedance...thin to a point is what we are after and what thickness would that be? that is most unclear.....even with evidences of only 2% temp droop, it still can be inconsistent....

Doer · Mar 22, 2015

Within the margin of error, I have to add. These are a bit pricy, so I rather add bit more and suffer the slight degradation. It's a trade off.

apoulin · Mar 22, 2015

Jeeyah said:
Damn. I've been getting Z2's on Digikey. What was I thinking?

lol yea I got me a z2 from there too....I'm definitely a more informed buyer now!

SupraSPL · Mar 22, 2015

Abiqua said:
Yes, but what is the thermal impedance? That is the kicker....if we were truly after the lowest impedance, we would go away from the paste imho.....too many inconsistencies...I know SupraSPL has great data, but more paste does not equal less thermal impedance...thin to a point is what we are after and what thickness would that be? that is most unclear.....even with evidences of only 2% temp droop, it still can be inconsistent....

What could perform better than a thin layer of PK3 paste?

DiY LED - Cree CXA3070

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member

Well-Known Member