• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 33 subscribers
  • Views 662 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.

DP83867IS: Required clock spec and tree structure for dealing with Jumbo packets

Takashi Onawa
Genius 13200 points
Part Number: DP83867IS

Hi Team,

I'm understanding DP83867 can support Jumbo packets without any register changes.
but I think it should need more accurate clock sync to avoid FIFO overflow and underflow on Phy and Mac.
Fundamentally, IEEE802.3 defines +/-100ppm clock tolerance for ref clock in a system. But this requirement only allows up to 1518byte packet communication with max IFG(Inter frame gap). In order to deal with Jumbo packet such as 9k byte, more accurate clock source should be required.   

Could you tell me the following points to design safety system?
1. How much accurate clock does it need for 9k packets transmission safely?
2. Are there any figures showing internal clock architecture of DP83867?
3. Which clock should be used as system ref clock? Or, is separated clock architecture acceptable(Using different clock sources on Mac and Phy)? 

Regards,
Takashi Onawa

  • 0
    TI__Mastermind 18425 points
    Hello Onawa-san,

    For our internal Jumbo packet testing, we used a crystal with +/-50ppm frequency tolerance. This test was performed as a spot check and not for characterization. The internal clock architecture is confidential IP and cannot be shared with customer. In SGMII and RGMII mode separate clocks for MAC and PHY are acceptable. It is also acceptable to provide clock from MAC to PHY, if the MAC can support this feature.

    -Regards,
    Aniruddha
  • 0 in reply to Aniruddha Khadye
    TI__Genius 13200 points
    Hi Aniruddha-san,

    OK, I understood about the test condition to check Jumbo packet communication and disclosure polity of clock IP.

    Can you estimate maximum data length with +/-50ppm tolerance ref clocks and connected by RGMII?
    I think it should be depended to internal FIFO size, internal clock structure and clock tolerance but we can not calculate it due to our disclosure polity.

    Let us share if you have any app note on this.

    Regards,
    Takashi Onawa
  • 0 in reply to Takashi Onawa
    TI__Mastermind 18425 points
    Hello Onawa-san,

    At +/-50ppm, packet size up to 13kbyte should be fine. Again, these results are from lab measurement and I would recommend customers to perform their own testing.

    -Regards,
    Aniruddha