Whats the best thermal compound?

didn’t know it is toxic!

but certainly found out it is a bugger to get off of just about anything.

I use a credit card (or similar piece of flat edge plastic) to scrape off excess, and acetone to clean up.

hah! But if we weren’t OCD, we’d just buy a Craftsman LED torch from Walmart and be happy with it, wouldn’t we :slight_smile:

I’d wager that most of the single emitter lights on here would benefit greatly from better thermal management, given how much heat they’re producing in such a small place. You only have to look at the output drop from switch on to +15min to see how ineffective they are at dissipating heat. I’m not saying cleaning up just one thermal interface (between star and pill) is going to magically solve that problem, but it’ll help.

All this has really got me thinking about the Jacob a60 and ke-5 I just nodded, on both I sanded then wire wooled the anodizing off and polished the stars, now perhaps, I should go further next time.

I’m also thinking I need to get an hd2010 to play with…

you sanded off the anodizing from the outside?!

Check their MSDS (material safety data sheet)… it’s rather benign. Don’t spread it on crackers, though.

Pretty much all thermal compounds perform the same! The main difference between a good and poor compound is the long term stability of the carrier fluid used to make up the paste.

Lots of interesting info in this thread, for my application i want to replace the star in this photo

with this in xm-l2 when it becomes available

So i want the best thermal transfer compound made by man, economically achievable

Sorry, I’m thinking everyone knows what I’m on about, on both the Jacobs and the ke-5, there is no pill, the light engine is part of the host, as such, its anodized just like the rest of the host, so I sanded and polished under where the star will sit, so the star was pasted direct to the host alu, rather than there being an anodized layer between the base of the star and the aluminium of the light engine, now I see, I should have also polished the base of the star flat too, something I’ve wondered about.

if you’re not willing to drill a couple holes and screw it down then you’ll need an adhesive one

slightly cheaper on ebay or walmart, but i like newegg

Lapped might be superior when done professional. With home tools? I'm unsure..

How could soldering have any negative effect? You already have a layer of solder between LED and PCB, but the layer between PCB and heatsink would have 10times the area -> better heat conduction. If you have a 1" pipe at the source it doesnt matter how many 1" pipes you add in the path, even if some are 3"..

Actually aluminium PCBs would be great if you could solder the LED directly on the aluminium, without the dielectric layer and all that stuff inbetween. Thats the main advantage of copper pcbs. I think even brass PCBs would be better if its a good alloy.

And I dont think the gold layer has any influence on thermal conductivity at all. Its there so the copper doesnt oxidate. Also, even those copper PCBs from LED-tech are uneven because they are punched out..

@ChimpOnABike: CPU heatsinks arent lapped because you need both parts for lapping. They are sometimes sanded pretty good, but often far from a mirror-like finish. Its an easy improvement actually.

Something with powdered diamond in it. Diamond powder is actually not too expensive. Check Ebay:
Shop by Category | eBay%28dust%2Cpowder%29&_sacat=3213&_odkw=diamond%28dust%2Cpowder%29+mesh&_sac=1&_osacat=3213&_from=R40

If you’re looking for diamond based paste: Amazon.com

ah, yes, I dont’ own either.

probably a good idea to remove.

I’ll be using some mother’s polish as a last step next time I do a pcb…

well, I have lapped machined parts professionally, but I don’t think someone else needs to have been paid to do a good job. it does take some patience.

there isn’t any fluid mechanics involved here, so fluid in pipes isn’t a good analogy.
i’m not sure how to explain it differently. Maybe we’ll have to agree to disagree.
the thermal conductivity of solder is lower than the metals involved, so putting it in between them reduces thermal conductivity if they have very good contact before soldering - what happens before or after that can change the temperature gradient in operation, but not the properties of the materials at the junction.
I guess we’ll have to agree disagree.

you can solder aluminum, but it requires a different solder, or at least a different flux, and it is a PITA

any material will have ‘some’ influence but it’d be near impossible to measure
anyway, I didn’t mean to imply that’s why gold shouldn’t be removed…I meant you don’t want the copper to tarnish…which would surely reduce contact area as well as thermal conductivity…

anyway, let me know when you find copper PCBs, I’d be happy to test this scenario.

I meant with the proper tools, not necessarily paying someone to do that..

Its the easiest I could think of.. I'm not good at calculating, but as there already is solder involved right under the LED, it cant really get worse. Maybe someone who knows how to calculate that stuff can chime in..

Metal to metal is of course superior, but if someone doesnt have the tools or patience, soldering is as good as he can get. Agreed? :D

I know, I did it before.. but as you said, not easy and therefore (I guess) rather expensive to produce. So they dont..

Well, gold is worse than copper, so.. :P

And yes, its there so the copper wont tarnish and its easier to solder to. But if you plan on using the copper stars from LED-tech, you better sand them, because they arent much better (regarding how even they are) than all those aluminium stars.

You guys gave me an idea on how to lap the insides of the pill very flat

I currently lap my stars on a flat piece of glass using wet/dry sandpaper progresively up to 2000 grit.

For the pill I am thinking on using a steel bolt like this

And lap the bolt’s head flat on a glass using wet / dry sandpaper up to 2000 grit and then glue wet / dry sandpaper to the bolt’s head and lap the inside of the pill up to 2000 grit using a drill, IMHO it should produce really good results, what do you guys think?

I’m familiar with the calculations, although admittedly not as familiar as when I passed thermodynamics w/ an A grade 16 years ago

The problem with calculating this is, interfacial thermal resistance is empirical.

I like this stuff, so I spent a couple hours last night looking for test data. ( No mystery why I’m single :stuck_out_tongue: )

I couldn’t find a study of ‘copper to copper’ vs ‘copper to solder’ interfaces. The closest I could find, in different studies, was that a milled copper to copper interface (1.6um roughness…not lapped), with no thermal compound, is has an order of magnitude lower thermal conductance than a copper-solder-copper sandwich where the solder is <30um thick and has above average (for solder) thermal conductivity. Of note, a copper-copper interface that is ground (3.6um) is 3x worse than milled…

So…I find that data interesting, and pointing in the direction of solder being superior…but ultimately inconclusive…

If you’re good at soldering :stuck_out_tongue:

I definitely think soldering is easier - although it doesn’t preclude lapping…in fact the copper solder sandwich study used lapped copper discs….

I picked up a good tip from that study. calculate the amount required to cover the contact area with 10um layer, use a little excess, press the parts together while still hot to squeeze out excess.

anyway, maybe some day we’ll get to test this. unfortunately, the studies I found use a laser for heating and very accurate measuring equipment…which to me implies, even if we get the chance to test, we’d be hard pressed to tell the difference w/ ‘amateur’ test equipment :bigsmile:

Lets recap, I was hoping for some consensus or side by side testing, but so far it seems to be a contest between

http://www.frys.com/product/6274350
http://www.intl-outdoor.com/shinetsu-x237783d-thermal-grease-5-pcs-p-480.html
same product, different numbers

http://www.wakefield.com/Catalog/tabid/92/CategoryID/110/ProductID/362/Default.aspx

http://vendor.parker.com/Groups/Seal/Divisions/Chomerics/Chomerics%20Product%20Library.nsf/24eb4985905ece34852569580074557a/c72bd4ba4470f5ff852570ce006d53a0/$FILE/Thermal%20Grease%20T650-T660.pdf

and

http://www.amazon.com/Innovation-Cooling-Diamond-Thermal-Compound/dp/B0042IEVD8

finally, according to this don’t use cheese
Thermal Paste Shootout!!

Also a suggestion to use as little compound as possible just to fill in air gaps (if the compound is more conductive then the metals then this advice may be mute?)


Can somebody wade through these and figure out the best 3?

sorry for distracting your thread 0:)

yes as far as use as little compound as possible, the compound will not be superior to metal.

if you’re not going to screw down the PCB, then you will need adhesive compound.

the technique (or at least my technique) is different for adhesive compound.

Since I flatten/lap the surface, I apply grease-like compound to both surfaces, rub it in to eliminate air, then scrape it off with something flat.

For adhesive, I apply to both surfaces, rub it in, then clamp the two parts together, so that excess squeezes out, and pressure is maintained while the adhesive sets.

i’ll look at the links then reply more :beer:

the last two links did work for me

the first two links are not adhesive.

the only adhesive thermal compounds I’m familiar with are fujik and arctic silver adhesive.

Start at the top of the chart you linked, search the product name (ie the products that don’t say they’re a grease) and when you find one that’s adhesive, and above arctic silver adhesive in the chart…you’ve got your answer :slight_smile: