LT4275 Cannot deliver power and ethernet signal together

hi, I am working design an LTPoE++ PD which require 90W power. I am using LT4275 as PD controller and ENC28J60 as ethernet controller. The goal is making a pi compute module(raspberry pi CM3) powered and connected to the internet. right now, the ENC28J60 can make it connect to the internet without using LT4275 and PoE feature, and LT4275 can support the power with no problem. However, when put them together, the ethernet part stops working. I check the backward compatibility and known it not work with other types of PSE, so we get a 90W PoE Injector which supports bt, but it still not working. I also try to change CLS and CLS++ resistors to reduce the power requirement, it also not working. do anyone can give a help?

  • 0
    •  Analog Employees 
    on Sep 24, 2018 8:03 PM

    Hi,

    To understand your problem; ethernet data stops working with the LT4275. Are you using an LT4275 demo board, or is this a custom design? Same question with the ENC28J60, is this a custom design? A schematic and block diagram would be very helpful.

    Does the data pass through the LT4275 board while it is unpowered, and if so are you powering a load during this test? Not all ethernet magnetics are rated for PoE, especially 90W PoE. LT4275A set to 90W should be used with a LTPoE++ 90W PSE. What 90W PoE Injector are you using to test this setup?

    Best Regards,

    Eric

  • 0
    •  Analog Employees 
    on Sep 25, 2018 10:47 PM in reply to weizhen1883

    Hi,

    I have a few notes about your schematic, but nothing that would prevent data transfer.

    • PHY is not isolated from the PD and cable ground, as is usually recommended. The cable needs to be isolated from all accessible external conductors and meet an 1.5kV isolation requirement for IEEE compliance.
    • 0826-1X1T-GH-F is only rated for IEEE 802.3at (25.5W) and not LTPoE++.
    • The center-tap TRCT2 should be floating, not connected C40 as per ENC28J60's datasheet.
    • D5 and D3 are rated for 600V instead of 100V. The power loss in these bridges at LTPoE++ 90W will be very high.
    • Q33 should be changed, do not use SSM3K341. This is a 60V FET, while a 100V rating is recommended. This FET is used as a hot swap switch so it has an SOA requirement. SSM3K341R has insufficient SOA. 
    • Remove C52 and change C53 to 100nF to maintain total capacitance at 100nF.
    • PWRGD and T2P pins should be used by the application.
    • What is the purpose of D8?

    I have not tested this injector, but it states LTPoE++ 90W so it should work. Have you tried getting data through this injector while it is connected but unpowered?

    Your schematic doesn't show how +48V/9?? is used to power the ENC28J60. What converter are you using to go from 48V to 3.3V? I imagine you also have some way of locally powering the PHY. Can you share this portion of the schematic? Can you remove D8, place a 3k resistor across C54 as dummy load and see if data works with PoE but while not using PoE to power the ENC28J60?

    Best Regards,

    Eric

  • 0
    •  Analog Employees 
    on Sep 27, 2018 6:46 PM in reply to weizhen1883

    Hi,

    PoE Power Sourcing Equipment, PSEs, often provide 52V - 57V, and there is allowed 10V or so of voltage drop across a 100 meter cable. The expected PD operating range is commonly 37V to 57V. A8498 is only rated for 50V absolute max input, so it is not suitable for this application. The additional 220uF input capacitor on this page should be included in your inrush calculations, so C53 should be doubled to 100nF.

    Can you confirm that you are getting a clean (low ripple) 5V rail from A8498 and 3.3V rail from AP2112 while operating with PoE?

    Best Regards,

    Eric