DP83822IF: Link never goes up unless RX cable is reconnected or switch/media converter is power cycled

Hi Louis,

Are there any communication issues? I see link is up with 0x1[2] = '1'.

DP83822 does not use auto-negotiation for 100Base-FX. This auto-negotiation flag is not relevant for communication in fiber mode.

Thank you,

Evan

Hi Evan,

Thank you for the suggestion! The link partner (switch) only uses 100M Fiber mode. 

Actually, I tried resetting PHY in the same ways as you mentioned:

0x0[15] = '1' (PHY software reset)

0x1F[15] = '1' (PHY hardware reset)

However, none of this resulted in link up. 

BUT.. I finally found a way to get the link to go up i.e. by initiating a Digital Restart:

0x1F[14] = '1' (PHY digital restart)

Do you know why I need to restart PHY?

Also, as you can see, there is no communication issue since 0x1[2] = '1'. Once the link goes up, everything is perfect; no false carrier counter, no pending interrupt, etc.

Thanks,

Louis

Hi Louis,

There may be some signal quality issue causing fiber fault detection to trigger, preventing link from coming up until PHY's state machine is restarted.

Please try this register configuration to test link on start-up:

0x40[13,6,5] = '111'

Thank you,

Evan

Hi Evan,

Looks like you are correct but can you help me understand here:

I wrote '11' to 0x40[6,5] and nothing happened. But as soon as I wrote '1' to 0x40[13] it shows continuous messages: 

Interrupt FALSE_CARRIER_HF received. MISR1 = 0x82FF. FCSCR = 0xFF

False carrier event counter is also full.

Upon closer look, earlier, when I initiated Digital Reset, apparently the message shows Link is Up - 10Mbps/Half - flow control off

When I wrote '0' back to 0x40[13], interrupt messages stopped and Digital Reset can get the Link Up (but with 10Mbps/Half).

Is it even possible with the 100base FX mode? There shouldn't be auto negotiation for fiber mode, right?

Thanks,

Louis

Hi Louis,

Please share the PHY register dump in the working and failing case. There is something unexpected happening on auto-negotiation - I would like to confirm PHY is set for fiber with auto-negotiation disabled.

Register dump will allow me to confirm PHY configuration and any auto-neg advertisements being sent from link partner side.

Thank you,

Evan

Hi Evan,

This is the PHY register dump. I wrote a script with phytool to get generate this dump. I will ask firmware engineer to generate it in a more official way. But i think it is good for now to see what is going on. I had to trim some registers that I didn't see in the datasheet and are showing 0x0000. Also, I don't know how to read values from registers 0x3000 and up.

Louis

The dump before is for the failing case. This one here is from the working (10Mbps/Half):

Thanks,

Louis

Hi Evan,

Yes, sorry you asked this before, there is a communication failure during link down where ping never goes through. But after link up, communication is fine, ping is always successful.

In both failed and working case, I can see 62.5MHz waveform out of PHY's TX differential lines.. weird, that means it is already working in 100base, right?

I am trying to work with the firmware engineer to understand the PHY initialization better during kernel initialization.

I also tried this:

${dir}/mdio-tool w $dev 0x1F 0x8000
${dir}/mdio-tool w $dev 0x00 0x8000
${dir}/mdio-tool w $dev 0x00 0x2100
${dir}/mdio-tool w $dev 0x04 0x0101
${dir}/mdio-tool w $dev 0x00 0x0000
${dir}/mdio-tool w $dev 0x0A 0x4020
${dir}/mdio-tool w $dev 0x12 0x00FF
${dir}/mdio-tool w $dev 0x13 0x00BF

But it won't get that link up.. 

Can you give me a procedure / list of steps of what registers are essential to get fiber mode to work?

Thanks!

Louis

Hi Louis,

Thanks for confirming. For Fiber mode, enabling FX_EN, SD_EN, and disabling auto-negotiation through straps or registers is sufficient.

Is TX transmitting at 62.5M or 125M? In forced 100M fiber mode, I expect TX lines to transmit at 125M.

Regarding your initialization script, please include 0x1F = 0x4000 as the final step, as you have validated digital reset to bring link up.

Assuming this modified script is robust for link up and communication, is this acceptable for your end application?

Thank you,

Evan