LED Array anomaly

yomamaphat

Newly Enlightened
Joined
Oct 22, 2006
Messages
8
I have an array of 6 K2s and 4 Royal blues. They are mounted to a heatsink using JB Kwik (Fast curing JB Weld). I'm having heat problems, I'm driving the LEDs at ~366mA. It seems I'm using more heatsinking than other people that are driving their LEDs harder and don't seem to have heat issues.

So I was looking for possibly causes, like not having my slugs electrically isolated. So I'd run my array and test voltage between the anode and the heatsink.
Here's the weird part though: When the array is cold and I've just fired it up I get ~8mV across there. I can get more voltage sticking my probes to my tongue. But when I let the sucker get hot, 10-20mins, I start seeing over 3 volts across the anode and heatsink. When it cools everything is back to normal.

I measured resistance in hot JB Kwik and it doesn't seem to be affected by temperature (Still insulative as ever).

Any ideas on this? I'm lost...

Thanks
 

chesterqw

Flashlight Enthusiast
Joined
May 9, 2005
Messages
1,968
Location
singapore,jurong
umm... as far as i know, that JB kwik isn't a thermal compound or thermal epoxy!

jb kwik will not be able to work there....
 

yomamaphat

Newly Enlightened
Joined
Oct 22, 2006
Messages
8
The problem does not seem to be with the JB Kwik in conducting heat anyway. I know JB Weld has been successfully used by a user on this forum.

The fact that the heatsink does get hot means the JB Kwik is doing its job. I was never in question as to whether it conducts heat, it does seem to very well.

My question is about the voltage across the heatsink and anode. Negligible when cold and ~3V when hot. As I said before I measured resistance across hot JB Kwik and it does not seem to conduct electricity dependent on temperature.

If voltage being present across the heatsink and the anode over time indicates the JB Kwik transmitting electricity then this is what I would like to know. Or can the slug and heatsink build up a little capacitance some how?
 

Kinnza

Enlightened
Joined
Nov 15, 2005
Messages
223
Location
Spain
I dont know the JB weld, but some thermal conductive "dielectric" epoxies, like Arctic Silver, are slighty capacitative, and produces strange effects when contacts are close between. Its compatible with the increasing voltage with higher temp. I dont know if its the cause, but check the JB caracteristics.
 

evan9162

Flashlight Enthusiast
Joined
Apr 18, 2002
Messages
2,639
Location
Boise, ID
thermal expansion can cause electrical contact to be made. If the JB weld has a smaller coefficient of thermal expansion, then the slug could be expanding into the heat sink, making electrical contact. Sounds like a classic slug isolation problem.
 

frenzee

Enlightened
Joined
Sep 4, 2006
Messages
216
I would venture to guess that thermal expansion might be culprit.

Attaching Luxeons to a bare heatsink can be real tricky. When you first attach things and test the connections, it might seem fine and everything looks isolated. When it dries and you test it again - surpirse! You might get a voltage drop. Or everything might be fine and when thigs start to heat up, those pesky Luxeon slugs seem to somehow make contact with the heatsink. This has happened to me using both Arctic Silver epoxy and JB Weld. JB Weld has no metal particles in it, so it couldn't possibly conduct electricity at any temperature. So I think in your case thermal expansion is causing things to shift and parts to come into contact.

The way I got around this problem is by coating the heatsink (or if you dare, the slugs) with a thin coat of high-temperature primer, preferably the sandable variety which has suspended ceramic particles - preferably Silicon Carbide. Ceramics are dielectrics with respectable thermal conductivity numbers and given the thinness of the layer, they shouldn't interfere too much with the thermal path.
 

MrAl

Flashlight Enthusiast
Joined
Sep 9, 2001
Messages
3,144
Location
New Jersey
Hi there,

You should use a product made for thermal applications whenever possible.
I have used PC-11 myself with very good results, but i have no idea what
the long term thermal stability is because it's not made for conducting heat.

Also, the heat sink needs to be large enough to handle all those LEDs.
As a min, 1 square inch of heatsink area per 1 watt of LED power.
For example, 1 Luxeon 1 watter should have at least 1 square inch.
You might get away with less than this rule of thumb, but everything
gets hotter.

Perhaps you can post your heat sink size...
 

Kinnza

Enlightened
Joined
Nov 15, 2005
Messages
223
Location
Spain
MrAl said:
Also, the heat sink needs to be large enough to handle all those LEDs.
As a min, 1 square inch of heatsink area per 1 watt of LED power.
For example, 1 Luxeon 1 watter should have at least 1 square inch.
You might get away with less than this rule of thumb, but everything
gets hotter.

Perhaps you can post your heat sink size...

When no space limits, as in flashlights, an active cooling works great, either directly over the heatsink, either by a aircooled system (a cooper wire just below the slugs, and attached hermetically to a fan of the right CFM. For just 13-14watts like your array, a small fan (2") would do the job nicely, for less than 1watt.
 

yomamaphat

Newly Enlightened
Joined
Oct 22, 2006
Messages
8
My heatsink is a slab of Al 12"x2.5"x0.25" with a Pentium 4 heatsink ( W/many Al fins & a large Cu base) JB-Kwik'ed to that. I have a 40x40x20mm 12V fan on that. When the fan is on I can run it for ~20 minutes before it gets to the point where you don't want to touch it too long.

Does it sound like I should be having heat problems with this heatsink?

Thanks for the ideas so far!
 
Last edited:

MrAl

Flashlight Enthusiast
Joined
Sep 9, 2001
Messages
3,144
Location
New Jersey
Hello again,

Well lets see...those K2's are something like 4 watts each right?
That means 40 watts of heating power.
The surface area is 12x2.5=30 square inches on one side partially covered
by the devices. This means the surface area may or may not be good enough
depending on if the back of it is exposed to air too...
If the back is exposed and it is mounted vertically it should be ok as there is
60 square inches minus the area covered by the devices themselves (which is
probably good enough).
If the back is not exposed it probably wont be good enough and the thing
would get very hot. A fan would help here as i guess you found out.

From what it sounds like, maybe the heat sink is mounted flat instead of
vertically. This would mean it requires a fan.

Just one question i have...
When you glued the devices down to the heat sink did you make sure to press
the device down and move it around to makes sure two things occurred:
1. The back of each device had enough compound to completely cover it
2. The layer of compound was kept to a very very thin layer
By moving the device around a little while the compound is still wet, the stuff
gets a chance to completely 'wet' the back of the device and also makes it
spread out so the layer is very very thin. Pressing down while moving it
around helps a lot to thin out the compound layer to keep it thin. Thin things
conduct very well.
 

yomamaphat

Newly Enlightened
Joined
Oct 22, 2006
Messages
8
MrAl: I looked at PC-11 on their website earlier today and I believe it only handles 200°F. JB Kwik handles up to 300°F and JB Weld handles up to 500°F with the highest tensile strength of all three (Not good when trying to remove the LEDs!! :) ). Probably the best thermal conductivity and highest resistance too.

Frenzee: From the MSDS sheets available on JB Welds website I found JB Weld and JB Kwik do have metals in them. Both contain 10-20% "Iron Powder" Normal JB Weld also has 1-5% Titanium dioxide but since the Ti is bound with Oxygen it does not conduct. I'd guess the Iron powder could conduct, it doesn't say it's oxidized or anything, but JB Weld is labeled on their website as an insulator. There are other metals too but most of them are not in pure elemental form so they probably don't conduct either.

I'm going to ditch the JB Kwik and redo my array with normal JB Weld. Some of my array that "leaky" when hot, after popping the LEDs off it looked like a reasonable layer in most cases of JB Kwik. Had I pushed the K2 down I'm pretty sure the slug wouldn't have been isolated. I tried on each one to "float" them on a uniform layer of ~.5mm thick, but that's pretty hard. Usually I'd have a corner pushed a little too thin here and there. Hopefully using normal JB Weld will remedy this.

If any of you have used JB Weld without any prior treatment (extra coatings, etc..), did you push the LED down into it? Or try to float it? I floated my K2s on a layer of JB Kwik and apparently still had heat production through the slug, possibly from thermal expansion. I might spread a thin layer of JB Weld with a razor or credit card and let it cure then put the K2s on that.
 

enLIGHTenment

Enlightened
Joined
Feb 5, 2005
Messages
814
Location
Ottawa, Canada
S478 Pentium 4 coolers are designed to cope with chips that must be kept under 165 degrees C even when putting out over 120W of heat. They'll be hard pressed to manage 40W worth of K2s which would rather run at much cooler temperatures.

Try using a better heatsink. Or get some XR-Es...

If you're dead set on K2s, you can now get K2 stars with electrically isolated MCPCBs:
http://www.luxeonstar.com/item.php?id=2212&link_str=330&partno=05027-PW12
 

MrAl

Flashlight Enthusiast
Joined
Sep 9, 2001
Messages
3,144
Location
New Jersey
Hi again,

Yoma:
I dont think you understood what i was trying to say before.
When you apply your LEDs to the heatsink with epoxy you need to press down
on the LED as the epoxy is curing. i used two small plastic spring clamps one on
each side, to keep pressure on the LED against the heat sink metal as the
epoxy cured. This makes for the thinnest layer possible. Keep in mind that
the only reason the layer should be there for is to take up the air spaces.
If you apply the epoxy to the back of the Luxeon and then put it on the heat sink
and then apply a little pressure and then slide it around, maybe in a little circle,
while applying more pressure. This makes for a nice thin layer of epoxy as opposed
to a thick layer, which doesnt conduct as good as a thin layer.
 

yomamaphat

Newly Enlightened
Joined
Oct 22, 2006
Messages
8
MrAl: My whole issue is not heat conduction through my JB Kwik, the reason for this post is voltage leakage through possibly through too thin a layer of JB Kwik. If I push my K2s down into JB Kwik I'm almost positive the slugs will not be electrically isolated.

enLIGHTenment: This thread has become rather long and up there someone said I was running 40W worth of K2s, but I'm running 10 @ ~350mA and ~3.6V, so only about 12.6W.

My whole question was really about the voltage across the anode and heatsink climbing to ~300mV only when the array heats up and I was wondering if this is due to the slug somehow contacting the heatsink or possibly caused in some other way (capitance building, etc...) The consensus when actually considering my question seemed to be thermal expansion of the slug during heating causing it to come into contact with the heatsink.
 

TMorita

Enlightened
Joined
Sep 5, 2006
Messages
506
Well, it's obvious that you're seeing an example of the Peltier effect!

You should try reverse biasing the your anode/heatsink and seeing if it helps cool the LED.

:)

Toshi
 

MrAl

Flashlight Enthusiast
Joined
Sep 9, 2001
Messages
3,144
Location
New Jersey
Hi again,

Oh ok that's not that much power heating then.

If you are worried about electrical isolation then you should use an
approved method to isolate the LEDs, especially when so many are
involved.
If you want to test for leakage, get a digital meter and start making
some measurements. It may help to run one LED at a time and make
measurements and compare readings between LEDs.
 
Top