The Endpoint Quality for Voice
It can be argued that, in the absence of notable network-induced impairments,
the QoS for voice resides in the endpoint. A complex array of technologies is
required for high-quality interactive voice communications [10]:
■■ Network echo cancellation
■■ Acoustic echo cancellation
■■ Noise cancellation
■■ Automatic gain control for transmit and receive level
■■ Voice activity detection for bandwidth efficient transmission
■■ Comfort noise generationAs we will show in the next section, the global Internet can support adequate
to excellent performance for VoIP, with the exception of some less-developed
regions in the world.The Internet Performance
As the bandwidth in the Internet core increases, and broadband Internet access
is delivering ever higher speeds, so does the Internet performance increase
with regard to packet loss and delay.
Figure 18.3 illustrates how the monthly average packet loss over the Internet
has decreased over 10 years by an order of magnitude and is at present in the
0.1 percent to 1.00 percent range except in a very few regions of the world.
Most global ISPs guarantee average packet loss lower than 0.5 percent but
actually deliver average packet loss in the 0.1 percent to 0.2 percent range over
their networks. Going back to the codec performance in the presence of packet
loss in Figure 18.2, it is obvious that average packet loss in the 1 percent range
will not adversely affect codec performance in any way. Packet loss bursts may
degrade quality for voice, but we have not seen any published data for Inter-
net burst packet loss.
Figure 18.4 shows the global average delay measured over periods of 24
hours and 30 days, respectively.
These measurements prove that because of the very high-speed routers
deployed on the Internet, the average delays are close to the delay of the speed
of light in fiber-optic cables.308 Chapter 18