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.

TPS2372: PoE switch not recognizing PD device and providing power

Part Number: TPS2372
Other Parts Discussed in Thread: BOOST-PSEMTHR-007, TPS23880EVM-008, TPS23880,

I have built a smaller footprint version of the TPS2372 evaluation board, however when I connect to a POE switch the switch does not recoginze the device and will not provide power.

Connecting other PoE devices to the switch has no problems.

I have also connected a dumb poe injector (48V straight dc input) and the device works.

Autoclass is pulled low.

The schematic is based on the reference design. Currently I have nothing connected to the output, and the output capacitor is 100uF. I have also tried every combination of class resistors, AMPS_CTL, and autoclass headers.

I have multiple devices made with the same board and schematic, and some of them do actually work, however all the connections on the non-working devices seem good, and I have replaced the TPS2372 chip multiple times.

When viewed with oscilloscope, when the ethernet wire is inserted into the jack, it looks like there is an determining device class, however it looks like it is unsuccessful.

  • Hi Oleh,

    Could you repost the schematic? The link is broken.

    Additionally:
    1. Are you intended to implement Autoclass in your design?
    2. What PSE switch have you tried it with?

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer
  • This is the schematic:

    Yes, the intention is to have it work with autoclass, but I have headers on all the option pins so that I can reconfigure them.

    The PoE switch we use is this one: www.amazon.ca/.../B075G2PM17.

  • Hi Oleh,

    Nothing jumps out from your schematic, but Autoclass is a new feature for IEEE802.3bt, so a .bt PSE is required.

    Our TPS23880EVM-008 daughter card & BOOST-PSEMTHR-007 are orderable EVMs are a great place to start. You can order one & let me know if you are still having difficulty with classification.

    Additionally, if you could show me the rest of your schematic for your input, that would be helpful.

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer
  • the input schematic

    The class resistors are also included, and I tried using them as well, with autoclass disabled.

    Also, we have two working devices based on the same schematic and layout with that PoE injector.

  • Hi Oleh,

    Is this the switch you are using?: www.amazon.com/.../B075G2PM17

    Are you able to power on with your board configured to class 4, autoclass disabled, & plugging into the 30W port of that switch?

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer
  • That is the switch we are using. I tried that configuration with no success.

    The voltmeter output read around ~30V.

    Edited: Correct voltmeter reading

  • Setting it to Autoclass, and connecting the MPS_DUTY to VSS, and connecting it to the 60W Hi-PoE port sort of worked. Connecting a load (100 ohm resistor) to the output repeatedly initially just dropped the voltage. However at one point the voltage increased to 48V, and the PoE light came on. When the load was disconnected, the voltage stayed at 48V, and the PoE light on the switch stayed on. The load could be re-connected to the output with no ill effects afterwards.

    When the PD device was unplugged and plugged back in, the same problem as above appeared again (30V output, connecting load dropped voltage) and repeatedly connecting the load as with the previous semi successful attempt did not duplicate the earlier result.
  • Hi Oleh,

    Autoclass will not work with the 60W port on this switch because it is not compliant with the IEEE802.3bt standard. In order to test Autoclass successfully, you must have a .bt PSE that is configured to support Autoclass. ( TPS23880EVM-008 & BOOST-PSEMTHR-007)

    Regarding the output voltage: After classification, the PD goes into inrush current limit in order to charge the bulk capacitor, C4. During this period, no load should be connected so that all of the current can flow into the capacitor. When a load is attached during this period, the output voltage cannot be maintained because the bulk cap is not full charged. After that point, the load can be attached successfully.

    Regards,
    Tom Amlee
    PoE PD Apps Engineer

  • I have tried disabling autoclass and using the CLSA and CLSB resistors to set the class, with the same result. The output voltage stayed at 30V.

    The 100 ohm load was connected in parallel to C4 after the capacitor was charged, usually about 30 seconds after plugging in the PD into the PoE switch.

    Do you know what could be keeping the switch from increasing the voltage to 48V, even when there is no apparent load on the system?

    note: I have the MPS_DUTY pin connected to VSS directly, since when it was left floating, the output voltage was only about 13V at no load. Connecting it to VSS through the 53.6k Ohm resistor also keeps the output voltage at about 30V
  • Hi Oleh,

    Could you post a waveform of Vdd-VSS & RTN-VSS during & after the detection, class, & inrush startup sequence when the board is configured for class 4 & connected to a 30W port?

    Should be a time scale of 20ms/div to 100ms/div to see it all.

    - Tom

  • Interestingly when I went to measure the waveform, the negative terminal of the scope actually sparked when I connected it to VSS. I then checked the VSS to ground (from a local grounding post, tied to the ground pin on the plug of the PoE switch), and it was around 42V. When I shorted VSS to ground momentarily, VSS dropped to 0V, and the PoE switch began providing power and worked as expected.
  • Hi Tom,

    Actually I found something weird when I went to view the waveforms. When I connected the negative terminal of the scope there was a tiny spark. Investigating further, I found that relative to true ground, the VSS terminal was around ~40V.

    Also what happened when I connected the negative terminal of the scope to VSS sometimes the device would then power up, the PoE light on the switch would come on, and the device would then output 48V, and work as expected.

    Unplugging the device and plugging it back in would return it to the previously inoperative state, though after repeated attempts it would turn on sometimes, I would say about 10% of the time, where you have to unplug it, plug it back in, connect the negative terminal to VSS a bunch of times, and maybe on the tenth time of unplugging it and plugging it back in would it begin working.

    When the device is working, the voltage between VSS and GND is about 0

    Also, I used just a wire connected to GND, and used it to tap VCC repeatedly. With just a wire instead of the scope probe, after a few taps the system always turns on.

  • Hi Oleh,

    That spark you are seeing is the ground probe of one channel connected to RTN shorting to the ground probe of another channel connected to VSS while there is a voltage difference between the VSS & RTN grounds. The pass FET is intentionally not fully on (thus large Vds across pass FET) in order to maintain inrush current limit.

    When you short across the pass FET, RTN & VSS are brought to the same potential & 48V can be seen on the VDD-RTN output. The problem is that your board is unable to get out of inrush & be operational on its own to show this voltage on the output.

    Could you post the waveform of VDD-VSS & RTN-VSS during startup? Both ground probes should be connected to VSS, before powering on. Channel 1 connected to Vdd. Channel 2 connected to output ground (RTN).

    Regards,
    Tom Amlee
    PoE PD Apps Engineer
  • The spark happens when I connect ground from a grounding post (not anywhere on the circuit) to VSS, not when I connect VSS and RTN. 

    the waveform below explains what I mean. Blue is VSS, yellow is VDD, both reference to the ground post. Purple is math of VDD-VSS, and everything is offset to down so that it fits on screen. Sorry for the poor quality image

  • Hi Oleh,

    All measurements (RTN to VSS & VDD to VSS) should be referenced to VSS. For safety, nothing should be connected to an earthed grounding post.

    Could you repost as a jpeg? Due to our network security policy, I am unable to access your cloud storage link.

    Thanks,
    Tom Amlee
    PoE PD Apps Engineer

  • Here is the picture that's in the link

  • Hi Oleh,

    The detection, class, & power up are not compliant, so there is no guarantee your PD will be recognized. I recommend testing your PD with our IEEE802.3bt compliant TPS23880 EVM( TPS23880EVM-008 with BOOST-PSEMTHR-007) & seeing if your PD is able to be recognized.

    For more information on compliance with the IEEE 802.3bt standard & autoclass, see our training series here: www.ti.com/.../support-training.html

    Regards,
    Tom Amlee
  • Do you know why would the PD would become recognized after repeatedly shorting the VSS to a grounding post?
  • Oleh,

    I am not sure. Earth grounding VSS may have damaged the PSE or PD.

    Regards,
    Tom Amlee
    PoE PD Apps Engineer
  • The behaviour of this board hasn't changed, and the switch works with the tps2372 evaluation board as before, so I don't think any of the parts were damaged.
  • Here is a more clear picture of what is going on: 

    I have also edited the previous post. 

    And this is the same switch with a working board:

    I have also switched the tps2372 chips between  the working and not working boards, and the previously working board, with the chip from the previously non working board, continued to work as expected, and the previously non working board, with the chip from the working board continued to not work. 

  • Hi Oleh,

    Could you clarify what these channels are on these waveforms images?

    It appears that you have found that the ICs are fine, but there are differences between the boards.

    You can follow the attached procedure with each of your boards & find if the detection & class currents are within compliant ranges. If your detection current is outside the specified range, then size your detection resistor, accordingly. Additionally, detection capacitance (input caps across Vdd-Vss) should not exceed 0.12uF.

    PSIL006A_test.docx

  • The channels are VDD and VSS, with Yellow being VSS

  • Hi Oleh,

    Just a quick comment, both signals should have their ground probes attached to Vss. The yellow channel should be RTN measured in reference to VSS. You want to be able to see the pass FET between RTN-VSS maintain the inrush current limit.

    It should look similar to the figure below:

  • Unfortunately it seems that my osiliscope's probes aren't floating, and so connecting the negative terminal of the probe to VSS actually brings it down to 0, causing the VDD-VSS voltage to be always >40V. Having one probe on VDD, and one on VSS is sort of a workaround for this, so that I can see the real difference between the two.
  • Oleh,

    Are you unable to connect both ground connections of the 2 oscilloscope channels to VSS & see that they are referenced to the same voltage? The oscilloscope we use has a shared ground plane between the channels.

    We are measuring RTN (not VSS), because we are measuring the Vds of the pass FET between RTN & VSS.

    We want to make the measurements at the test points below:

    The 2nd channel, RTN referenced to VSS, will eventually be 0V when the hotswap FET turns on after inrush. Additionally, PG will go high after successful inrush.

    Regards,

    Tom Amlee

    PoE PD Apps Engineer

  • Yes, the osilliscope I we have does not have floating leads, so CH1 - and CH2 - are always at 0V, and when connected to VSS cause a current from VSS to CH1- or CH2-

    That is why I had to have CH1+ connected to VSS, and CH2+ connected to VDD

  • Hi Oleh,


    Regardless of the oscilloscope issue, the main issue is finding a compliant PSE & properly configuring your PD to use autoclass. My recommendation is this:

    1. Order TPS2372-4EVM-006 & configure for autoclass
    2. Test the PD EVM with a PSE EVM compliant with the IEEE802.3bt standard (required for autoclass)
    1. Make the necessary changes to your design to be similar to how TPS2372-4EVM-006 is configured

    If your issue has been resolved, feel free to click the green button. If it is still in works, feel free to reply. Otherwise, if I do not hear back from you in a week, I will be closing this thread.

    Regards,
    Tom Amlee
    PoE PD Apps Engineer