DP83826E: DP83826E MSE (Mean Square Error) via register 0x225

Hi Jan,

Can you send a picture of your setup?

Please also read register 0x10 and let me know the value. 

Thanks,

David

Hi David,

Value of register 0x10 is 0x6F15.

My current connection is far from the ideal. I use DP8325EVM with RMII connection to STM32H7 MCU. Connection is via wires, that means impedance matching will be pretty poor. At MDI side I have removed transformer from EMV and I have connected RJ45 connector with integrated magnetic. MDI connection is via wires. It means not proper impedance matching and maybe crosstalks. But although I have such "horrible" connection of PHY, Ethernet communication works reasonable well for my surprise.

MSE may to be interesting metric to dig deeper into link quality. By TI were told to me that DP8325 and DP8326 should beehive at the same way in terms of MDIO registers. But maybe this statement was not correct and it not count extended registers like a MSE.

Regards,

Jan

Hi Jan,

Let me see if I can replicate the issue in lab. 

Other than reading the MSE register, are there any other issues? You said throughput is satisfactory?

Thanks,

David

Hi David,

Let me briefly summarize.

Our initial design was done with KSZ8081RNA, but we have found poor performance at long Ethernet cables (90m-100m). That is reason we have try to use TI PHY. Even with not properly connected DP83825 we achieve better results than with KSZ8081RNA. New PCB version with DP83826 is under manufacturing now. We selected DP83826 with RMII connection because its package better suites for us than DP83825. Performance of our testing hardware with DP83825 is not excellent, but acceptable especially with such poor wire-based MDI and RMII connections. We expect that performance will be much better with properly designed PCB.

Let's talk about other issues. If we not talk about multiple typos at documentation for DP83826 (which I already reported by another channels), I saw weird values reported by PHYSTS Register (10h) few times. I saw set bit 3 (MII Loopback Status) to 1 without any settings of loopback mode. I as also saw that sometimes bit 0 (Link Status) were set to 0, but after 100ms repeat reading was again set to 1. But I hope that this issues with register 0x10 are somehow related to our PHY connection, and we will not see them at real PCB.

Regards,

Jan

Hi Jan,

How often does do you see the weird value in register 0x1? Can you share a few register dumps (0x0 - 0x1F) while data is being sent/received? We can look for bits that are changing. 

Thanks,

David 

Hi David,

Weird values at 0x10 register I saw only few times. For this moment this is not big deal for me. Because I think this was somehow related to poor connection MDI, RMII which I currently have. I think after right connection this will disappears.

Registers when PHY link is UP and ~12Mbps data transfer.

reg 0x00: 0x3100
reg 0x01: 0x786d
reg 0x02: 0x2000
reg 0x03: 0xa140
reg 0x04: 0x1e1
reg 0x05: 0x45e1
reg 0x06: 0x5
reg 0x07: 0x2001
reg 0x08: 0x0
reg 0x09: 0x0
reg 0x0a: 0x100
reg 0x0b: 0x0
reg 0x0c: 0x0
reg 0x0d: 0x0
reg 0x0e: 0x0
reg 0x0f: 0x0
reg 0x10: 0x4615
reg 0x11: 0x10b
reg 0x12: 0x20
reg 0x13: 0x0
reg 0x14: 0x0
reg 0x15: 0x0
reg 0x16: 0x100
reg 0x17: 0x65
reg 0x18: 0x480
reg 0x19: 0x8c00
reg 0x1a: 0x0
reg 0x1b: 0x7d
reg 0x1c: 0x5ee
reg 0x1d: 0x0
reg 0x1e: 0x102
reg 0x1f: 0x0

Regards,

Jan

Hi Jan,

Okay, yes I am guessing it may be related to the poor MDI and MII connection. Let me know if the issue becomes more apparent and you can grab a register dump when the weird case occurs. 

As for now, I am seeing that link has not dropped, and there are no receive errors in the register dump you have shared. 

I am still checking on the MSE register and will let you know when I hear back.

Thanks,

David 

Hi David,

OK, let me know when you will have results.

Regards,

Jan

Hi Jan,

Yes, hopefully tomorrow or early next week.

Thanks,

David

Hi Jan,

Sorry for the delay. 0x218 is the correct MSE register. This will be corrected in the next revision of the app note.

Thanks,

David

Hi David,

Thanks for your answer. I will test and let you know.

Regards,

Jan

Hi Jan,

Okay, let me know.

Thanks,

David

Hi David,

I am testing with testing hardware with DP8325EVM. Register 0x218 shows values at expected range. It seems its somehow works, although I expected much worse values. Values of MSE register are at range 320 - 517 (excellent link quality according table 3-7 from snla4230). When I will have ready my "real" hardware with DP8326, I will test this too.

Jan

Hi Jan,

Okay, yes let me know.

Thanks,

David

Hi David,

I have finally ready new hardware with DP83826. Performance at long cables is very good. At 122m long cable no RX Errors, no FCS errors and MSE at range 150 - 550. Performance is much better than KSZ8081KSZ8081RNA.

Jan

Hi Jan,

Thanks for updating, the result is great to hear. I guess earlier issues was related to poor MDI and MII connection. Do you have any further questions?

Thanks,

David

Hi David,

Yes, I think this was related to poor MDI and MII connection.

With my new hardware with DP83826 I saw one thing. When is DP83826 set to auto-negotiation but opposite side have set fixed speed. This cause communication speed issues and packet lost (<1%). I would ask if this is expected behaviour of DP83826.I read some articles about duplex mismatch issue when auto-negotiation is enabled at one side only, but I never seen this before than now with DP83826.

Thank for answer.

Jan

Hi Jan, 

Yes, may be duplex mismatch. Maybe you have seen it already, here is a good article on that: http://www.ethermanage.com/ethernet/pdf/dell-auto-neg.pdf

Can you check the duplex status on each side?

Thanks,

David