I’m not sure if when you say signalling you use the term ‘signalling’ to mean the entire call or if you actually mean the signalling used to setup and clear down calls. I’ll assume that you mean the signalling to setup and clear down calls.
With regards to UDP based VoIP signalling i.e. SIP, unless you are planning to route your signalling over a separate IP network/path then you do not need to know how much bandwidth is used for signalling. The reason being that the voice CODEC will always use (even based upon todays low bandwidth CODEC’s) more bandwidth than the signalling. As you will not be signalling at the same time as voice is being transmitted if you have sufficient bandwidth for the voice you will have sufficent bandwidth for the signalling.
There are certain rare scenarios which create exceptions to this. However, for almost all existing deployments and requirements, these do not apply. These exceptions are:
1) If the signalling is used to
2) If the signalling is constantly
in use (i.e. T.120).
3) If large number of
supplementary services are used
during the call.
For basic voice the CODECs will use more bandwidth than the signalling.
With regards to TCP based signalling – usually H.323 but H.323 can use UDP for call signalling (and SIP can use TCP) most of the above still applies but you will also have the overhead of the TCP keepalive packets (for H.323 you may have two lots of TCP packets – those for H.225.0 and those for H.245).
With regards to the bandwidth used for each voice channel it will depend upon the CODEC used, CODEC frames per IP packet and if voice activity detection (VAD)is used.
Do you know which CODEC you plan on using? If not what level of quality are you looking for?