• Resolved

DP83825I: Base100-TX half duplex transmit only

Prodigy 70 points

Replies: 9

Views: 66

Part Number: DP83825I

Dear support,

We are developing a specific streaming protocol and we would like to use an ethernet phy as a transceiver. We have a limitation by hardware, we can use only a single differential pair what unsupported by 100base-tx.

Our solution would be the following:
-connect the phys RX pins together, terminate TX pins.
-control the data flow by manual MDIX configuration
-disable autonegotion

We made a board for testing and the receiver is working fine, the data was captured properly. But the transmitter part does not send any data (except idle bytes) if there is no active link.
With only one differential pair, only one of the phy will have detected link. The transmitter should send data even if there is no link.

My question is,
-Can I set some register to send data even if there is no link?
-Can I imitate a link?
-Is the link mandatory for transmission? For the receiver is understandable because of the scrambler, but for the transmitter why?

  • Hi Peter,

    What is the register configuration you are using to disable auto-negotiation? If there is no link, you can try using PRBS mode to generate and transmit random data. 

    In our compliance testing we also terminate our TX and RX lines in forced 100BASE-TX and are still able to generate signals 

    You can refer to our datasheet on how to enable PRBS or review our app note on how to generate 100BASE-TX compliance signals 

    Also, just for my understanding out of curiosity, how is this new protocol going to be different than single pair ethernet?

    Thanks,

    Cecilia

  • In reply to Cecilia Reyes:

    Included my register configuration and a schematic about the setup:
    drive.google.com/.../1rjNzk1rN3SZl0B8WSa1nW7wpeEGMVgL4

    >If there is no link, you can try using PRBS mode to generate and transmit random data.
    I could, but the device is able to send data if there is link so i think there is no need to test with PRBS

    >In our compliance testing we also terminate our TX and RX lines in forced 100BASE-TX and are still able to generate signals
    The problem is, we are not only using the 100base-TX in forced mode, we are using it with only one differential pair. To answer your next question as well, as I know the 100base-tx is full duplex only so there isn't any single pair ethernet by standard for 100Mbit. Our protocol will handle the data directions as well and negates the possibility to collision occur.

    I have shown two test in my included schematic:
    Test a) would be the working solution. Thy physical RX pins connected together and the data direction is controlled by manual mdix on both side. During the test, there was a link on the receiver side, and wasn't on the transmit side. Also, the transmitter didn't transmit any data except idle.

    Test b) On the transmitter side, the RX and TX was shorted together. Thats why the transmitter detected the link and started to transmit valid data.

    So, the link state is a requirement for the transmission. Its specified by the standard and tested on my device as well.
    Can I bypass the link state check in the phy or can I force the link on somehow? What is the requirement for the link to turn on? Its specified by the manufacturer, not the standard.

  • In reply to Peter Pazmandi:

    Hi Peter,

    The link to the google drive is not working unfortunately. Could you please share the files through the thread here?

    Thanks,

    Cecilia

  • In reply to Cecilia Reyes:

  • In reply to Peter Pazmandi:

    Hi Peter,

    I spoke to my team and it turns out there is a register setting to force link. I will need to confirm in the datasheet or if there is another setting not mentioned. I will let you know my findings in 1-2 days. 

    Thanks,

    Cecilia

  • In reply to Cecilia Reyes:

    Hi Cecilia!

    >I spoke to my team and it turns out there is a register setting to force link.
    Thank you! Thats the one what I will need. Please inform me if you know the specific register address and its details.

    >I will need to confirm in the datasheet or if there is another setting not mentioned.
    As I saw it wasn't mentioned in the datasheet. I think the register itself undocumented.

    Thanks,
    Peter

  • In reply to Peter Pazmandi:

    Hi Peter that is correct it looks like it is not in the datasheet. Let me reach out to our internal validation team to see if they can share the value with me.

    Thanks,

    Cecilia

  • In reply to Cecilia Reyes:

    Hi Peter,

    Bit 13 of register 0x40 is a RW. I believe setting to 1 should force the link.

    Let me know if that works for you.

    Thanks,

    Cecilia

  • In reply to Cecilia Reyes:

    Hi Cecilia!

    >Bit 13 of register 0x40 is a RW. I believe setting to 1 should force the link.
    That was the solution, its working.

    Thank you for the help.

    Thanks,
    Peter