Codec Bandwidth/kbps Comments
G.711 64 Delivers precise speech transmission. Very low processor requirements. Needs at least 128 kbps for two-way. It is one of the oldest codecs around (1972) and works best in high bandwidth, which makes it a bit obsolete for the Internet but still good for LANs. It gives a MOS of 4.2 which is quite high, but optimal conditions have to be met.
G.722 48/56/64 Adapts to varying compressions and bandwidth is conserved with network congestion. It captures ranges of frequency twice as large as G.711, resulting in better quality and clarity, close to or even better than with PSTN.
G.723.1 5.3/6.3 High compression with high-quality audio. Can use with dial-up and with low bandwidth environments, since it works with a very low bit rate. It, however, requires more processor power.
G.726 16/24/32/40 An improved version of G.721 and G.723 (different from G.723.1)
G.729 8 Excellent bandwidth utilization. Error tolerant. This one is an improvement over others of similar naming, but it is licensed, meaning not free. End users indirectly pay for this license when they buy hardware (phone sets or gateways) that implement it.
GSM 13 High compression ratio. Free and available in many hardware and software platforms. The Same encoding is used in GSM cellphones (improved versions are often used nowadays). It offers a MOS of 3.7, which is not bad.
iLBC 15 Stands for Internet Low Bit Rate Codec. it has now been acquired by Google and is free. Robust to packet loss, it is used by many VoIP apps especially those with open source.
Speex 2.15 / 44 Minimizes bandwidth usage by using variable bit rate. It is one of the most preferred codecs used in many VoIP apps.
SILK 6 to 40 SILK has been developed by Skype and is now licensed out, being available as open-source freeware, which has made many other apps and services to use it. It is a base for the newest codec named Opus. WhatsApp is an example of an app using the Opus codec for voice calls.

When businesses understand that VoIP should be embraced and not rejected then the benefits just start rolling in . To embrace VoIP the business should really understand how it works and what they need to do to give it the best chance of working . Talking to a VoIP/cloud/hosted telephony specialist such as Phoenix Link ( www.phoenixlink.co.uk ) is a good place to start.

Voice signals , frequencies are  analogue….. not digital. Voice communication can only go onto a data network if it is converted into a digital equivalent that can be converted back into “voice” when it reaches its destination.

The conversion of audible sound into data is called quantization. The process of quantizing voice is typically done by sampling the sound at eight thousand times per second, assigning one of eight bits to each sample. This yields a 64-thousand bit-per-second data stream that can be reproduced at the other end.

The additional added packet information (overhead ) required to get the voice data stream across the data network when there are several simultaneous conversations can cause issues on the network , and therefore voice is generally compressed unless there is lots of available bandwidth or the voice data network is a separate network to the general data network.

Compression as a technology in not something that is new – it has been used for many years with fax machines. Voice communication can be compressed into a much smaller data stream ( than the original ) with almost undetectable loss of quality. Compression standards are expressed as CODEC ( Coder / Decoder ) , of which a common one is called G.729 which cuts the actual voice data stream down to 1/8 of the original size (the overhead remains the same). The highest quality voice has the CODEC G722 although the most common CODEC for voice is G711 which uses 64 kbps over bandwidth ( plus overhead)

A basic element of a data network is the ability to recover from lost packets by tracking and retransmitting any portions of the data that do not arrive intact and on time. The packets that handle this are a variety called Transmission Control Protocol, or TCP. These TCP packets cannot be used for voice communication. It would result in poor quality when missing packets were re-sent and inserted into the sound at the other end in the wrong place. It’s much better to just drop any defective or missing packets and listen to the part that does arrive as expected. Remember, you have 8,000 samples per second. You can lose some and still understand what is said. To accomplish this, voice uses exclusively packets designated User Datagram Protocol, or UDP. These packets are “best effort” delivery.

A subset of UDP packets is used to further enhance the delivery of voice packets in a timely manner: Real-time Transport Protocol, or RTP. This streaming-specific protocol improves the quality of sound by giving special attention to packets that may arrive late or out of sequence (called jitter).

In summary ; as long as the data network is configured correctly , with sufficient bandwidth and the right CODECs are used there is no reason why a user would even realise that their conversations are taking place over VoIP rather than the PSTN