Technical data

Architecture and System Requirements
EN4442EN00EN_0002
© 2007 Siemens AG
Siemens
14
For telephony applications, voice is sampled every 125 μs at 8bits. It would not make sense to
package each individual 8 bit sample into an IP packet. Instead, for G.711 encoding at least
240 8 bit samples are combined into an IP packet.
It takes 30 ms to acquire 240 voice samples.
The Sampling Time is defined as the time it takes to sample a specified number of voice
samples that are sent out in one IP packet.
It is obvious that the longer the sampling time, the smaller the IP overhead gets. Thus, in order
to minimize bandwidth, the sampling time will have to be increased.
However, increasing the sampling time will increase the overall delay!
For IP line gateways the sampling time can be set between 10 ms and 90 ms depending on
codec type.
2.5.3 IP Overhead
The IP overhead includes the Real Time Protocol (RTP), User Datagram Protocol (UDP) and
IP headers as well as the Ethernet framing and additional octets for QOS Tagging.
A typical IP overhead value for G.711 is 30% (see Table 2-1 IP Overhead and BW per
Connection for more details).
2.5.4 Required Bandwidth per Connection
The following table shows the required IP bandwidth and IP overhead per voice connection as
a function of Sampling Time and encoding:
>
Important
There is always a tradeoff between minimizing the delay versus minimizing the
required bandwidth.
Encoding Sampling Time IP Overhead BW per Connection
G.711 10ms 88% 120.0 Kbps
20ms 44% 92.0 Kbps
30ms 29% 82.7 Kbps
40ms 22% 78.0 Kbps
50ms 18% 75.2 Kbps
Table 2-1 IP Overhead and BW per Connection