WAN PHY Experiments

One of the objectives of the IEEE 802.3ae is to be able to use existing SONET/SDH networks for the transport of 10G Ethernet over the WAN.

This requires layer 1 attachment to the LTE (Line Terminating Equipment). 10G Ethernet packets need to be encapsulated into SONET frames and adapted to the OC-192 bit rate (which is a little less than 10Gbit/s). No conformance to SONET jitter and clock specifications has been done to allow for low cost solutions.

Attachment to the LTE is as follows:

The T6 uplink (developed in the project for early testing) is connected to the WAN PHY:

We wanted to use a Xenpak WAN PHY. However, the Xenpak transceiver is an Ethernet specific market and very few of the transceiver manufacturers have a Xenpak WAN PHY in their roadmap (XFP is a more promising form factor) and so the availability was not guaranteed for when the project wanted to do the tests.

The solution chosen by the project is to create a 10G Ethernet WAN PHY board and attachment kit to allow it to be connected in place of the Xenpak:

IEEE 802.3ae specifies a relaxed bit rate – 9.95328Gbuad +20 ppm. The desycnhroniser embedded in the LTE will tolerate a 20ppm offset, but this may cause a large number of pointer adjustments. Each of these will generate an alarm, which is passed to the operator console. For carrier class systems, this is simply not acceptable. Our WAN PHY avoids this problem, making it suitable for deployment in a real world environment.

The solution is to implement a well controlled bit rate on the 10G Ethernet WAN PHY to allow the desynchroniser to operate in its normal conditions 9.95328 Gbaud +-4.6ppm

This was then used in the interoperability test with CANARIE.

The equipment used was:

• Ethernet switch – BATM T6 equipped with ESTA 10G uplink and 10G Ethernet WAN PHY
• LTE – Cisco ONS 15454 (owned by CANARIE)

The results were that the status is correctly transmitted from one equipment to the other in the case of a degraded line or failure. Full duplex data transmission was achieved between the two equipments

• -4.5 dBm to -17 dBm on RX : BER < 2.10^-15 (15 hours, no error)
• -18 dBm on RX : BER ~ 10^-12

This highlighted the fact that there is an lack of interoperability between the IEEE 802.3ae and the ITU-T G.691 standards

• LTE jitter specifications do not map 10G Ethernet jitter specifications
• LTE does not necessary cope with 20 ppm relaxed bit rate accuracy

However, the project design bypassed this problem and achieved successful interoperability with a Cisco 15454.