VoIP hybrid designs at net edge require synchronization

The migration from circuit to packet will be implemented in stages. In the meantime, hybrid circuit/packet environments will exist for many years.

Because of synchronization issues, the marriage between the two communications methodologies will require careful engineering design. In circuit switching networks, sometimes referred to as connection oriented networks, a physical path is dedicated to a single connection between two end points in the network for the duration of the connection. All resources on that circuit are unavailable for all other users.

The main advantage of these networks is that they are ideal for communications that require data to be transmitted in real time. Applications well suited to circuit switching include landline telephony (live audio and video) and conventional cellular radio. Two disadvantages of a circuit switched network are that bandwidth is not efficiently utilized and is static.

Packet switching routes data using addressed packets such that a channel is occupied during the transmission of the packet only, and upon completion of the transmission the channel is made available for the transfer of other packets. Each packet is transmitted individually and can follow different routes to its destination. Once all the packets arrive at the destination, they are assembled and rearranged if necessary. Although packet switching is essentially connectionless, a packet switching network can be made connection oriented by using a higher-level protocol. For example, TCP, makes IP networks connection oriented.

Packetizing data allows the same path to be shared among many users in the network. This is one of the main advantages of packet switching. Packet switching is flexible and available bandwidth can be used more efficiently, provided some amount of delay is acceptable during transmissions. Packet switching was originally designed to carry non real time data and applications include email, the Internet, automated teller machines and credit card validation.

Although packet switching is more flexible, efficient, cost effective, and faster than circuit switching, there are disadvantages associated with the technology including unknown, uncontrollable and variable delays and packet loss.

In this environment, it is critical that Voice over Packet (VoP) applications meet the challenge of transporting legacy voice traffic over packet networks. Since voice packets need to be assembled and possibly rearranged at the destination, long and variable delays can result causing unnatural or unintelligible speech. Delay also causes echo. Echo is the signal reflection of the speaker's voice from the far end telephone equipment back into the speaker's ear. Echo becomes disruptive when the round trip delay is greater than 50milliseconds.

Another challenge is corrupted and/or lost (dropped) packets during transmission. Data is not time sensitive and lost data packets can be retransmitted. Voice, however, is real-time so lost voice packets cannot be simply retransmitted. Corrupt and/or lost packets voice packets result in clipped speech, voice distortion, dropped calls, and 'pops' during conversations.

Synchronization key

Proper synchronization reduces the number of lost packets during transmissions. Cell loss ratio (CLR) is the ratio of total lost cells (or packets) to total transmitted cells (or packets). CLR is proportional to the long-term accuracy of the timing reference in the networks - the better the synchronization level, the smaller the CLR.

For ATM networks, ITU-T I.356 specifies a cell loss ratio of less than 3X10-7 for the stringent Class 1. For IP networks, ITU-T Y.1541 Quality of Service (QoS) Classes 0 through 4 specifies a cell loss ratio of less than 10-3. Because IP networks do not guarantee service, they will usually exhibit a much higher incidence of lost voice packets than ATM networks. In current IP networks, all voice packet frames are treated like data. A cell loss ratio of less than 10-3 is not stringent enough for VoIP applications and a proposal for another IP QoS class with tighter cell loss ratio is under consideration.

To understand this problem it is useful to look at how a generic hybrid packet/TDM network works. It is assumed that the voice and voice band data calls originate from a private enterprise network using both TDM and packet technologies and complete over the PSTN (TDM) network. During transition from one PBX to another within a corporate network the voice and voice band data traffic streams are converted to packets. Following packet transport, the packets enter a TDM network through another gateway where the packets are buffered and converted to octets in the TDM streams. The TDM network provides the remaining connection for termination at the destination.

It is possible that the timing relationships between the two streams differ. While both may have timing references with long-term accuracies of 1X10-11 (stratum 1), one network may also have a free running clock with a stratum 3 accuracy (4.6X10-6).

In both cases, de-jitter buffers in the packet gateways will absorb the phase change until there is a need for a buffer adjustment either because the buffer has grown too large or the buffer has shrunk to the point where the intended range of delay variation cannot be accommodated, and too many losses occur.

There are currently no synchronization standards and/or guidelines for hybrid packet/circuit networks. But it can be shown that packet loss ratio is proportional to the long-term accuracy of the timing reference. The better the synchronization level, the smaller the packet/cell loss ratio. The CLR is very small when both networks have Stratum 1 timing and CLR requirements in ITU-T I.356 and ITU-T Y1541 are met.

The most effective way to reduce packet loss due to synchronization is to maximize voice performance in an ATM or IP network. This means that the gateway one must receive timing from an independent stratum 1 source, such as GPS and the second gateway must also receive timing from an independent stratum 1 source, such as GPS, or from the TDM network, provided it is stratum 1

Lack of synchronization also effects modem connections. Modem connections cannot be made and modem disconnects will occur when offsets exceed 225 parts per million. In a network where the two IP gateways and access server are not synchronized the offset of the internal oscillator of the network access server can reach 17.5X10-5 much worse than even stratum 4 requirements of 3.2X10-5. When the offset reaches 22.5X10-5, the modem connection was lost.

When the network was synchronized with GPS primary reference sources, the short-term wander was 1.15X10-6 sec peak to peak and the modem connection was easily made and not dropped.

Important as they are to the proper operation of VoIP, there are currently no synchronization standards or guidelines that exist for hybrid C2P networks. However, packet performance requirements are stated in ITU-T I.356 for ATM networks and ITU-T Y.1541 for IP networks.

As a result it is still up to the engineer to maintain a sufficient level of synchronization between the networks to satisfy the packet performance requirements, ensure voice quality and prevent modem disconnects.

See related chart