Though it’s a rarity on the baseball diamond, carriers, utility companies, and real estate developers are increasingly talking about the “triple play.” Delivering voice, video, and data services over a fiber access network will help traditional and new providers attract, retain, and dig deeper into the wallets of their customers. But delivering voice service over a packet-based fiber network can be a challenge. Circuit emulation services over packet (CESoP) is an emerging technology that can support carrier-grade voice lity to complete the triple play.
Ethernet PONs are an emerging access technology that supports the delivery of triple play services over fiber. Based on the Ethernet 802.3 protocol, EPON bridges the access gap between the residential or business environments dominated by 10/100-Mbit/sec Ethernet LANs and the service provider’s central office or point-of-presence connection to the public network.
Data, which is already Ethernet-based in the business LAN or residence, will be carried effectively across an EPON. The extra bandwidth capacity of EPON compared with a DSL or T1/E1 line will be advantageous for services such as virtual private networks (VPNs). Similarly, IP-switched digital video will benefit from the increased bandwidth, multicast, and broadcast capabilities of EPON.
However, voice service-whether provided over an analog POTS line or through a T1/E1-connected PBX-will not be as easily served by a packet-based network. That’s because the provision of real time voice services over a packet network presents a series of challenges not present with “normal” data traffic.
A voice over packet implementation must minimize latency and pay close attention to specifications such as ITU-T G.114 “One-Way Transmission Time” and Telcordia GR-303-ILR. Voice latencies are helped by the plentiful bandwidth and quality of service (QoS) mechanisms in the packet network but are more keenly affected by the delays suffered through packetization, voice processing, and jitter. For example, late or lost packets result in a noticeable reduction in voice quality. Packets may be lost or delayed because of queuing delays, buffer overflow, insufficient queue resources in an Ethernet switch, a routing change, or a more serious impairment such as a link or path failover. To help compensate, the voice over packet receivers must be equipped with packet-loss concealment schemes.
The synchronous nature of voice presents challenges as well. A critical part of delivering T1/E1 service to a PBX is ensuring that it’s synchronized to the PSTN; otherwise, the framers in a PBX will incur periodic buffer slips that reduce service quality. Similarly, if a POTS line receiver is not synchronized with the sender, buffer slips will occur and appear as a burst of lost packets.
Two distinct types of voice service will be considered here: analog POTS and T1/E1-connected PBXs. CESoP technology, based on standards from the ITU (Y.1413), Metro Ethernet Forum (MEF 8), and MPLS-Frame Relay Alliance (MFA 8.0.0) as well as draft standards from the IETF PWE3 working group allows circuit-switched services to be carried across a packet-switched network.
Previous PON flavors were based on ATM technology that included AAL1 support for high-margin T1/E1 services. The absence of this support within EPON equipment means businesses must look elsewhere for fractional, leased-line, or private-line T1/E1 access. Manufacturers have primarily relied on VoIP to deliver voice services over packet-based equipment. VoIP adds unnecessary complexity and cost to the network and is ill-suited for some T1/E1 services.
In comparison, CESoP enables equipment vendors to support fractional, leased-line, or private-line T1/E1 services and deliver carrier grade voice over an EPON. CESoP involves packetizing the TDM circuit traffic or voice samples, including both the data and signaling, as either unstructured private-line data or structured N×64-kbit/sec voice channels. These packets are transported across a packet network such as an EPON. At the far-end, the ingress packets are smoothed using a receive jitter buffer. The TDM circuits or voice samples are then extracted from the packets and played out onto the TDM circuit.
An optical-network unit (ONU) for business use provides many 10/100-Mbit/sec Ethernet interfaces but traditionally does not support T1/E1 services. An office typically has many digital handsets connected to a PBX, which is in turn connected through a T1/E1 line to the PSTN.
CESoP is a natural choice to service this market with EPON since it’s specifically intended to provide T1/E1 and fractional T1/E1 (N×64-kbit/sec) services over packet networks. By adding a CESoP block, as shown in Figure 1, the ONU has been upgraded to support fractional, leased-line, or private-line T1/E1 services. CESoP meets the numerous challenges of providing voice and T1/E1 services over a packet-based access network. Given the advantages of trunking channels, end-to-end latency may be as low as 125 µsec (one frame) or as large as 1 msec (eight frames). In a non-bandwidth-constrained EPON, latency should be less than a few milliseconds upstream and even less downstream.
The receive jitter buffer latency is matched with the EPON and in the order of a couple of milliseconds in the upstream direction and less in the downstream direction. In total, one-way latency is ~5 msec upstream and ~3 msec downstream. Measured lab values were 1,900 µsec upstream and 800 µsec downstream, even under loaded data traffic conditions (with one-frame packetization).
Given that packet networks are asynchronous themselves and/or asynchronous to the PSTN, the timing and synchronization aspect of T1/E1 service has been incorporated into the CESoP standards. Timestamps and sequence numbers are used in the CESoP packet headers to transfer timing information from the PSTN to the customer premises equipment (CPE). CESoP meets the required ITU-T and ANSI standards for timing distribution to provide T1/E1 service to CPE.
With CESoP, in a PDH environment each T1/E1 connection may be independently timed from a different clock source. In addition, each direction of an individual T1/E1 connection may also have independent timing. In the event of late or lost voice packets, the CESoP connection will replace missing packets with appropriate “filler” data to minimize the effects on the voice quality.
Best performance of CESoP connections across an EPON may be achieved in several ways. Selecting a low frame-per-packet value will reduce overall latency. Using a managed QoS-aware switch and enabling QoS in the EPON equipment ensures CESoP traffic is prioritized over data traffic. Sizing the CESoP interworking function jitter buffer to a value that is well fitted for the packet delay variation (PDV) on the packet network is also important-too small and there will be packet discard, too big and there will be unnecessary latency.
An ONU may use a few different approaches to support residential voice service. Through its Ethernet interface, it may be connected to a residential VoIP phone. Alternatively, through one or more POTS interfaces, it may use either VoIP or CESoP to provide traditional analog phone service. Figure 2 illustrates an ONU implementation using CESoP to provide POTS phone service.
The choice of VoIP or CESoP is dependent on the application, complexity, and cost. In comparison with VoIP, CESoP is a very simple mechanism to transparently carry a few voice channels from a residential customer to the PSTN (see Figure 3).
VoIP will typically implement three main blocks-voice processing, packet processing, and control and signaling. The voice processing function consists of echo cancellation, compression, and tone detection and generation as well as silence detection and suppression. The packet processing function consists of conversion between TDM and packets (packetization), implementing the packet-switched network protocol stack and providing a jitter buffer to compensate for PDV and clock recovery (optional). The control and signaling consists of telephony functions and packet-switched network call control such as H.323 or MGCP (media gateway control protocol).
CESoP removes the voice processing block entirely, significantly reducing the cost and complexity of the hardware and software in the customer equipment. CESoP tunnels the voice channels back to the PSTN-connected optical line terminator, avoiding the need for local intelligence to handle call control processing and gateway signaling functions.
With the abundant bandwidth available in EPON, for simplicity a CESoP connection may be permanently established to carry all the voice channels provided by an ONU. For an ONU with 32 analog POTS lines, the CESoP connection bandwidth may be on the order of 2.5 Mbits/sec when taking the packet header overhead into consideration.
The use of CESoP over EPON enables reliable and cost-effective voice services as part of the triple play offering for residential and business customers. CESoP may be used to provide fractional, leased-line, or private-line T1/E1 service for business customers or voice channels over POTS lines for residential customers. For business customers, the ability to provide T1/E1 service to a voice PBX is critical to completing the triple play package. A CESoP connection seamlessly delivers voice services with the same performance quality of a pure TDM network.
Peter Meyer is an applications engineer at Zarlink Semiconductor (Ottawa, Ontario).