Heatsinking your regulators?

[pseudolobster]

Lately I've been running into problems with heat and my drivers...

I'm using the LM1117T in a TO-252 package, and even at 300mA they get far too hot to touch. I read in the datasheet that they have an auto thermal shutdown temperature of 150c, and at 400-500mA it seems I'm exceeding this. After about ten seconds the laser starts to fade, and if I move it around I can see it pulsing, it creates a dotted line. According to the datasheet they recommend a leaving a small patch of PCB around it empty so the copper clad acts as a heatsink, but from what I can tell there's not a whole lot of copper there.

In my experiments I've been soldering chunks of copper onto the regulator to act as a heatsink, upwards of the size of a penny.. Yet without fail, the copper heats up really quickly, and after 30 sec or so the chip drops out of regulation and starts oscillating.

I've attached a picture of my test setup, it's a little hard to make out, but the thing on the top of the stack of power supplies is a PC fan controller with 4 thermal probes, it's reading 91c, after about 20 seconds of running at typical LOC currents, and the thing I circled in green is the chunk of copper I've soldered onto the regulator... I'm not sure it should be so hot.

So, for anyone who builds drivers, what do you do to heatsink your regulators? This just isn't going to cut it.

 

[Meatball]

I hear ya man!
If you don't have a chassis to mount to, than you can use some use some heatsinks designed for these smaller packages. I don't know what the TO-252 package looks like, but you can use these heatsinks for the 220 and 202 packages.

TO-220 Heat Sink - RadioShack.com

TO-220/TO-202 Aluminum Heat Sink - RadioShack.com

My lm317s in the TO-220 package gets quite warm but it always survives at 420mA even without the heatsink. Any random, flat piece of aluminum with lots of surface area can usually be bolted to the IC. If all else fails, you can always use a fan, or a peltier device. Does this help any?

 

[Bionic-Badger]

According to the datasheet, the power dissipated by your regulator is:

Pd = ((Vin - Vout) * Iload + Vin * Iground

From the circuit topology, Iground = 0, so the second term disappears.

Assuming a Vin-Vout drop of about 3V, and Iload of 500mA, that would be around 1.5W of power that regulator is dissipating.

Your measured thermal temperature rise is about 90C - 25C = 65 degrees. We'll use this as your maximum temperature rise, T_rmax, though you probably want it even lower right?

So your junction-to-ambient thermal resistance is:

Theta_ja = T_rmax / Pd = 43.33 C / W.

According to page 10, anything below 103 C / W requires heat sinking.

With the TO-252 package, the closest Theta_ja rating they've tested for is 47C / W (4C more than yours!) which requires a 1x1 inch square of 2oz copper (2.8mil, penny is ~50mil) as a heatsink, attached directly to the regulator's heatsink (i.e. top-side). One inch is slightly larger than the diameter of a US quarter.

So you'll have to see where your heatsinks fall in line with the above. I'd personally just get a different LM317 regulator package, especially since that TO-252 is skirting its maximum current (0.5A) anyway, and you might as well if the heatsink is going to be so large to compensate. The 1.5A varieties get hot too, but at least they have a screw-hole to mount a finned heatsink to.

 

[Bionic-Badger]

Oh whoops, you're using the LM1117, not the LM317-type. That just changes the Vdrop. So the Vdrop = Vin-Vout will be about 1.25V @ 500mA, so 0.625W instead of 1.5W. That's much better.

Typical C/W for that package is 92C / W, and you're at 65 / 0.625 = 105 C / W. If you want something not-so-hot, like 70C (122F), that would be about 72C / W, you can consult the heatsink table. The 72C/W heatsink patterns don't look too bad.

Also, did you check to see that your input voltage is just barely above Vload + 1.25V? Every bit of voltage above that heats it up more.

 

[Apaka]

Hi boys,
This is my first post.
Try use the video card memory heatsink!

Apaka

 

[ElektroFreak]

I just saw somewhere on here where someone had taken the little heatsinks that are on the back of 6X diodes while they are still in the sled and attached them to SMD regulators like that using thermally conductive adhesive.

 

[bugattidude]

i used a 1 n a half by 1 n a half by a half inch thick chunk of aluminum that i got from my metals class.. sometimes thats not even enough to keep the regulator cool at the current i have it at

 

[Bionic-Badger]

You can always buy some dedicated heatsinks for those types of chip packages.