Heatsinking surface mount high power MOSFETS

So why do they exist at all?

When the board provides enough cooling that you don't need a heatsink, therefore saving you the cost of a heatsink, plus hardware, plus assembly. Plus, you get to pick-and-place them in one step along with all the other SMTs, without the extra through hole place + wave solder steps.

I am a little surprised at the cost disparity though. That's a pretty good difference, considering it basically represents a one step punch/forming operation.

In theory, you might also be paying for solder-proof packaging, whereas through-hole devices don't have to withstand soldering heat at the package. The expense would come from low-moisture or higher heat tolerance materials. "In theory" meaning: they probably use the same goo in all packages, SMT or otherwise.

Tim

-- Deep Friar: a very philosophical monk. Website:

Case study using Digikey pricing. Manufacturer's parts are STP80PF55 (TO-220) and STP80PF55T4 (D2PAK). The surface mount part is $4.14 and the thru hole is $2.54, volume pricing doesn't change the disparity. Specs are identical except for package.

Even if they were priced the same, the only reason I would design for D2PAK is if it is a switching application and current levels are way below the 80A max since even with the very respectable 18 mOhm on resistance, the device would have to dissipate around 28 W of heat at

40 A and 100% DC. However, that would only be after checking pricing and specs for a lower power device commensurate with the actual power levels involved.

Choosing D2PAK over TO-220 would be absurd for a high power application. The awesome efficiency you mention also applies to the TO-220 case. Also, heatsinking the TO-220 device would be a fairly standard exercise.

>

IMO, because SMT is in, thru hole is out.

.

In other words, a low power application. I already stated this in an earlier post. However, what is dubious is the use of a high power device for a low power application or the use of a surface mount device for a high power application. All these suggestions to use special boards, cut out openings etc. make no sense if the TO-220 case equivalent is available which is still usually the case.

od

That may be a Digikey thing for this item. Typically, both packages cost the same based on my observations.

>

I guess we'll need an answer from someone in the know, then... Jeorg, do you use SMT or THT and why?

Let me guess... SMT where assembly is costly (US, EU), THT where assembly is 'free' (China, et al.)? I seem to recall a post about that. So, it's all relative to location and other process costs... TO-220's may be cheaper to assemble in China than in the US, making them an even better deal.

What do you mean by "low power"? 5-10W dissipation (requiring no special boards or cutouts) isn't much dissipation, but with the low Rds(on), you can switch 200W easily. That's not what I'd call "low power". If your heat budget is really small (like what D from BC was asking about recently, where even the FWB was getting too hot), you don't really care about heat sinks, because each component might be dissipating a watt, but all of them together gets toasty.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

You're missing the point. This is about knowing when the THT component is the better choice even when the SMT equivalent is available. So you can have a predominantly SMT board with some THT components. Example, we probably won't see SMT headers replacing THT ones anytime soon. Similarly, size may make it more worthwhile to use THT electrolytic caps instead of SMT ones. High wattage resistors/devices, depending on dissipation requirements, are another example.

can

Switch>

LOL, but I have seen them!

And supposedly, Tek used SMT BNC jacks on their not-so-cheapass TDS scopes... oops...

Why anyone would buy SMT mechanical components, I don't really know. I've also heard of SMT pushbuttons, which doesn't really sound all that awful, but they'll fall off eventually.

Is class D linear? ;-) Depends on how you count it, I guess. :)

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

I dug up some random IR part in that package. They specd 300W at tC=3D25degC. Now granted that won't be easy, but I guess the on- resistance is such that if you dumped enough current into the switch to achiece 300W, it would not fail due to electromigration. I don't think the spec is a lie, but it does seem pretty useless unless they were thinking very small duty cycle applications, say a flash tube.

Yeah, I've seen SMT IGBTs rated for that... crazy stuff...

Pulsed operation is definitely something they'll do well. Keep pulses small enough that the heat basically stays in the silicon and you don't have to worry much about raw cooling power, only the average (i.e., transient RthJC is teensy). I've seen D2PAKs in plasma TVs, where big amps and volts are switched for a couple microseconds, to light up a row or whatever it is they're doing. Of course, plasma TVs run hot, and usually fail because of it. But that's consumer engineering, not the cozy overrated stuff we'd like to see.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

.

good

e.

ls.

or

e the

get

Of course you do need to waste "floorspace" making the copper heatsink. The app note should give you a theta JA based on the copper on the board. At least that would be useful data. ;-)