With convergence becoming a reality, many enterprises don’t see economic viability in running separate networks for voice and data. While running a packet-based network is great for data, there are many problem issues on the voice quality front. Sandeep Sharma identifies the issues and provides some solutions

The move towards packet-based networks is driven by the fact that data is increasingly replacing voice in terms of traffic in networks. This is primarily due to the convergence of the public telephone network and the Internet. Operating two separate networks for data and voice transmission does not make economic sense for operators. Therefore, they are increasingly looking at consolidating their networks which are optimised to carry both data and voice in a single network. And since data is most efficiently carried in packet networks, it is not surprising that the integrated network for both voice and data are packet-based networks.

But the transfer of voice over packet networks raises many issues relating to the quality of voice. It is primarily because the packet network is logically optimised to match the requirements of data traffic. Secondly the access links that are dedicated to voice in traditional circuit-switched networks may be shared between voice and data in a packet-switched environment. Apart from this internetworking between circuit-switched PSTN (Public Switched Telephone Network), PBXs, (Private Branch Exchanges) and other networks such as wireless networks also put forward quality management issues. Primarily, end-to-end voice quality in packet transmission depends upon the speech codec (Coder-Decoder) used, end-to-end delay across the network and variation in the delay, packet loss across the channel and echo control.

To achieve high quality end-to-end voice transmission it is important to understand these issues and find out ways and means to reduce the bottlenecks that arise due to these factors. For instance, selecting the right codec is quite essential. This is because codec performance includes the baseline quality and the performance with voice impairments present. A codec essentially converts the analogue voice signal to a digitised bitstream at one end of the call, to its analogue form at the other end of the call. In telephone networks generally two techniques are used—waveform coding or CELP (Code Excited Linear Predictive) coding. As far as codecs are concerned the main delay is the packetisation delay. Generally packet transmission offers the flexibility to use different codecs as needed. In choosing a codec for a particular call or application, there are several considerations like the compression rate needed, the desired voice quality, the delay that the codec adds to the connection, how well the codec allows missing packets to be smoothed over, etc. But when packet loss is introduced these codecs show different amounts of degradation and depend heavily upon the packet loss concealment algorithm. There are many effects of packet loss. Some of the major effects are end-to-end delay, processing delay, propagation delay, buffering delay, etc.

End-to-end delay
The end-to-end delay, also known as latency, is the time between the generation of a sound at one end of a call and its reception at the other end. Delay causes two different impairments. First, as the delay increases echo becomes more noticeable, and secondly, when the delay becomes long enough, it disrupts conversation dynamics, making communication difficult. While impairments such as echo and noise can be reduced, little can be done to lessen the delay caused by transmission media or packetisation of analogue voice signal by codecs. In the conventional PSTN, the largest part of the end-to-end delay is the propagation time of the transport medium.

Packetised voice also encounters significant processing delay and additional delays created by queuing and jitter buffers. To minimise these queuing and propagation delays, network processing must be streamlined and packets carrying interactive voice communication must be given priority over data packets.

Processing delay
Even though processing delay is much less than delays created by queuing and propagation delays, these cannot be ignored when it comes to achieving acceptable voice quality. Processing delay includes the time taken for encoding and decoding speech, collecting the voice data into packets, etc. When we look at propagation delay, it is associated with sending a signal over a substantial distance. For instance, a fibre optic trunk imposes a propagation delay of about 5 microseconds per kilometre. By controlling the topology of the network, such types of propagation delays can be reduced by ensuring that packets take the most direct routes.

Buffer issues
Buffering also adds up to affect the quality of voice in packet networks. Buffers are used for queuing at routers and to control packet arrival time at the decoder, and data waits in the buffer for processing and propagation. Since voice playback speed must be constant, a jitter buffer is used to remove variation in the flow of packets to the decoder. In cases where congestion control at the network nodes is implemented, the jitter can be fairly controlled. But in cases where there is uncontrolled jitter it causes impairment in conversation. For instance, longer delays cause simultaneous starts and awkward silences. It then becomes quite difficult for the parties to interpret each other.

Echo problems
Apart from these, echo is also a major cause for concern as far as voice quality in packet networks is concerned. Echo in the network results from the coupling between the transmit path and the receive path, which causes the outgoing speech to be sent back to the talker. Echo that is inaudible in the circuit-switched network may become quite noticeable in packet networks because of the increasing delay (time between the original voice and the return echo). Interconnections between packet networks and circuit-switched networks are especially susceptible to echo impairment. The delay associated with packet transmission contravenes the engineering assumptions of
circuit-switched networks. Therefore echo control at the interface between the networks is essential to protect users at both the ends from hearing the echo. Some of the echo-reduction techniques that can be used individually or in combination with others are echo cancellers, echo suppressors and loss level planning.

An echo canceller is a device that looks for an echo (a delayed signal on the return path that is strongly correlated with the signal seen on the incoming path) and uses an adaptive filter to model the echo and then subtract it from the return signal. Whereas an echo suppressor or voice switch detects a signal on the incoming or outgoing path and switches attenuation to the other path to reduce the level of any returning signal. An echo that arises due to shorter delays can be managed effectively by introducing loss in the path. Loss planning is a key strategy in the management of network echo.

Lost packets
Packet loss is a major aspect that determines the quality of voice in packet networks. When we take a look at traditional TDM networks, a call is assigned a physical connection between end-points, and the circuit remains dedicated to that channel for the duration of the call. However, in packet networks voice, fax or data is broken into small samples or packets of information. Each packet is assigned a header that identifies where the packet is going and contains the information for reassembly when the packet arrives at its destination. Each packet travels independently and the travel time varies for individual packets. Therefore, unless the network is precisely matched to the peak traffic load packets sometime fail to arrive at the destination. These lost packets do not matter when data transmission is concerned, but for an application like voice it creates a lot of problems. If packets do not arrive in time at the destination they create gaps in voice communication that can result in clicks, muting or unintelligible speech. In a network running without call admission control, and without quality of service (QoS) protocol being enabled, packet loss is uncontrollable in the event of congestion. The consequences of congestion depend on the type of network, the proportion of voice and data traffic, the number of hops, and the duration of the event.

Some of the strategies that can be used for minimising lost packets are implementation of QoS protocols, call admission control, adaptive jitter buffer, sending duplicate data, concealing missing data, etc. Implementation of QoS protocols in network devices hastens the transmission of voice packets at various gateways and routers, minimising jitter and the resultant lost packets. In networks with a high proportion of voice traffic, call admission control can prevent congestion by limiting the number of calls that can be active through various nodes in the network. Whereas when a voice packet arrives at the destination it is held in the jitter buffer until the decoder is ready for it. In the process, packets arriving late are discarded. An adaptive algorithm can be used to adjust the jitter buffer delay as the packet loss rate rises and falls. This adjustment helps minimise the number of late packets when the system gets congested and avoids adding unnecessary delay when congestion eases. Apart from these, sending redundant data also corrects voice packet loss. For this, information is copied into the next packet in the sequence and is used if the original packet is lost or delayed. As far as concealment of lost packets is concerned, it can camouflage gaps in the output signals using the simplest techniques. This will require a little extra processing power, but more sophisticated techniques can restore speech to approximately original levels.
While choosing the right technology for voice in packet networks, all these factors need to be kept in mind. A fair understanding of these quality factors would provide a solid grounding for selecting voice quality targets to meet business needs and for planning and provisioning a network to achieve these targets.

The author is the manager, Product Marketing for VoIP & Intelligent Internet at Nortel Networks.