The large majority of voice services today are carried over legacy telco networks. This legacy is the result of many decades of telcos deploying mainly TDM-based services, from enterprise to residential to cellular.

The residential voice market is finally being penetrated at significant levels by cable operators offering VoIP services. In commercial markets, organic growth has started, but combined voice and data still only accounts for around 2 percent of total business spending. The reasons for continued telco dominance in commercial voice are many: incumbency (including an existing sales force), network capacity requirements, cost of deploying infrastructure, and old perceptions about cable's performance, reliability, and availability.

The old "standard" for business and commercial voice has been the T-1 line, a standard that was defined for telco networks. But T-1s cost more to deploy, offer limited capacity, and are costly to maintain in a decaying copper network. Ironically, outdated telco-based standards are still used to measure the acceptability of new service offerings, thus artificially extending the telco lead in business and commercial services.

The new standard is Carrier Ethernet (CE). A packet-based alternative to circuit-switched networks, Carrier Ethernet provides a set of end-to-end standards that allows packet-based traffic to meet all the needs of current users of circuit-switched network services. In doing so, Carrier Ethernet also defines a strategic, standards-based path for MSOs to level the playing field with the telcos in the business services market–at just the moment when the telcos are shifting to this new network to reduce cost and improve service to end users.

This article will show the performance and capacity requirements that Carrier Ethernet must meet in order to achieve its potential as the new standard for voice. In addition, a new architecture for HFC, namely fiber-to-the-curb (FTTC), will be used to illustrate how aggressive cost objectives and capacity requirements can be met using this platform.

Core requirements for CE internetworking

The leading body for standardization of Carrier Ethernet is the Metro Ethernet Forum (MEF). Metro Ethernet Services standards do not specify the physical layer for the access network, but rather, the service provided. Carrier Ethernet services can be provided over fiber, copper, coax and wireless. CE can support T-1 services using Pseudowire Equipment (PWE).

Table 1
Table 1: Performance specifications.

The MEF 8 Implementation Agreement is the standard used when providing a service that is handed off to another operator or to another system within a given MSO. In MEF 8, the key requirements may be summarized as follows:

  • Network availability >99.95 percent
  • Latency, or Frame delay <25ms
  • Ethernet Frame Delay Variation <10 msec
  • Ethernet Frame Loss defined as 2.5 x 10-6 Frame Error Ratio (FER). This is equivalent to a payload BER of 10-6, selected because an end-to-end PCM-encoded voice signal tolerates a 10-6 BER in the TDM circuit (Source: SR-TSV-00275, Issue 1, March 1991).
  • Maximum end-to-end wander (PSTN to TDM CPE) is 18 microseconds over 24 hours and 8.4 microseconds over 900 seconds.
Requirements for cell backhaul

Cell towers have the most stringent requirements for voice applications, with latency and jitter being the key factors. The MEF specifications of 25 msec latency and 10 msec jitter provide a suitable requirement for business voice services, but these are not sufficient for all cell networks, as shown in Table 1.

The limiting factor in the provision of backhaul circuits for the cellular market is differential delay, with CDMA having the strictest requirements. Differential delay is the difference in path length (msec) from one base transceiver station (BTS) to an adjacent BTS. This difference in delay may be due purely to distance, but can also be caused by different physical media (e.g. copper has a shorter delay than fiber due to fiber's refractivity). To achieve soft handoff from one tower to the next, the differential delay in paths may not exceed 8 msec in the case of the most demanding cell switches, although for some vendor equipment, figures as high as 20 msec are acceptable. Since the 8 msec limit includes PWE encoding on the order of 3 msec (with low jitter buffer settings), the Ethernet circuit must meet a requirement of 5 msec latency.

Thus, for business class voice, a one-way end-to-end latency of 25 msec is suitable, but for cellular backhaul, a 5 msec Ethernet or 8 msec TDM latency is required. Similarly, since jitter (frame delay variation) impacts maximum latency and call handoff performance, 5 msec delay variation is the limit for cellular backhaul, while 10 msec as specified by the MEF is an acceptable figure for business services.

Thus, MEF specs can be used as common specs for carrier-class voice and data, while cell towers add incremental application specific requirements. Performance specs are summarized in Table 1.

Quality metrics

Quality of service has typically been measured through availability, BER or Severely Errored Seconds, and Error Free Seconds. With the advent of VoIP, Mean Opinion Scoring (MOS scoring) has taken on new importance. Measured on a scale of 1 to 5, with 5 being perfect, MOS scores of 4.0 or higher ensure that normal users perceive no difference between a TDM and a VoIP delivered call. MOS scoring is similar to the ITU approach to subjective assessment of video quality.

In terms of quality measures for telco copper T-1 circuits (access portion), specific ANSI/ITU specifications apply (see Table 2).

Table 2
Table 2: ANSI/ITU T-1 quality metrics.

An Errored Second is defined as a one second interval with one or more bit errors. A Severely Errored Second is defined as a onesecond interval having a bit error ratio of 10-3 or worse. If either condition (ES or SES) is encountered, this is considered a degraded condition.

For business voice or data, a BER of 10-6 is considered adequate. This is well above the performance threshold required for VoIP quality and is transparent to applications using TCP and many other protocols. Once again, however, for cellular backhaul, more stringent performance criteria may be requested. Mobile network operators typically request a BER of 10-8 measured over a 30-day period.


The most fluid component of the quality equation is availability, defined as follows: a circuit is unavailable when the BER is worse than 10-3 for a period of 10 or more seconds. The ideal for availability has been 99.999 percent, referred to as "five-nines." However, availability is also a business decision, noting that availability and price are related. Within certain bounds, the customer will accept a lower availability if the price is right or if alternatives are limited. For example, the copper network providing T-1 service to cell towers only provides 99.9 percent availability, on average. Cell operators want 99.999 percent, but today, they get much less. The same want versus need gradient exists everywhere in the market. Availability, thus, is often more the basis of competition than a strict requirement.

Capacity requirements

The capacity usage for HFC nodes in typical business districts has been documented previously in CED (see March 2006 issue, Click Here), which is shown here (Table 3) for reference.

Table 3
Table 3: Telco capacity usage in HFC node areas.

A glance at these numbers gives a strong indication that an increased amount of fiber will be needed to serve the most attractive nodes.

HFC as the Carrier Ethernet vehicle

The first bit of very good news for cable operators is that long fiber runs that were constructed many years ago have essentially unlimited capacity at a very low cost. The use of wave division multiplexing (WDM) and Ethernet transport allow existing fibers to be leveraged for huge amounts of capacity. Using 10 GbE transport via CWDM, one pair of fibers can supply 80 Gbps to a node.

The second bit of very good news is that optical Ethernet capacity is inexpensive. The CWDM multiplexing equipment and the Ethernet interface converters (XFPs) cost about $0.50 per Mbps (symmetric). In a few years the price will drop to $0.10/Mbps.

How then, to get the most delivery power out of this raw capacity with the lowest cost while meeting all performance and capacity requirements? One option is to deepen fiber from node-depth to curb-depth and leverage coax with an HFC FTTC network. Since HFC plant is also being used to carry voice and data services, it is becoming increasingly important to avoid disruptions to the HFC network. The overlashing of new fibers in the trunking portion of the coax network not only extends the reach of the existing fiber but also preserves the legacy network without disruption to existing services.

Table 4
Table 4: Fiber trunk construction cost model.

The cost of fiber overlash varies depending on the amount of aerial vs. buried plant, and the buried plant can vary in cost based on the absence or presence of useable conduit. Table 4 shows a method to calculate the cost of fiber trunk construction.

Carrier Ethernet in the HFC network

It has been shown that very high-speed, low-cost Ethernet can be transported to the HFC node location via existing fiber, and that deep fiber trunks (FTTC) can be cost-effectively built. The task now is to use Carrier Ethernet transport equipment in the access network to distribute packets across this fiber, all the while meeting the stringent performance requirements described for voice services above.

Because the access network is, by definition, outdoors, outdoor Ethernet switches are used at various aggregation and tapping points that interconnect the new fiber trunks. Just below the coax tap level, on the drop cable, symmetric Ethernet flows can be combined with cable TV signals up to 1 GHz and carried on the existing coax drop cable. The Ethernet traffic rides above 1 GHz, and the cable TV traffic rides below 1 GHz (technically 1.002 GHz). Thus, 100 Mbps symmetric traffic can be carried along with the existing or expanded cable TV spectrum. Note that using 1 Gbps fiber trunking allows each subscriber to have dedicated, not shared, 100 Mbps symmetric Ethernet traffic.

Figure 1
Figure 1: Carrier Ethernet-over-HFC.

The cost-per-Mbps of the complete end-to-end FTTC system, including equipment and labor, can reach about $5 in the next few years.

The cable advantage

Carrier Ethernet, and the standardized equipment that meets the MEF specifications, allows cable operators to compete for voice services while gradually migrating to all-IP. Cable already has the best deep fiber network with vast amounts of low-cost Ethernet capacity available. The key to success is to deploy an access network solution that takes full advantage of cable's fiber network by delivering capacity and performance at the lowest possible cost. The sweet spot for the HFC network is FTTC, which uses existing fiber to the node, new fiber trunks, and existing coax drops where available with the option to deploy new fiber drops as needed.

Compare this to the (FTTP) PON and (FTTN and FTTC) DSL systems being built by telcos today. The telco challenge with PON is cost; equipment and labor come in well above $2,000 per customer. The telco challenge with DSL is performance; maximum speed is 25 Mbps within certain distance limits, and some portion of that must be used for video services.


The Carrier Ethernet movement is critically important for MSOs. A strategy to reduce infrastructure cost and improve service delivery to end users, Carrier Ethernet will drive a new, higher bar in the market for business services. It also serves to level the playing field for MSOs seeking to compete with telcos.

MSOs enter the competition with a fiber network that is much deeper than the telcos'. With xWDM and optical Ethernet, the raw transport capacity of the MSO network is massive and provides the basis for dramatic growth in revenues.

To unlock the full potential of the existing fiber network, MSOs need a cost-efficient access network solution that can deliver maximum capacity deep in the network while meeting all performance criteria.

Outdoor Ethernet switches deployed in a fiber-to-the-curb (FTTC) topology make maximum use of cable's fiber advantage, meet all known requirements, can be deployed without disruption, and are very affordable.