Connect with us

For the ultimate heatsinking...

Discussion in 'Electronic Design' started by Tim Williams, Aug 28, 2009.

Scroll to continue with content
  1. Tim Williams

    Tim Williams Guest

    Would you consider soldering TO-220s directly to a heat spreader?

    I wonder what happens to the die after all that heating and warping...

    I'm considering using "powIRtab" devices in a project. 150A diodes are
    nice, but they're claiming 150W in a TO-220 package, which makes me uneasy.
    And how are you supposed to grease them? Something conductive
    (silver-filled) I suppose...

    Tim
     
  2. Jim Yanik

    Jim Yanik Guest

    I've seen TO-220's soldered to PCBs.Commercial products.
    Arctic Silver.
     
  3. Tim Williams

    Tim Williams Guest

    Indeed. And if it's not quite flat on that heatsink, thermal resistance
    goes way up. So you put heatsink goo inbetween, maybe with a pad of some
    sort (not sil-pads, but maybe mica for this). Now you depend on the screw
    for electrical connection, and thermal resistance is higher than polished,
    greased surfaces should be.

    If you haven't looked at what a powIRtab is, this might be confusing.
    I'm not concerned about moving the heat. Ten of these things, along with
    ten IGBTs, will be mounted on water cooled heat spreaders. I'm concerned
    with removing the heat from the die as efficiently as possible

    Tim
     
  4. Tim Williams

    Tim Williams Guest

    Yeah, but these are neither package or current limited. I just wonder about
    over 100W from that size device, and how best to deal with it. Since I need
    a mechanical and electrical connection to the tab, I'm thinking soldering is
    a good idea.
    Well, I'm already intent on using 7 in parallel to handle the reaction of a
    big stinking buck converter: 120V supply, 1kA max. output (it's a constant
    current feedback loop). Figure 1kA and 40V as a worst-case scenario. If
    that's 1kA at up to 66% duty cycle for the diodes, it's 667A / 7 per diode,
    or a bit under 100A average, and about as many watts per diode (Vf ~= 1V).
    They show up as "cheap" at Mouser. About a quarter the price of a
    comparable module-style device (e.g. HFA135NH40).

    Seriously, what idiots make, and buy, those things? Parallel is way
    cheaper. I haven't seen, say, a 600V 100A dual IGBT module that's less than
    four times the price of 4 x TO-247 IGBTs, either. Is all the silicone goo
    they fill them with really worth that much?

    But anyway, if they'll dissipate the power, that's good enough for me.

    Tim
     
  5. Tim Williams

    Tim Williams Guest

    Even better, the powIRtab thing has a *solid* leadframe. So, like, you
    could bend the lead end down in a zig-zag so it's level with the backside
    and tack it to a board. It would look like a D2PAK with a whole ribbon lead
    (and extra tab..).
    I don't think they make anything like that? Bismuth solder (pricey!) melts
    at some really low temperature, like 120C. Ordinary 60-40 isn't too bad.
    Just don't use lead free shit, eh?

    Tim
     

  6. Silver filled chip die epoxy. No, not a suggestion for die attachment,
    a suggestion for attaching each device you place on that spreader.

    If isolation is an issue, you need a thin, electrically insulative film
    or "layer". Hopefully one that has good thermal characteristics as well.

    The epoxy takes up all the gaps, just like the old cream (paste) used
    to do.

    Still, that is what the industry uses for die-to-carrier and
    die-to-heatsink attachment, so it has a lot of research behind it.

    Have to bake at 180° F for four hours to cure. Very good numbers after
    that.
     
  7. Not the answer you want, but is there any way to use a few more devices
    and run them at reduced ratings? If the cost of the individual devices
    is not high, this might be the most pragmatic answer - and it might give
    you a good reputation for bomb-proof designs.

    (I presume that proving you can use devices right on their limits is not
    the real object of the project)
     
  8. Jasen Betts

    Jasen Betts Guest

    the surface mount version of TO220 seems to handle it OK.
     
  9. Jim Yanik

    Jim Yanik Guest

    to which Arctic product are you referring?
    the epxoy adhesives or the thermal compound(s)?

    And why?

    I note neither is meant to be electrically conductive.(I was referring to
    the "grease" comment.)
     
  10. DarkMatter

    DarkMatter Guest


    The gak comment is retarded.

    Both the silver filled thermal compound, used widely by the PC builder
    folks (AND SUCCESSFULLY), and the silver filled epoxy, used universally
    by the chip fab industry, are the best on the market. One does not see
    "copper filled" pastes. So let the science boys come up with it and then
    USE it. If you are a skeptic, you are likely too stupid to be working in
    this industry, because you are a STUPID skeptic, whereas a smart engineer
    is an intelligent skeptic.
     
  11. Nemo

    Nemo Guest

  12. Hammy

    Hammy Guest

    Yes I've been noticing amperage ratings in the 100's when searching
    Newark,Digikey.

    Here is there testing procedure from an-1140

    "IR defines what can be called the "ultimate current" for power
    packages on discrete products. This ultimate current represents the
    largest current any given package can withstand under the most
    FORGIVEING of setups for heat management.

    The BENCH SETUP used in measuring the ultimate current at
    International Rectifier is full immersion of parts in a
    nucleated-boiling inert fluid."

    The approach adopted here is more than adequate to SHOWCASE the
    ability of new Silicon platforms. The key word in this approach is
    SHOWCASE.

    To actually utilize our parts in an application that targeted
    currents at the level of the ultimate current would likely be costly
    and impractical." Gee no shit!

    Full pdf - Continuous dc Current Ratings of International Rectifier's
    Large Semiconductor Packages

    http://www.irf.com/technical-info/appnotes/an-1140.pdf
     

  13. IDIOT! The heat gets from the generation point to the pipe *through*
    the SPREADER! A "pipe" only has a single, thin line of tangency
    otherwise.
     
  14. legg

    legg Guest

    Due to it's single offset mounting screw, TO220 is a notoriously poor
    package, if a predictable mechanical or electrical interface to the
    entire mounting surface area is required. Clamps are usually required
    to distribute mounting forces equally across the mounting surface.

    For direct soldering, you should follow the recommended reflow
    soldering profile provided by the mfr. If a large heatsink is
    involved, maintaining this profile will require extreme controller
    energy inputs due to the specific heat of the heatsink material's
    mass.

    Soldering the devices to smaller intermediate structures that DO offer
    reliable mating to a heatsink is usually more practical, as is
    demonstrated in most higher-powered commercial semiconductor modules.

    Without soldering, both thermal and electrical contact to aluminium or
    other hard copper can be improved marginally by introducing soft
    copper foil stock. Alternately an intermediate soft copper sheet can
    be introduced for direct electrical pick-off. Passivation of the
    surface of the copper is required to reduce oxidation in service.

    This will not overcome the physical problems of TO220.

    RL
     
  15. Rich Grise

    Rich Grise Guest

    I can't speak directly at this, but considering that there are surface
    mount components which are _intended_ to be soldered down by the heat
    pad, and that people can do reflow at home in a toaster oven, I'd say
    go for it - at least once - but be careful that the heating is even
    so that there are no drastic delta-T's in the assembly, and just reflow
    them. Presumably silicon doesn't significantly degrade at the temp. of,
    say, 63/37 eutectic solder.
    Well, if they're soldered to the spreader, you don't need grease there;
    from the spreader to the heat sink, I've always had impeccable results
    with DC-340, which is basically silicone grease with some thermally
    conductive oxide embedded in it. It looks like that zinc oxide stuff that
    lifeguards smear on their nose, and it's about as tenacious as Vaseline. :)

    Good Luck!
    Rich
     
  16. legg

    legg Guest

    The basis for rating a semiconductor has to start from somewhere -
    historically this assumes an 'infinite' heatsink maintaining case
    temperatures at 25degC. How this is achieved in a test site is
    irrelevant. IR simply suggests a method using conventionally available
    hardware and materials that assist to overcome the increasing
    difficulty of making real measurements under the theoretical
    condition.

    This is only attempted to investigate the increasing influence of
    internal bond-wire limitations at higher currents - the effects of
    which will not follow behavior of previous data that assumed all
    losses were in the die.

    RL
     
  17. Hammy

    Hammy Guest


    You are right they do have to have a standard baseline comparison.

    It's really no different then claiming a FET can dissipate 200W;
    obviously not practical unless your immersing it in a
    "nucleated-boiling inert fluid" (I had to say it "nucleated-boiling
    inert fluid") to maintain a case temp of 25C.

    I've just been noticing higher and higher current ratings.
     
  18. Mycelium

    Mycelium Guest

    It goes beyond that. You want ZERO internal contact, so even though
    there is no dust, if you cut your finger, you do not want one of these
    pads making contact with your internals.

    In fact, if you cut your finger WITH a pad, it is dangerous as well.
    ANY internal contact is bad... VERY bad.

    So NO, dipshit, it is NOT just the dust that is a danger.
     
  19. Mycelium

    Mycelium Guest


    Beryllium, its salts, and many of its compounds are dangerous
    carcinogens.

    The metal itself is fairly safe, when fresh and new. I would not want
    to even be around ANY of its oxides though.

    So, the "danger" of the DUST is Berylliosis. The dangers of other
    Beryllium based mediums is as bad, if not worse than Berylliosis.

    Just because you were aware of THAT particular affliction does not mean
    that it is the only affliction associated with the compound.
     
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

-