DP83822IF: DP83822 Communication issue with fiber port

Hi,

I believe you have Signal Detect of Optical Transceiver connected to DP83822 ( LED_1). What is the Optical Transreciever signal detect Polarity ?

can you confirm the strap configuration on RX_ER ? We need it to be Mode 3( if RGMII disabled) or 4 ( if RGMII is enabled)

Regards,

Geet

Hi Geet,

Thanks for the support! In fact, customer didn't use the Signal detection(As you can see they pull LED1 pin to high forever). They don't detect the signal from optical module.

For the RX_ER, when we do the test, we let this pin open, that should be Mode 4 based on our datasheet.

Thanks a lot for your support!

Best regards

Sulyn 

Hi Sulyn,


We need to do both :
- Connect signal detect from Optical Transreciever to LED_1. Check the polarity is aligned with default polarity of signal detect on DP83822.
- Configure the RX_ER strap to mode 4.

Regards,
Geet

Hi Geet,

The truth in customer side is they can't use the LED_1 to detect the signal from SFP, they don't want to check the status but make it always "signal detected" by pulling it high forever. This is customer's requirement.

As for the RX_ER mode,we left this pin float which is Mode 4 yet.

Based on customer's requirement, is these some software solution to avoid(Don't consider Rest solution)? Thanks.

Best regards,

Sulyn

Hi Geet,

Can you help give feedback for the question, as it's in the MP stage, so it's important to support on this, thanks! If not, they will switch back to BCM5421 solution quickly.

Best regards,

Sulyn

We recomend to enable signal detect. Else, on fiber removal/connection, phy needs to be soft reset.

Regards,
Geet

Hi Geet,

Thanks for the support! So If we recommend customer to do soft reset, how can customer know when to do the soft reset? Which registers should they monitor to determine when to do the soft reset after plug out the cable?

Best regards,

Sulyn

Hi Geet,
Another question that I want to double check with you if customer don't use the LED1 pin as signal detection and they also set the SD_DIS(RX_ER set to Mode 3), what will be the problem during communication, based on customer's feedback, the communication is not stable, sometimes it's ping successfully but sometimes not, can you also help give some suggestion for this? Thanks.

Best regards,
Sulyn

To answer first question : Phy does not have any indication to indicate need for soft reset. Since cable plugging/unplugging is a manual event, it shall be handled at System level.

Setting RX_ER to Mode 3 shall enable the Signal Detection. If LED1 pin ( Pull High/Low) will get reflected on the link status. With no cable unplugging, this communication shall remain stable.

Regards,
Geet

Hi Geet,

So for the first question, do you mean these is no way that PHY can know that the cable was plugged out and plugged in, so customer can do the soft reset when get this trigger. Can you help think of some available solution for customer to go ahead? Thanks.

For the second question, as showed in datasheet, when RX_ER set to Mode 3, the SD_DIS =1, which should mean that Signal detection was disabled if my understanding is correct, but you say it shall enable the Signal detection, can you help double confirm this? Because customer set it to Mode 3

Best regards,

Sulyn

Hi Sulyn,

Yes, soft reset is needed to re-sync the phy.

SD_DIS=1 : It's an error in datasheet. We are working on update of this field. You can read this field as SD_EN field.

Regards,
Geet

Hi Geet,

Thanks for the information, we are now planning to use bit 9 of 0x0012 register to judge and do the soft reset to re-sync the phy, detailes as below:

Monitor 0x0012 register, when the bit 9 of 0x0012 was detected as 1, then will do soft reset to re-sync the communication. This works well after several times test to avoid the unstable communication issue after plug out&in, but I need your help to check if this solution ok to use and any other potential problem that may have?

            phy_data = phy_read(phydev,  0x0012);

               if( phy_data & 0x0200 )

               {

                              printk("phy reset\n");

                              phydev->link = 0;

                              phydev->speed = SPEED_10;

                              phydev->duplex = DUPLEX_HALF;

                             

                              phy_write(phydev, 0, 0x8000);

                              udelay(100);

                              dp83822_config_init(phydev);

                              phy_write(phydev, 0, 0x2100);

               }

 

Another import question that need to make clear that during the test, the register 0x0001 seems always showed 0x784D even when the fiber cable was moved(bit 2 is 1), which should no link, but it still showed link is good？Why it is like this? Is this because of the SD_DIS was EN_DIS like you mentioned?

 

Best regards,

Sulyn

Hi Sulyn,

Yes, Usage of Signal Detect signal from Optical Transceiver helps fix the wrong Link-Up status in Register 0x0001. DP83822 usage the signal to reset the Link-up state machine on the signal. 

On usage of bit 9 : why you choose this interrupt bit ? Are you observing this interrupt bit as n when the cable is removed and re-inserted ?

Regards,
Geet

Hi Geet,
Thanks for the support! For bit 9 of register 0x0012 is because we found when cable was plugged out, the bit will be set to 1 every time, even when the cable was plugged in again before do the soft reset, that's why we use it and it works well after test. So not sure if this is a good way to solve this issue?
Is these any potential risk and we have other solution to solve this issue by software?

Best regards,
Sulyn