• State
  • Locked Locked
  • Replies 9 replies
  • Subscribers 33 subscribers
  • Views 314 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.

DP83869EVM: Connected back to back through RGMII

Esteve Cruanyes
Prodigy 30 points
Part Number: DP83869EVM
Other Parts Discussed in Thread: DP83869

Tool/software:

We are developing a media converter that has to be able run one port at 100 and the other at 1000 (also 100-100 and 1000-1000). The built-in media converter mode of the DP83869 does not work for us because it is either 100-100 or 1000-1000.

So I've come up with the idea of connecting 2 DP83869s via the RGMII interface and configure one as RGMII to Copper and the other RGMII to Fiber. First of all, is this a feasible solution?

I have two DP83869EVMs and I have connected them together like this:

RX_CTRL -> TX_CTRL

TX_CTRL -> RX_CTRL

RX_D3 -> TX_D3

TX_D3 -> RX_D3

RX_D2 -> TX_D2

TX_D2 -> RX_D2

RX_D1 -> TX_D1

TX_D1 -> RX_D1

RX_D0 -> TX_D0

TX_D0 -> RX_D0

RX_CLK -> GTX_CLK

GTX_CLK-> RX_CLK

GND -> GND

For the moment I am trying to make everithing rin at 1000, so straps are configured like this:

Copper board: OPMODE[2..0] = 000

Fiber board: OPMODE[2..0] = 001

Now I connect the copper board to a raspi running iperf server, the fiber board to a commercially available media converter and this media converter to another raspi running iperf client.

With this set-up, connection is not established, although some activity on the LEDs of both boards indicates that they are talking to each other, since the LEDs blink simultaneously.

I have read that in this scenario clocks need to be delayed. I have tried all combinations on both boards of register RGMII_CTRL[1..0] (RGMII Transmit Clock Delay and RGMII Receive Clock Delay) without success.

What might I be doing wrong?

Regards

  • 0
    TI__Mastermind 39795 points

    Hello,

    While the connection scheme may look okay, one concern I have is the idea of mixing speeds between port. For proper communication, everything on the chain must be at the same speed. If you mix speeds, what happens to the 1G packets when the communication drops to 100M? 

    Sincerely,

    Gerome

  • 0 in reply to Gerome Cacho
    Prodigy 30 points

    Hi Gerome,

    I am aware that this might be an issue. But, as I said, for the moment I am trying to run everything at 1000 and it does not work.

  • 0 in reply to Esteve Cruanyes
    TI__Mastermind 39795 points

    Hello,

    There are signal integrity concerns trying to get jumper wiring between EVMs due to the speed of the RGMII (125MHz DDR in 1G). You may probe these pins to verify RGMII VIH/VIL/Setup/Hold times are being met.

    Sincerely,

    Gerome

  • 0 in reply to Gerome Cacho
    Prodigy 30 points

    Greetings,

    I have probed with the oscilloscope lines GTXCLK, TXD0 and TXD1 of the FO board:

    I can see the 125 MHz clock, however, data lines have no activity at all (I have probed all of them TXD and RXD). This is weird since I can see the activity LEDs flashing on both boards synchronously, which to me, would mean there is some communication between them both...

    Thanks

  • 0 in reply to Esteve Cruanyes
    TI__Mastermind 39795 points

    Hello,

    It is odd that one of the datalines (*blue) is showing higher than GND. I would expect this to be rail to rail signaling based upon VDDIO. It is uncertain if this is due to probing or overall setup issue. Trying to jumper wire from EVM is strongly not recommended due to the parasitics that appear compared to copper traces.

    Sincerely,

    Gerome

  • 0 in reply to Gerome Cacho
    Prodigy 30 points

    Hi Gerome.

    I don't fully understand your first statement. Why is it weird that it is higher than GND? It is precisely at 3.3 V (VDDIO). Shouldn't this signal be changing between GND (0) and VDDIO (1)? To me, this is the confusing thing, that it is not switching at all. Actually, none of the data signals are; some are stuck at GND and others at VDDIO.

    Thank you for clarifying

    Regards

  • 0 in reply to Esteve Cruanyes
    TI__Mastermind 39795 points

    Hello,

    I find it odd that there is ringing that is nearly 1Vpp on the blue and red lines. I would suggest evaluating with a prototype that has trace connection between MAC and PHY to confirm that the jumper wires are not causing any issue.

    I would also be curious if the signal is the same on the MAC board without the jumper wire as the PHY is just acting as a HI-Z input.

    Sincerely,

    Gerome

  • 0 in reply to Gerome Cacho
    Prodigy 30 points

    Good morning,

    Not really sure about what you mean by MAC. I have just 2 PHYs connected back to back. One is in RGMII-Cu and the other is RGMII-FO. Which one do you consider the MAC.

    Thanks

  • 0 in reply to Esteve Cruanyes
    TI__Mastermind 39795 points

    Hello,

    If this is utilizing the EVMs, I would be concerned about the signal integrity from jumping between two boards. Please ensure that the communication is okay on the RX line of one board without jumper compared to TX line on the opposite board with jumper. Again, I would strongly suggest this be evaluated via traces compared to jumper wiring due to signal integrity concerns.

    Sincerely,

    Gerome