May 23, 2012

Tips for Improving Internet Call Quality

Voice that runs over Session Initiation Protocol (SIP) trunks, of course, is formatted in voice over IP (VoIP) packets. VoIP traffic that flows over SIP trunk connections traverses the public Internet en route to its destination somewhere on the public switched telephone network (PSTN). The Internet leg of the trip can introduce packet loss, jitter and latency - characteristics with the potential to degrade call quality. Degradation is most likely if traffic moves on and off of different ISP networks managed by separate operators.

The first two impairments might be resolved by the voice gateway that connects the Internet to the PSTN or even by the IP phone if they are not too severe. Latency, however, cannot be resolved by the gateway or IP phone. So what can you do to maintain high-quality VoIP sessions?

Make Call Center Traffic a Priority

UCTNMay21-CallCenterArt.jpgA subjective but standard measure of voice quality is the mean opinion score, or MOS. The MOS ranges from 1 (bad) to 5 (excellent) quality. Different applications can tolerate varying levels of voice quality.

One area where voice quality is very important is the call center - the "voice" of the company to the consuming public. Callers might interpret poor voice quality to mean that they are not valued by the company, for example. Poor quality also can lead to errors, lengthened call times and disappointed or frustrated customers.

Most call centers prefer the MOS to average 4.0 or better. So a common best practice in the call center is to use a codec for uncompressed voice, partly because it introduces less delay and distortion as packets move from an IP network to the PSTN.

Call centers using VoIP over SIP trunks to connect to the PSTN operate with digitized voice. There are two common digital voice standards: G.711, which supports 64Kbps uncompressed voice, and G.729, which supports 8Kbps compressed voice.

G.711 is usually the favored standard for the call center. It supplies higher voice quality, requires less complexity and can reduce transcoding (conversion) delay.

Bandwidth Savings vs. MOS

The total bandwidth required including packet overhead for G.711 is about 83Kbps, while G.729 consumes about 27Kbps including the packet overhead. The bandwidth consumed will depend on the voice packet size used by the SIP trunking provider's implementation.

While G.711 eats up more capacity, it delivers an average MOS of 4.1; G.729 delivers a MOS slightly better than 3.9. So the enterprise must decide whether both MOS values deliver acceptable quality or whether they'd prefer to trade some bandwidth for slightly higher quality. Whichever approach is chosen, the SIP trunk subscriber should independently monitor the voice quality as well as read the provider's report on the average MOS to make sure calls are consistently of an acceptable grade.

What About QoS?

One of the recommendations of IP telephony vendors is to implement quality of service (QoS) - marking IP packets with higher priority than data packets on the enterprise network is the most common QoS tool - to ensure voice quality. But this is problematic when traffic gets merged with the general public's traffic in the Internet.

The enterprise can implement QoS for voice through the Session Border Controller (SBC) connected to the SIP trunk. Once the voice packet moves onto the Internet, however, it is treated with the same priority as data packets. Inbound transmission will have lost any QoS capability when packets are received from the Internet.

One recommendation is to keep SIP trunk utilization levels at 80%. It is also recommended that the voice and data access over the SIP trunk be logically or physically separated to eliminate bandwidth contention between voice and data packets.

SIP trunking to the PSTN will continue to grow because of a strong voice quality track record as well as the cost reduction when compared to T1 and Primary Rate Interface (PRI) ISDN access alternatives. The good news is that as the Internet continues to mature and operators use the latest traffic engineering and performance management tools to control traffic flows, Internet services are doing a better and better job of handling latency-sensitive traffic like VoIP.

This TechNote is brought to you in part due to the generous support of:



"G.711 ... delivers an average MOS of 4.1; G.729 delivers a MOS slightly better than 3.9" Where do these figures come from? In my experience it is very difficult to get G.729 to deliver 3.9, while G.711 will often give 4.5 or better.

Steve, the figures of 4.1 MOS for G.711 and 3.9 MOS for G.729 are from the experiences and measurements performed by IntelePeer, a major SIP trunking provider for their network connections.

This includes some good considerations for companies looking at SIP trunking, especially when using an over-the-top SIP trunk provider.

But do keep in mind that for many Enterprises, the SIP Trunk will be delivered over private MPLS connections that fully support QoS and will be backed with Service Level Agreement (SLA) guarantees.

An interesting thread, thank you. Firstly MOS scores, while essentially a score obtained by taking the average of human listener scores, i.e Mean Opinion Scores, is more often nowadays performed by perceptual analysis tools, such as P.862 PESQ and P.863 POLQA on computer systems designed to predict the MOS scores. Scores for narrowband voice range between 1 and 4.5 (even though the true P.800 MOS scale measures 1-Bad to 5-Excellent, human nature will often not score a perfect quality with a 5, and 4.5 was found to be the Mean maximum). Measurment with a test tool generally involves direct connection to the telecoms circuit in question and measuring the score end-to-end. Consequently this configuration measures the transparency of the circuit and does not include the customer handsets and environment. The full picure is given when the measurment uses the Customer handset, the environment acoustics and ambient noise. This is where the quality perception really can drop, just as we humans have difficulty with the noise on calls from call centres, bad use of handset positioning and even in some cases the language. P.863 POLQA has been designed to allow acoustic measurement and new HD-voice services, so is the method to measure voice quality going forward. Using acoustic measurement processes with a Head and Torso simulator (HATS) will also give better representation of call centre performance and should allow better design for them in the future.

Graham Rousell's statement, "bad use of handset positioning and even in some cases the language." is very accurate and, in my opinion, makes all of these other measurements worth very little until these problems are overcome.

Search Webtorials

Get E-News and Notices via Email




I accept Webtorials' Terms and Conditions.

Trending Discussions

See more discussions...

Featured Sponsor Microsites



Please note: By downloading this information, you acknowledge that the sponsor(s) of this information may contact you, providing that they give you the option of opting out of further communications from them concerning this information.  Also, by your downloading this information, you agree that the information is for your personal use only and that this information may not be retransmitted to others or reposted on another web site.  Continuing past this point indicates your acceptance of our terms of use as specified at Terms of Use.

Webtorial® is a registered servicemark of Distributed Networking Associates. The Webtorial logo is a servicemark of Distributed Networking Associates. Copyright 1999-2018, Distributed Networking Associates, Inc.