VoIP Voice and Fax Signal Processing

11.1 VOICE COMPRESSION AND BIT RATE OVERVIEW

In VoIP, voice samples are collected from a telephone interface at an 8-kHz sampling rate. These samples are interfaced to the processor using a pulse code modulation (PCM) time division multiplexing (TDM) serial interface. Every sample on the PCM interface of the VoIP system uses either 16-bit linear or 8-bit A/μ-law. Voice modules operate on linear samples, and logarithmic A-law or μ-law 8-bit samples are converted to linear 16-bit samples. Modules such as echo cancellation, G.711, G.729AB, and G.723.1 compression process linear samples. For every linear input sample of 16 bits, G.711 gives an 8-bit μ-law or A-law sample. G.729A operates on a block of 80 sample (10-ms duration) frames and gives 10 bytes of compressed payload at an 8-kbps compression rate. G.723.1 operates on a block of 240 sample (30-ms duration) frames and delivers 24 bytes (6.3 kbps) and 20 bytes (5.3 kbps) depending on the selected rate.

A compressed voice frame is created into a packet with required headers to ensure end-to-end delivery on an IP network. As an example, G.729A takes 80 samples of input and gives 10 bytes of compressed frame in 10 ms. On the Ethernet interface, 10 bytes of basic payload is created as a total of 88 bytes. On the DSL interface with point-to-point protocol over Ethernet (PPPoE), the same 10 bytes of G.729A will use a total of 106 bytes to make the payload utilization 9 (10/106 = 9) percent. The headers on top of compressed payload are ...