• State
  • Locked Locked
  • Replies 8 replies
  • Subscribers 34 subscribers
  • Views 288 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.

DP83867IR: Delay between SFD and MDI

Eric Opitz
Prodigy 70 points
Part Number: DP83867IR

Tool/software:

Dear TI Team,

I'm implementing a time synchronization scheme using the DP83867IR. My aim is to measure and correct for the transmission delay from device A to device B. The DP83867IR is capable of detecting the SFD in the transmitted packet and provide it on a GPIO. To measure the line delay, I want to make the following measurement:

T1 - Device A Tx SFD

T2 - Device B Rx SFD

T3 - Device B Tx SFD

T4 - Device A Rx SFD

Tline = ((T4 - T1) - (T3 - T2)) / 2.

My problem is that the Tx SFD is detected in the PHYs PCS and there is an additional time before the first byte is transmitted on the wire (MDI). Similarly, there is some unknown time between MDI and Rx SFD detection.

If both devices are DP83867IRs then I could consider those unknown delays as part of the line delay. However, If the PHYs are different and have different internal latencies, this would lead to a poor synchronization.  

The only values I can find in the datasheet are the Tx/Rx Latencies (88ns and 288ns), but those are far greater than the delay that I have measured between device A Tx SFD and device B Rx SFD (e.g. with short cable 248ns).

How do I find out the internal delays? Can you provide a value?

I'm using the RGMII Interface and 1000BASE-T.

Best Regards,

Eric Opitz

  • 0
    TI__Mastermind 25955 points

    Hi Eric,

    I will check it with the team and get back to you on this later this week.

    --
    Regards,

    Hillman Lin

  • 0 in reply to Hillman Lin
    TI__Mastermind 25955 points

    Hi Eric,

    I check with the team internally. When the PHY received the signal on the MDI lines, the Rx SFD will trigger simultaneously. When the PHY received the signal on the MAC side, the PHY will trigger Tx SFD simultaneously. The main delay between Rx SFD and Tx SFD would be the latency between the MAC to MDI or MDI to MAC.

    --

    Regards,

    Hillman Lin

  • 0
    Prodigy 70 points

    Hi Hillman,

    Unfortnuately this does not answer my questions: What are the internal delays? Can you provide a value?

    If I understand you correctly by MAC side you mean the RGMII input or output. You are saying, there is no delay between the RGMII input and the SFD detection. But the DP83867IRs datasheet clearly states that there is a difference between the RGMIIs TX_CTRL signal and the SFD signal - why else should I use the SFD for accurate timestamping? To test this, I have measured the time between the RGMII signals (Device A TX_CTRL to Device B RX_CTRL): 392ns and the time between the SFDs (device A Tx SFD to device B Rx SFD): 248ns. So clearly, there is a delay between RGMII input/output and the SFDs.

    Furthermore, you state that in the receiving PHY there is no delay between the signal arriving on the MDI line and SFD detection. I find this hard to believe as the signal on the MDI line is encoded and it would be difficult to detect the SFD byte in the encoded signal.

    Therefore I ask again: What is the delay between Tx SFD and the MDI line? And what is the delay between the MDI line and the Rx SFD?

    Best Regards,

    Eric Opitz

  • 0 in reply to Eric Opitz
    TI__Mastermind 25955 points

    Hi Eric,

    I will check it with the team on the measurement again.

    Can you let me know what is the setup when you are performing this test?

    --

    Regards,

    Hillman Lin

  • 0
    Prodigy 70 points

    Hello Hillman,

    Devices A and B both use a DP83867IR PHY. Both are using the RGMII interface. There are communicating in 1000BASE-T mode. The output pins are configured to output the Rx and Tx SFDs on two GPIOs (configuration via registers 0x172, 0x134, and 0xE9). Device A is explicitly configured as clock master.

    The cable has a length of about 60cm, so I would expect about 2ns delay for the cable.

    Do you need any more information on the test setup?

    Kind Regards,

    Eric Opitz

  • 0 in reply to Eric Opitz
    TI__Mastermind 25955 points

    Hi Eric,

    We bring up this topic with the whole team internally today. We have confirm that there is no variation in delay between the MAC signal to TX SFD GPIO and MDI signal to RX SFD GPIO between the parts. The only possible variation between parts to parts are the latency between MDI signal to MAC signal through the due to the encoding process. Hopefully this clarify your concern.

    --

    Regards,

    Hillman Lin

  • 0 in reply to Hillman Lin
    TI__Intellectual 1510 points

    Hi Hilman, do you have any feedback here?

    Thanks and regadrs

    Alex

  • 0 in reply to AlexW
    TI__Mastermind 25955 points

    Hi Alex,

    We have confirm that there is no variation in delay between the MAC signal to TX SFD GPIO and MDI signal to RX SFD GPIO between the parts. The only possible variation between parts to parts are the latency between MDI signal to MAC signal through the due to the encoding process. Hopefully this clarify your concern.

    --

    Regards,

    Hillman  Lin