Circuit Emulation Services over Packet offers the traditional circuit emulation service over a packet network. The service offers TDM trunking over a range of transport protocols including Internet Protocol (IP), MPLS and Ethernet.

In traditional TDM circuits, the clocking information is inherent in the transportation layer. In Packet Switched Networks (PSN) the timing information has to be specifically transported. Zarlink has implemented a method (patent pending) to transport the timing information over the packet network.. This method embeds the timing information in the packet header and may be used over multiple IP router and switch hops. The resulting timing performance can meet ITU-T G.823 and ITU-T G.824 timing requirements.

This technology is finding its place in the access market. Thus markets that can use this technology are cell-site backaul, Metro Ethernet equipment, DLCs, MTU/MDU equipment, data communications equipment, and EPON technology. The CESoP technology offers low cost per channel solutions that can take advantage of CESoP's low latency.

End-to-end latency is the sum of the TDM-to-packet latency with the packet network latency and the packet-to-TDM latency. The TDM-to-packet latency is comprised of the packetization delay, or the amount of frames in a packet, and Zarlink's intrinsic processing delay of less than 125 us. The network latency is dependant on the type of packet network, with ranges of less than 1 miliisecond on a dedicated network to around 3 milliseconds on a metro Ethernet network. The packet-to-TDM latency includes Zarlink's intrinsic processing delay of less than 375 us. Overall, and end-to-end emulated circuit may have a latency under 1 millisecond.

The markets that CESoP is aimed at use either dedicated networks or managed networks. These networks provide low latency and good PDV (packet delay variation). The fact of the matter is that QoS exists in the data communications equipment. The Zarlink products support QoS mechanisms such as Weighted Fair Queuing and Strict Priority. In this way, the voice can be prioritized over the network and the latency is kept very low.

In the case where a data packet starts to transmit just as a voice packet arrives to an Ethernet switch, the latency over a fast Ethernet port will be 125 microseconds and 12.5 microseconds over a Gigabit Ethernet port. At most this is just one frame and will not affect the QoS at all. Delays in TDM based systems are in the order of 250 microseconds or longer.

Packet size for a voice packet is variable. There is a trade off between packet efficiency and latency. Packet efficiency is the ratio of the header size to the payload size. Latency is the time that it takes to create a packet, transmit it, and receive it. In the case of voice, one wants low latency and thus a small payload size. Efficiency wants to increase the payload size. Typically in CESoP, the payload size is of the order of 4 E1 frames or 128 bytes. This represents a latency of 500 microseconds. For 8 E1 frames this would be 1 millisecond of latency. Thus it can be seen that the latency is very low over the Ethernet.

The circuit emulation service does not discriminate by traffic type. Thus any Nx64 kbps, fractional T1/E1, T1/E1 or T3/E3 service may be carried over the CESoP connection. This might be Frame Relay service, or voice service or video service. There is no limitation as to the service carried.

Testing has shown that the packet loss rate would have to be very high to have an affect on the quality of service. For unstructured service, tests have been performed that indicate that at a 1% packet loss rate, there is little degradation in framer timing or voice quality. This is also dependent on the number of frames in each packet. If a packet contains 1-4 E1 frames, then the timing quality impact is negligible as the receive framers do not lose frame lock, or timing synchronization at all. If the number of E1 frames is increased per packet (>4), then framer timing errors start to occur. This stands to reason, as there is more voice information in the packet, and when one is lost, it has a greater impact.

For structured service, the packet loss would affect the voice performance only. The impact would be in the voice quality, as voice samples would be lost. With a 1% packet loss rate, the number of lost samples depends on the packet payload size. If the payload contained a small number of frames, then the impact of the loss is minimal. In the previous example, with 4 E1 frames per packet, the loss would be 4 consecutive samples. The Zarlink product can compensate for the lost packets at the receiver end, by transmitting the last sample, or an all 1s pattern. These 4 samples represents a dropout of 4*125 microseconds of voice or 500 microseconds of voice. This is such a short time that the hearer would not notice this drop.

There would be no impact in the timing, as the Zarlink timing recovery performs very well under packet loss conditions. Thus the timing is maintained even at high packet loss rates. However, most dedicated or managed communication systems today have packet loss rates in the order of 10-5 to 10-6 packets. Thus the impact on quality of service is very small.

CESoATM supports Constant Bit Rate (AAL1) service and Variable Bit Rate (AAL2) service. CESoP is similar to AAL1 service in ATM. VoIP service is similar to ALL2 service of ATM. With reference to AAL1 service, one major difference is in the cell size versus the packet size. The cell size is fixed at 53 bytes and the payloads are 46 or 47 bytes. Thus the payloads are limited. The packet size is variable. The user can define and optimize the length of the payload to meet the requirements.

Another difference is has to do more with the network operation. The ATM cell header contains the address of the next switch in the network. The IP address in the packet contains the final destination address. Thus the network behaviour is different. Thus as the packet traverses the network, there is no modification of the addresses needed, as in ATM.