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.

TPS2378: TPS2378 getting hot

Part Number: TPS2378
Other Parts Discussed in Thread: TPS2372-3EVM-757, TPS23861, , TPS23861EVM-612, TPS2372

Hi,

I have a PoE circuit for LED Lighting that uses the TPS2378 as the PD interface. The 48V is then taken to a LED constant current circuit to power my LED. A LED constant current driver is connected to the 48V supply from the TPS2378.

  1. If i connect the PD to the PSE with the LED connected and the CC driver enabled, then the LED comes on but the TPS2378 gets uncomfortably hot.
  2. If i connect the PD to the PSE with the LED disconnected and the CC driver enabled, then the TPS2378 is thermally ok. When i then connect the LED, which comes on, the TPS2378 stays cool.
  3. If i carry on with 2 above and turn the CC driver OFF and then ON, then the TPS2378 gets uncomfortably hot again. I don't disconnect the PoE cable for this.

My (limited) understanding of this is that, if the load on the TPS2378 tries to draw more than 140mA at start up (80mS) then the TPS2378 goes into current limit mode which is why its getting hot? So i need to delay turning on the 'load' until 80mS has passed?

However, this doesnt explain what happens when i turn the CC driver OFF and then ON as in (3) above. This I will need to overcome as i plan to control the LED CC driver ON or OFF remotely.

I would be grateful for any guidance on this.

  • I should add, that as i described the status in my post above, I was not making use of the CDB pin, which is probably at the route of my problem. But im not sure that CDB would help in scenario 3?
  • Hi Howard,
    For #1, I agree with you that the PD can't get past inrush and so is getting hot.
    for #3, I sounds like the OFF then ON is causing the PD to go into current limit (maybe because of the startup of the DCDC).

    What is the total power of the load? You may want to look at the input current to see if this is true.

    Also, have looked at the TPS2372-3EVM-757? This uses a PD that has an internal 80ms delay to solve your #1 issue. It also uses a LED driver if you want to compare solutions under the same load conditions you're currently getting failures for.

    Thanks!
    Regards,
    Darwin
  • Hi Darwin,

    I had a look at the connection information on the PSE (TPS23861 based) and it tells me that the class is only class 1. I use TI-commander-poe for this and the i2c interface.

    Rcls is 63R4 and Rden is 24k9 so im stumped why the class is only 1. Detect is showing as resistance valid.

    The class 1 will explain why i cant get the power down the line that i need.

    Give that there is only the RJ45, Magnetic, Bridge and then TPS2378 im not sure what could be causing this?

    Howard

  • Further to this, ive just plugged in TI eval board tps2378evm-105 and this shows up as class 1 also on the Ti commander software.

    Im going to double check everything on a different PSE tomorrow.
  • If the TPS2378EVM was not modified and class resistor is 63.4 ohms with a ohm meter, then it might be a PSE issue. Please let us know what you find after looking at a different PSE. Thanks!
  • Hello Darwin,

    I ran all myPD  boards on the TPS23861EVM-612 connected to my PC running TI-Commander-POE

    • Our proprietary PCB with 2x isolated 12V supplies was found and correctly identified as 2x Class 4 devices on one port
    • TPS2378EVM-105 was correctly identified as Class 4
    • My TPS2378 PCB which my original post was about was correctly identified as Class 4

    I have just built our 8 port PSE which uses 4x TPS23861 and connects to my PC running TI-Commander-POE

    • Our proprietaty PCB with 2x isolated 12V supplies was found but incorrectly identified as a class 2 and a class 1
    • TPS2378EVM-105 is incorrectly identified as Class 1
    • My TPS2378 PCB which my original post was about was incorrectly identified as Class 1

    So my PSE is not able to correctly identify the class of the attached device.

    My PSE was based on the TPS23861EVM-612 with some additions/changes:

    • It has 8 ports as opposed to 2. (On each 8 pin port we have two switched power channels)
    • There is onboard lightening protection
    • We had a footprint issue with the FDMC3612 which meant i needed an alternative. We fitted a IRF5015 device which was a close match. However there are differences - Vdss is 150V, Rds is 31mR, Qg is 36nC and Ciss is 2300pF.
    • I fitted CMChokes in the supply to the centre taps of the magnetic.
    • The LEDs within the RJ45 are placed across the centre taps of the magnetic.

    My questions regarding the incorrect Class identification -

    1. Could the different FET be responsible?
    2. If not, what else should i look at? Even with the incorrect Class identification, power is supplied to the PD, albeit limited.
    3. Is the Kelvin sense used in the Class identification?
    4. Is there somebody i can send my schematic to for review - (iI'm not allowed to post it on a public site)

    Thanks

    Howard

  • Hi Howard,

    Yes I think looking at the schematic will help. I recommend sending a private message through E2E and attach the schematic. I would particularly be interested in looking at the FET, sense resistor, and PoE output connections.

    Thanks!

    Regards,

    Darwin

  • email sent - i had your address from previous correspondance
  • Hi Darwin,
    In an ealier post of yours you said
    "Also, have looked at the TPS2372-3EVM-757? This uses a PD that has an internal 80ms delay to solve your #1 issue. It also uses a LED driver if you want to compare solutions under the same load conditions you're currently getting failures for."

    Does this mean that the CDB pin on the TPS2378 should trigger an external delay (80mS) to ensure inrush is over?

    Does CDB going high impedance not indicate inrush is over and that i can power up my downstream devices?

    Howard
  • Hi Howard,

    That's correct, the TPS2378's CDB only releases when the PD has finished inrush (charging the bulk cap) but does not tell whether the PSE is still in its inrush phase. You're correct that the there will need to be an extra delay before going into normal operation current.

    In the past, the load would provide this delay (via microcontroller delay); however, as we see more and more designs needing it to be implemented in hardware, we have integrated this delay in our newer TPS2372 PD.

    Regards,

    Darwin