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.

TPS23756: Question on Class Regulator

Part Number: TPS23756
Other Parts Discussed in Thread: , TPS23752, TL431

Hi TI,

What is Class Regulator mentioned in page-16 of TPS2375x_datasheet? There are detection, classification... in PoE handshake between PSE and PD. Is the Class Regulator referred to to the Classification?

Regards,
Kok Khuan

  • Hi Kok,

    Thanks for reaching out!

    Are you asking the classification in the following paragraph?

    From my understanding: when APD is high, the classification regulator will be disabled and hotswap could not be turned on. But when PPD is higher than PPD2 voltage, the classification regulator still could be turned on. This may be useful for a PoE and adapter oring circuit and ensure hitless transition. Or just directly, PPD can enable classification regulator at a different VDD-VSS voltage (other than 15.5 - 20.5V range).

    While the function block diagram does not show detailed information. I will see if I could run some tests on EVM.

    Best regards,

    Diang

  • Hi Diang,

    Yes.
    I mocked up circuitry below and found out that the class regulator (classification) still works even PPD is connected to VSS through a 10kOhm resistor. This contradicts to what TI mentions that classification regulator will be disabled when APD is high and PPD is lower than PPD2. Do I understand correctly?

    Regards,
    KK

  • Hi KK,

    Thanks for your information.

    From datasheet I would think classification should be disabled when APD-RTN = H and PPD-VSS= L.

    I will run some tests in EVM and get back to you in 2-3 business days. 

    Best regards ,

    Diang

  • Hi KK,

    I did a test using TPS23756EVM. From the test results we could see PPD = high can turn on Classification when APD = high. 

    1.

    Vadapter = 20 V

    VDD-VSS = ~20 V

    APD = 4.2V

    PPD = 0 V

    CLS = 0V

    2. 

    Vadapter = 20 V

    VDD-VSS = ~20 V

    APD = 4.2V

    PPD = 20 V

    CLS = 2.5 V

    Best regards,

    Diang

  • Hi Diang,

    We did experiment here too and the result is same as yours. Just could not understand why PoE input can supply almost immediately when Aux input is Off, with no classification measured when RJ45 is disconnected and then connected on the board at Aux input ON initially?
    APD high, PPD high:

    APD high, PPD low:

    Another question is why TPS23756 shut down the converter although APD is triggered low fast enough to transition from Aux input to PoE input?

    We did an experiment where getting external 45V supply to the PoE input instead provides the same output dip as below.

    Regards,
    KK

  • Hi KK,

    Thanks for your test. We will reply you in 2-5 business days. 

    Best regards,

    Diang

  • Hi KK,

    Thanks for the tests.

    When PPD is high than PPD1, the PD can turn on the internal hotswap. This allows a low input voltage between VDD-VSS.  

    When PPD is high than PPD2, the PD can turn on Classification when APD = high. 

    When APD is used. APD = high will make the DC adapter higher priority than PoE input by turning off the internal hotswap.

    When APD is high -> low, and with a valid VDD-VSS PoE input, the internal hostwap is off -> on. During internal hotswap turning on, the PD's PWM controller will be hold off.   

    Best regards,

    Diang

  • Hi Diang,

    PPD high mentioned in my both question above means PPD > PPD2.

    Does this mean that TPS23756 is not capable for seamless failover from Aux input to PoE input when Aux input is On -> Off because the PD's PWM controller would be hold off during internal hotswap turning on? As per my understanding, when Aux input is On, the internal hotswap is hold off and it is just the PD's PWM controller does not turn on fast enough when internal hotswap is turning on.

    For my previous first question, PoE input would need around 3s for detection, classification, inrush... before being a main supply to PD. However, with or without the classification voltage measured on Rclass, we do not see the 3s. Shouldn't the PoE input go through classification before being the main supply to PD?

    Regards,
    Kok Khuan

  • Hi KK,

    I pervious tested TPS23752 for Adapter (at APD) -> PoE transition. In that case, the PWM controller was turned off when APD is H->L and PoE is turning on its internal hotswap. I saw your test showed a similar behavior that Vout dropped during that period.  

    For the PSE-PD power on time, it will need detection -> classification -> turn on + inrush. If PPD pin is used, as well as PPD is H (keep APD = L), the internal hotswap can be turned on. If you are using a passive PoE injector, it can directly give a voltage to PD to turn on the internal hotswap (UVLO is reached or PPD is met) without detection & classification. 

    Best regards,

    Diang

  • Hi Diang,

    Understood from your and my testing that the PWM controller will be moving from On -> Off -> On when the input is transitioned from Aux to PoE. Please correct me if TPS23756 does not behave in this way.

    For 2nd question, I don't think we are using passive PoE injector because when we blocked the PoE detection, where Aux input is connected to PoE input, PoE input requires around 3s to recover to around 48V when the input is transitioned from Aux to PoE. However, when we block classification (no voltage measured at Rclass) by having PPD pin connected to VSS through a 10kOhm resistor, the output does not require 3s to recover. Instead, it requires around 25ms to recover to nominal level. Do you know why?

    Regards,
    Kok Khuan

  • Hi KK,

    Thanks for your reply. We will reply you in 2-5 business days. 

    Best regards,

    Diang

  • Hi Diang,

    Do you have the answers for the 2 questions?

    Regards,
    KK

  • Hi KK,

    With the holidays this week, many device experts are currently out of the office. They will look into this and provide you a response when they return. Please expect some delay accordingly.

    Thanks,
    Field

  • Hi KK,

    Thanks for your patience. 

    "Understood from your and my testing that the PWM controller will be moving from On -> Off -> On when the input is transitioned from Aux to PoE. Please correct me if TPS23756 does not behave in this way."

    Yes, it will work as this way as power source is Aux -> PoE and APD is H->L.

    "For 2nd question, I don't think we are using passive PoE injector because when we blocked the PoE detection, where Aux input is connected to PoE input, PoE input requires around 3s to recover to around 48V when the input is transitioned from Aux to PoE. However, when we block classification (no voltage measured at Rclass) by having PPD pin connected to VSS through a 10kOhm resistor, the output does not require 3s to recover. Instead, it requires around 25ms to recover to nominal level. Do you know why?"

    Could I assume the test shows:

    - When PPD is used and able to set H, the AUX -> PoE transition (meanwhile APD is H->L) makes ~3s interrupt on Vout  

    - When PPD is connected to VSS and make PPD = VSS, the AUX -> PoE transition (meanwhile APD is H->L) makes ~25ms interrupt on Vout 

    The below diagram shows when AUX is there and hotswap disconnects the PoE. If when Vpoe is low (before detection and classification), but C_det is high since hotswap body diode is forward bias when AUX is used. It may turn on hotswap instantly when APD drops considering VDD-VSS is beyond UVLO. Then Vpoe needs detection and classification while C_det and C_buck is discharging by the load. Then you probably could see Vout recover soon but later it drops due to Vpoe needs to be able do detection and classification when C_det drops to ~0 V. 

    If with a passive PoE injection, there could be no second drop consider there is no need for detection and classification.

    Best regards,

    Diang

  • Thank you, Diang, for the answer on question-1.

    For question-2, below test shows:
    1. When PPD is used and able to set H, the AUX -> PoE transition (meanwhile APD is H->L) makes ~25ms interrupt on Vout

    2. When PPD is connected to VSS and make PPD = VSS, the AUX -> PoE transition (meanwhile APD is H->L) makes ~25ms interrupt on Vout

    Above 2 tests are done with ethernet LAN cable unplugged and plugged into the board RJ45 connector before AUX -> PoE transition is enabled.
    In 1st test, at PPD > PPD2, class regulator operation is finished when the ethernet cable is plugged into the RJ45 connector before AUX -> PoE transition is enabled. This is why it makes sense that only ~25ms interrupt on Vout happens, where detection, classification are done before AUX -> PoE transition.

    In 2nd test, at PPD = VSS, class regulator operation is disabled when the ethernet cable is plugged into the RJ45 connector before AUX -> PoE transition is enabled. Why does only ~25ms interrupt on Vout happen because classification are not done before AUX -> PoE transition?

    An experiment was done on the PSE where it does not provide supply to PD when Rdet is disabled at PD side. As I understand, PSE will not provide supply to PD if classification is not done too.

    My circuit is as below.

    Regards,
    KK

  • Hi KK,

    Thanks for your update. 

    According to your correction, no matter PPD is enabled or disabled the AUX -> PoE transition (meanwhile APD is H->L) always makes ~25ms interrupt on Vout.

    In the circuit you drawing, did you make the APD refer to VSS? Normally, APD should be refer to RTN and PPD should be refer to VSS. I feel if APD is set in the location you shows, there could be failed during detection consider R divider current will cause detection current fails.

    Is your target making both PoE->AUX change and AUX->PoE change no Vout drop? If so, I feel you can oring AUX and PoE at the RTN ground and do not use the APD pin. In such case, these two sources will be both on and they are in redundancy. However there is no priority can be set. The higher voltage one will source the load.

    Best regards,

    Diang

  • Hi Diang,

    "According to your correction, no matter PPD is enabled or disabled the AUX -> PoE transition (meanwhile APD is H->L) always makes ~25ms interrupt on Vout."
    Yes, this is right.

    For the circuit that I provided, it is my bad that it is incorrect. It should be more towards Figure-33 as captured below.

    Regards,
    Kok Khuan

  • Hi KK,

    Sorry that the PoE team has traveling and multiple works in list recently. Please expect a delay on the E2E reply. 

    Best regards,

    Diang

  • Thanks for informing, Diang.

  • Hi KK,

    Sorry for the inconvenience and thanks for your patience.

    Best regards,

    Diang 

  • Hi KK,

    Thanks for your patience.

    If your target is making both PoE->AUX change & AUX->PoE change no Vout drop at full load, there are two general ways you can do: 

    1. Put the AUX on the Option3: It is easier to implant in circuit and easier to set adapter priority (you can make the TL431 resistor divider ratio change when AUX is inserted to lower the PoE's output voltage). Maintain power signature can be realized by enable a dummy load for PoE when AUX is inserted too. While the adapter needs to have 1500-Vrms isolation to meet the standard.

    2. Put the AUX on the Option2: it is more complex in circuit and the you could not guarantee the AUX has a higher priority than PoE. You cannot use APD pin in this case. A dummy load is also needed to main power signature. 

    Best regards,

    Diang

  • Hi Diang,

    Recapped my question:
    At PPD = VSS, class regulator operation is disabled when the ethernet cable is plugged into the RJ45 connector before AUX -> PoE transition is enabled. Why does only ~25ms instead of 3s interrupt on Vout happen since classification are not done before AUX -> PoE transition?

    Regards,
    KK

  • Hi KK,

    Thanks for your follow-ups.

    When adapter is removed and APD voltage drops below threshold, the handshake won't be disabled anymore. If the Vout has a relatively large capacitor, and the PSE (if not a poe injector is used) happens to sending detection signal in time, it could give a relatively short Vout interrupt time.

    Could you measure the VDD-VSS, ADP-RTN and Vout-GND voltages during your PoE & adapter transient? So we can directly see the behavior.

    Best regards,

    Diang