IP voice quality: Getting a handle on echo control

Voice remains an important revenue stream for operators, as well as the most complex media type to control and deliver with guaranteed quality. The allocation of a heavy complement of dedicated channels for voice -essentially a case of throwing bandwidth at voice transmission - is the principal means of ensuring that voice quality is retained to a consistently high degree. For operators, this necessitates tying up valuable bandwidth for voice-only transmission, rather than being able to use the network to its fullest capability.

VoIP applications represent a particularly challenging technical environment - particularly in the context of echo control. VoIP has to provide the end-user with the appropriate level of communications transparency, involving mechanisms that are completely unnoticed during the "call". This requires the deployment of technologies such as automatic level control (as in ITU Recommendation G.169) and Noise Reduction to restore natural, comfortable speech.

The service network offering must therefore include a set of features that improve the degree of transparency experienced by the end-user with the connection. Obviously, different levels of operators will have different agendas- second-tier operators, for example, will be more interested in just enabling the initial connection to be established and maintained. For first-tier operators, however, the emphasis will be on making that connection not only fast but also efficient, easy to use, and with the highest voice quality, in other words, transparent.

Voice quality can vary on VoIP networks tremendously, due to a variety of different factors - from the gateway equipment, to the phone systems being utilized, the client software and the carrier infrastructure. In order to understand the degree of complexity that ensues, there are two groups of impairments to take into account.

First, there are the packet-specific impairments, which include the inherent issue of delay (latency),delay variation (jitter), packet loss and coding distortion. All of these impairments affect voice quality either directly by interfering with the coding of the speech or indirectly by adding delay and noise.

The second group of impairments are those which affect all networks, not merely those with packet backbones. They include echo, variation in signal levels and noise levels. When the sum of these effects are combined, the overall result creates a considerably more complex scenario.

Product engineers need to focus on the control of impairments independently of the core transmission scheme. The main objective is clearly the removal of echo, but it is the manner in which this is accomplished which is critical. The speed and effectiveness of the echo canceller has to be judged alongside the way in which the equipment handles other application-specific but quality-related tasks - such as level control and noise reduction techniques. It is important to ensure that the echo canceller algorithm meets G.168 (2000) or later. Echo cancelling schemes should be tested against as many real-world signal types and network configurations. Echo cancelling is easy to do well most of the time but very hard to do well all the time. Operators need good performance all the time to avoid costly customer complaints and churn.

Deregulation has meant that the controls surrounding the maintenance of the optimum signal levels in a network are being removed, resulting in a growing problem for operators. ITU Recommendation G.169 is providing the basis for ensuring that this problem becomes controlled more effectively in the future, setting out the requirements for the control of signal level variation through a range of equipment. The echo canceller represents an important tool in this respect, monitoring levels as they come into the network and enabling a range of level control options to be deployed to compensate for variations.

The impact of noise goes beyond simply improving the signal-to-noise ratio (SNR) of the received speech since noise also impairs the functioning of the voice-encoding algorithms that are used. When noise is more speech-like, the impact on the voice-encoding algorithms can be even more pronounced. The resulting received speech can be quite distorted, and the varying degrees of accuracy in encoding the background noise can in itself be annoying.

As with signal level control, the advanced approach to this scenario involves applying sophisticated application-specific software to a high-performance echo cancellation platform. This optimizes the canceller's capability to cope with the varying characteristics of a call, while combining it with a range of additional functionality. Testing with real-world scenarios is critical. White noise is easy to generate in the lab but not indicative of real-world performance.

One thing is very clear in the VoIP domain: Nothing is particularly easy or straightforward. There exists a myriad of avenues and options open to operators as they define their strategies for the short, medium and long term. By 2010, we are likely to have largely completed the transition to packet-switched networks per se, with the migration involving a significantly more important role for IP switching. However, the road to achieving this is paved with pitfalls. Signal processing is extremely tricky - pick the wrong options and an operator faces an unenviable volume of field complaints and the resultant marketing fall-out.

In the end, one point is undeniable: Operators have to improve their time to market for new services, combined with lowering their operating overheads in order to survive. There is no doubt that the move to IP at Layer 3 is a fundamental requirement to accomplish this.

See related chart

However, the rewards for having IP transport only come several years down the line, and there is a short-term requirement to maintain existing revenue streams - namely voice. It is clear that customers will not remain with a carrier that provides inferior voice quality - nor, in many developed countries, will they switch as the cost differential continues to drop. Call quality is clearly one of the most vital differentiating factors in any business strategy. This is the price that operators will have to pay - investing in an infrastructure that ensures that the precious revenue streams from carrier-grade voice can be preserved for the future.