Why Does IPSLA UDP Jitter Display 4.06 as the Maximum MOS Score?
Overview of MOS scores
In order to understand why IPSLA currently displays the maximum MOS on the UDP Jitter test as 4.06, you need to understand a little background about MOS scores in general.
A MOS score stands for Mean Opinion Score.It is used as a quality indicator in the delivery of voice (VOIP) packets over a data network. Because there are many different network conditions that can affect the performance of a network as well as many different types of Codecs used to convert analog voice to digital voice and vice versa, MOS scores are used as a measuring stick to help find the best balance between Voice compression (determined by the type of Codec) and available bandwith.
Related Pages
MOS scores are displayed as a value between 0 and 5 with 5 being highest (indistinguishable from face-to-face quality)
Because no phone has ever sounded exactly like a face-to-face converstion the maximum scores are never a perfect 5. Uncompressed voice (G.711 codec) scores a maximum of 4.1 and compressed voice codecs score below that point depending on how much quality is sacrificed for additional bandwidth efficiency. G.729 codecs, for example, have a maximum MOS of 3.92 but are 8 times more efficient over the network than G.711 voice streams.
So in summary, the maximum MOS is determined by the quality of voice that any particular Codec is able to deliver on a perfect network.
Producing Maximum Scores
Don't forget, the developers of Mean Opinion Scores wanted a number that refects the quality of the voice that is being transmitted, not the quality that could theoretically be achieved on a perfect network.
(Leaving the effects of Quality of Service (QOS) aside for the moment in order to make this next point) If you were running VOIP over a very busy network with insuffucient capacity to forward packets in a timely manner then the type of codec would become less significant. The quality of the converstation would suffer as a result of the impairment experienced in transit.
For this reason, MOS scores fluctuate depending on how much impairment any given voice sample experiences in transit. One second the sound could be fine and the next it is distorted. If you were taking MOS samples every second then you would capture this fluctuation with higher scores when the sound is good and lower scores when the sound is poor.
Putting all the theory together - IPSLA 4.06 Max Scores
When you set up an UDP Jitter test using IPSLA the maximum possible score is 4.06. In other words, the codec that Cisco emulate in those tests would theoretically score 4.06 voice quality when a packet traversed the network experiencing no or insignificant levels of impairment.
As the IPSLA probes are sent, impairment levels are measured based on the RTT latency, jitter and any packet loss experienced by the probes. This data is plugged into a formula and a resulting Mean Opinion Score is the result. If the probes experienced no impairment or insignificant levels of impairment the score returns at 4.06 bacuse that is the maximum quality that the emulated codec could possibly achieve.
Any more significant levels of variation to RTT latency, jitter or packet loss would affect the quality of the voice and in turn the score. In my experience, IPSLA scores of 3.8 or less can be detected by users so for a better user experience you should ensure that scores are constantly around the max.
