Connect with us

Using a PCB as a heatsink

Discussion in 'Electronic Design' started by Mike Noone, Feb 14, 2007.

Scroll to continue with content
  1. Eeyore

    Eeyore Guest

    Really ? I've heard it claimed thar FR4 counducts not a lot worse than
    aluminium.

    Graham
     
  2. Eeyore

    Eeyore Guest

    I certainly agree about surface finish.

    I spent some time too on 'turbulent air flow' heat sinks. They work far better
    than ones where the air flow is laminar.

    I've *removed up to 30% of the material in force cooled heat sinks and got a
    result that's 20% *better* !

    Graham
     
  3. Eeyore

    Eeyore Guest

    That too. I like to have some of them around surface mounted regulators for example.

    With some holes you can get convection 'through' the pcb. I've done this where I've had
    pcb mounted vertical heatsinks for the likes of TO-220 devices.

    Graham
     
  4. John  Larkin

    John Larkin Guest

    It's like resistors in series. A couple of mils of solder mask in
    series with inches of air. And the mask material is a much better heat
    conductor than air. So the mask adds a small fraction of a percent to
    overall theta, not enough to matter. It probably improves radiation a
    bit, more than it impedes conduction.

    John
     
  5. John  Larkin

    John Larkin Guest

    Are you saying that infrared photons are the mechanism by which heat
    is transferred from a pcb surface to the surrounding air? Or that IR
    is the way heat is transferred "within the hot medium"?
    Thermal design by hunch and opinion is like any other design by hunch
    and opinion: probably wrong.
    I never mentioned micro air-gaps under large IGBT's, so you would be
    in error to say I ever did, much less that I got it wrong. Hell, I
    don't think I've ever seen an IGBT in person.

    John
     
  6. Terry Given

    Terry Given Guest

    I've never looked at solder mask in this respect, but I did the calcs
    once on powder-coating. Its a surprisingly good blanket (after all, its
    plastic). The conclusion was: dont powder coat heatsinks. So we anodised
    it instead - "it" being a flat Al plate

    disclaimer: I had no idea they were going to be powder coated until I
    saw one; some marketing guy decided he wanted everything black.....

    Cheers
    Terry
     
  7. RHRRC

    RHRRC Guest


    Yes - highly creditable source you refer to.
    According to their blurb they are 25 times brighter than a Krypton
    bulb.

    Facts: from luxeon datasheet typical output for 1W emitter at 350mA
    (typical Vf 3.42V) for a white typical 45 lumen, minimum 30 lumen: for
    a warm white typical 20 lumen, minimun 14 lumen. all at Tj=25degC.
    Derate down to 68% of these figures at Tj=120 degC. Linear
    interpolation between these figures.

    The bulb they are 25 times brighter is (say typically) a KPR102
    flashlamp bulb -700mA, 2.4V, 15 lumen.

    According to their blurb the batteries last 10 to 15 times longer

    According to their blurb the led last 100,000hrs.
    - better let Lumileds know since they claim much less.

    In practice they run the led (very variably) in the 150mW region *not*
    the 1W or more the poster requires.
     
  8. MassiveProng

    MassiveProng Guest

    You're kidding right?

    Al is one of the best thermal conductors. That's why we use it for
    heat sink media.

    Don't see many FR4 or G10 heat sink fins out there.

    No. PCB substrates are specifically designed to be rigid,
    non-conducting (electrically) mediums. As a rule, most electrical
    insulators are also fairly poor thermal conductors. Of course, silica
    is one exception.

    The fiberglass component moves heat very slowly. The epoxy binders
    make it a bit better. The material is also meant to withstand high
    temperatures, like ovens and frying transistors. The cladding
    adhesive releases before the board substrate burns.
     
  9. MassiveProng

    MassiveProng Guest

    Yes, which is why both added vias (large with small annular rings),
    as well as staggering the placement heights would make this design
    work quite well.
    Yep. even slow moving fans work if the design is done right. All
    depends on power levels of course. It gets harder to say these things
    as the heat actually really does start being exhibited by a part for
    any given case size. Air or not. But for low power dissipation
    levels, it is very easy to abate small thermal generations.
     
  10. MassiveProng

    MassiveProng Guest

    IR is the engine for all heat. When you pass current through a wire
    to make a soldering iron's tip hot, it is IR energy that heats up
    those mediums, and it is IR energy that "conducts" into the air
    streaming up off that hot tip.
    There was no hunch, but it seems we do have a chump in the thread
    now, chump.

    It is MORE than a fraction of a percent, and it is YOU that issued
    the hunch. Accusing me of it makes YOU the chump.
    It is the SAME kind of thing, mr. resistors in series.
    OK, dipshit... Think HUGE HOT FET. The interface between the FET
    and the sink is crucial.

    Having the mask on a PCB meant for abating heat is NOT optimal, and
    represents MORE than a "fraction of a percent" of degradation.

    Hunch that, hunch boy.
     
  11. MassiveProng

    MassiveProng Guest


    If you make sure to have it grit blasted before the anodize, you'll
    maximize its thermal efficiency. Matte black IR paint (not just any
    shit) will make it radiate AND thereby convect to the surrounding air
    even better still, further maximizing the capability of the sink.

    The finish extrusion typically places on the Aluminum is usually too
    smooth for my tastes, but I used to make fully traceable Black Body
    calibration sources for NIST, so what do I know?
     
  12. Fred Bartoli

    Fred Bartoli Guest

    MassiveProng a écrit :

    So you should have no pb giving us some emissivity figures and back of
    the envelop convection/radiation ratio at usual temperatures. Right?

    I tend to prefer figures to "taste".
     
  13. MassiveProng

    MassiveProng Guest

    Our tests were such that a matte finish surface of any media
    radiated IR better than a polished or glazed surface will. It is also
    in just about any CRC physics book around. Surface emissivity
    listings tend to refer to surface quality and many times give two
    figures. Polished, and a rougher surface quality figure.
    I prefer common sense to your need for hard documents or formulae.
    Pretty basic physics.

    It all depends on how the face being observed is finished.

    We cut grooves in ours so nothing "looking at" the black body source
    was ever actually pointed at a perpendicular surface, and a device
    "looking at" the surface could also be at other than a perpendicular
    angle and still get the same reading.

    We had others where the target was heated air that were essentially
    a coil of nichrome heating wire wrapped around a basketball
    (literally) then doped up with fiberglass/silica "paper mache" <sp>.
    After that dries, we split the halves, and remove the ball, and apply
    more silica medium to the inside of the sphere., then the two halve
    get assembled back together with a 2" silica tube about ten inches
    long on it. That assembly gets baked. We suspend that in am
    insulated cabinet, and provide NIST with a source that can do 3000 C
    at an emissivity of 0.995, and hold/keep temperature within a few
    degrees of setpoint.
    That's pretty awesome!

    On the benchtop sources, getting a uniform temperature across an 6"
    plate sized surface with a single heating element source is not an
    easy task. The ingot is several inches thick, and by the time it (the
    heat source) soaks out to the front surface, all but the last half
    inch at the outer edges of the ingot are evenly radiating. Especially
    since it gets recessed in a case so that no local air currents play
    upon it.
     
  14. John Larkin

    John Larkin Guest

    That's demonstrably silly, and useless besides. Air is almost
    perfectly transparent at thermal wavelengths (easy to demonstrate) so
    radiation is not the mechanism by which a hot object transfers heat to
    the surrounding air... any IR leaving a hot surface zips right through
    the air without heating it.

    Example: a square cm of blackbody-quality pcb surface, 25 deg C above
    room-temp radiative ambient, radiates only about 17 milliwatts of
    thermal IR. That's tiny compared to convection.

    John
     
  15. Eeyore

    Eeyore Guest

    Marketing !!!!!!!!!!!

    Aaaaarrrgggghhhhhhhhh !

    Graham
     
  16. MassiveProng

    MassiveProng Guest

    You need to give the actual package you are using as the coupling
    characteristics play into it.
     
  17. jasen

    jasen Guest

    I had the impression that alumina and AlN were better,
    silica is one ingredient in the fibres of FR4
     
  18. Rich Grise

    Rich Grise Guest


    Do you mind if I ask from whom you heard this? It doesn't sound very
    realistic to me.

    Thanks,
    Rich
     
  19. MassiveProng

    MassiveProng Guest

    Take a hike, dipshit.
     
  20. Eeyore

    Eeyore Guest

    Some dim and distant application note.

    Graham
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-