My Voice Codec Reference

The below is a summary of digital voice standards from the ITU. These standards are used in the speech codecs (coder/decoders) that sample voice and convert it into digital code or back to analog. I compiled the below from the ITU and Cisco.

Codec            Sampling
G.       Coding     Rate
Number   Method    (kHz)   Bit Rates (kBps	MOS	Compression Delay(MS)
G.711    Mu-Law PCM   8    64		4.1	0.75
G.711    A-Law PCM    8    64		4.1	0.75
G.723.1  MP-MLQ       8    6.3		3.9	30
G.723.1  ACELP        8    5.3		3.65	30
G.728    LD-CELP      8    16		3.61	3-5
G.729    CS-ACELP     8    8		3.92	10
G.729a   CS-ACELP     8    8		3.7	10

CallManger Express supports the G.711 and G.729 codecs. While it may seem advantageous at first to use g.729 throught the enterprise, we typically use g.729 on WAN links and G.711 on LANs.  Most voicemail system can only use G.711 , so DSPs should be reserved to transcode G.729 to G.711 if needed.  With Cisco products we typically use hardware devices called Digital Signal processors or DSPs to encode and decode voice.  Since G.729 uses codebooks rather than sampling to encode voice it is more resource intensive and uses more DSP resources.  Another potential disadvantage to using using these low bitrate codecs is if your network design requires mutiple (tandem) encodings.  An example is if a G.729 RTP stream is encoded two times the MOS drops from 3.92 to 3.27 and if the signal is encoded a third time the MOS drops to 2.68.

MOS or Mean Opinion Score is a commonly used subjective scale to judge voice quality.  Listeners grade voice quality on a  scale that ranges from 5 (excellent) to 1 (bad). A score of 4 is generally accepted as toll quality.

Cisco provides a DSP Calculator to help with configuring DSPs for Cisco router.  It is available at this link for registered Cisco users.

 

0
Your rating: None