The codec plays one of the most important roles in Voice over IP. It determines voice quality and how much bandwidth is needed for a call. The rule of thumb is the higher the voice quality, the more bandwidth is required for the call. Several different codec standards exist that allow network designers to choose the best compromise between sound quality and bandwidth requirement.

Here’s how it works. Your voice, an analog signal, is sampled several thousand times per second and converted into a digital signal. If the codec calls for compression, the signal is then compressed to significantly reduce the bandwidth needed for transmission. The compressed signal is put into packets (each packet typically contains about 20-30 ms of audio) and sent over the IP network to its destination, where it undergoes a reverse process to be converted back to an analog speech signal.

VoIP codecs are designed specifically for the frequency range of the human voice. If you tried to pass a music sample through one of these codecs, the sound quality would not be very good.

VoIP codecs are defined and standardized by the ITU, International Telecommunications Union. Each uses a different algorithm to sample, compress, and packetize voice signals. The table below shows the main codec types with their respective bitrate and voice quality. Voice quality is represented by a MOS (Mean Opinion Score). MOS is a popular subjective measurement of voice quality, also defined by the ITU. The voice sample is judged by a group of individuals and then given a score on a scale from 1 (bad) to 5 (excellent).


Compression delay (also known as packetization delay) is the amount of time the codec needs to do its job. Cisco suggests that as a general rule you should aim for a packetization delay of no more than 30 ms.

Some degree of compression is usually required for networks. Take DSL service, where upstream links are usually limited to 128 kbps. If a VoIP provider uses G.711 (uncompressed codec), the actual bandwidth needed for a single channel is about 100 kbps, meaning VoIP providers could only offer a single voice channel over the link. By compressing voice, the provider can now offer multiple channels over the same link.

Codecs can contain other advanced tricks to optimize voice quality and bandwidth use. Some codecs use PLC, or Packet Loss Concealment, to compensate for lost packets. With PLC, codecs try to minimize the impact of packet loss by filling in the gaps with audio that’s acceptable to the human ear. There are other methods for dealing with packet loss, but PLC is the most common.

Some codecs also support silence suppression. Essentially, this follows the rule of “don’t send any data if no one is talking.” This method is far more efficient than the traditional phone system (or PSTN) which keeps a full channel (in both directions) dedicated to a call regardless if anyone is talking or not.

See also:

• Bandwidth Requirements
• Low-Bandwidth Solutions
• Asterisk: Open-Source PBX

Request a Page