AM5749: Custom board: PRU1 not working

Roland Quentin

Part Number: AM5749
Other Parts Discussed in Thread: DP83640, DP83620

Hi,

We developed our custom board with the processor AM5749. On this board we have ethernet port to manage industrial communication:

We are able to detect those ethernet links, see the following post: https://e2e.ti.com/support/processors-group/processors/f/processors-forum/977155/am5749-idk574x-pr1-ethernet-hardware-modifications

The phy we're using are DP83640.

PRU2 is using here a single-mac mode, so we need 5.4 kernel to use it. SDK 6.3 is on 4.19 so we must update to 5.4 or backport patches, no problem for that. PRU2 is working after those modifications.

PRU1 are not working, drivers detected but no ping. When we look the phy register status we can see "MDI pairs swapped" (Receive on TPTD pair, Transmit on TPRD pair). We tried to use the PRU1 on the IDK board, after changing the resistors allowing this. It's working well.

So our study went to PHY and routing.

There is multiple links about DP83640 or DP83620 PHY on TI. Best idea resulting of this is to change the PHY to another.

For the routing we've compared the two ports on PRU1 and the one on PRU2:

We've checked the following document to verify routing:

https://software-dl.ti.com/public/hpmp/sitara/eth_sys_hw_am_devices/presentation_html5.html

https://www.ti.com/lit/an/snla079d/snla079d.pdf?ts=1614867581989&ref_url=https%253A%252F%252Fwww.google.com%252F

Routing seems clean. Maybe top/bottom vias and routes length are too much.

Regarding those investigations, we decided to change the PHY for another on our next board revision. Is there something else we should try? Any idea on TI side to explain why PRU1 are not working?

Best regards,

Quentin.

over 3 years ago

0 Robert Eschler over 3 years ago

TI__Guru 51635 points

Does PRU1 and PRU2 use the same PHY (DP83640)? Is there any difference between the PRU1 and PRU2 PHY configurations (hardware or software)? Have you doubled checked to ensure you've selected the correct PRU1 transmit/receive pins on the AM57x device?

0 Roland Quentin over 3 years ago in reply to Robert Eschler

Prodigy 55 points

Indeed PRU1 and PRU2 are using the same PHY. Hardware configuration are the same (except the routing lengths and the localization on the pcb). For software configuration, the only difference is the single-mac mode for PRU2.

0 Robert Eschler over 3 years ago in reply to Roland Quentin

TI__Guru 51635 points

My understanding is that is not a data problem, rather the PHY is not establishing any link? If so - does that point to an issue with the PHY? Is that why you asking if PHY change is recommended?

0 Roland Quentin over 3 years ago in reply to Robert Eschler

Prodigy 55 points

We can indeed think the link between AM5749 and DP83640 seems good. Which means the problem must be between PHY and Ethernet port. We are not sure of that. I asking if a PHY change is recommended because of three links I found on TI forum:

- https://e2e.ti.com/support/interface-group/interface/f/interface-forum/961570/dp83640-comunication-issue-with-dp83640-ethernet-phy/3555142#3555142

- https://e2e.ti.com/support/interface-group/interface/f/interface-forum/810821/dp83620-direction-detection-mdix-issues-wrong-direction-detected

- https://e2e.ti.com/support/interface-group/interface/f/interface-forum/462123/dp83620---link-detection-problems-with-short-cables

For two of them the issue was close after being solved offline. So I can't know what solved it. For the last one, the PHY was changed and fixed the issue.

0 Romain Naour over 3 years ago in reply to Roland Quentin

Intellectual 516 points

We also tried to use each PHY on the PRU1 using the single-mac mode with the 5.4 kernel that was able to use the PHY on the PRU2.

By dumping the PHY registers between the PRU1 and PRU2, we noticed that the MDIX bit was set only on the PHY on the PRU1.

0 Robert Eschler over 3 years ago in reply to Romain Naour

TI__Guru 51635 points

What is the latest status on this item? Any update?

0 Roland Quentin over 3 years ago in reply to Robert Eschler

Prodigy 55 points

Our plan is to change the PHY to see if it helps, as explained in my previous message regarding the difficulties seen in several posts on the forum. The next board revision will be manufactured for February 2022. Before this date, we didn't planned to continue the tests because we spent many hours on it and we study everything we had in mind. That's why I wanted a feedback on what we did and maybe ideas of other tests to perform. If you have ideas, please let us know to try to perform them and maybe made last minute changes on the manufacturing plans we sent to our sub-contractor.

0 Robert Eschler over 3 years ago in reply to Roland Quentin

TI__Guru 51635 points

If no additional debug/discussion is planned, please close this thread. As additional testing is done, a new thread can be started.

Processors

Processors forum

AM5749: Custom board: PRU1 not working