Two standarst using in VoIP telephony are the the IETF standard SIP and the ITU standard H.323. In the very beginning the H.323 was the most popular one, but with time it has been beaten by SIP. Nevertheless, many of the providers do use H.323 in their core backbones, while callers do not even know that their POTS calls are made over VoIP.
One of the typical problems of VoIP telephony is the wasted bandwidth used for packed headers. Usually to send a G.723.1 5.6 kbit/s compressed audio path requires 18 kbit/s of bandwidth based on standard sampling rates. The difference between the 5.6 kbit/s and 18 kbit/s is packet headers. Silence suppression and header compression are used for bandwidth optimization. This can save up to 35% on bandwidth usage.
VoIP trunking techniques such as TDMoIP can reduce bandwidth overhead even further by multiplexing multiple conversations that are heading to the same destination and wrapping them up inside the same packets. Because the packet header overhead is shared between many simultaneous streams, TDMoIP can offer near toll quality audio with a per-stream packet header overhead of only about 1 kbit/s.
Where VoIP travels through multiple providers’ softswitches the concepts of Full Media Proxy and Signalling Proxy are important. In H.323, the data is made up of 3 streams of data:
- 1) H.225.0 Call Signaling;
- 2) H.245;
- 3) Media.
So if you are in London, your provider is in Australia, and you wish to call America, then in full proxy mode all three streams will go half way around the world and the delay (up to 500-600 ms) and packet loss will be high. However in signaling proxy mode where only the signaling flows through the provider the delay will be reduced to a more user friendly 120-150 ms.