Gigabit Ethernet Switch Module: Link Up & Down

Thread Starter

Hafiz Haja

Joined Apr 23, 2019
4
Hi,
We are designing an Ethernet switch module using Microsemi chip VSC7422. With the same design circuits 5 modules has been assembled out of which first 4 modules are working fine with all ports 1Gbps links are up continuously. We are facing link issue in the 5th module alone. We have checked power rails and clock input to switch IC. Tried replacing the chip VSC7422. Everything is similar to first 4 modules. But the link is going up and down continuously for both 1Gbps and 100Mbps speed. Link is establishing in 10Mbps speed (Link LED is stable), but when tried to ping the switch most of the time request time out error is coming and randomly pinging is happening. We cannot suspect magnetics/transformer since all the ports of the switch are behaving same. When the 1Gbps Ethernet DATA-A-P & DATA-A-N is probed we are getting some pulses when the link is trying to establish, but again going to zero since the link is going down. Attached the signal waveform. We are unable to find out any issue and we are stuck where to debug.

Regards,
Hafiz
 

Attachments

Thread Starter

Hafiz Haja

Joined Apr 23, 2019
4
Hi,
Adding to the above information. We tried probing the PHY output to magnetics primary side of both working module and non working module. It is observed that in the non working module, the pulse outputs are having same amplitude. Whereas in working module there is pulse amplitude modulation. What can be the reason for this to happen?. Attached the signals for reference. If anyone got any idea kindly support.
 

Attachments

BobaMosfet

Joined Jul 1, 2009
2,110
Most people don't realize that before any part of the TCP/IP stack is involved, ethernet I/C's try to communicate at a lower, electrical level. And it is at this level that they try to communicate to determine capabilities of each other on either end of the line. This is what the PHYS module does. Once it is successful, it passes that information up to where the TCP/IP stack can utilize it.

In the non-working module the amplitude is likely the same because it's just sending the same signal over and over, trying to establish contact. While in the working module, you're seeing actual communications (more complex).

Have you checked your cables to make sure they are good? If you have a really get ethernet tester, I'd put that on the other end of the line and see what it says- it should be able to talk to your chip if it's going to work.
 

Thread Starter

Hafiz Haja

Joined Apr 23, 2019
4
Hi,

We have probed the PHY output without connecting any cables (other end open). We are using two IC's in the module, VSC7422 and VSC8522. VSC7422 is the Ethernet switch with 12 port PHY integrated in it, whereas VSC8522 is another 12 port PHY chip.

In the working module, both chips generates amplitude modulated pulses every time even though cable is not connected for establishing link with another device.
But in non-working module switch IC VSC7422 PHY output only is generating same amplitude pulses, whereas VSC8522 PHY output is always amplitude modulated.

Cables are tested fine and are ok with other working switch modules.

Regards,
Hafiz
 

BobaMosfet

Joined Jul 1, 2009
2,110
Still reviewing. Meanwhile, what happens if you put an RJ-45 loopback jack on the port and then TX/RX on it- that will isolate everything to the port itself. See section 4.3.7.6 in the datasheet for built in testing that is ideal for a loopback.
 

Thread Starter

Hafiz Haja

Joined Apr 23, 2019
4
Hi,

We have tried both internal loopback test and external loopback test. Both are working fine. Then we are suspecting the 25MHz reference clock input to the chip. We are using one clock source to drive Ethernet switch chip VSC7422 and Ethernet PHY chip VSC8522. Clock input to both chips are supplied via voltage divider as per datasheet recommendation. Attached the clocks probed at clock source output and clock input (voltage divided) to VSC7422 of both working module and non-working module. It seems clock jitter is very worse in non working module. But can't suspect what will cause this jitter problem. Because clock output is still clean and only the clock input after voltage divider is having jitter. Is there any chance that the chip VSC7422 is causing problem. We tried a new IC also.

Regards
Hafiz
 

Attachments

Top