1. NETWORK TRANSMISSION DESIGN FOR THE MOVE TO IP
Impacts on CPE suppliers
Telecom's current planning of the transition from today's PSTN to a "New Generation Network" has shown that there are some factors that require the support of the CPE industry.
Some of these issues resulting from the use of IP (Internet Protocol) were dealt with briefly in Newsletter No 119 under the title "Voice over Packet Operation". This article explains some of the background issues. It is expected that follow-up articles will be published in future Newsletters.
What is happening to public networks?
Within the telecommunications industry worldwide there is a lot of interest in the transition from circuit switching to packet switching for "New Generation Networks". The current 64 kbit/s circuit switched PSTN provides high quality transmission performance, but it does not cater easily for broadband delivery. On the other hand, an IP network provides for bandwidth on demand, but it is not yet necessarily going to provide all the features and quality of transmission, which have been developed over many years for the present generation of public networks.
Having efficient high quality voice networks networks already in place, incumbent public network operators have been relatively slow to adopt IP technology. Although private IP networks are already being installed, it appears that transmission performance is not always taken into full consideration. Often, the question of potential cost reduction may have a higher priority. In addition, while the PSTN is circuit switched and offering high quality transmission for "any to any" connections, a sub-standard private network is generally not too obvious because the vast majority of calls to or from that network are with PSTN users. However, end-to-end voice transmission quality is not necessarily going to be acceptable should calls take place between two sub-standard private networks.
IP networks have generally been designed for carrying data, which is very reliant on error-free delivery, but not so critical as regards delivery times. If errors occur, a packet is simply re-sent. Voice has quite different needs in that the odd data error or lost packet is acceptable, provided conversation is natural and users can interact as they do now on circuit-switched networks, without echo or distortion.
This article deals with the question of voice transmission performance and the design issues that have to be addressed in an IP network environment.
Key parameters
There are five main transmission parameters that need to be taken into account when assessing voice connection quality. These are:
* Loudness - determined by the characteristics of the CPE
* Noise - circuit noise and room noise
* Echo - as the combination of signal reflections and transmission delay
* Delay - in the sense of absolute delay, with no echo or variations in that delay
* Equipment impairments - typically introduced by digital signal processing
When reviewing its National Transmission Plan in 1997, Telecom confirmed the separate limits it had set on each of these principal transmission parameters. Limits were also set on access line characteristics and the Loudness Ratings of all voice CPE and internal network design, such that the overall performance met ITU Recommendations. Other New Zealand network operators have also generally complied with this plan and it has worked well to provide good voice quality in the circuit-switched network environment.
Introduction of IP technology significantly increases two of these voice-quality parameters: delay and equipment impairments. The challenge is to control these factors for voice calls, such that transmission quality is not unduly impaired, particularly in the transition stages from a circuit-switched network to a predominantly packet based network.
Loudness
The Loudness Ratings(LR's) of all voice CPE have been defined in Telecom's PTC specifications and Telecom requirements align with those recommended by the ITU some years back. As published in TNA 151: 1997, the Telecom Transmission Plan, a Send Loudness rating of 8 dB and a Receive Loudness Rating of 2 dB for digital telephones are regarded as optimum.
The analogue LR's, published in clause 6.2 of PTC 200, are RLR = -6.5 and SLR = 5. At first sight, these seem quite different, but this is adjusted by the T pad (0.5 dB) and the R pad (6 dB) used in the exchange switch and by the traffic-weighted mean circuit loss of 2.5 dB. The overall optimum requirements for analogue and digital telephones are thus aligned.
Noise
With the old analogue switches now removed from Telecom's network and most aerial lines replaced by cable, circuit noise is no longer an issue for most voice calls. If anything, the most interfering form of noise these days is the ambient noise in an office or home environment. Both are, to some degree, under the control of the customer should they be regarded as a nuisance.
Echo
Except for a very few special cases where cancellers must be provided, the transmission delays due to the longest geographic path distances experienced within New Zealand are at about the limit for operation without echo cancellers. It has been a long-established principle that "any party introducing significant extra delay is required to provide echo cancellation". While "significant" has not been defined earlier, a figure of 5 ms one-way path delay should be regarded as a valid guideline. This applies independent of how the extra delay is caused.
Because IP digital signal processing inherently causes delay (or "latency", as it is commonly termed) adding significantly more than 5 ms to any existing path delay, echo cancellers will be required in all cases where calls enter or leave an IP network.
PSTN echo cancellers are currently installed only at the International Gateway exchanges, where they cancel any signals reflected from 2-wire points within the New Zealand network. The international exchanges in other countries follow the same practice, such that our customers do not hear their own voices as annoying echo.
Other than for the access lines, our circuit-switched PSTN is now virtually "all digital", using 64 kbit/s (ITU Rec. G.711) transmission throughout. As a result, speech from one analogue access line to another undergoes only one encoding process to digital and one decoding process back to analogue. Although this introduces minimal distortion and delay, worst case path delays are at the limit. Since IP operation will further increase transmission delays within New Zealand, Telecom expects to install cancellers at all of its own interfaces between the existing switched circuit network and its new generation IP network.
The need for cancellers is even more significant should low bit rate encoding be used by private networks and systems connected to the PSTN.
IP operation can introduce further delays for a quite different reason. One of the aims of IP transmission is to carry both voice and data over the same paths in order to gain economies through the use of one network rather than two separate ones. Path delays can be minimised while traffic is low, but should traffic reach the route capacity (such as during large data file transfers), voice packets may be delayed. This also leads to degraded speech quality. Thus, unlike the circuit-switched environment, where network congestion can prevent a call being connected, an IP network can complete the call in congested conditions, but subject it to degraded quality. Steps thus have to be taken to ensure the quality of service by suitable traffic engineering and by ensuring priority delivery of speech packets.
Equipment impairments
Equipment impairments arise from signal distortion arising from the digital encoding of the speech packets. To minimise transmission costs in private networks, there is a lot of incentive to adopt low bit rate encoding of speech signals, but this also leads to increased distortion. This may be exacerbated by additional coding conversions (transcoding) as a call traverses one or more public networks on its way to another customer or even to another private network using a different encoding protocol. A related problem is that any such transcoding also introduces additional delays, impacting on echo and delay issues.
Assessing voice transmission quality
With several public network operators and a large number of private networks in this country alone, the number of different networks carrying an individual call can easily reach four or five. This is further complicated for international calls, the routing of which may also involve several overseas networks. Each network could well use different encoding processes, so the question of assessing voice quality over a complicated multi-network call is itself a problem. Ensuring that the desired end-to-end call quality is maintained can be even more of a problem - unless all parties accept that they have to set some minimum (and preferably common) standards and not only limit the potential impairments, but share them on an equitable basis.
The assessment issue is addressed by the ITU in what is now known as the "E-Model". This is a computational model for use in transmission planning, which recognises that the complexity of modern networks is such that individual transmission parameters can no longer be considered separately.
ITU-T Rec. G.107, published in May 2000, allows the overall effect on voice quality for a given set of call conditions to be assessed in the form of a transmission rating factor "R". Telecom is currently using the E Model for establishing its network transmission architecture for the transition from circuit-switched to IP transmission.
The E Model takes into account the five key parameters described above, by allocating "scores" to the quality level of each process involved in a call to assess the overall R rating.
R-values are around from 90 for most calls entirely within the PSTN. However, if the R value drops to 60 or below for a call, most users would describe such a call as "poor" quality. In particular, ITU-T Recommendation G.109, September 1999, states that "R-values below 50 are not recommended".
What needs to be done to maintain call quality?
Telecom's analysis of the E-Model has revealed that there are some immediate moves that can be made at minimal cost to prepare for the transition to IP transmission. The recommended use of digital public network interfaces for PABX systems and private networks, discussed in Item 2 of this Newsletter, is one such move.
Analysis of the E Model reveals that call quality can degrade quite significantly if the LR's are set either higher or lower than the optimum values. Getting the LR's right is a relatively low-cost move, as it can be argued that it costs no more to design for the optimum than for any other LR values. Also, since our PTC 200 requirements align with the ITU Recommendations, it would seem only reasonable for all suppliers to aim for these optimum levels.
Since the overall aim is to have a zero transmission loss IP network in order to carry all forms of data and voice at any required bandwidth, the CPE is now even more significant as the determining factor for speech quality across networks, especially under marginally acceptable call conditions.
It is clear that the introduction of VoP technology has the potential to generally lower overall voice quality. Any impairments will probably be worst during the next few years, when voice calls will be carried over a mix of circuit-switched and packet-switched paths.
Important messages to CPE suppliers
* 2-wire analogue CPE Loudness Ratings need to be optimised to the maximum extent. The current ratings are SLR = 5 +/-1 and RLR = -6.5 +/- 1.
* Digital CPE Loudness Ratings need to be optimised to the maximum extent. Optimum levels are SLR 8 dB, RLR 2 dB
* ISDN network interfaces should be used for all PABX systems and private networks wherever practicable (see item 2 below).
* Echo cancellers shall be used in all customer equipment which introduces a delay of more than 10 ms on voice calls to or from the PSTN.
2. DIGITAL PABX's AND PRIVATE NETWORKS: RECOMMENDED PUBLIC NETWORK INTERFACES
With the impending moves towards progressively replacing today's circuit switched PSTN with an Internet Potocol (IP) public network, it is timely to review the current range of network interfaces offered by Telecom for PABX's and private networks.
Current network interfaces
There are currently three different classes of network interface for voice applications:-
a. 2-wire analogue (the interface used for most types of single line CPE), as covered by specification TNA 102;
b. Basic rate and primary rate ISDN interfaces (covered by our specifications, PTC 131, PTC 132, TNA 133 and TNA 134); and
c. 2 Mbit/s Digital Trunk Interface, commonly known as "DTLM" and supporting a Telecom service known as "Digital Voice Access" (DVA). This is a Telecom-specific line signalling protocol, covered by our PABX PTC specification, PTC 107.
Future use of these network interfaces
2-wire analogue interface
The 2-wire analogue interface has the disadvantage that it introduces the variable losses of the analogue line and the fixed loses of the T and R pads used in the Telecom switch, plus any reflection losses due to mismatches at the 2w/4w point. This interface is so commonly used, both here and overseas, that it will continue to be supported in an IP network environment. However, it is far from optimal for a private network.
Because of the above transmission impairments, 2 Wire analogue interfaces are NOT now recommended for PABX systems or for private networks. However, 2 Wire analogue interfaces may need to be provided if the systems do not offer either basic or primary rate ISDN trunk cards (such as a small analogue system or an earlier digital PABX being re-installed), or Telecom is unable to provide basic rate ISDN to the site where the system is to be installed.
ISDN
In those service areas where the two ISDN options under "b" are available now, it is expected that they will be continued under the new IP network environment. What is not certain at this stage, is whether basic rate ISDN interfaces will be made available in those areas where this service is not currently provided.
Larger PBXs, subject to PTC207 or likely to be networked, should use ISDN interfaces wherever possible. This ensures that the PABX or private network 0 dB reference point is at the same transmission level as that of the PSTN.
Telecom is able to provide primary rate ISDN at almost all sites likely to be requiring this service.
DTLM/DVA
What is certain is that option "c" is now obsolescent. It is expected that the service will soon be withdrawn for new customers and it will not be available on future IP networks. Few PABX suppliers have supported this interface so far, so its non-availability is not expected to cause any serious problems.
Where a PABX installation already has a DTLM interface and additional trunks are required, Telecom's remaining stocks of the relevant exchange equipment can be made available to support such additions. Nevertheless, it is recommended that the system be converted completely to primary rate ISDN operation, wherever this is practicable. This will avoid the need for an overall trunk interface replacement at a later date.
OTHER NETWORK INTERFACES
Needless to say, the new generation public IP network is likely to provide additional types of interface, such as Ethernet and xDSL. While its introduction is still probably 2 - 3 years away, it would be of interest to Telecom planning staff to learn whether CPE suppliers have any future preferences or even current plans to use interfaces other than ISDN.
Any suggestions or comments on this issue will be welcome.
Conclusions
ISDN is the strongly recommended interface for all new digital PABX installations. This interface is even more necessary for PABX's which are to be networked and for those which are capable of being networked. For optimum transmission performance, ISDN offers the benefits of a digital 4-wire interface and wide availability, not only today but also in the future. There are side benefits too, as can be seen from the list of features available with this service.
2-wire analogue interfaces will be available, but they are NOT recommended. They should be limited to those smaller systems without networking capability and to those situations where ISDN interfaces cannot be used.
3. LEVELS AND PAD SETTINGS FOR PABX SYSTEMS
As explained above, the use of a 4-wire network interface, as provided by ISDN, improves transmission performance and simplifies the level setting. Given that the long term situation will be for the zero level reference point to be extended from the public network into the private network, such that the levels delivered to users are entirely dependent on the CPE at both ends of the call, it makes sense to have some default pad settings in any interim network interfaces, such that the system will evolve to the optimum settings with minimum need to alter the initial settings.
The concepts of standardising default settings to deal with both analogue and digital telephones, as well as different network interfaces, are illustrated in the
draft Supplement to PTC 109 which is currently published on the Access Standards website in pdf format.This supplement covers 2-wire analogue interfaces, as well as ISDN, notwithstanding our strong recommendation that only ISDN be used. Although many manufacturers provide an optional 2-wire analogue interface to their systems, it is hoped that installers will now keep analogue interfaces to the minimum in the knowledge that there are better ways to do it.
As explained above, the move to IP operation has placed increased emphasis on setting optimum levels in terminal equipment and using telephones of optimum Loudness Rating. It can be argued that as there is no additional cost in setting the right values, rather than the wrong ones, all suppliers should aim to prepare for the new IP world by introducing the correct settings now.
4. VOICE OVER IP TRANSMISSION ENCODING ALGORITHMS
There is often a need to make savings in the operating costs of private networks, especially those incurred with leasing circuit capacity for inter-node tie lines. This has led to the use of low bit rate encoding for many private networks, despite the resultant transmission quality penalties.
IP networks present a similar situation, except that it is a question of bandwidth rather than circuits. With data and voice sharing the same transmission paths, it is feasible to reduce the bit rate for voice, if only on a temporary basis, whenever the combined traffic is approaching the network capacity.
As explained in Item 1 above, reducing the bit rate usually increases delay due to the more complex digital signal processing required. There is also the issue of increased distortion. Both factors impact on the overall quality of a call.
Although it requires more capacity, the use of 64 kbit/s encoding to ITU Recommendation G.711 is strongly recommended.
This form of encoding introduces minimal delay and distortion and, in fact, is not even considered an "impairment" under the E Model. From Telecom's viewpoint, it means that the impairments introduced by a private network terminating a national or international call will be minimised, especially if the party at the other end of a call has connected via some sort of private network using low bit rate encoding.
DOUG BURRUS
Manager
Access Standards
