Heatsink, just don't drop it on concrete - Page 2

Do you have a question? Post it now! No Registration Necessary

Translate This Thread From English to

Threaded View
Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  The flatness is what counts.  Coplanarity is what is important  
when mating the heat source to the sink.

  No, surface texture at that level has ZERO effect for or against  
heat flow.  The interface is where most is lost, and that is about  
the interface medium.  Air = no good.  So we put fillers in. sil-pad  
tabs or liquid 'paste'.
  The surface quality of molded ceramic is fine grained, and  
presents exactly ZERO difference.

  If it were actually 'rough' you would have a valid point.  But the  
profile of the texture of a molded ceramic face is far too fine to  
have any effect against thermal flow.

  IF it was ground with 36 grit. MAYBE.  But the surface quality of  
ceramic has zero effect.  Far more important is a tight, flat  
interface.  Whereas the 36 grit ROUGH surface WOULD impede thermal  
flow.  But not really a lot.  You said it "wrecks it" but it doesn't  
really. Non-coplanarity does.

  You could find out with your IR imager.  The difference is  
practically nil until the roughbess gets pretty far up there.

  We had a guy who insisted on soldering the big FETs down before  
mounting the heat sink.  If they were individual sinks sure, but a  
gang of 12 FETs across the edge of a board there to mate with a heat  
sink poses problems with his method.  We were frying FETs within  
seconds of use.  I showed him that we needed to mount the heat sink  
and FET packages onto the PCB and THEN solder the FET legs.

Re: Heatsink, just don't drop it on concrete
On Thu, 20 Jun 2019 14:38:25 +0000 (UTC),
snipped-for-privacy@decadence.org wrote:

Quoted text here. Click to load it

Both roughness and flatness matter for good heat conduction. Thermal
grease is not actually a very good heat conductor, and it needs to be
thin to minimize theta. If the contacting surfaces aren't flat and
smooth, that hurts. We machine aluminum extrusions flat for high power
stuff, because they tend to be wiggly as-extruded. Our ceramic AlN
insulators are lapped to basically optical limits. Transistor packages
tend to be very good.

Thermal silicone grease has ceramic filler particles, which may help
conduction but increases gap. They compress to below 100 microinches,
which is good. I tried some diamond filled grease, but the particles
were so big that they increased the gap, with no net benefit.



--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

    Yes, but what you fail to note is the actual degree of  
roughness.  I used to polish stainless steel plates so I know what  
surface quality is.  I also know about the interface mediums used  
between heat sources and heat sinks.

  Flatness matters, but once one is below a certain degree of  
surface smoothness, there is no difference in conduction.


Quoted text here. Click to load it

  Yes, but it comes down to mils, not microinches.  The 'matte'  
finish on a perfectly flat A1N insulator has been perfectly fine for  
decades.

Quoted text here. Click to load it

  Oh boy!

Quoted text here. Click to load it

  Yes.  Most credible companies mill mating surfaces for the heat  
source to mate to the heat sink.


Quoted text here. Click to load it

  By you guys or the maker?  I'd bet immeasurable difference between  
an off the shelf A1N insulator and an "optically lapped" version.  
It is serious overkill with a nil net gain result.

Quoted text here. Click to load it

  Not really. A copper slab plated.  That 'slab media' is extruded  
and shear cut when they are building those FETs and transistors.
It is very flat, but not in any way an "optical quality lapped  
surface".  You have taken your obsession a bit too far and no longer  
have a picture of what is happening.

Quoted text here. Click to load it

  SOME thermal grease has ceramic filler particles.  Some have  
Silver.  Some have other media.  Some have NO granular media at all.

Quoted text here. Click to load it

  As long as electrical conduction is not an issue, a silver filled  
thermal paste or epoxy is best.

  Why do I say this?  Because THAT is what the entire microchip  
industry uses for their die attachment.

  Their stuff is NOT polished to a mirror finish either.

  There is a point at which any 'returns' or gains are minimal or  
even immeasurable.  You are worrying about a molded ceramic  
insulator surface quality that is far and well below the profile  
level which causes a conduction differential.

  You have problems looking at the bigger picture stuff, and  
apparently at the micro scale as well.

  Zero gap is zero gap.  Flat, co-planar mating is what is required  
MOST.  A mirror finished surface intrfaces no better than a surface  
with a 15 micron pebble matte finish. You are well below the point  
where it matters or even shows a difference.

  Also, not that the tabs have a hole that is offset from the  
conduction surface.  That means that the clamping force on it causes  
it to want to tip a bit away from co-planar mating.  There are a lot  
of mounting clamps out there now, which clamp against the main body  
of the part. That too is important.


Re: Heatsink, just don't drop it on concrete
On Thu, 20 Jun 2019 15:35:18 +0000 (UTC),
snipped-for-privacy@decadence.org wrote:

Quoted text here. Click to load it

Silver epoxy is used for its electrical conductivity.

Quoted text here. Click to load it

One thing I do is measure thermal conductivity and then do what works.

https://www.dropbox.com/s/0pr4hq6euhjn4yr/DSC02169.JPG?raw=1

https://www.dropbox.com/s/9vt61si6taj2ivw/DSC02177.JPG?raw=1

That cuts through a lot of theorizing. And creative data sheets.

Quoted text here. Click to load it

This uses clamp mounting of the power fets onto a very flat copper
heat spreader, with no insulators.

https://www.dropbox.com/s/nmyb9pz0qpma2xa/Amp.jpg?dl=0

The entire heat sink is the amplifier output node, insulated from the
chassis. 17 kilowatts peak power for NMR gradient drive.




--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  Silver conducts heat better than copper does.  Silver filled epoxy is  
for conduction of BOTH heat and electrons.

  On CPUs at Intel, it is used specifically for its thermal conduction  
qualities.

Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  You thinking that a matte finish ceramic is going to conduct better  
with a mirrored surface is ludicrous.

  Maybe if it was pebbled like a hip joint replacement, but at the  
surface texture it has...  not one single bit of difference.

Re: Heatsink, just don't drop it on concrete
On Thu, 20 Jun 2019 16:33:46 +0000 (UTC),
snipped-for-privacy@decadence.org wrote:

Quoted text here. Click to load it

The flatter, the better, whatever scale you measure it on.


--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: Heatsink, just don't drop it on concrete
On Thursday, June 20, 2019 at 12:38:22 PM UTC-4, John Larkin wrote:
Quoted text here. Click to load it

Within the law of diminishing returns.  "Better" is not always better.  

--  

  Rick C.

  ---- Get 1,000 miles of free Supercharging
We've slightly trimmed the long signature. Click to see the full one.
Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  He is experiencing science observer overkill.

  I know for a fact that the difference is so small as to not even  
be measurable.  The term nil works here.

  One can take ten of the FETs and do a 'thumbprint' job with them  
on a known flat surface (an inspection block), and you will find  
that even they are not all as flat as you might like to think they  
are.  We even "honed" some to prove it.  Aside from the copper  
oxidation problem with the now exposed bare copper, they actually  
conduct heat away better.  But the difference is so slight as to not  
matter at all.

Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  You are incorrect.  *I* said flatness is important.  YOU ranted on  
about the surface condition.

  Flat is better.  Polished?  Does not matter.

  The flatness matters.  The roughness will only matter AFTER it  
rises above a certain level on the profilometer.

  You jacking off at the mouth about a flat face with a matte finish  
comprised of 15 micron surface undulations causing less cooling  
efficiency is a joke.  Period.

  The flatness is ALL that matters once the surface quality is past  
a certain point.  That point is way up in the "36 grit" scratch  
profile range, not anywhere near the surface profile of a matte  
finish ceramic face with 15 micron bumps.

Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

  We used those too.  They rely on the spring characteristics of the  
clamp material.  I think they could have made them slightly thicker,  
and produce about 15 or 20 lbs more force aginst the tab.

  Nice design where the plate offset matches the leg heights.  Easy to  
build and then solder.

  Ours was perpendicular to the PCB and was only 2kW but that was  
continuous.  We ended up using a sink with no fins.  It was a  
rectangular tube and allowed better air management and worked fine for  
the task.

Re: Heatsink, just don't drop it on concrete
wrote:

Quoted text here. Click to load it

An AlN insulator over a water-cooled aluminum or copper block is
pretty good for cooling vs capacitance. Flowing water directly onto
the fet tab would be a lot better.


--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: Heatsink, just don't drop it on concrete
John Larkin wrote...
Quoted text here. Click to load it

 Yep, I'd have much less struggle with my thermal  
 problems, if my whole circuit was in an oil bath.


--  
 Thanks,
    - Win

Re: Heatsink, just don't drop it on concrete
wrote:

Quoted text here. Click to load it

Oil is nasty. If you get water on your jeans, it evaporates.

I was thinking about a cavity and some water channels machined in an
aluminum block. The Cree would push into a cavity with an o-ring seal.
Water would flow on the bottom of the fet. Cooling would be awesome
and capacitance minimal.

My average drain voltage is low (I'm making narrow positive pulses) so
electrolysis might be managable. One could bond a very thin AlN slab
to the fet tab too.




--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Re: Heatsink, just don't drop it on concrete
On 2019-06-18 07:45, John Larkin wrote:
Quoted text here. Click to load it

Not a problem if people dress accordingly :-)

https://www.tokyohive.com/upload/2011/02/00delete11143.jpg


Quoted text here. Click to load it

Really hoppy India Pale Ale might also work.

--  
Regards, Joerg

http://www.analogconsultants.com/

Re: Heatsink, just don't drop it on concrete
On Tuesday, June 18, 2019 at 7:45:38 AM UTC-7, John Larkin wrote:

Quoted text here. Click to load it

The compatible garments are wool; oil just diffuses in, doesn't
leave a spot.   Wool shirt is great for machine shop (but roll up the
sleeves before using power tools, of course).

Re: Heatsink, just don't drop it on concrete
On Monday, June 17, 2019 at 9:40:44 PM UTC-4, Winfield Hill wrote:
Quoted text here. Click to load it

Not that it really matters, but do you have a price point for the  
X-chapters?  (Seems like less sales volume than AoE3, so more $.)  
And will it be reading for beach reading this summer?  
(before July 4th say.)  

George H.  
Quoted text here. Click to load it


Re: Heatsink, just don't drop it on concrete
On 2019-06-17 18:40, Winfield Hill wrote:
Quoted text here. Click to load it

That almost cries for oil cooling. Of course, when I did that with a  
larger setup the container began to weep. I must have botched one of the  
solder seams.

--  
Regards, Joerg

http://www.analogconsultants.com/

Re: Heatsink, just don't drop it on concrete

Quoted text here. Click to load it

Fluorinert is dielectric and a lot less messy. (and a lot more  
expensive)  Not better than water thermally, but far better than  
water around electrical circuits!

Re: Heatsink, just don't drop it on concrete
wrote:

Quoted text here. Click to load it

Every once in a whie one of my people will suggest oil cooling or
potting. A few hours alone locked in a dark closet seems to help them.


--  

John Larkin         Highland Technology, Inc

lunatic fringe electronics  


Site Timeline