What do I need to know about voice quality?
VoIP voice quality - the differences are grave
Voice quality is a challenge for internet telephony. Experienced providers of Voice-over-IP (VoIP) telephony use a variety of methods to ensure interference-free telephony. The quality in these networks of experienced providers is high. However, it should also be noted that there are providers who offer internet telephony on the market, which unfortunately is ”cheap” in the double sense of the word. While a ”shaky connection” may just be acceptable to private customers, it can become ”catastrophic” for business customers: Annoying cracking and noise caused by poor telephone connections can make efficient communication considerably more difficult or, in the worst case, even impossible. Therefore, when switching to an internet telephony solution, companies should strictly ensure that their provider can guarantee high VoIP voice quality.
Voice transmission requires a continuous stream of data to prevent dropouts, delays or reverberation effects. This is in direct contrast to the internet protocol, which allows the data stream to be broken down into individual data packets and the transmission of these packets via arbitrary and, above all, different paths. Traditional telephony is based on a single connection, through which the entire voice transmission takes place. Internet telephony, on the other hand – Voice over Internet Protocol (VoIP) – is based on a large number of transmission paths, whereby each individual data packet is determined separately as to which way it takes. This can cause data packets to pass each other up, get lost or arrive in very different quality. These effects, which are hardly noticeable when setting up a website or sending an e-mail, can have a devastating effect on voice transmission. The potential consequences: Failures, delays and fluctuations.
VoIP is as good as ISDN, but the network…
In the pioneering days of VoIP, people were happy, if at all, to hear a voice, comparable to the early years in mobile communications. Since the market spread of internet telephony, however, it has to put up with the comparison with the classical ISDN quality - and the voice quality of ISDN was and is undisputedly very high. To avoid any misunderstandings: From a technological point of view, VoIP offers the same high voice quality as ISDN, but the performance of VoIP can only be as good as the transmission network allows. This applies both to the internet connection and of course to the company’s own network. Exactly this network quality was not guaranteed in the VoIP initial years - and unfortunately it is still not available for all VoIP providers.
But today’s good commercial VoIP providers have developed and implemented a variety of quality assurance procedures for internet telephony. Modern VoIP systems are generally robust, even when faced with higher loss rates for voice data packets, and work with sophisticated mechanisms for error concealment, so that interferences are not obvious and calls can be made in good voice quality.
The extent of the differences on the supplier side is shown by a recent test conducted by the renowned trade magazine connect, which compared a total of eight providers of virtual telephony systems. toplink and Deutsche Telekom occupy the two leading positions in terms of voice quality.
Minimum requirements for VoIP
In the case of VoIP, the greatest possible care must be taken to ensure that fundamental requirements for the structure, sizing and operation of IP networks are met. This applies not only to Internet telephony, but to any kind of real-time applications. The error-free transmission of real-time communication, such as voice, is a considerable challenge for any IP network compared to traditional data traffic. VoIP reacts sensitively to errors in the transmission network, which is particularly pronounced by negative effects on voice quality. Speech quality is defined as the subjective perception of speech information. The objective parameters that influence the speech quality are the packet delay times (delay), the variance of these jitter times and the packet loss during transmission. Each VoIP packet is basically transported in a race against other IP packages. Buffers are available at many points in the network, the extent of which depends on the respective network load. This means that the packet throughput time varies from package to package. If a packet is transmitted for too long, it reaches the receiver too late for the phone call or gets lost, plus packets can be lost due to bit errors or overfilled buffers.
Error compensation is essential
In reality, there is no permanently error-free IP data transmission. This is not provided for in the networks and the respective standards. The internet Protocol is characterized by the fact that it anticipates network disturbances and searches for the best available channels in a faulty network. This concept leads to the fact that, due to loads in the network, long packet transmission times, high jitter and high loss rates occur again and again. Even in high-performance networks, a packet loss of 3 to 5 percent can occasionally be observed. The transmission times also vary greatly, often a packet delay of up to 150 ms is classified as a good transmission. For a short time, these values can be significantly exceeded. However: If 5 per cent of the call is not transmitted at all and the remaining 95 per cent is transmitted extremely slowly, the call is of course no more fun.
To compensate for packet losses, error compensation methods such as retransmission are used. However, real-time applications such as telephony also present considerable challenges for error compensation. Frequently, corrections do not provide sufficient time for a new transmission or, in certain procedures, a change to multiple transmissions leads to a self-reinforcement of the network load.
Another frequently used method is the concealment of errors. In simple terms, a data packet that is lost on the way is replaced by a substitute signal on the receiver’s side. This substitute signal can be generated in different ways. Noise suppression is particularly popular. The term may sound peculiar, but in practice the procedure works quite well, because the brain of the receiver tries to continuously replace the missing information itself. This method is often also used in mobile phone networks. The effect is significantly better than a clearly recognizable interruption. Repetition of packets at the receiver is also a common procedure. The missing packet is replaced by the last still correctly received packet. Optionally, the signal can still be attenuated with a waveform, which fades it in or out. In some cases, the following packet is already received after the missing packet is used for signal detection. The receiver tries to interpolate between the last and next package with intermediate values. For this purpose, the packet must be received early after the missing packet, or a large jitter buffer causes the receiver to delay the packet. The effect for the listener is often amazingly good.
Sufficient bandwidth is required
Ultimately, it is important that sufficient bandwidth is always available for telephony in the network. For example, reservation of a minimum bandwidth for VoIP as well as prioritization functions in switches such as class of service, port and service prioritization are helpful here. Depending on the labeling of the voice packets, quality of service mechanisms then intervene in the switches in order to give voice transmission priority over less time-critical data such as e-mails.
An experienced VoIP provider has optimized its network infrastructure and uses all methods to guarantee high voice quality.