Voice Quality Considerations
From Bandipedia
When considering VoIP for the enterprise, it is important to understand voice quality metrics. Voice quality is measured by the Mean Opinion Score, a standard developed by Bell Labs in the mid-1950s.
Voice quality is directly affected by the QoS metrics provided by the IP network transport metrics found in all Tier 1 Network Service Provider (NSP) Service Level Agreements, specifically, packet loss, delay, and delay variation (also referred to as jitter). Packet loss causes voice clipping and drop-outs, which can be objectionable to the callers. Many CODEC algorithms can correct for up to 30 ms of lost voice with concealment algorithms that mask many of the lost voice samples. This means that the loss of two or more consecutive 20-ms voice samples will often be noticeable, a problem that is compounded when using a compression CODEC such as G.729a. When using G.711, intermittent packet loss is generally not an issue. Assuming a random distribution of packet loss, a drop rate of just one percent in a voice stream will result in an unconcealed loss once every three minutes, on average. Tier 1 SLAs guarantee a packet delivery rate of between 99% and 99.9% (less than 1/10 of 1 percent loss).
Packet delay, measured as round trip delay, greater than 200 ms causes impaired interaction for a voice conversation. The loss in interaction occurs because one party may begin speaking before the transmission from the other is received. The result is collision in conversation. People using similar communication channels must wait to ensure that the other person has finished before speaking in a mode similar to that which occurs with telephony poor quality international calls. The standard for VoIP specifies 150-ms one-way mouth-to-ear delay budget for high voice quality for an international connection, obviously the lower the better.
Delay variation, better known as jitter, is the difference in the delay between packets, and the adaptive jitter buffers within IP telephony devices automatically compensate for 20-50 ms of jitter. If jitter exceeds these bounds, then loss will occur, impacting voice quality as described above. Current experience indicates that if jitter is held to within 10-20 ms, then voice quality is acceptable. For example, if one packet requires 100 milliseconds (ms) to traverse the network from one point to another, and the following packet requires 102 ms to make the same trip, then the delay variation is calculated as 2 ms. Here again, Tier 1 NSPs guarantee through their SLAs jitter that is anywhere from 25ms to as low as 1ms.
By using differentiated class of service such as the Expedited Forwarding (EF) class for VoIP traffic loss, delay, and delay variation are minimized, thus assuring the quality of the voice conversation.
What do you think about this page?
Comments are appreciated and assist in building this site.
We welcome comments, question, and suggestions in the following manner:
1. Click here to leave a comment.
2. Send an e-mail to wiki@bandwidth.com
3. Call 919-297-1069 and record your comments
