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.

DP83826E: No contact with ethernet port

Part Number: DP83826E

Hello,

We're trying to get a connection with Ethernet but we can't seem to get it to work.

We have been able to get a connection to the PHY but we aren't able to get a connection to the Ethernet port. 

The PHY is working in BASIC mode, RMII Master mode & Auto Negotiation Enable.

Under here you can see the schematic.

Does anyone know what we might be doing wrong?

  • Hello,

    Are you able to get access to the PHY's registers? Trying to understand if this is a MDI issue or a MII issue. It appears you are using a mag-jack. The center tap of the magjack appears to have the incorrect capacitors. Do you see some signals toggling on MDI as well as on RX pins?

    Sincerely,

    Gerome

  • Hello Gerome,

    We are able to read out the registers but the auto-negotiation doesn't seem to work.

    Yes, we are using a mag-jack.

    I changed the schematic to be up to date with some modifications we did but I see i made a mistake while doing this. I updated the schematic. Is this what you were talking about for the wrong capacitors? Or do you mean that the value is wrong?

    On de ETH_MDIO net i do see a couple of signals at starting up but nothing after that. On the ETH_RXD0 and ETH_RXD1 I don't see any signals.

  • Hello,

    It appears that PHY does not have link as evidenced by Reg 0x1 value. Is the ethernet cable plugged in? What is the link partner?

    In addition, I believe your driver is setup for DP83848. Could you double check that MAC portion is alive by seeing signal on RX_CLK? 

    Sincerely,

    Gerome

  • Hello Gerome,

    The ethernet cable is plugged in. It is connected to an prosafe 24 port gigabit switch with is connected to our network.

    The driver was originally meant for the DP83848 but my colleague this shouldn't matter to much because the registers from both PHY's should be the same. We are also  not changing much in the registers.

    We do have a clock signal.

  • Hello,

    Can you try testing throughput when utilizing feedback?

    I would like to try MII feedback of PHY to isolate the MII section. This is done by setting Reg 0x0[14].

    Sincerely,

    Gerome

  • Hello Gerome,

    We were trying some other things that sadly didn't work and eventually tried the feedback. We are not getting any data back with this enabled either. Then it seems to be a problem in the communication between the ESP32 and the PHY right?

    Sincerely,

    Ramon

  • Ramon,

    This tells me that the issue currently isn't seen on the PHY as MII loopback is the shallowest loopback on MAC side. Please ensure that register 0x16 is set to the appropriate speed as per datasheet directions. However, if this has already been done, the issue is not on PHY related.

    Sincerely,

    Gerome

  • Hello Gerome,

    We have managed to get the loopback working. The ESP was expecting an "link" with the ethernet port before is started sending packages. My colleague changed the software so it always sees a (fake) link. Now we are getting the packages back that we are sending in loopback mode. So it then seems to be a problem between the RJ45 connector (we are using the LPJ16211DNL) and the PHY. Would you know what the problem might be here?

    Best regards,

    Ramon

  • Hi Ramon,

    Insertion loss looks to be a little higher on this magnetic than our recommendation (-1.2dB vs -1dB recommended).

    I would recommend now that the PHY is alive to try connecting with link partner to see if it is linking up.

    Can you try the following?

    - Connect to PC, see if link is up

    - Connect to second instance of board, see if it links up

    - I see that auto-neg is on, and that 10/100 full duplex advertisement is on. If the above don't give link, try expanding the options to also advertise 10/100 half duplex as well to see if the link partner is having issues sending the wrong advertisement.

    - Try out different cables, ~1m and ~10m+

    Sincerely,

    Gerome

  • Hello Gerome,

    We've tried this but it doesn't seem to help.
    We have also tried to configure the print so the ESP32 sends an clock to the PHY which also didn't seem to work but we did notice by accident that if we load the firmware for the print without a crystal onto a print witch a crystal it eventually does get an (unreliable) Ethernet connection. My ques is that the clock the ESP32 is making sometimes lines up with the clock of the PHY so it can communicate for a short while. This means there is a problem in the clock signal. The clock signal looks fine to me. Could the problem maybe be a phase change? And is there a way to shift the phase?

    Best Regards,

    Ramon

  • Hi Ramon,

    I am unfamiliar with your use of the term "print". Which clock signal is the ESP32 sending the PHY? I see the XTAL on XI and this should be the only signal the PHY is getting clockwise. It then would send its RMII 50MHz clock out to the MAC (ESP). Could you check the PPM of the XI input? Is the xtal conforming to the datasheet requirements for XTALs?

    Sincerely,

    Gerome

  • Hello Gerome,

    Sorry, print is an other word for PCB in Dutch. This might not be something in English.

    The way the PCB is set up to make it sort of work is that the PHY is using a 25MHz crystal as it's clock. It is configured an Master so it sends out a 50MHz signal to the ESP32. The ESP32 has firmware loaded on it so it is not listening to the 50MHz signal from the PHY and making its own 50MHz signal. This firmware was meant to be put on a PCB that had the PHY configured in slave mode but i loaded it on the wrong PCB. This way we were able to get an Ethernet IP response. The PCB with the PHY in slave mode also didn't work.

    For the crystal we are using the ECS-250-10-37B-CTN-TR. I believe this complies with al the requirements.

    Best regards,

    Ramon

  • Ramon,

    So to clarify, you are saying with ESP with Master config and PHY in Master config, you are able to get IP response? For PHY, can you read 0x467 and 0x468 in this case to confirm it is in RMII master? Please be aware that this most likely will need extended register access. Can you also confirm, that PHY has link via reg 0x1 in this case?

    How does these registers vary when Slave config is given to ESP?

    This is very odd as the communication shouldn't work between two RMII masters.

    Sincerely,

    Gerome

  • Hello Gerome,

    Correct, but very unreliably.

    By chance I've also just noticed that we are able to get an IP address with the ESP32 as slave as well. It just happens so sporadically and with a long delay that I haven't seen it before. We are able to get an Ethernet HW address reliably in both cases but we're just very rarely getting an IP address.

    Address 0x467 gives 0x94, 0x468 gives 0x193 and 0x1 gives 0x7849 before I get an HW address and 0x786D afterward. When I configure the ESP32 as slave these registers are the same.

    Best regards,

    Ramon

  • Hi Ramon,

    It may appear that the issue may be outside of the PHY's control at this point. The PHY looks to be doing its job by obtaining link.

    However, I do want to check; how long does it take to get linkup? What cable type and length are you using? And what is DP83826's link partner?

    Sincerely,

    Gerome

  • Hi Gerome,

    The linkup I usually get within 20 seconds of starting up the PCB, but I can be trying for 30 minutes before getting an IP address. At the moment I'm using a category 5e cable of 1,5m. It's trying to connect to a prosafe 8 port gigabit switch gs108. But I have been trying different cables and link partners.

    Best regards,

    Ramon

  • Ramon,

    So with other cables and link partners, you see the same behavior of long link? What if you connect the PCB with an exact replica of itself, what is the linkup time?

    Sincerely,

    Gerome

  • Hi Gerome,

    Yes, with other link partners and cables I see the same behavior. I get a link up + HW Address pretty quickly but getting an IP address takes very long. At the moment I only have one working PCB, so connecting it to a copy of itself I can't test at the moment. I can see if I'm able to fix the other PCB next week to test this.

    Best Regards,

    Ramon

  • Hi Ramon,

    If PHY is able to link up quickly, then unfortunately the issue lies somewhere else. The PHY is able to do its job by linking up and completing that portion of the signal chain. 

    Sincerely,

    Gerome