• State
  • Locked Locked
  • Replies 5 replies
  • Subscribers 62 subscribers
  • Views 394 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM317: LM317S/NOPB Issues with Thermal Analysis

OhmMyGod
Prodigy 120 points
Part Number: LM317


Tool/software:

Hi everyone,

I'm working on a design where I replaced an LM317HVK STEEL (TO-3) with an LM317S/NOPB (TO-263). I'm seeing higher-than-expected temperatures on the TO-263.

Here's the situation:

  • Original Design: LM317HVK STEEL (TO-3) without a heatsink.
  • New Design: LM317S/NOPB (TO-263) with 1 sq. in. of 1 oz copper on both the top and bottom layers of the PCB for heat dissipation.
  • Load: 2W (Vin = 28V, Vout = 20V, Iout = 0.25A) for both designs.
  • Measurements:
    • TO-3 case temperature: 52°C
    • TO-263 case temperature: 76°C

 As I understand it, the junction temperature should be the same for both solutions, and the TO-3 package should be closer to the junction temperature (Rjc = 2.3°C/W). However, observed TO-263 temperature is significantly higher.

Appreciate if you could shed some light on be contributing to the higher temperature on the TO-263 and why does TO-3 based solution experience significantly lower temperature?

  • 0
    TI__Guru 50025 points

    Hi OMG,

    It's a bit difficult to compare the expected top case temperatures because the Rθ parameters are known to be highly layout dependent. It would be a better comparison if the LM317HVK STEEL had a ΨJT parameter listed in the datasheet as the LM317-N does since the junction temperature could be estimated with higher precision. With that said, there does seem to be a large discrepancy when these should have similar thermal performance based on the similar RθJA. 

    I am a little surprised that the TO-3 has such good thermal performance when not using any heatsink (I haven't supported this legacy package in some time). It's usually only the larger packages that can achieve sub-40C/W and especially sub-30C/W with good heatsinking. You mentioned the layout for the LM317S; can you share a screenshot of that layout and the layout of the old design as well? What was the ambient temperature for the tests that you shared? 

    What is the max ambient temperature for your application? 

    Regards,

    Nick

  • 0 in reply to Nick Butts
    Prodigy 120 points

    Hi Nick,

    Agreed on expected similar performance, which is why we are trying to get to the bottom of it and determine the root cause for the discrepancy.

    Attached is the screenshot for the new layout. New layout has 1" x 1" polygon on all 4 layers.

    The legacy board is a 4-layer board with no heatsinking provision on any of the layers.

    Ambient temperature is 25C for the measurements, with Max ambient temperature target of 60C.

  • 0 in reply to OhmMyGod
    TI__Guru 50025 points

    Hi OMG,

    The layout looks fine. 

    Here's what I'm thinking. The junction temperature for the LM317-N, from your measurement, is estimated to be 76C + 2W * 10.4C/W = 97C, and if your board was similar in thermal performance to the JEDEC High-k board from which the thermal metrics are derived (which I think your board is similar and probably a little better due to the top-layer relief), the junction temperature would be estimated to be 25C + 2W * 41C/W = 107C. So, I think the measurement and junction temperature estimate for the LM317-N looks pretty consistent. What I'm less sure of is whether the thermal metrics for the legacy TO-3 package are accurate. The datasheet for the LM317HV-MIL does mention that the thermal metrics assume no heatsink, but as I mentioned in my last comment, only large package with good thermals and large thermal pads can achieve 35C/W, so I wonder if the RθJC(top) (or the RθJA for that matter) are actually as low as they seem in a real application. The heatsinking for the TO-3 package would be mostly due to the copper traces connecting to the pins, so I suppose if the power traces were large then the heatsinking might be good enough to function in your application, but I doubt that the TO-3 package actually performs better with no heatsink than the TO-263 with the level of heatsinking you have on your new board. I suspect that the RθJC(top) for the TO-3 package is a lot larger than it appears to be. 

    By the way, how did you make the thermal measurements? 

    Regards,

    Nick

  • 0 in reply to Nick Butts
    Prodigy 120 points

    Hi Nick,

    We've used a thermal camera to measure the temperature with a black tape attached to the TO-3. Also, TO-3 is warm but not hot. TO-263 based solution is hot to the touch.

  • 0 in reply to OhmMyGod
    TI__Guru 50025 points

    Hi OMG,

    Got it. I still suspect that the TO-3 is not performing as well as it appears to be since the TO-263 since the TO-263 thermal metrics align with the measurement. Perhaps a test you could do is to load the device and increase the input voltage until the device reaches thermal shutdown. Then you can directly compare the thermal performance of the 2 packages. 

    Regards,

    Nick