DP83869EVM: 1000Base-T to 1000Base-X mode not working in setup

Hi Liam, 

Liam Borg said:

if there is a way to figure out why the dev board doesn’t work in the described configuration and fix it

Could you check register 15h?
Is link loss pass through enabled?
Lastly, try writing 0x2053 on 53h register. 

Liam Borg said:

Why the commands shown above don’t properly configure the device 100Base-TX to 100Base-FX

What exact is the issue you are running? Is there no link up?
I tried this setup on the EVM and I see the packet transfers. However, I noticed that the LED functions are not exactly the same as the hardware strapping setup.

Best,
J

Hi J

To the first point:

register 0015 is 0000

link loss pass through is enabled as register 006E is 1800

register 0053 was 2055 and i wrote it to 2053 but nothing happened.

To the second point, when i run the command to change it to 100Base-TX to 100Base-FX the device doesn't work at all there is no link up. Note that it works in 100Base-TX to 100Base-FX mode if configured by straps initially.

Kind regards

Hi Liam, 

1. Is the media converter linking up on both fiber and copper? Does disabling link loss pass thru feature improve this? Also, if you directly hook the optical fiber to the network from the media converter, is there any issue?
2. I realized while recreating this on the EVM that the LED configurations are different if you follow the datasheet SW 100M media converter configuration so copper link doesn't show on LED0. Do you not see copper or fiber link up? If it is copper, could you check 01h register and if it is fiber could you check C01h register? 

In 1000Base-T to 1000Base-X mode in the above setup the LED 0 and 1 blink on power on the turn off after a few seconds. If i disable link loss LED0 doesnt change, but LED1 stays on, establishing a fiber link.

I've noticed that the device sometimes works When i configure it to 100Base-TX to 100Base-FX mode via the registers, however it is inconsistent even if i do the exact same setup.

When it does work:

Register 0001 is: 7969
Register 0C01 is: 614D

When it doesnt work, none of the LEDs are on. checking the registers i get:
Register 0001 is: 7949
Register 0C01 is: 6149

kind regards

Hi Liam, 

Thank you for the information. 

On issue 1:

Liam Borg said:

I want to note that if I switched the positions of the E2845 and the DP83869EVM, or if I replaced the DP83869EVM with the E2845, the system would work at 1000Mbps.

Based on this, it sounds like the fiber link on 1G media converter is good, and it seems to be the issue between the EVM and the dock. 

Liam Borg said:

In 1000Base-T to 1000Base-X mode in the above setup the LED 0 and 1 blink on power on the turn off after a few seconds. If i disable link loss LED0 doesnt change, but LED1 stays on, establishing a fiber link.

So, based on this, is it correct for me to understand that copper and fiber link was established and when you pull the copper cable out the copper link is lost but not the fiber link? 
Would it be possible for you to show registers 01 and C01h on these cases also?
Also, is the laptop reporting that there is an ethernet connection through the dock, and how long is the ethernet cable?

Based on the previous post, it doesn't seem like the PHY is receiving any symbol errors and it doesn't respond to the internal tuning configuration. 

To debug, we could try two things:

1. Could you do 0053 2054? This will increase our tuning by 1. 
2. Another thing is to see if we can put the EVM in RGMII to copper mode and then do a reverse loopback to see if the PHY can send the receiving packets back to the dock. 

On issue 2:
How often does it work, how easy is it for you to reproduce this issue on your end? When it works, do you configure the media converter mode via register and then replug the cables, or are the cables plugged in after the register configuration happens?

Also, by default 100M media converter mode has link loss pass through disabled. Therefore, you have to enable it by writing 0 to bit 3 and 0 of registers 1ECh. Could you try and see if this helps?

Please let me know.

Best,
J

J said:

So, based on this, is it correct for me to understand that copper and fiber link was established and when you pull the copper cable out the copper link is lost but not the fiber link? 
Would it be possible for you to show registers 01 and C01h on these cases also?
Also, is the laptop reporting that there is an ethernet connection through the dock, and how long is the ethernet cable?

Sorry I'm not quite sure what you mean i never pulled out the copper cable, i just tested the device with disabled and enabled link loss. 

With link loss enabled:

Register 0001 is: 7969
Register 0C01 is: 6149

With Link loss disabled:

Register 0001 is: 7969
Register 0C01 is: 6169

No when in this mode the laptop doesn't report an ethernet connection, the cable is 1m

J said:

On issue 2:
How often does it work, how easy is it for you to reproduce this issue on your end? When it works, do you configure the media converter mode via register and then replug the cables, or are the cables plugged in after the register configuration happens?

Also, by default 100M media converter mode has link loss pass through disabled. Therefore, you have to enable it by writing 0 to bit 3 and 0 of registers 1ECh. Could you try and see if this helps?

It works somewhere between a third and half the time. I have the cables plugged in then i configure the registers.

After adding the configuration of 01ECh to my script i got it working every time consistently. I then tried the original script a few times an that was working consistently as well. I cant explain why it is working today and yesterday, it does take some time to get a connection perhaps before i was carless and disconnected it too early.

I think this resolves my register configuration question, now the only thing that is left is why the device cannot connect to the docking station in 1000Base-T to 1000Base-X mode 

Hi Liam, 

Good to hear that 100M Media converter mode is working now. I apologize for the confusion across multiple documentations and we are looking to consolidate all information into one document for readability. 

Going back to issue 1:

Thank you for the information on link status. I didn't realize link was never up on the either side. 
I am unsure why the link is not seen on either end. 
Since it is 1G, could you try doing a force link up? You can write 0x5448 to reg 10h. 
If this works, it will narrow our issue down to auto-negotiation process. 

Best,
J

Hey sorry for the delay,

I tried to do that, it did not work, did i have to write anything else or was that all?
Register 0010 is: 5448

Kind regards

Hi Liam, 

The only write was 5448. Bit 10 forces 1G link on DP83869. 

Liam Borg said:

In 1000Base-T to 1000Base-X mode in the above setup the LED 0 and 1 blink on power on the turn off after a few seconds. If i disable link loss LED0 doesnt change, but LED1 stays on, establishing a fiber link.

Based on previous experiment, it is clear that the copper connection between the EVM and the dock is an issue. Forcing the 1G link does not help. I do not think we tried restarting auto-negotiation so could you please try that? You can set bit 9 of 00h register high to restart auto-negotiation on the PHY. 

Another thing we can try is could you set the EVM into RGMII to copper mode and see if the EVM and the dock links up by default?
I want to see if this link is generally causing an issue, or if this is an issue only for the media converter mode. 

Lastly, we have seen link margin issues on short cables for our sister device DP83867. 
Could you run these scripts below to see if they enable the link?

Please let me know. 

Best,
J

J said:

Based on previous experiment, it is clear that the copper connection between the EVM and the dock is an issue. Forcing the 1G link does not help. I do not think we tried restarting auto-negotiation so could you please try that? You can set bit 9 of 00h register high to restart auto-negotiation on the PHY. 

I tried this but unfortunately it did not work:
Register 0000 is: 1340

J said:

Another thing we can try is could you set the EVM into RGMII to copper mode and see if the EVM and the dock links up by default?
I want to see if this link is generally causing an issue, or if this is an issue only for the media converter mode. 

When i did this LED_0 turns on (10/100M/1G Link-up: Stable ON) and stayed on showing that the EVM linked to the dock

J said:

Lastly, we have seen link margin issues on short cables for our sister device DP83867. 
Could you run these scripts below to see if they enable the link?

For first script: I wrote the script and ran it unfortunately it did not work either
LED 0 and 1 flash then LED 2 flash, this repeats in a cycle sometimes they flash at the same time

For second script: This did not work either, LED 0 and 1 flash then LED 2 flashes but not link

Kind regards

Hi Liam, 

Thank you for testing these. 

Let me discuss this internally for more ideas to isolate the root cause. I will get back to you tomorrow. I apologize for the delay. 

Best,
J

Hi Liam, 

We had internal discussion and it looks like there is copper link initially in the media converter mode, but the link gets dropped for some reason. With it, the fiber link is also dropped when link loss pass through feature is enabled. 

Could you provide us a register dump of 00h to 1Fh and 6Eh when the device is in media converter mode and RGMII to copper mode when both are connected to the Dock? Also, would it be possible to do a multiple read of register 01h and see if the value changes? The link status bit is latch low so it will not clear until you read that the value is low if the link is dropped and re-negotiated. 

Please let me know. 

Best,
J

Okay, i didn't know what you meant for multiple reads of register 01h so i read it at the start and end of the script.
I want to add i was using Wireshark to view packets between the DVM83869 and the PC, and i can give you key information if you want, but not going to show all the packets here due to information risks.

For the device in media converter mode:
register dump.txt file is open...
Register 0000 is: 1140
Register 0001 is: 7969
Register 0002 is: 2000
Register 0003 is: A0F3
Register 0004 is: 0181
Register 0005 is: CDE1
Register 0006 is: 006F
Register 0007 is: 2001
Register 0008 is: 7001
Register 0009 is: 0300
Register 000A is: 0000
Register 000D is: 401F
Register 000E is: 0000
Register 000F is: F000
Register 0010 is: 5048
Register 0011 is: 7802
Register 0012 is: 0000
Register 0013 is: 9DC2
Register 0014 is: 29C7
Register 0015 is: 0000
Register 0016 is: 0000
Register 0017 is: 0040
Register 0018 is: 6150
Register 0019 is: 4444
Register 001A is: 0002
Register 001E is: 0012
Register 001F is: 0000
Register 006E is: 1800
Register 0001 is: 7969
Register 0001 is: 7969
End of file.

For the device in RGMII to copper mode:
Register 0000 is: 1140
Register 0001 is: 796D
Register 0002 is: 2000
Register 0003 is: A0F3
Register 0004 is: 01E1
Register 0005 is: CDE1
Register 0006 is: 006F
Register 0007 is: 2001
Register 0008 is: 7001
Register 0009 is: 0300
Register 000A is: 0000
Register 000D is: 401F
Register 000E is: 0000
Register 000F is: F000
Register 0010 is: 5048
Register 0011 is: 7C02
Register 0012 is: 0000
Register 0013 is: 1DC2
Register 0014 is: 29C7
Register 0015 is: 0000
Register 0016 is: 0000
Register 0017 is: 0040
Register 0018 is: 6150
Register 0019 is: 4444
Register 001A is: 0002
Register 001E is: 0012
Register 001F is: 0000
Register 006E is: 1000
Register 0001 is: 796D
Register 0001 is: 796D
End of file.

Hi Liam,

Thank you for the register dump. Just to confirm, is link loss pass thru disabled when the dump with the media converter mode was generated?

Best,

J

Yes it should be disabled, the link loss jumper is not on the board

Thank you. Let me review tomorrow and get back to you.

Best,

J

Hi Liam, 

It looks like there is auto-negotiation error happening when the EVM is in the media converter mode. 
We definitely do not expect this since RGMII to copper mode does not throw an auto-negotiation error. We are going to investigate this more internally and update you. 
Also, please note that the US office is going to holiday next week so the response will be delayed next week. 

Best,
J

Hey

I built laid out my own board using the DP83869, and tested it in the same configuration and same setup as before and it worked in unmanaged more (just with straps). Sorry for the work but i think there may be issues in the layout of the dev board, or something individually wrong with the dev board i have. 

Kind regards

Hi Liam, 

Great to hear that it works on your board design! Unfortunately, it seems like this can be an issue with the device on the EVM itself. If you would like to verify, I suggest doing a ABA swap with the device on the board. 

Please let me know. 

Best,
J