Maker Pro
Maker Pro

Using a PCB as a heatsink

M

Mike Noone

Jan 1, 1970
0
Hi - I'm going to be making a board that has a large number of high
power (1W) LEDs. I'd ideally like to use the PCB as a heatsink to get
rid of all the heat. How much surface area/watt do I need to allow?
Thanks!

-Mike
 
B

Boris Mohar

Jan 1, 1970
0
Hi - I'm going to be making a board that has a large number of high
power (1W) LEDs. I'd ideally like to use the PCB as a heatsink to get
rid of all the heat. How much surface area/watt do I need to allow?
Thanks!

-Mike

How about
http://www.alibaba.com/catalog/11083603/Aluminum_Based_Copper_clad_Laminate_Sheet.html

If things get really hot?



Regards,

Boris Mohar

Got Knock? - see:
Viatrack Printed Circuit Designs (among other things) http://www.viatrack.ca

void _-void-_ in the obvious place
 
E

Eeyore

Jan 1, 1970
0
Mike said:
Hi - I'm going to be making a board that has a large number of high
power (1W) LEDs. I'd ideally like to use the PCB as a heatsink to get
rid of all the heat. How much surface area/watt do I need to allow?

How many ? Use some simple logic and work out how hot your pcb is going to get.
Hint: 10W is a lot in an enclosed space.

Graham
 
M

MassiveProng

Jan 1, 1970
0
How many ? Use some simple logic and work out how hot your pcb is going to get.
Hint: 10W is a lot in an enclosed space.

Make small PCBs for each LED (1 sq cm ea or less), and mount those
above the driver board like a third of a cm or more with pins or stiff
wires, and the airgap will do a lot of cooling with a bit of flow.
Leave both sides of the tiny boards fully cladded on one or both sides
(of course negate shorts), and leave them bare ie no mask so they
radiate their heat better. Heavy gauge copper or SPC wire (solid)
would allow some heat conduction as well.
 
R

Robert Lacoste

Jan 1, 1970
0
Mike Noone said:
Hi - I'm going to be making a board that has a large number of high
power (1W) LEDs. I'd ideally like to use the PCB as a heatsink to get
rid of all the heat. How much surface area/watt do I need to allow?
Thanks!

Hi Mike,

Power leds (1W and above) definitively need an aluminium subtrate (SMI type
PCB, ie PCB with the base epoxy replaced by aluminium), and this SMI needs
to be efficiently fitted on an appropriately calculated heatsink. Even
Alumium PCB alone is not sufficient, especially if you have several LEDs...
A good reference is Luxeon's "Thermal Design" application note available
here : http://www.luxeon.com/pdfs/AB05.pdf

If you can't design your own SMI board then the only reasonnable solution is
to by "level 2 LEDs", meaning LEDs already mounted on a small SMI, and to
fix these small SMIs on a good heatsink.

Cheers,
 
E

Eeyore

Jan 1, 1970
0
MassiveProng said:
Make small PCBs for each LED (1 sq cm ea or less), and mount those
above the driver board like a third of a cm or more with pins or stiff
wires, and the airgap will do a lot of cooling with a bit of flow.
Leave both sides of the tiny boards fully cladded on one or both sides
(of course negate shorts), and leave them bare ie no mask so they
radiate their heat better. Heavy gauge copper or SPC wire (solid)
would allow some heat conduction as well.

All good advice but you have to get the heat out of the enclosure eventually.

I unintentiomally left some paprework on top of my digital 'set top box' yesterday
and was astonished to see how warm it got. I had some picture breakup too which
alerted me to it.

The other important thing is to check how hot it gets *after several days* !

Graham
 
J

John Larkin

Jan 1, 1970
0
Make small PCBs for each LED (1 sq cm ea or less), and mount those
above the driver board like a third of a cm or more with pins or stiff
wires, and the airgap will do a lot of cooling with a bit of flow.
Leave both sides of the tiny boards fully cladded on one or both sides
(of course negate shorts), and leave them bare ie no mask so they
radiate their heat better. Heavy gauge copper or SPC wire (solid)
would allow some heat conduction as well.

Bare copper has a thermal infrared emissivity near zero; it's hard to
find anything that radiates worse.

John
 
M

Mike Noone

Jan 1, 1970
0
How many ? Use some simple logic and work out how hot your pcb is going to get.
Hint: 10W is a lot in an enclosed space.

Graham

Probably about 10 maximum. It won't be too enclosed though...
 
R

RHRRC

Jan 1, 1970
0
Make small PCBs for each LED (1 sq cm ea or less), and mount those
above the driver board like a third of a cm or more with pins or stiff
wires, and the airgap will do a lot of cooling with a bit of flow.
Leave both sides of the tiny boards fully cladded on one or both sides
(of course negate shorts), and leave them bare ie no mask so they
radiate their heat better. Heavy gauge copper or SPC wire (solid)
would allow some heat conduction as well.

Since this will obviously not work why go to all that trouble just to
destroy some leds?

You will need an mcpcb mounted to a heatsink.
Do the sums and work out the required theta J to A. Thence with the
knowledge of the led theta J to C the 'heatsinking' ability of the
mounting is revealed.
Look at the mcpcb data then calc the heatsink required.
 
M

MassiveProng

Jan 1, 1970
0
Bare copper has a thermal infrared emissivity near zero; it's hard to
find anything that radiates worse.

I do not agree with the number you "selected" to use, but yes, any
polished surface will have a lower emissivity. I forgot to mention
grit blasting the PCBs to make the finish matte.


Bare copper with a matte finish has a higher emissivity than any
type of copper under an insulative blanket of solder mask.

Heat conducts through the device leads or package, to the PCB copper
and substrate, and then radiates to air. If it has an added layer of
solder mask, that radiation will be less.
 
M

MassiveProng

Jan 1, 1970
0
Since this will obviously not work why go to all that trouble just to
destroy some leds?

You will need an mcpcb mounted to a heatsink.
Do the sums and work out the required theta J to A. Thence with the
knowledge of the led theta J to C the 'heatsinking' ability of the
mounting is revealed.
Look at the mcpcb data then calc the heatsink required.

There is also conduction cooling through the large leads mentioned
into the main PCB as well as their radiating surfaces. This works.

One can also buy a bit of sheet copper, and solder a little U shaped
piece onto each PCB, or across all them, linking them together.

For a mere LED. Yes... it will work.
 
M

MassiveProng

Jan 1, 1970
0
Since this will obviously not work why go to all that trouble just to
destroy some leds?

You will need an mcpcb mounted to a heatsink.
Do the sums and work out the required theta J to A. Thence with the
knowledge of the led theta J to C the 'heatsinking' ability of the
mounting is revealed.
Look at the mcpcb data then calc the heatsink required.

That's funny. Looks to me like exactly what these guys did with one
of their one watt LEDs... you know Luxeon...

They even put it in a little plastic tub.

http://www.amazon.com/TerraLUX-LED-Replacement-Maglite-Flashlights/dp/B00068VAP0
 
J

John Larkin

Jan 1, 1970
0
I do not agree with the number you "selected" to use, but yes, any
polished surface will have a lower emissivity. I forgot to mention
grit blasting the PCBs to make the finish matte.

Do you really grit blast pcb's?
Bare copper with a matte finish has a higher emissivity than any
type of copper under an insulative blanket of solder mask.

I'd have to try that. I'm a bit dubious. But radiation is a minor heat
dump mechanism at the sorts of temps that pcb's usually run at.
Heat conducts through the device leads or package, to the PCB copper
and substrate, and then radiates to air. If it has an added layer of
solder mask, that radiation will be less.

Heat doesn't raidiate to air, it conducts and convects to air. Heat
radiates *through* air.

And a solder mask adds a minute amount of thermal resistance, not
enough to worry about.

John
 
E

Eeyore

Jan 1, 1970
0
John said:
Do you really grit blast pcb's?



I'd have to try that. I'm a bit dubious. But radiation is a minor heat
dump mechanism at the sorts of temps that pcb's usually run at.

I'd have thought so.

How about chemically blackening the copper ? Or a black solder resist?

Heat doesn't raidiate to air, it conducts and convects to air. Heat
radiates *through* air.

I'm sure the main mechanism for heat removal will be conduction/convection. In that
respect a few holes in the pcb will allow extra air flow.

And a solder mask adds a minute amount of thermal resistance, not
enough to worry about.

I'd have thought so myself.

Graham
 
M

MassiveProng

Jan 1, 1970
0
Do you really grit blast pcb's?
If its function is to be used as a thermal radiator.

With HV, we have even grit blasted the shiny surface of HV caps. and
diode packages. It makes the surface grab potting compound better.

Potting detachment is a main cause for multiplier stage failure in
HV supplies.

But for this thermal application, yes, bare PCB (no mask) and if the
foil is to shiny, a bit of blast makes it matte, or a bit of 320 grit
emory cloth.

Both before any population takes place.
 
M

MassiveProng

Jan 1, 1970
0
I'd have to try that. I'm a bit dubious. But radiation is a minor heat
dump mechanism at the sorts of temps that pcb's usually run at.


Most of it at this level will/would get handled by thermal
conduction cooling via the leads into surrounding substrates.

That is why I think this would work so well. It has two soaking
elements instead of one, and if they were all on the main PCB the
whole mass could warm and actually degrade the thermal abatement
desires. PCB substrates are fiberglass and epoxy, and don't move
thermal energy very fast.
 
M

MassiveProng

Jan 1, 1970
0
Heat doesn't raidiate to air, it conducts and convects to air. Heat
radiates *through* air.

Heat radiating from a hot surface conducts into the air. Using the
single term radiates to air is fine. IR does the work either way,
from within the hot medium. How do you think heat conducts to air
down at that little boundary layer? Does it shake hands with the
surface? No. IR is the thermal engine of ALL matter. We wouldn't be
here otherwise.
And a solder mask adds a minute amount of thermal resistance, not
enough to worry about.

It DOES matter when what WE are after is a maximized system for
removing heat.

Shall we discuss micro-air gaps under large IGBTs? Next thing,
you'll be telling me that doesn't matter either. There is an entire
industry that says you're wrong.
 
M

MassiveProng

Jan 1, 1970
0
I'd have thought so.

How about chemically blackening the copper ? Or a black solder resist?


With emissivity the apparent color is not as important as the
surface quality.

We painted our black body calibration sources with a matte black
paint and got near .98 plus emissivity. Average Human skin is also at
..98

Both are due to surface texture. The bare aluminum ingot has a smooth
surface, and some of the IR generated inside when it gets heated get
reflected back inside. When it has the matte finish, it has many many
more angles to radiate from the surface at. They are little tiny
scratches that add up to more actual surface.

IR and a smooth, polished , flat surface do not get along. It
reflects back inside the medium, and any that does radiate only
radiates from the surface in a perpendicular ray. The matte surface
allows the heat in the medium to radiate away from its surface at
several angles of incidence. A denser IR "flux" emanates.
 
M

MassiveProng

Jan 1, 1970
0
I'm sure the main mechanism for heat removal will be conduction/convection. In that
respect a few holes in the pcb will allow extra air flow.


In this case, if anything, s bunch of plated vias would assist in
the conduction cooling process, but not much on convection in a box
with no forced air. With forced air, it would add a little.
 
Top