Some supplement.
why sometimes you will see the G.729A have 20byte or 40byte pure data in one RTP package? because in this case the Voice Packetizer output the RTP data after two/four 10ms, so in fact it have 2/4 frames/package, it will save bandwidth when compare to 1 frame/package because now 2/4 frames share the same IP/UDP/RTP header;