lm317 and heatsink thermal calculation dead end

Discussion in 'General Electronics Chat' started by Efron, Sep 5, 2011.

  1. Efron

    Thread Starter Active Member

    Oct 10, 2010
    Hi all,

    I've looked at many posts in this forum about lm317 and its applications but never found the answer to the following thought:

    According to the datasheet, the lm317 is rated to provide up to 1.5A for a Vin-Vout <= 15V. Above 15V (approx) the maximum current that the lm317 can provide decreases.

    So, if we have Vin-Vout = 15V and Iout = 1.5A ; the maximum power through the lm317 will be 15x1.5 = 22.5W.

    For a thermal resistance calculation for the heatsink I've got the following standard inputs:
    Ta = 25°C
    Tjmax = 125°C
    Rth(jc) = 4°C/W (datasheet)
    Rth(heatsink) = to be calculated

    With the data given above, the maximum thermal resistance must not exceed 4.44 °C/W = (125-25)/22.5 = Rth(jc) + Rth(heatsink)

    Therefore Rth(heatsink) must be under 0.44 °C/W. Is this value possible to have?

    The result is even worst if using lm350, which allows up to 3A instead of 1.5A. In this case, the power through the lm350 will be 15x3 = 45W, which leads to a total thermal resistance of 2.22 °C/W but with a Rthjc = 4 °C/W (according to datasheet), the minimum value for Rth(heatsink) will never be reached ! >> the only solution is to share the power with additional device (two lm350 in parallel or additinal power transistor).

    Why the lm350 is rated to provide 3A when Vin-Vout = 15 if it cannot support a single configuration layout with heatsink ?

    Thanks in advance for your support,

  2. JingleJoe


    Jul 23, 2011
    Those are maximum ratings, they shouldn't really be used at the max ratings all the time. What are you making that really needs so many amps anyway?
  3. Adjuster

    Late Member

    Dec 26, 2010
    It is quite common for a device to have to be run at less than its ultimate maximum ratings to avoid overheating.
    It would be fairly unusual to have a maximum ambient temperature as low as 25°C to start with. The environment outside the equipment may not be air-conditioned, and a practical design will often have to cope with considerably higher temperatures, particularly when a device is operating inside equipment containing other heat sources.

    The device may thus have to be considerably de-rated to allow for the highest external ambient temperature, plus the temperature rise within the equipment, plus the temperature rise of a reasonably-sized heat sink on top of that!

    The net result is that a higher rated device may be required than might be expected at first sight. It may be cheaper overall to use a heavier device with a lower internal thermal resistance, if it allows a smaller heat-sink to be used.
  4. #12


    Nov 30, 2010
    I'd suggest you try an LM338k with theta JC of 1.4 to get the current you need, but it seems you don't need current. You want to know why the manufacturers rate things the way they do.

    Problem is, I'm horrible at mind reading.
    Efron likes this.
  5. t_n_k

    AAC Fanatic!

    Mar 6, 2009
    Picking up on #12's point, Efron should also note that the LM350 package type influences the thermal resistance Rth_jc value.

    According to the datasheet I looked at ...

    The 'T' package has Rth_jc=4°C/W [max] whilst the 'K' package has Rth_jc=1.5°C/W [max]

    I also imagine one could achieve a thermal resistance of 0.44°C/W using a copper water cooled heatsink.

    There is also the problem of achieving a good thermal resistance value at the contact interface of the the device and the heatsink surface. There is an art encompassing just that small matter.
    Efron likes this.
  6. Efron

    Thread Starter Active Member

    Oct 10, 2010
    Thanks everyone for your quick reaction.

    I agree with '#12' and 't_n_k', thank you for your suggestions. By shifting the Rthjc from 4 °C/W (lm317 - T package)to 2.3 °C/W (lm117 - K package), the required Rth(heatsink) will be about 2°C/W instead of 0.44 °C/W (if using lm317 - T package).

    Well, that's a little bit better but the heatsink will still be quite huge :-(

    At the end, I guess the best option is to use a by-pass darlintong transistor (with better heat characteristics) and to keep the lm317 as the regulator function with low power on it.

  7. iONic

    AAC Fanatic!

    Nov 16, 2007
    A bit of calculating by another....
    See Here

    I suppose it is prudent that if you are going to approach the Mac value then a better part choice would be to step up one notch. They also recommend the LM338
  8. bertus


    Apr 5, 2008
    PackratKing likes this.
  9. Efron

    Thread Starter Active Member

    Oct 10, 2010
    Hi iONic,

    I already went throuhg this web page, where you can see a small heatskink capable of 21 °C/W. However, according to my calculation, I would require between 0.5 and 2 °C/W!
  10. iONic

    AAC Fanatic!

    Nov 16, 2007
    Thought you might have....
    So it's time to realize that the 317 is a big hungry bear and the nice slim fox, the switching regulator is better at the game your playing.

    There's two ay to go: rip the giant heat sink of the back of the old Heath-kit stereo amplifier and attach it to the 317 or use a switching regulator. I suppose it also depends on the climate zone you live in. You may opt for the extra radiant heat. Not trying to bad-mouth you. If the Vin-Vout is 15V, your gonna fry some eggs.