• State TI Thinks Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 1 answer
  • Subscribers 33 subscribers
  • Views 5911 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.

DP83822I: Detecting loss of signal in SFP applications

t1moh
Prodigy 50 points
Part Number: DP83822I

We are Implementing a 100Base-FX 100Base-TX interface using a COTS SFP module.
In 100Base-TX application we use a SFP module that has a built-in TX/FX media converter.

While the interface works perfectly with both modules while the media is connected, we have been unable to detect loss of signal, i.e cable disconnected, with the TX module.
We have SFP LOS output connected to LED_1 of the DP83822 PHY. The voltage on the pin swings from 0.17V to 3.3V(we use 3.3V supply for SFP and the PHY) when we disconnect the RJ45 from the SFP-TX module. Reading the specs, there should be not problem with the signal levels.
We have also ensured that the PHY COL and RX_ER straps are correct reading the 'Strap Latch-In Registers' and that '100Base-FX Signal Detect Polarity' &  'LED_1 GPIO Configuration' are correct.
Reading BMSR and/or PHYSTS while connect and disconnect the media does not result in any change in either of the registers. Our assumption is that the PHY ignores the LOS as it sees itself synchronized to the media converter.

Is there a way to detect loss of signal reliably in SFP applications using the PHY register interface?

  • 0
    TI__Mastermind 26320 points
    Kindly share the strap settings you have using for Signal Detect

    RX_ER strap mode 4, SD_DIS = 0 i.e. SD_FX detection is enabled.

    Regards,
    Geet
  • 0 in reply to Geet Modi
    Prodigy 50 points

    Here is what we read from PHY register '0x0467 Strap Latch-In Register #1'

    Register 0x0467:08C3

    According to the spec SNLS505A – AUGUST 2016 we have:
    RX_ER: MODE=1, i.e RGMII_EN=0 and SD_DIS=0
    COL: MODE=3. i.e FX_EN=1 and PHY_AD0=1

    We took some educated guesses about the internal structure of the register file inside the PHY and discovered that the state of LED_1 is readable from undocumented register 0x046A as bit 8.
    We'll use this in our prototype to detect 'media disconnected' for now. Reading the register we get some false negatives on bit 8, but we figured out how to filter them out reliably.
    This method works with the PHY chip revision we have, but we may have to route LOS separately to the MCU unless there is a officially supported way to detect LOS through the PHY register interface. 

  • 0 in reply to t1moh
    TI__Mastermind 26320 points
    Hi,

    Can you try moving it to RX_ER strap to Mode 4 ?

    Regards,
    Geet
  • 0 in reply to Geet Modi
    Prodigy 50 points

    Just to clarify.

    LOS detection works fine reading the BMSR bit 2(Link Status) if we use a 100Base-FX module. The problem occurs if we replace it with a 100Base-TX module that has a built-in media converter.
    Connecting/Disconnecting the cable does not change the value of the Link Status bit. Removing the entire 100Base-TX module clears the Link Status bit as expected.

    We cannot see how changing RX_ER strap to Mode 4 would resolve the issue. This seems like a bug in the DP83822.
    We can make an allowance for this problem by resorting to reading the bit 8 of the undocumented register 0x046A if a 100Base-TX module is present. With 100Base-FX we read just the BMSR bit 2.

  • 0 in reply to t1moh
    TI__Mastermind 26320 points
    Hi,

    Your original post mentioned that "SFP LOS output connected to LED_1 of the DP83822 PHY is not getting detected.

    Reading BMSR and/or PHYSTS while connect and disconnect the media does not result in any change in either of the registers. Our assumption is that the PHY ignores the LOS as it sees itself synchronized to the media converter"

    However In your last post you are mentioning its working fine.

    I am bit confused .... I assumed you are using DP83822 in 100-FX mode and having SFP module connected to it. If it's not matches with your usecase, kindly provide a block diagram to align and further debug of your issue.


    Regards,
    Geet
  • 0 in reply to Geet Modi
    Prodigy 50 points
    SFP LOS output connected to LED_1 of the DP83822 PHY is not getting detected with 100-Base-TX(copper SFP module) only.
    These modules have a build-in media converter that converts the PHY interface to be compatible with fiber transceiver interface.

    There is no problem with 100-Base-FX(fiber) module "LOS detection works fine reading the BMSR bit 2(Link Status) if we use a 100Base-FX module".

    Our development HW is DP83822 EVM with SFP HW populated.
    If you are unfamiliar with SFP, here is a link to SNIA document database. www.snia.org/.../specifications
  • 0 in reply to t1moh
    TI__Mastermind 26320 points
    You mean, when you are using DP83822 in 100Base-TX mode, SFP LOS signal is not been detected by DP83822 and update Link Status accordingly ?

    Regards,
    Geet
  • 0 in reply to Geet Modi
    Prodigy 50 points

    We use the DP83822 in FX mode with both SFP modules. That's what the built-in media converter in 100Base-TX is all about. It makes the module to look like an FX module to the DP83822.

    This arrangement allows us to use the same HW for both modules and even replace the modules on-the-fly as is the intention of the SFP concept.
    We do not seem to be getting anywhere with this discussion. We'll sort this problem out in out target HW.

    Thanks for your efforts.