2. VOICE TRANSMISSION QUALITY AND THE ITU E MODEL
As readers will have gathered from the last two Newsletters, we are very conscious of the need to maintain the present quality of voice transmission despite the increasing use of packet-based private networks, digital cellular networks, and the transition to "New Generation" packet-based public networks. Unless overall transmission planning is fully considered and appropriately controlled, margins will be eroded to the point where increasing numbers of calls will be considered unsatisfactory.
Telecom's recent advertising campaign has focused on the importance of "availability" and, strictly, a service is not really "available" if its quality of performance does not provide for effective communication. While Telecom has control of the quality of its core network and sets requirements for each CPE item, just how that CPE is used and how private networks are set up is largely under the control of other parties. The same principles apply, of course, to other New Zealand public networks and what is connected to them.
With deregulation of telecommunications worldwide, just what elements comprise an end-to-end call are almost infinitely variable. Nevertheless, the aim is to ensure that the highest possible percentage of voice calls provide effective "conversation quality", even in what is likely to be "a very data-centric new world".
Newsletter No. 130 explained the increased significance of "loudness" for voice calls and the need to aim for the optimum loudness rating limits for telephones as a preparatory step towards maintaining voice quality in a VoP environment. There are many other factors affecting the overall quality rating value "R", some of which will be under Telecom's control, but the same principles apply in that the design of CPE and all private and public networks involved in a call needs to consider the impact of these components on the overall quality of end-to-end performance.
In Newsletter No. 129, I explained the basic concepts behind the "E Model", which the ITU describes as "A computational model for use in Transmission Planning". I made only a brief mention to the significance of the "R" values computed under this model. The following explains the position in a little more detail.
The ITU have carried out extensive investigations into transmission quality issues over the years and such measures as Quantisation Distortion Units (QDU's), Mean Opinion Scores (MOS) and "R" units have been developed through a combination of formal objective testing and subjective (actual listener-talker combinations) testing. These tests take into account not only the technical and equipment parameters likely to be experienced, but also the highly variable characteristics of human users. After all, human conversations take place between all types of people, with a wide range of vocal clarity and volume, and a wide range of hearing ability. The net result of these tests is that for any level of assessed quality, some percentage of users will find conversation difficult (described by ITU as "Poor or Worse" - "POW"), while another percentage of users will regard the same call as acceptable (described by ITU as "Good or Better" - "GOB").
Under Telecom's National Transmission Plan, these assessment and design objectives are used "to minimise the number of customers experiencing difficulty on any connection through the network, and to maximise the number of connections falling within the customers' range of preferred losses".
Needless to say, there will always be a small residual group of people who have difficulty even communicating face to face under otherwise ideal conditions. A 100% GOB score is thus unlikely for any telephone conversation, no matter how good it is. Network designers have to limit the number of calls likely to encounter "worst case" conditions by setting suitable limits on all elements of a call, especially the main ones described in Newsletter No 129.
With our present wholly digital circuit-switched core network, the great majority of national and local calls between PSTN customers would have an R value of 90 or more. However, this is not always the case. For example, with connections involving telephones having Loudness Ratings at the limits of acceptable performance, or a long analogue line at both ends (especially if one or both of the users are subjected to line noise), the R value might drop back to 70 or so.
In comparison, PSTN calls to digital cellular networks and private networks using voice compression or digital processing of some sort will incur some additional impairments, such as increased delay and distortion from digital encoding processes. The majority of calls involve a PSTN user at one end at least and, as long as the PSTN offers high transmission quality, private networks can "get away" with introducing transmission impairments without the call quality falling to an R value of 60 or less. However, it is important that designers do not continue to assume that the whole of any margin currently existing in the PSTN can be used up elsewhere. The growing use of IP-based private networks, with their inherent additional delays and encoding impairments, will mean that the probability of calls between two such private networks will rise.
The relationship between R values and customer satisfaction is illustrated in the following table, derived from ITU Recommendation G.107. Note that the two percentages for any R value do not add to 100%, as there are always some users who are not particularly concerned about the call quality at the defined levels.
As a comparison, "Mean Opinion Scores" (MOS) are also shown.
