Dave explains how to attach an SMD power transistor or regulator to a case to use as a heat sink in this design tutorial. And in the process talks about thermal design, the electrical/thermal analogy, and thermal vias.
This is Part 15 of the µSupply Power supply design series. Other videos are here:
https://www.youtube.com/playlist?list=PLBF35875F73B5C9B5
Forum: http://www.eevblog.com/forum/blog/eevblog-744-smd-thermal-case-design-supply-part-15/'>http://www.eevblog.com/forum/blog/eevblog-744-smd-thermal-case-design-supply-part-15/
SilPad datasheet: http://www.bergquistcompany.com/pdfs/dataSheets/PDS_SP_A2000_12.08_E.pdf
PowerPeg thermal PCB pin/turret: http://tem-products.com/index.php/powpeg.html/
Saturn PCB Toolkit calculator: http://www.saturnpcb.com/pcb_toolkit.htm
EEVblog Main Web Site: http://www.eevblog.com
The 2nd EEVblog Channel: http://www.youtube.com/EEVblog2
Support the EEVblog through Patreon!
http://www.patreon.com/eevblog
EEVblog Amazon Store (Dave gets a cut):
http://astore.amazon.com/eevblogstore-20
Donations:
http://www.eevblog.com/donations/
Projects:
http://www.eevblog.com/projects/
Electronics Info Wiki:
http://www.eevblog.com/wiki/

Hi back in video 644 which I'll link in down below if you haven't seen it. I showed these uh project cases and how you can do front panels and things like that. but I also mentioned about how you can get heat out from a regulator in these things. Now these are a split case like this and I won't go over the whole thing.

but I basically talked about if you had say, a To220 package like that mounted on your board, you could actually flip it upside down like that. bend the leads in uh, the opposite direction to what you would normally bend them for for a through hole package. and if you happen to have the right size case which this one is, then you can see that it just touches the bottom of there. There's enough room to put in an insulating seill pad in there to uh, isolate your regulator tab from the case and you can dissipate the power from your regulator or your power transistor or whatever it is into your case.

and you can effectively use your case as a big heat signal. or I mentioned also how you could dissipate heat internal as well with just a regular heat sink like freestanding like that. or you know it might be a flatter one like that provided that um, it didn't touch the top of your case for example. So if you wanted to keep your the Uh tab of your voltage regulator or power transistor electrically isolated from the case which is usually the case, no pun intend.

uh, usually what you want, then well, you can do that. But the problem with internal heat sinks is that, uh, usually you're going to have a seal case like this. it's generally not going to have a fan on it so all that heat can build up inside and especially if you're using, say, a battery powered product. Uh, you know you don't want your batteries inside to get hot and things like that, so you know, really, it's not a great thing.

so dissipating the heat out to your case can be a fairly decent option. So I showed how to do that with a To220. But what if you wanted to use a surface mount package like a Dpack? Uh, there are lots of advantages to these. They could be more readily available, they could be cheaper, you know, through hole is not gone the way of the dodo yet.

But uh, you know, generally, uh, you might have a lot more availability in these modern uh, surface mount packages like this dpack or something else. And of course you can pick and place them. They're brilliant for Pick and placing and uh uh so that you don't have to have the extra step. the manual step of having to, uh, manually, you know, bend the leads on your Uh Power transistor or your regulator and do that.

You need a you know, you need a human to sit there and actually do that sort of thing and then hand solder the thing in. That's really annoying, but how do you get the heat out of these surface mount parts to the case? Yeah, you can come along and still use your internal heat sink like that. you can get like surface mount Heat sinks that also work with pick and place machines. They come along boom and they place them down provided that they aren't you know a huge amount of mass, otherwise they tend to fall off.
If it was a big heavy thing like this relatively, then you know it' typically fall off the head of the pick and place machine. it wouldn't have enough suction to you know, pick it up and then bring it over and dump it down. But you can get surface mount heat things like that which sit uh which can be pick and placed and then Reflow solded. but once again, you got all the heat being dissipated in your case cuz you'd have the end on it here and there'd be no forced air uh through the thing cuz you wouldn't.

You know you generally wouldn't have a fan on something this small so it's not too hard with to220 packages like that. Uh, for example, especially if you got the right height case which this one is and you can sort of. you know it. sort of.

you got to have Force down there so you might have to screw it or something like that. Um, but still, you can. You know? fair? Not too Without too much trouble, get the heat out of that case. It's a little bit un professional and messy and a bit sort of like cheap product uh type.

So it's it's a bit more professional to do it better with surface mount parts and have a proper solution that I'm going to show. Now, of course, you could actually flip your board upside down like this. You could have your parts mounted on the bottom side like that, but then of course you got a double-sided uh load. For example, you can have all your parts on the bottom and put your uh board on the upper half of the case.

Um, either way, you do it, but you can't sort of flip. You can't do the same trick and flip this one over like that it and get the heat out that way. it just doesn't work and you can't uh, wedge the plastic case of that uh transistor up against here because that it's just way too much thermal resistance. You got to have metal to metal contact to get all that heat out.

That plastic is just going to be horribly horribly inefficient. So how do you get your heat out of your little surface mount power transistors or Regulators to the case of your product? Well, it's a good question, and there's probably you know a few solutions to this. now. one solution might be of course, a surface mount uh, heat sink like that internally and which actually came up to the top of your case or the underside depends on which way you want to.

uh, mount it and then you can actually put screws there and get the heat through. and well, that's fine and dandy. But how do you electrically isolate your tab? So for transistors like this, that tab on the end there and bottom that gets Reflow solder down to the board. It's going to be the collector of the transistor.

Take a look at the data sheet here, so you don't want your collector of your transistor connected to your case. That could really ruin your day. It could short out, um, something else cuz you might have the collector uh, for example, connected. You know it's not going to be ground Now of course, if we had a, uh, like a 7805 voltage regulator for example like we do here, this one is, then this tab is actually connected to the middle pin which is actually ground and okay, if you want to connect your ground.
If you do want to ground your case, then that's fine. You can just put your screw in there and you don't need any insulating washer in there. So in probably the majority of cases, you want to isolate your tab of your regulator or your transistor from the case. So you've got to have some sort of insulating material microw washer, or a um, seal pad or something like that.

You could have multiple devices. you want to isolate them from each other as well cuz they're generally not going to be electrically connected. So this is how I would uh, probably do it as a first past now I've had David 2 mock up a little animation in solid Works here and you can see. isn't it quite neat? And we've got our power transistors over here or both electrically isolator like that.

You can see that they've got both different islands of copper so that keeps them all separate so you can have as many isolated devices as you want and you can see that I've got what are called thermal vas I'll go over these in a second either side of the each power transistor there and they are what transfer heat back from the top side of the board cuz we're using topside component load here through to the bottom side of the board where we can actually get a heat uh, spreader bar or a heat transfer bar to the case. So at this stage it pays to actually go to the Whiteboard So please excuse the crudity of this model. Didn't have time to build it to scale or to paint it. Not as good as David 2's 3D model, but I've redrawn it here.

Now you may be thinking Dave I think you've shorted out these two power transistors because if we've got this uh, red metal bar in here like this and we've got these vs. the electric, the tab of the power transistor here is connected to our big copper uh, plane there with all these thermal Vas going through the bottom side. Now a Thermal Vea is just another name for a regular Via except it's not used to carry current. although it can still do that.

Of course it's just just a regular copper via through the board from the top side of the board to the bottom side of the board. But in the case of a thermala, it's designed to actually transfer heat energy from the top side of the board to the bottom side of the board. So anyway, so we've got a duplicate pad on the bottom side of the board. with all these vs.

it's exposed copper, hasn't got any solder mask over it as you saw in the 3D model, so that's going to make contact to our metal bar down the bottom here and it's going to make contact to the metal bar over here one. We've just shorted out our power transistors so there is one thing missing from the 3D model which we didn't show. You need a S pad in here? Sil pad is like a trademark. I Think an insulating uh washer like old school stuff is made out of mic the new stuff.
The seal pads are like flexible, sort of tear prooof um kind of like a rubbery type thing, but they're designed to transfer heat rather efficiently from one device to another whilst actually providing electrical isolation. So that's where you would put the seal pad. You could either have just one little one there and one little one there or just put it right across the bar like that and this could be a bar. It could be a big block.

it can be Whatever you could do this in the middle of the board doesn't have to be on the edge like this. Do it wherever you want, you can see how. Now if we put an insulating seal pad in here, then we can transfer the heat from the surface mount device on the top, through the thermal vas, to the bottom, to the Copper on the bottom side of the board and then through to our heat uh, bar or heat transfer bar, heat spreader, whatever you want to call it and then through to the case down here. You don't need an extra seal pad between uh, the case and the bar down here.

CU You've already electrically isolated your devices like that. So Bingo We've got a neat professional solution for getting heat out of SMD devices through to an external case which you can use as a heatsink. Not huge amounts of power, but you know if you only need to get rid of a couple of Watts or something like that you know bit of annoying amount of heat then this is quite a decent solution for getting that heat out. But I know what you're thinking.

We've got some loss in these thermal veas which I'll explain in a second. Yes, you could actually Mount these power devices on the other side of the board of course. or or flip the whole board and mount all your components cuz it's cheaper to Mount Your components on one side of the board with a pick and place machine if you have to put them on the top and the bottom. Well, that's a two-step process at your assembly house.

It's going to cost you a bit more so you want to avoid that if possible. So you could actually Mount these devices on the bottom and then have a cut out in your thermal bar like that so that you know your device just sat down there. so you've effectively got like a couple of thermal bars just over the vaa and your device could sit nicely in even a little cutout like that. but you probably wouldn't do that.

You just have multiple blocks like that. And of course, the other trap for young players. your screws. They're metal.

You don't want to accidentally connect this through to your like for this screw to touch your copper pad here. so you want a big bit of isolation right around there like that so that, uh, your screw can go through and hold your board down because you need to apply pressure. It's important to actually apply pressure. on your seal pad, otherwise you're going to get a pretty piss poor contact and your heat transfer is going to be absolutely horrible.
So you need to screw that down. so there's a bit of, uh, consistent pressure between the board and the heat bar and also you would screw it up from the bottom as you saw on the animation there as well to get good thermal contact between your bar and your case. And if you wanted to Guild the lily a bit and make sure you had the best possible solution cuz you've only got two screws like this and you're going to get uneven surfaces on not only your bar, but your case as well. They're all rough surfaces, you might add some.

you might squeeze a bit of, uh, thermal compound under there as well. Now this step here is just uh. repeating a video I did Way Way Back in the Eev blog, but it's worth including here. I'll link it in down Below way back in the early days it was and it's showing the equivalence between electrical design and thermal design and how you can calculate uh, temperatures in your system, the temperature of your transistor, the temperature of your case, and all sorts of things.

Now the great thing about thermal design and looking at it and thinking about it like this, is that you already know it. It is just basic. Electric Theory You're used to series resistors voltages, and currents. Very easy to calculate and it's a direct analogy.

It's not a fudge. it actually works now. Uh, there are three rules here. Current in the electrical analogy is equivalent to power.

In the thermal analogy, the resistance is still a resistance, but instead of being ohms, it's resistance in resistance. In quote marks in degrees C per watt, it's the thermal resistance of the Heat syn of the device of the Vaa or whatever it is. And voltage is equivalent to temperature. So if we've got we can, uh, model this as an electrical equivalent.

Let's say our device here is uh, producing 10 watts for example. Then that 10 watts is equivalent to 10 watts of current flowing through all these devices. CU You can see how they got them stacked. We've got the actual Junction the semiconductor Junction inside the trans.

So if you look up the data sheet for your power transistor or your regulator, it'll have a term cord. It'll have the thermal resistance uh, in degrees C per what? that little symbol there is a Theta. So they will have Theta Junction to case. So I.E the thermal resistance between the semiconductor Junction inside the transistor and the tab on the transistor that's going to have a specific thermal resistance.

When that package dissipates x amount of power, it's going to produce a voltage across it, which is actually a temperature. So we can actually get temperature across each particular device here and then all your other components in the system are also going to have a specific thermal resistance the next one after we get from the transistor. I'm ignoring the copper. The copper will have a thermal resistance too, but it can get quite complicated so I'm just assuming there's no loss in the in that copper itself.
The next one is the via. The heat actually has to remember that 10 watts of heat or whatever it is has to transfer through the VR It's going to have a specific thermal resistance, and then it's got to get through that seal pad that we put in there. That seal pad will have a thermal resistance. look up the data sheet for it.

It'll tell you what it is typically. uh. And then we're going to have the thermal resistance of the bar. Here, it's going to be pretty low.

It's a nice big chunky bit of aluminium, but it's still something that you have to consider in there. And then we've got the thermal resistance of the case. Now the case is our heat sink. In normal thermal design, you'd look up the data sheet for the heat sink.

It's going to give you a Theta value in degrees C per watt. It's going to give you a thermal resistance, so uh, but that will depend on whether or not you've got forced air across it, whether or not it's radiating, whether or not it's radiating into free air Usually they're specified into free air, but they might may also have a Uh, a thermal resistance specified for a specific amount of air blowing over the heat sink in a certain way. And then we've got our ambient temperature, which is equivalent to adding a voltage down here because remember, voltage is equivalent to temperature. So 22 C that's going to be our ambient temperature.

Say if it's you know room temperature here in the lab. then you can actually go through and calculate based on your thermal resistance and based on the amount of power you have flowing through I.E the amount of power you dissipa in your transistor. You can actually calculate the temperature rise of the case. You can calculate what temperature that case will get to by just uh, you know simple Ohms law stuff, the power multiplied by the thermal resistance of the case and you can get the case temperature and likewise you can then the as you can see the temperature builds up at each stage until you get to a specific higher temperature at the junction and a semiconductor might be rated for maybe you know, say 120 C absolute maximum Junction temperature.

Well, you got to ensure that you don't exceed those sort of specs. but of course you wouldn't want your trans to get up to 120. So you you know you calculate you do all this. you design the size of your case, You design the size of your spreader bar, the number of Vs, and all sorts of things you know.

and you can do some good rough ballart calculations here if you really want to thermally Model A A complex. this is a, you know, even though this is a relatively simple example. if you really wanted to model this properly, you got to do finite element analysis. And there's software packages which cost you tens of thousands of dollars to to do or attempt to do this sort of uh modeling.
But you can really get good results with just simple electrical equivalence like this beauty. So how do we calculate the thermal resistance of our Vas here? You remember we got all these multiple Vs in here. It's not particularly easy, but you can get some nice ballpark calculations now. One program I Love using I Highly recommend you use uh, the Satin PCB toolkit for anything to do with PCB uh design.

It happens to do uh, calculate the thermal resistance of Vas like this and you punch the numbers in. For a 1 mm diameter hole on a 1.6 mm double-sided PCB with uh, typical 1 o uh copper 1 o plate in on which is 35 microns on your hole, then it's going to be about 49. I'll round it to say 50 C per watt for that one V So that's pretty horrible. 50 C per watt.

So if you only stuck one VR in there and tried to get the heat through to the other side side here, even if you had one watt, if your power transistor dissipating one watt flowing through, here, you had 50 C per watt. Well, you're going to get a 50 C temperature rise in that VR Awful. So what do you do? Well, you put multiple ones in parallel cuz remember all this is electrical equivalent. So if you have a look over here, then if you've got multiple V's like that each one has a 50 C per wat thermal resistance.

Well, your wack 3, well, your wack 2 in parallel. For example, Bingo 25 C per wat 3 and so on. It goes down and down. If you put a matrix of uh V's like this, which is typically what you'll see on the board, you might see you know, nine like this, or you know, like a 4x4 16 either side or something like that surrounding the thing.

You may even see some thermal VAs on the bottom of the device. But if you do that, there's pros and cons both ways. Is like the uh solder when you put the solder paste on it can uh Wick down through the board and that can. There can be issues with that, so let's not go there.

but because as I briefly mentioned before, there is going to be some loss in the copper spreading the copper itself like the 1 oz copper you got on your PCB, there's going to be some loss in that and so it's not simply a matter of just paralleling m. It gets a bit more complicated than that, but the basic uh result you're going to get out of it is it's not going to be a linear thing. it's going to tap off like this. So if this is thermal resistance in degrees C per wat uh versus the number of Vas down here.

The more that you put in parallel like this, you know if you start putting 20 30 of them in parallel, you get diminishing returns. So and I've seen some data that shows maybe you know, like like 10 or 12 vs. or something like that starts to become, you know, fairly. Optimum Uh, in terms of you know, you can't just gild a Lily and put 100 in there.
it. It doesn't really gain you much. and of course, uh, it's going to be a a trade-off as well with the uh diameter of the VR. Generally you know, maybe half a millimeter or 1 millim Vas might do the job you don't want to do.

little uh, tiny, you know., 3 mm ones or something like that. so maybe you know half mm8 mm. Probably a good Park to bpk to use for thermal vas. Now of course there are other ways to do it.

Uh, I've done a mail bag where you can get these like uh, surface mount turret things which go through your board and they can extract power directly from the back tab of your power device. And if you're really designing a critical thing that might be a uh, good way to do it or a good additional uh thing. For example, you could also have a large hole or cut out in your board and then actually have a metal uh, you know, Rod or a standoff or something going up and directly contacting the back of your surface mount device in there and it gets all nasty. And anyway, I Think this is not a bad solution for dissipating.

You know, like a few Watts or 5 Watts or something through the case. You wouldn't use this for like you know, 100 watt audio amp or something like that? Then you'd be using big beefy uh Power devices like you know, to220 packages that all in parallel that are bolted directly to the side of the shazzy. Because ultimately you want as fewer things in series here as possible. So if you can go straight from the case straight onto the heat sink here and avoid the V and the seal pads and the Uh and the heat transfer bar.

and if you can take all those out of the equation, then you're going to be operating in a much better, lower temperature environment. You're going to get the heat away from your Junction where it's being dissipated. This is, by the way, that's the temperature. The Junction I forgot in there.

Then you're going to extract the heat from your Junction as efficiently as possible. But when you're dealing with surface mount devices in a little case like this and it all gets a bit harder. Once again, better if you got a fan in there to Fan Force stuff, but in a little handheld thing, you know you're not going to do that. if you're dissipating.

you know, just a couple of Watts or five Watts or something like that. So there you go. That's just some basics of doing Smds many ways to skin a cat here. So if you you got your favorite technique, add it in the comments below.

Hope you enjoyed it as always. If you want to discuss it, uh, leave YouTube comments, blog comments or jump over to the Eev blog Forum link specifically for this video which will be down below and if you like it, please give it a big thumbs up. Catch you next time.

Avatar photo

By YTB

17 thoughts on “Eevblog #744 – smd thermal heatsink design – supply part 15”
  1. Avataaar/Circle Created with python_avatars Alex says:

    Hi Dave, excellent video as usual. One question, what would you do with the vias if they increase the stray inductance and cause voltage overshoot in a switching circuits?

  2. Avataaar/Circle Created with python_avatars Gary Russell says:

    I have 2

    Scosche HD4004 HD Speakers

    4 oms 30 wat RMS 120 Peek

    will the Lepai LP-2020TI drive both?

  3. Avataaar/Circle Created with python_avatars Tom Legrady says:

    It seems to me if you are using thermal vias leading to the case, it's best to do that in the middle of the board, so heat spreads out in two directions on the case. Putting it at one end distributes heat in one direction along the case (efficient) and the other direction into the air (good insulator).

  4. Avataaar/Circle Created with python_avatars hesperaux says:

    Really awesome series. Super helpful for me, but I would love to see it finished. It's been 4 years. Is Dave 2 still around? Thanks for what you've already done, Dave. Regardless of what you've done so far here, the thinking out loud, analysis, testing, mechanical concerns, etc., are so useful for everyone. That being said, we all hope you finish it up.

  5. Avataaar/Circle Created with python_avatars gearstil says:

    Great video!

  6. Avataaar/Circle Created with python_avatars Andy Plater says:

    Thanks for pointing out this video Dave. I've been in the dark on this aspect of product design for years and this is really great.

  7. Avataaar/Circle Created with python_avatars Wouter says:

    Have you finisched the powersupply?

  8. Avataaar/Circle Created with python_avatars iwtommo says:

    This is a goldmine of heatsinking info. Thank you so much, this has helped a ton

  9. Avataaar/Circle Created with python_avatars Stanley Seow says:

    Today, i learn something new … SIL-PADS …. thanks

  10. Avataaar/Circle Created with python_avatars E J says:

    Finish This! Please!

  11. Avataaar/Circle Created with python_avatars Dexter says:

    Gosh, you do yap on a lot……..anyhoo, where is 16-20?

  12. Avataaar/Circle Created with python_avatars FrozenHaxor says:

    [*] µSupply

  13. Avataaar/Circle Created with python_avatars gmanj88 says:

    Here's an idea for the case mount heatsink option with the to-220 as Dave started with: Include a hole in the board (or drill) to fit a machine screw through. You can then screw down your regulator via a tapped blind hole in the enclosure! Better heat transfer B-)

  14. Avataaar/Circle Created with python_avatars Ted van Matje says:

    Dave! Thanks mate! you've taught me more in 22 minutes than any book. I'm an absolute 'noob', when it comes to surface – mounted tech. seeing that I still 'piddle about' with valves and through – hole tech, the wish to take that step and move 'over' to surface mounted, was hindered by the thermal issue. watching this video has made it all crystal clear! happy days!
    on a side note:- over this week, I finished watching your psu project vids. The insight on how it's done by a professional – from start to (near) finish, isn't just enlightening but something us hacker-types don't get a chance to see often.
    put bluntly; "fookin' marvelous!" is the easiest way to express it.
    Dave, you're a legend mate! cheers for posting and taking the time and effort.

    new project idea for you: design/build a cranial plug (ala "The Matrix") so we could download all that knowledge.
    don't forget to add the "kung foo program" though 😉

  15. Avataaar/Circle Created with python_avatars Mirko Mueller says:

    Most effective heat transfer with isolation at the same time is made by using thin plates made of diamond. Only problem is to actually buy them.

  16. Avataaar/Circle Created with python_avatars Kevin Amadiva says:

    nice video. got something valuable concerning heat dissipation. especially for SMD. thanks big time.

  17. Avataaar/Circle Created with python_avatars Richard Lynch says:

    Any word at all when the PSU is going to be available? In any form!

Leave a Reply

Your email address will not be published. Required fields are marked *