Thermal Grease Application Method

HydroDC

HydroDC

Well-Known Member
Evidently, there are many differing ways to apply thermal grease when attaching an LED to a heatsink. Dots, x's, lines, full coatings (thin or otherwise), push down, let the screws in the holder do the pushing, etc. The info I've found (more often in relation to CPUs) lacks consensus as to which approaches are best. Finally, it's not clear that attaching a CPU to a heatsink has relevance to attaching a COB.

Perhaps it's not that important that a best method be found if virtually any way is nearly as good as any other. Maybe completeness of coverage, thinness of coating and micro air bubbles are of minor concern. But in case it does make a real difference in heat transfer and thus in COB performance and longevity, identifying the better methods would matter.

I'm finishing up a DIY using Ideal holders and CXBs and would welcome help in finding out about those better methods or if they really matter.
 
SupraSPL

SupraSPL

Well-Known Member
I recall @stardustsailor had a situation where a large air bubble in the center of a CXA3070 caused damage to the center of the COB. So the paste can make a significant difference. I prefer to err on the side of using too much paste rather than using too little because you can squeeze the excess out by pressing and the layer will be very thing either way.

I think it was @Doer that suggested I spread the paste on the COB rather than using dabs. I tried it and it does work well and reduces how much paste I am waste. I have also noticed that Prolimatech PK3 can get thick if the temp is in the low 60s Fahrenheit, so making sure PK3 is not too cold can help getting a thin even layer.

Some diyers are using stencils I believe. Reduces the waste even further, maybe less messy but I do not expect any difference in performance considering that polishing the heatsink surface made no difference and stock heatsink pastes perform just as well as PK3 for our application.

And finally, I have checked to see if increasing pressure on the COB affects the thermal transfer and in every test the result was negative. So no need for strong torque setting on the screws etc.
 
coolbreez1

coolbreez1

Well-Known Member
I put the paste onto the COB its self, full coating of thin lines, even it out as much as possible, then place it onto the heat sink, push down and jiggle it back and forth ever so slightly to make sure it spreads out evenly. I think the biggest issue is full coverage of the die, no bubbles or air gaps, not so much thickness, assuming no extremes of way to much or way to little.
 
bicit

bicit

Well-Known Member
I followed the old recommendation for pentium 4 CPU's... a single glob of paste about the size of a pea, and then 'massaged' it into place and lock it down...

A stencil sounds a lot more precise.
 
HydroDC

HydroDC

Well-Known Member
salmonetin said:
...a example vid..

saludos
Click to expand...
Actually, I watched this video a few days ago and it's illustrative of why I asked about the "right" way to do this. Ignoring the screening method which shows a square of grease that looks to my eye to be more like icing a cake than applying thermal grease, the video shows the guy snaking a globby line on the COB and later shows 2 COBs, one with too much and one with not enough. Nothing at all about spreading the grease or how much you should end up with so there isn't too much or too little or really anything I found very useful. I don't think "more is more" is at all the right answer.

Thanks to all who replied. I'm going with the view expressed by many that a thin even coat on the back of the COB followed by a wiggle on the heatsink is the way to do it - and I'll go with "less is more" as to amount.
 
salmonetin

salmonetin

Well-Known Member
...from my point of view... ...i like the stencil method... same technnic for solder paste on reflow soldering...

...but for cobs led... the paste its on the component side...

...but its only my opinion...

saludos
 
Last edited:
heckler73

heckler73

Well-Known Member
I have also burned a multichip because of improperly applying compound, thinking it would squeeze out.
After two weeks, the chip burned. Upon taking it off, I saw a large streak where the air pocket was.
I was also lazy and applied it too thick.

So I fell back on old knowledge of better methods and whipped out one of these for the replacement application:


Haven't had a problem since...keep it thin...like paper.
 
Doer

Doer

Well-Known Member
bicit said:
I followed the old recommendation for pentium 4 CPU's... a single glob of paste about the size of a pea, and then 'massaged' it into place and lock it down...

A stencil sounds a lot more precise.
Click to expand...
Well, this is all about efficiency. And let @alesh tell us about total impedance. Thickness does matter. CPUs are not in the same league as this, IMO. if we can't get it squished to the thickness masking tape then we are throwing away efficiency with too much thermal impedance.

On a cpu I could care less. :) Don't want to smoke the puppy with air gaps is all I am concerned about.
 
stardustsailor

stardustsailor

Well-Known Member
First step : Choose (with caution ) which thermal paste should you use ....

For example,by personal experience :
P5202814.JPG

Gelid GC extreme : Voted as the BEST thermal paste by PC overclockers all over the world .
Reality : Crap ... Only for a single reason ...
This paste is almost impossible to be applied at a paper-thin layer .
It's like trying to apply a chewing gum .

ProlimaTech PK-3 :
Same as Gelid GC-extreme ,altought a tad easier to work with ..

Arctic Silver 5 : Oldest TIM from the three ....
But .....Piece of cake to apply into an ultra-thin layer ...
My own favourite ...

My advice :
Use a thermal paste that can be applied easily .
Forget about their "thermal conductivity " figure...
It really means nothing at all ,if you can't have an ultra thin layer of the TIM used ....
 
bicit

bicit

Well-Known Member
Doer said:
Well, this is all about efficiency. And let @alesh tell us about total impedance. Thickness does matter. CPUs are not in the same league as this, IMO. if we can't get it squished to the thickness masking tape then we are throwing away efficiency with too much thermal impedance.

On a cpu I could care less. :) Don't want to smoke the puppy with air gaps is all I am concerned about.
Click to expand...
Actually, practices such as sanding the heatsink and CPU as well as the quality of thermal paste do make rather large differences in operating temperature. Especially when overclocking.

Cobs, not so much.

Though I tend to agree with the 'less is more' approach.
 
stardustsailor

stardustsailor

Well-Known Member
Stardustsailor's Thermal Resistance Case to Heat sink notes :

- Flat & smooth surface of heat sink
- As thin as possible thin layer of TIM .Preferably the " thermal paste " type .Even ordinary -cheapo pastes will do ,
as long the layer is kept as thin as possible ...
-High Contact pressure between COB and heatsink .
( aka " sandwiching out the air bubbles trapped within and squashing the metal abnormalities of the surfaces in contact ..)

From AN30 of Bridgelux :

" Thermal Interface Material (TIM)

Although the substrate of the Vero LED module and the machined heat sink appears to be smooth, surface imperfections exist and when the LED module is mounted to the surface of a heat sink these small imperfections trap air and restrict heat from moving efficiently into the heat sink. The use of thermal interface materials, such as thermal greases (pastes), adhesives, phase change materials (PCMs) and conductive pads/films are recommended to improve the contact and heat flow from the substrate to the heat sink.
As shown in figure 8 above, the TIM improves the heat transfer by filling in the air voids between the array and the heat sink. There are 3 resistances in the TIM, the resistance through the TIM material and the contact resistance (remaining air layers) between the TIM surface and array and between the TIM surface and heat sink. The contact resistances dominate when the TIM does not fill in the voids such as with at thermal pad. A thermal grease or phase change material will generally fill more of the voids and have a thin bond line, thus, the contact resistance is less and the material resistance will dominate. For other materials such as pads, graphite materials, and metal TIMs the contact resistance will dominate and it will be necessary to test the TIM in order to determine the thermal resistance through the material to be sure the thermal requirements are met.
Although the contact resistance cannot be easily calculated, the thermal resistance of the TIM material can be estimated by calculating the conduction resistance through the TIM material. Equation 8, is only an estimate and should be used for materials where the contact resistance is known to be low such as that of grease or a phase change material.

Equation 8 shows the calculation for the thermal resistance of the TIM material:
RTIM = t / (k* Ac)
Equation 8: Thermal resistance of an interface material
Where:
t is the thermal interface material bond line thickness (BLT) (m)
k is the thermal conductivity (W/m-K)
Ac is the heat source contact area (m2) "



*(((( m / m^2 ....=>
from 0,000xx to 0,00000x (<<=BLT ) / { from 1 to about 12 (<< = K) * 0,0xx ^2
(<<= constant :Case pad surface area ) } .... ;-)
What do you choose as priority for lowering thermal resistance of TIM ?
Low BLT or High K ? :wink: ))))



" TIM suppliers will often state the thermal performance as a function of area with units such as C–cm2/W. When stated in this form, calculate the thermal resistance by dividing by the heat source contact area Ac.
RTIM = RTIM per Area / Ac
Equation 9: Calculate thermal resistance of an interface material when resistance is a function of area
Where:
RTIM per Area is the thermal resistance as a function of area (C–cm2/W)
Ac is the heat source contact area (can be assumed to be the Vero Module substrate area) (cm2) "

Types of TIMs ,blah blah ,blah ....
And further down,we can read :

" TIM Selection
When selecting a thermal interface material many factors must be considered. These include the TIM thermal conductivity, contact resistance, operating temperature range, cost, manufacturability, electrical conduction, and the ability to control the bond line thickness (BLT).
TIM thicknesses for LED arrays typically range from 0.050 mm to 0.500 mm, depending on the type of TIM selected and material properties of the TIM. When selecting a TIM ensure that the thickness of the material is sufficient to fill gaps between the substrate and the heat sink while at the same time minimizing the final bond line thickness (BLT). Thermal conductivities and contact resistances of interface materials vary from product to product. Also, while the bulk thermal conductivity of the TIM material is important, in many cases the contact resistance (air trapped in the interfaces) is the main contributor to the overall thermal resistance, so it is best to test the TIM in the application to be sure it is performing as expected. "

For More : Bridgelux Application Note 30

Cheers..
 
Last edited:
Doer

Doer

Well-Known Member
Also, while the bulk thermal conductivity of the TIM material is important, in many cases the contact resistance (air trapped in the interfaces) is the main contributor to the overall thermal resistance, so it is best to test the TIM in the application to be sure it is performing as expected. "
-----------------
Who is still thinking big air bubbles here? :)

If you butter just the COB or just the sink, that's not really the point.
My point is a wet to wet surface. Sanded aluminum, (not polished) wiil keep air trapped INSIDE the scratches, if we are not careful.

And the surface of the ceramic, is the same thing but much worse. These are pits and not filled at all, by a quick spread and squish.

Easy to see this. Take your Mark 1, single edge razor and try to scrap off ALL the TIM from the COB and the sanded sink. You can't. You will see where the material is filling the voids.

Scraping tight with razor blade is to me, a required step 1. The material you leave (can't scrape off) fills millions of tiny pits and/or scratches. It is spread across the surfaces and amounts to a layer of high thermal resistance due to microscopic air gaps.

So, 2 steps for both COB and sink.
1- scrape tight
2 - add TIM again (stencil on sink) and put a wet surface against a wet one.

Only then can we know we have done all we can to reduce air gaps.
And it does matter. You get practically nothing from a Peliter Junction without this, IME. It might not be as critical in COBs, obviously.
But, it is an easy extra step for Quality Assurance. :)
 
You must log in or register to reply here.
Top