| Quote: |
I know that it should be G. 711 since it uses the most bandwidth and its not compressed.... |
Actually G.711
is compressed. There's about 13 bits of sample resolution squeezed into 8 bits. Each byte has an exponent and value, like the way floating point numbers are represented in a computer. It's not that good, but the algorithm is very simple and requires practically no CPU, so it will live on forever.
As for whether G.729 is clearer: you can find research that says perceived quality is better (measured in the form of Mean Opinion Score, MOS). It is designed to represent/model speech specifically rather than "general audio". Like a lot of modern codecs (e.g. AMR, GSM Enhanced Full Rate, Speex) it uses CELP (or ACELP) which is a way to encode packets of data according to a parameter table that is meant to describe speech. To make things complicated there is a family of algorithms in G.729 including wideband (7kHz), Annex A which is a low CPU power version, and there are also add ons for silence suppression and comfort noise generation. You could probably make an argument in favour of G.711 or G.729.
The great thing about G.711 is that it's supported by just about everything, so this makes interop very simple. Many fancy codecs (incl. G.729) need to be licensed, as they are someone's intellectual property, which means that not all devices will implement G.729. The flipside is that 'free' low bitrate coders (like Speex and iLBC) are too new to be implemented in everything.
There's a lot to be said for simplicity if you have the bandwidth to spare...