VoIP – Bandwidth, Codec and Voice Quality

VoIP (Voice over Internet Protocol) has established itself as a dominant force in telecommunications. SIP (Session Initiation Protocol) is the application layer for creating and controlling the calls and the voice is governed by RTP (Real Time Transport). When we are setting up a contact center or converting to VoIP, we want to know how to take advantage of the VoIP without getting into the pit falls of the technology.

In the digital world, your voice conversation needs to be sampled every 125 micro-second (8000 times within a second). On the receiving end, such samples can be re-assembled and converted to analog for you to hear the voice again. If it is sampled any less, the voice will not sound less like the original sound. They are encoded and decoded in both ends by CODEC. Let us look at the most common codec G 711 and its bandwidth requirements.

Each sample represented by 1 byte (8 bits) results in 64 Kbps often referred to as PCM or Pulse Code Modulation (u-Law U. S. A. and A-Law in Europe are the two flavors). These samples will be wrapped with address and other decoding details before transmission over the network. G711 typically is set to package 160 samples (20 ms worth of sampled voice data) to minimize the amount of header data and thereby conserve bandwidth. There is an upper limit of 30 ms, above which the encoding delay will start impacting voice quality.

With all the overheads taken into account, the bandwidth required for the single voice conversation in G711 is around 85Kbps both up and down. There is the time delay for the voice to reach from one end to the other. Any delay over 100 millisecond may be noticeable and make conversations difficult. This is the overall time taken for the voice conversation to reach the other end.

There are other reasons for the delay as well. In the realm of VoIP, the RTP uses UDP to send packets. There is no guarantee packets will reach the other end or arrive in the same order as they are sent. The receiving end cannot endlessly wait till all packets are received. There will be loss of packets. There are some well developed tricks including the concept of Jitter and Jitter buffers to manage this. Jitter is the measure of the difference in time to transport packets. Ideally we would want them to be equal and manageable. A Jitter buffer will allow the packets to be held back to the maximum delay so that all packets delayed equally.

Asterisk supports G711 and within you call center you can setup every agent use SIP. The issues of bandwidth and delay are not an issue. A sample layout is provided in the Indosoft blog site. At Indosoft, we have been deploying our call center software Q-Suite from 2003.