The elements that influence the quality of interactive voice shown in Fig-
ure 18.1, from the left to the right are:
■■ Far end echothat is due to the feedback from the far end. A user having
an analog phone, as shown, may cause the feedback in the “4 wire to 2
wire” element, also called thehybridin analog telephony. In VoIP
adapters or phones, the equivalent of the hybrid is a digital signal pro-
cessing (DSP) application. Tail lengthis the amount of network delay to
the left over which echo is controlled in the hybrid. A typical value is 64
ms for a voice sampling rate of 8 kHz. Far end echo is compensated by
the echo canceller shown on the right (adaptive filter).
■■ The network can contribute with impairments to quality that will be
discussed in the following section on Internet performance.
■■ The coderand decoderform the codecand serve to convert the digitized
voice signal into a format suitable for transmission over the Internet,
which is RTP media packets.
■■ The near end echoshown on the right is also called the sidetone. A sidetone
is manifested most commonly when using multimedia PCs or laptops
without a headset. The sidetone and the far end echo can be compen-
sated for by using an adaptive DSP filter for the voice application.
■■ There are other sources of impairments, such as noise that sometimes
has to be locally compensated for. An interesting observation is the fact
that digital transmission over the Internet is practically noise-free, and
this makes users unsure if the session is still alive. For this reason, many
codec packages introduce comfort noise for the reassurance of the user.Internet Codecs
Most telephony codecs used at present by telephone company-provided VoIP
are part of the legacy ITU-T G.7xx series codecs designed for 3.1 kHz audio
bandwidth. These were first developed for the now defunct ATM-PSTN gate-
ways and are technically obsolete, with a few exceptions, such as the narrow-
band G.723.1 narrowband codec. There are more than 25 flavors of ITU-T
legacy codecs and, to our knowledge, all but the G.711 codec (which is based
on 50-year-old PCM technology) require license fees (which may explain their
longevity with the legacy telecom vendors).
State-of-the-art Internet codecs feature a wide range of audio bandwidths
that can deliver better-than-PSTN-quality voice and conference-room-quality
sound, while at the same time using less bandwidth and having higher resilience
to packet loss.
Quality of Service for Real-Time Internet Communications 305