# LM317 Heat Sink

Discussion in 'The Projects Forum' started by lijoostenk, Feb 11, 2013.

1. ### lijoostenk Thread Starter New Member

Jan 25, 2013
12
0
Im currently working on a circuit which operates at high currents. The circuit has an LM317 which requires an efficient heat sink for its operation under such conditions.
Please suggest a suitable heat sink ?

2. ### tshuck Well-Known Member

Oct 18, 2012
3,527
679
Determining a suitable heatsink requires a LOT more information that what you've posted.

Most often, people will shove a heatsink, and, if after X hours of operation, the device hasn't caught fire, it counts as a valid heatsink.

However, the science behind it requires knowledge about maximum operating range, ambient temperature, ambient air speed, how efficient is the convection cooling of the surrounding atmosphere, air density, and many others. There are rule-of thumbs for using heatsinks, but they are not exact, though work well.

So, what is the maximum current you are passing and the duration of that?

3. ### bmxerds New Member

Sep 4, 2012
16
0
You could use a heat sink out of an old computer if you have one on hand because they are fairly large but first you need to determine how much watts of heat your chip is despensing.

4. ### lijoostenk Thread Starter New Member

Jan 25, 2013
12
0
In my circuit, the LM 317 operating conditions are,
max. current = 350 mA for a duration of 20 minutes.
ambient temperature = 50 °C.
I need the thermal resistance and the configuration of the heat sink?

5. ### tshuck Well-Known Member

Oct 18, 2012
3,527
679
How much voltage are you dropping?

By the way, 350mA is not high-current!

6. ### #12 Expert

Nov 30, 2010
18,076
9,676
A heat sink can be anything from, "Glue a penny to it" to something off an old brain chip, fan included. Without the specifications, we can do nothing but guess.

7. ### lijoostenk Thread Starter New Member

Jan 25, 2013
12
0
The LM 317 regulator of my circuit drops about 10V.

8. ### #12 Expert

Nov 30, 2010
18,076
9,676
If it's a TO-220 it doesn't need a heat sink at all.

File size:
862.9 KB
Views:
82
9. ### bountyhunter Well-Known Member

Sep 7, 2009
2,498
510
dissipating 3.5W, allowable temp rise 70C, Total theta allowed less than 20C/W. You could use just about any small heatsink.

10. ### bountyhunter Well-Known Member

Sep 7, 2009
2,498
510
yes it does. A TO-220 with no sink is 60C/W so 3.5W would raise the temp to about 260C with 50C ambient.

11. ### #12 Expert

Nov 30, 2010
18,076
9,676
The datasheet I posted shows 19 C/W for Rja, page 2.

12. ### bountyhunter Well-Known Member

Sep 7, 2009
2,498
510
I see it. I absolutely guarantee that is wrong. It has to be a misprint.Theta J-A with no sink is ballpark 60C/W for any T0-220 or TO-263 package.

http://www.ucm.es/info/electron/laboratorio/componentes/Lm317.pdf

T package is TO-220.

FYI: I personally took thermal data for publication on TO-220 and TO-263 and the LOWEST theta J-A value I could obtain soldered to a 10" x 10" copper PCB was about 30C/W. For a small copper "tab" about one square inch it was around 40C/W.

• ###### ScreenHunter_01 Feb. 11 16.15.jpg
File size:
22.6 KB
Views:
24
Last edited: Feb 11, 2013

Sep 7, 2009
2,498
510
File size:
56.3 KB
Views:
19
14. ### #12 Expert

Nov 30, 2010
18,076
9,676
It's a common mistake to believe datatsheets. I guess it was my turn to get fooled.

15. ### bountyhunter Well-Known Member

Sep 7, 2009
2,498
510
Lot of misprints around. Most data sheets are done by marketing guys. I have no idea who screwed up the one you saw.... but I worked at nat Semi where the original version of the LM317 document was written. They are owned by TI now and that data sheet is the TI version. Probably have idiots doing the data sheets. They laid off all the people who knew what was going on there, including me.

16. ### #12 Expert

Nov 30, 2010
18,076
9,676
That's normal.

Apr 14, 2005
7,018
682
18. ### bountyhunter Well-Known Member

Sep 7, 2009
2,498
510
That's why I said it was 60C/W ballpark. Different departments at Nat Sem got slightly different values depending on who took the data, it also varies SLIGHTLY depending on die size of the device inside the package. But 60C/W is about right.