ETSI finds modest VoIP improvements - Global News Analysis

Telecommunications International, August, 2002 by Ken Wieland

According to a voice over IP (VoIP) speech quality test organised by ETSI early this year, there has only been a marginal improvement in average equipment performance compared to eighteen months ago.

The results of the test--which was held in Sophia Antipolis, France, in April 2002--were published in June. They are comparable with the previous speech quality test results for VoIP--organised by ETSI in September 2000--as the same test conditions were re-created.

"Using a scale of nought to five, with nought being really bad, we found that the overall average performance of VoIP speech quality was around 3.4, which is about a 0.1 improvement compared with September 2000," says Dr Jens Berger, co-author of the ETSI report on the speech quality test event.

As a point of reference, 3.4 is about the equivalent of a GSM call while a typical PSTN/ISDN call registers around 4.2/4.3 on the 0-5 scale.

"Of course, some manufacturers did a lot better than 3.4 and some did worse," adds Berger, "but, due to our commitment to anonymity, we can't name names."

Measurements were performed for one-way speech transmission (listening quality) as well as 'double talk' situations where background noise transmission performance was also evaluated. PLC (packet loss concealment), echo cancellation techniques and jitter buffer design/control were all put through their paces during the examination.

"There are lots of variables involved in VolP, but if you have a high-performance network with G.771 codecs [64Kbps transmission rate using PCM line code] it may be possible to get speech quality nearly up to PSTN standard," says Berger.

One of the tests carried out at the ETSI event involved comparing the PESQ (perceptual evaluation of speech quality) performance of the G.771 codec with the lower bit-rate G.729 codec on a like-for-like basis. It was found that G.771 scored in the 3.66 to 4.05 range (different network settings produced different results) while the comparable PESQ scorecard range for the G.729 codec was 3.21 to 3.59.

"Speech quality is just one of many issues surrounding voice over IP take-up," says Margaret Hopkins, associate consultant with UK-based Analysys. "You've also got to remember that many corporates use voice compression techniques over their networks [to optimise bandwidth usage] which can result in GSM voice quality anyway. I would suspect that the voice quality expectations of many enterprise-users have come down in recent years, particularly given the rise of mobile usage."

According to Hopkins, if corporates provision enough IP bandwidth for incoming calls (say, from clients) they can preserve PSTN-like voice quality. And for internal calls, they can save on their PSTN bill -- and perhaps accepting a lower quality call performance -- while still enjoying all the call features associated with TDM.

Analysys, which published a report entitled IP Voice Services: European Corporate Market Forecasts 2002-2007 in June 2002, predicts that the corporate voice over IP service market in western Europe will range between [euro]2.5 bn (low-case scenario) and [euro]7.4 bn (high-case scenario) by 2007, which respectively represents 15 per cent and 41 per cent of the total voice over VPN market. "The low-case scenario assumes a steady growth of 50 per cent year," says Hopkins. "The high-case scenario is if the voice over IP arguments for capex and apex savings [put forward by vendors and operators] have been proven to be credible and end-users can enjoy tangible financial benefits."

According to Hopkins, the corporate voice over IP market needs more independent testing to verify the cost-savings claims made by vendors if it is to really take off. There's also the issue of high-priced IP terminals. "The idea of 'hot-desking' [plugging in your own IP phone to whatever desk you may be working at] will be attractive for many workers but at around US$200 a time, IP phones are still way too high."