SMT Thermal Design Mistakes

This is the ubiquitous TO-220 through-hole package that I’m sure we’re all familiar with. This device in particular is the popular 1117 low-drop-out voltage regulator manufactured by ST. And the first thing you’re probably going to calculate when using one of these is the thermal performance and so the data sheet here gives us these figures for the package thermal resistances. With this here junction to ambient figure we can calculate the die temperature for a given amount of power or conversely the power at which the regulator reaches Tj max. So as you can see calculating the thermal performance of this package when free standing is very easy and the numbers you get are quite often very close to reality.

But of course the big advantage of using 00:48 - a TO-220 style device is the fact that it can be bolted to an external heatsink which allows it to dissipate significantly more than 2 watts or simply just operate at a lower temperature and extend its life. So this here is a pretty standard stamped copper heatsink from AAVID Thermology and looking at the data sheet for this we find a nice little graph here which is quite common amongst heatsinks although i wish manufacturers would use meters per second not US customary feet per minute. But using this line here we can find how hot our heatsink will get for a given amount of power dissipated with convection cooling only or if we have forced airflow we can use this other line to get a thermal resistance to ambient figure. Doing the calculations again but instead for this configuration we can easily show that this will reduce the die temperature at 2 watts to significantly less than Tj max. So doing this kind of thermal design calculations with this TO-220 through-hole part is pretty easy and straightforward.

Instead you should be worrying about 01:54 - making sure you have the correct mounting force or whether or not you’re placing it next to a wet electrolytic capacitor. But what about this same 1117 regulator but instead from Advanced Monolithic Systems and in this SOT-223 package how do we do the same calculations. Well looking at the datasheet again we see thermal resistance numbers but with this asterisk; thermal resistance depends on the size of copper area connected to the case tab. Okay that makes sense seeing as the tab is soldered to the PCB and it’s not like there’s a bolt hole you can connect the heatsink to. And so if we look further we see that they’ve given us a nice little table of thermal resistances for several different sizes of copper areas and well this looks easy enough to calculate and it actually is.

However, the problem comes when you actually come to apply 02:44 - this let’s look here at this PCB and well what here counts towards our copper area what parts will actually provide useful heat sinking I mean all this copper here in the ground plate probably counts right? There’s only these little tiny gaps between the copper areas here between the power rail and the ground plane and well I see regulators on PCBs like this all the time that just use the minimum footprint size with the ground plane surrounding it I mean the thermal resistance of the pcb can’t be that bad right? I mean look here we’ve got copper at 400 watts per meter kelvin and going down, oh, ohhhhhh… okay yeah this might be a problem. So this here is a PCB I designed back in 2016 I made it for fun to see if I could make my own multiplexed LED matrix and I’m pretty proud of it I was even able to get the entire LED matrix on one layer because I used the LED leads themselves as jumpers across the common anode lines useful little trick for you however today we’re interested in it because on the back of it we have the same AMS1117 regulator with a nice little PCB heatsink. Now at full brightness this regulator dissipates less than a quarter of a watt so this heatsink is plenty big enough. However I have here a spare PCB that I’ve repurposed as a thermal test board and I’ve been abusing it just a little bit. By the way if you’ve ever wondered what happens to a 1206 quarter watt resistor when you put several watts through it for a little bit too long yeah don’t do it.

04:24 - So we’re going to use this PCB to test our theory to see whether or not these little gaps actually do anything currently I have the board under the thermal camera now and I’ve got the output loaded down so it will dissipate about one and a quarter watts and when we turn it on we instantly see the temperature rise. Okay now that it’s reached steady state we can clearly see the outline of that copper area directly connected to the package tab so unfortunately that means the temperature drop across this little gap here is rather significant so that’s definitely going to have an effect on how much heat is conducted into this ground plane and it’s probably not as much as I would have liked. And of course if we push this device even further to dissipate several watts we can see this little heatsink is never gonna save it. And so this brings us to the part where I screwed it all up with my thermal design and roasted some LEDs. This here is a modular LED panel I designed back in 2018 before I had a thermal camera and before I knew these gaps were a big problem for SMT design.

05:37 - This thing is pretty simple it takes 12 volts and steps it down to a constant current for the LEDs with near zero flicker and the brightness is controlled with this potentiometer. I actually use three of these panels on a shared heatsink above me for filming and reference photography and at 10 watts per panel I get about a 1,000 lux here with them all at 50% which is more than enough. So overall I’m happy with them and all of the files are on GitHub if you’re interested although you might not want to use them after I show you the problem it has. If we look here at the LED side of the traces we see the pads of the LEDs are connected to these big traces and when designing it I paid extra special attention to length matching these traces to equally drive the LEDs. But I just assume these little gaps here between the traces in the ground plane wouldn’t be a big contributor to thermal problems I even had plenty of vias to transfer heat to the other side of the board but again here I have the same problem with traces and gaps in the ground plane.

I was assuming that this would be enough 06:46 - to transfer heat out of the PCB and into the heatsink and well I was kind of correct it doesn’t instantly catch fire but, I definitely could have done a lot better and not needed to actively cool the heatsink if I had designed it just a little different. So I have the LED side of the PCB now under the thermal camera and turning it on yeah we see the gaps are definitely a problem here as well. And now at steady state we can see this thing has big problems running at 100%. Now remember this thermal camera can only show us the temperature of the top of the phosphor of this LED and that’s at 112 °C that’s not the die temperature the die temperature is definitely over Tj max which is 115 °C so this LED is definitely having a short life. And again we see the heat isn’t conducted very well across this gap, heat is still transferred but it’s nowhere near as good as it could be taking a look at the other side of the PCB now and we can see the vias are doing a great job at transferring heat through the PCB to the other side however we have the same problem.

08:02 - And this side we’re relying on the copper area to spread out the heat so it can be transferred through more surface area to the heatsink and it’s definitely not doing a good job at that. So I hope you’ve learned something useful here so you don’t have to go through the same mistakes i did to learn this. So as always thanks for watching. :) .