DP83869EVM: FIBER-RGMII connected back-to-back RGMII-Copper

Hi,

I remember this case from our previous E2E thread.

It sounds like both boards are linking up on their respective MDI connections (LED_0 is lit), however what concerns me is the RGMII soldered connection. Were you able to get the MII loopback test working again? I remember this had worked at one point then stopped working.

Best,

Shane

Hi Shane,

I lost that old ticket thread - thanks for including the link of previous thread.

Yes, I performed the MII loopback test on 100Mbps; the EVM1 (Fiber-RGMII) was failing to detect the fiber (LED0 is off). I have set OP_MD[2:0]=010 it didn't detect but when I connected the strap to OP_MD0 (high), the LED0 lit up which is odd. For LED configuration, I had LED0 &2 to ground (low) and set LED1 high. The EVM2 board seems to be working properly - OP_MD[2:0] = 000 and all LEDs are set to 0.


Even for the 1000Mbps case, I have set EVM1 to forced fiber mode, but for EVM2 if I set it to forced 1000Mbps (RGMII-Cu), its failing to detect the host (LED0 with all other LEDs getting off).

[So that you know, for 100Mbps and 1000Mbps I have a separate SFP module for the specific speed.]

Thank you.

Fnu Premsagar said:

I have set OP_MD[2:0]=010 it didn't detect but when I connected the strap to OP_MD0 (high), the LED0 lit up which is odd.

If you are setting OPMODE[2:0] to 011 the device will be in RGMII-SGMII bridge mode. This repurposes the fiber connections into SGMII connections, which wouldn't be valid on the EVM. I wouldn't rely on LED_0 for the correct status in this case:

I see from the UMC-GA1F1T product description that it supports 1000Base-X, not 100Base-FX.

In this case, you should set the Fiber EVM to OPMODE[2:0] = 001. I believe this is the issue with linking up on the Fiber EVM side.

For the RGMII-Copper EVM, keep the working configuration you're currently using (all LEDs set to 0).

Best,

Shane

Hi Shane,

If you are setting OPMODE[2:0] to 011 the device will be in RGMII-SGMII bridge mode

Just for clarification, I have set OP_MD[2:0] = 010, where only OP_MD1 is high to set it to 100Mbps-RGMII operation.

I see from the UMC-GA1F1T product description that it supports 1000Base-X, not 100Base-FX

Yes, for 100Mbps I got separate module - SFP-FE-BX (100BASE SFP 1550/1310nm) AND ANOTHER IS SFP-FE-BX (100BASE SFP 1310/1550nm) . I have tested this connected directly to host without using DP83869EVM and I was able to ping properly using UMC-GA1F1T. (I connected two UMC-GA1F1T back to back with 100Mbps SFP module in it PC<->UMC1 (Cu-Fiber)<->UMC2(Fiber-Cu)<->Host-SK-AM62B-P1 with static IP, and I was able to ping)

For 1000Mbps, On EVM1 (Fiber-RGMII), OP_MD[2:0] = 001 for 1000Mbps base mode (tried auto and forced negotiation), and on EVM2 (RGMII-Cu) OP_MD[2:0] = 000.

When both boards turned on:
EVM1:
LED0 is in forced fiber mode (high), and LED1&2 are low
- LED0 lights up.
- LED 1 blinks synchronously with EVM2-LED2 and the host RJ45 LED, and 3 of them blink at the same time - mirroring
- LED 2 is off

EVM2:
LED0,1&2 - pin 47, 46,45 are low (grounded)
- LED0 lights up.
- LED1 is on
- LED2 blinks synchronously with EVM1-LED1 and the host RJ45 LED-mirroring

It appears that both links are stable on the MDI side of each PHY. Can you read register 0x0015 on both EVMs when you ping through the boards:

This measures the error counter on the MDI side of each EVM. For EVM1 this would be errors on the fiber interface, for EVM2, this would be errors on the copper interface. The reason I want to check this is that sometimes you can get a link on the MDI side and still have errors or corrupted data. Seeing the RX_ERROR_COUNT will help to rule out issues on the MDI side connections. 

If you see no errors (reg 0x0015 = 0 on both EVMs) and still cannot ping, the issue is likely with the RGMII interface. In this case the next step would be to run the MII loopback test again.

Best,

Shane

both board reads Register 0015 has: 0000
Let me run the loopback test again and I'll let you know

Hi Shane,
Not yet, I am getting a custom PCB prototype board with the DP83869HM chip on it and planning to test it so there won't be any issues with soldering or external jumpers coming off of RGMII.

Sounds good, you want to be sure:

• RGMII trace lengths are matched
• RGMII traces aren't too long
• There are no stubs in the signal trace

For details on layout guidelines, check out our Ethernet PHY Layout checklist.

Best,

Shane