TELECOM ACCESS STANDARDS NEWSLETTER NO. 149

July 2004

CONTENTS
1. CPE-GENERATED TONES FOR ISDN VOICE CALLS
2. LOUDNESS RATINGS FOR IP TELEPHONES AND HEADSETS
3. PTC 220 TESTING
4. SERVICE INTERFERENCE CAUSED BY POWER SUPPLIES
5. TELECOMMUNICATIONS RELAY SERVICE
6. ADSL DEVELOPMENTS
7. ADSL-POTS INTERFERENCE

Appendix 1. EVOLUTION OF VOICE TRANSMISSION STANDARDS

RETURN TO MAIN INDEX



1. CPE-GENERATED TONES FOR ISDN VOICE CALLS

  A number of cases have arisen recently where ISDN CPE (mainly on Primary Rate Access) has been found to be locally generating ringing tone on voice calls, rather than connecting the speech path to the network and leaving the terminating end of the call to send back ringing tone while the caller is waiting for the call to be answered.  This is causing a number of problems.  

It is standard Telecom network practice to let a call drop back to the originating exchange when the called number is found busy, where network congestion has occurred, or where the number is invalid.  In such cases, it is pointless to hold the whole call path when the originating exchange can simply send the appropriate tone or message direct to the caller.  In the same way, if the originating CPE has the capability of dropping an ISDN voice call and playing a call failure tone to the caller when it gets a signal that the call cannot be completed, there is no problem in this being done.  

However, it is not advisable to locally generate ringing tone while a call is awaiting an answer signal or some sort of network message from the terminating end. Unfortunately, it appears that some ISDN CPE has been set up to do just this. 

As a result, any network voice announcements sent in place of ringing tone are not being heard by the caller.  This can cause some confusion for the caller. For example, the network may be playing an announcement advising that the called mobile is out of the coverage zone while the originating CPE is playing locally generated ring tone!  

This current situation is further complicated by the introduction of “personalised ringing tones”, which are already being used in some overseas networks and likely to be introduced here in the future.  The generation of ringing tone within the calling CPE will prevent other customers from making the best use of this new service feature.  

It is recommended that CPE suppliers and maintainers check that their systems are not generating ringing tone at the “originating end” of a call for some reason.  Any that are found to be doing so should be modified.


2. LOUDNESS RATINGS FOR IP TELEPHONES AND HEADSETS

Several earlier Newsletters (Nos 129, 130, 131, 134, 135, and 144) have dealt with the subject of speech transmission performance and conversation quality. The first of these newsletters was published in 2001 and it would have been thought that suppliers would have “got things right” by now.

Nevertheless, we are still seeing a very wide variation in such basic issues as Loudness Rating performance. Not only are Send Loudness Ratings (SLR) “all over the place”, but Receive Loudness Ratings (RLR) with 30 or more dB adjustment between minimum and maximum settings are not uncommon. These, of course, allow the IP phone user to mask the effects of low receive volumes due to unnecessarily quiet SLR on another phone. As a result, the IP Phone user sees no problems, other than perhaps the need to adjust RLR from call to call. On the other hand, more and more PSTN users will experience difficulty hearing calls from IP phones unless their SLR’s are improved.

This situation is particularly disappointing in view of the ITU-T having published their E-Model and extensive explanations of the need to recognise those aspects impacting on conversation quality and wide international acceptance of the need to aim for an SLR of 8 dB and an RLR of 2 dB for all digital telephones. This is also expressed in the North American standard EIA/TIA 810A.

The basic principle we are working towards is that “all –digital IP networks” should be “transparent” as far as Loudness Ratings are concerned, with no losses or gains introduced within either public or private networks. This means that the CPE at each end sets the Overall Loudness Rating for any call. The simplest and most effective way to run the Next Generation Network is to have ALL CPE set to the optimum LR’s and thus have no need to make adjustments to compensate for variations.

Certainly, the E-Model allows for trade-offs between Loudness Ratings, packet loss, delay and jitter for the various speech encoding systems and these other factors can compensate to some extent for non-optimum Loudness Ratings. However, our message is “Get the LR’s right and there is more room for other factors to vary and still meet the desired standard of conversation quality”. This is obviously the best approach where LR’s can be clearly defined at little or no extra cost, whereas managing the other factors may be quite expensive.

This issue is being further complicated by the use of wireless devices, such as WiFi IP telephones, which also have “short” handsets. Currently, no test labs in this country have testing equipment which can produce authoritative LR measurements for these short handsets. Nevertheless, we can get a fairly clear indication of non-compliance from our present measuring equipment and this is always backed up by a “sanity check” - simply making calls from the device in question to a PSTN telephone which is known to comply with the optimum LR’s and making a subjective assessment of the loudness. While this does not give accurate measurements, it is exactly what the users will do and these subjective assessments are very consistent with the measured results.

Telecom is currently looking into the purchase of suitable testing equipment for IP telephones and networks. This is very expensive and unlikely to be purchased by local test labs in view of the low demand fro such testing so far. Nevertheless, we hope to be able to improve the accuracy of our testing and, hopefully, maintain the present high conversation quality of the PSTN for our customers during the transition to an “IP World”.

The attached Appendix No. 1 provides background information on the evolution of transmission standards, covering why we are so concerned about such an apparently small detail as “Loudness Ratings” in the new all-IP world.


3. PTC 220 TESTING

As mentioned above, Voice over IP (VoIP) testing to PTC 220 has been difficult to achieve so far, as no local labs yet have the ability to test at IP level and formally confirm that the private network zero transmission reference point is at the same level as that of the public network. As such, we have been reluctant to formally accredit any test labs to PTC 220. Nevertheless, PTC 220 Telepermits have been granted where we have been reasonably satisfied that the test results and other information provided have shown that the supplier is aware of transmission requirements and has the means to get them right.. Typically, suppliers have specified the Gateway settings to provide a zero gain/loss in both transmission directions, such that the LR’s of the associated telephones can be determined by testing at the network interface. This goes at least part way towards our objective voice transmission planning described in previous Newsletters. In addition, most transmission test results have been given the additional “sanity check” of practical subjective assessment.

Telepermits have been granted on assurance that the telephone device in question provides the correct levels at the public network interface. In principle, these interim arrangements have been no different in principle to the way in which we have dealt with proprietary telephones (SIT’s) and their associated PABX systems for many years.

Despite this interim measure, our long term goal is to be able to ensure “any to any connectivity”, especially when private VoIP networks are interconnected directly at IP level via IP public networks. At that time, the only really workable solution to avoid argument about which party should change its settings and by how much is to have common zero transmission reference points. We are currently investigating the availability of suitable commercial IP telephone testers as a means of attaining the necessary measurement capabilities at IP level.

Some test laboratories have done a lot of preparatory work in this area and it is only fair to recognise this. We thus propose to grant “interim accreditation to PTC 220” to those test labs which have shown that they have an acceptable test and measurement process to determine levels at the public network interface of the IP network Gateway device.

The first such “interim accreditation to PTC 220” is granted to ComTest, in Australia. While we are still dealing with this laboratory on some of the testing procedures for “full” PTC 220 testing, we will consider test reports issued by ComTest based on the above “system” approach. However, where the equipment concerned incorporates variable gain or loss settings, the actual settings used are to be included in the test report.


4. SERVICE INTERFERENCE CAUSED BY POWER SUPPLIES

We have had another case of switched mode power supplies (SMPS) causing noise on customers’ lines and causing degradation of dial-up modem performance. In this particular case, the noise to line passed PTC 200, which is focused primarily on Psophometrically-weighted noise measurement (this assesses the varying impact of noise at differing frequencies on the human ear)and its impact on telephone performance. It now appears we will need to set some additional limits to minimise interference to analogue modems.

As an interim measure, it is recommended that the suppliers of any CPE using SMPS should check the impact of connecting that CPE on the same line as a dial-up modem to make sure there is no significant reduction in the modem’s speed or ability to connect to an ISP. This should preferably be done by setting up a dial-up call without the SMPS CPE connected and checking the speed, then dropping that call and re-establishing the connection with the SMPS CPE connected with the modem.


5. TELECOMMUNICATIONS RELAY SERVICE

As recently announced in the press, the Ministry of Economic Development has announced that a consortium of Sprint and Counties Power have been chosen to implement this new service for hearing-impaired and deaf users. Initial service introduction is planned for mid-November.

The Ministry have advised that the TRS will probably “certify” any CPE that is shown to work with the Sprint system and this certification will cover any “special requirements” for the CPE to be used with this service. If so, these will supplement the relevant requirements of PTC 200. In the circumstances, potential suppliers of TTY’s or other CPE for this service are invited to submit applications for Telepermit in accordance with PTC 200, along with test reports indicating safety compliance. This seems the most pragmatic approach in view of the fairly tight timing of this project.

Local suppliers should note that test reports showing EMC compliance will need to be held in a technical folder in accordance with the outline given in Newsletter No. 141. Supplier Code Numbers will be needed for those suppliers not already holding one or not covered by the SCN of an Australian associate company.


6. ADSL DEVELOPMENTS  

Since Telecom introduced its “JetStream” and related ADSL-based services in 2000, telecommunications technology has been quietly developing a number of variants to the current technology, which are now being ratified by the ITU.  “JetStream” operates to ITU-T Recommendation G.992.1, which provides for up to 6.144 Mbit/s “downstream” and up to 640 kbit/s “upstream”.  Many modems and DSLAM’s provide higher speeds up to around 8 Mbit/s downstream and 800 kbit/s upstream, but the rate actually achievable is most dependent on the individual customer’s line loss. 

Our current Alcatel 7300 ASAM ADLT-L card) uses Software release 4.2.22, but we are now looking into the next generation of DSLAM

The more recent Recommendation G.992.3 uses the same spectrum for “ADSL2” which, in its basic “ANNEX A” form, provides relatively marginal speed improvements over the present system. 

However, ADSL2 does provide for much improved and standardised diagnostic data to be available from both the DSLAM and the modem, as well as a comprehensive power management system.  Instead of continuously running at full power as in the present ADSL, the DSLAM can drop back to a reduced “keep alive” power level while no data is being transmitted, or even further reduction to a “standby” power level during long periods of no actual data flow.  This not only saves energy consumption, but also reduces the level of interference on cables carrying a high proportion of ADSL lines.  

Two new Annexes to this Recommendation have also been introduced: “Annex L” and “Annex M”.  Annex L covers a long line version with higher downstream signal levels in the lower frequency channels 32 – 128, which are not so heavily attenuated on long lines.  Annex L is commonly referred to as Reach Extended ADSL, or READSL. In comparison, “Annex M” covers a higher “upstream” data rate version, by making use of some of the downstream channels.  This can only be achieved at the expense of downstream speeds and there are some concerns about its use causing interference to ADSL signals from other customers connected on the same cable.  

The other big step forward is ADSL2+, which is specified in ITU-T Recommendation G.992.5.  This version makes use of twice the bandwidth (the downstream frequency bands extend up to 2.2 MHz), but because of the even higher attenuation of these upper frequencies, this service is most suited to short lines.  ADSL2+ has a number of other significant advantages over the current system, including those introduced by ADSL2.

Dynamic Rate Adaption (DRA) is another feature of these next generation ADSL services. This allows the modem and DSLAM to reduce speed temporarily should there be bursts of noise on the line, then restore back to full speed when that noise has gone. This will be done without needing a full re-start. In comparison, should the current ADSL service be impacted by noise, the speed may drop to handle the lower signal to noise margin and then stay at the lower speed until the next time the service is re-started.

The various chipset manufacturers are currently carrying out inter-operability testing, but it should not be very long before ADSL modems are supporting these new versions.  Local importers may care to ask their suppliers when they might be making any of these changes.  Meantime, Telecom and Alcatel (Telecom’s DSLAM supplier) are keeping abreast of these developments and looking into the service implications.  More on this subject is expected to be provided in future Newsletters.


7. ADSL-POTS INTERFERENCE

With the present rapidly increasing take-up of ADSL-based services likely to accelerate to even higher rates in the future with Telecom’s objective of 25 000 broadband customers by the end of 2005, the issue of interference between ADSL signals and POTS CPE could become more significant. In past Newsletters and specifications, we have stressed the need for adequate filtering, either by using a Telecom-installed splitter at the network demarcation point (on the line side of all premises wiring and CPE) or line filters in each POTS CPE socket (including the SKY Digital TV decoder, each of which houses an analogue modem).

There are around 8 000 different Telepermitted items now, most of which were introduced to the Telecom network before ADSL was available here. As such, they have never been tested for either their impact on ADSL signals or the impact of ADSL signals on them.

To minimise the risk of service problems affecting their own equipment, it is suggested that all CPE suppliers include in their user instructions a brief “reminder” to customers that they will need to fit a line filter in front of the POTS device if they are subscribing to Telecom’s JetStream service. This should not only prevent or at least reduce any interference caused by the high frequency ADSL signals, but also help to ensure that the CPE item will not degrade their ADSL-based service.



Doug Burrus
Manager
Access Standards


APPENDIX 1

EVOLUTION OF VOICE TRANSMISSION STANDARDS

Transmission Plans have been developed over the past century or so with the aim of making telephone conversations as uniform as possible. That is, a person talking on any telephone in a normal voice should be able to heard comfortably by a person on any other telephone anywhere in the world. This principal was first recognised in the late 19th century. At the time there was no method of amplification, so losses had to be limited, leading to the use of very thick copper aerial wire on long haul trunks. By the 1920s amplification using valves made this aim more achievable, but until the introduction of digital transmission and switching which became widespread in the 1980s, compromises were necessary for all but local calls. This was due in part to the fact that much of the networks were 2-wire, and amplification can only be used to a limited extent to prevent instability occurring.

One of the developments which took place during the middle of the 20th century was the concept of Loudness Rating. This was an objective method of measuring the total end to end performance of a telephone connection literally from the mouth of one user to the ear of the other. A weighted loss of 10dB (overall loudness rating (OLR) =10dB) approximates two people one metre apart conversing in normal voices.


Analogue Transmission, Switching, Access and phones

In the analogue world the OLR for a call between customer A and customer B is made up of several components as follows:

The loudness ratings are divided amongst the networks as follows:

  • The loss from the mouth to the telephone jack via the microphone and telephone electronics is known as the Send Loudness Rating.
  • The loss in the access line is known as the Circuit Loudness Rating (CLR).
  • The losses in the network are made up of further losses in the transmission medium, and deliberately introduced losses (generally at 4-wire 2-wire interfaces for stability purposes).

In the analogue world the network losses varied depending on where the call was routed, and local line lengths tended to have considerable variation between the longest and shortest, and to reduce the variability, the phones themselves adjusted their own gain depending on line current which was an indication of line length. So the lower the current, the more gain the phone would insert, thus in part overcoming the losses incurred by a long line. Such phones were known as regulated phones and were common until the early 1990s.


Digital Switching and Transmission with analogue access lines and phones

With the move to digital transmission and electronic switching equipment in the 1980s, three significant changes took place.



Figure 1. Digital network with analogue access lines and phones

Note that there are still compromises because the local lines and telephones are still analogue, so there is some variability in the line loss, although this is reduced by a gradual upgrading of the local networks. The OLR is calculated as follows:

OLR = SLR(Customer A's phone) + CLR (Access line to customer A) + Network input loss (T Pad) + Network output loss (R Pad) + CLR (customer B access line) + RLR (customer B's phone).

In New Zealand this is implemented as follows (see Figure 1):

SLR (Analogue phone ) = 5 dB
CLR (average) = 2.5dB
T Pad at local exchange = 0.5dB
R Pad at far end local exchange = 6 dB
CLR (average) = 2.5dB
RLR (Analogue phone) = -6.5 dB

The total of all these losses added together is 10dB which is the desired OLR for a call.

Between the T Pad and R Pad there is a lossless network, and it is useful to define all levels relative to the level at this point. It is known as the 0dBr reference point. By international convention, all networks are interconnected at this point. The loss from the acoustic input to the 0dBr point is the sum of the phone SLR, the CLR and the T Pad value which is typically 5 + 2.5 + 0.5 = 8dB. From the 0 dBr point to the acoustic output is likewise 6 + 2.5 + -6.5 = 2dB. Given that networks from all around the world are interconnected at the 0dBr point, it follows that each network operator has to chose values for SLR and the T Pad, and engineer the access lines to a traffic weighted mean value of CLR such that they add up to 8dB, and also on the receive side the RLR and R Pad, such that when added to the CLR a value of 2dB is arrived at.


All Digital networks (ISDN, VoIP)

The final evolution from a transmission point of view is the move to all digital networks, including the phone and the access line.


Figure 2. All Digital network

This was first realised in the 1980s with the introduction of ISDN, and the same transmission principals also apply to VoIP. This produces the following changes:

As digital access will coexist with analogue access as the evolution proceeds, the 0dBr reference obviously must be common to both. This leads to digital phone loudness ratings of SLR = 8dB and RLR = 2dB. These loudness ratings are in fact specified for VoIP networks by the major standards bodies including the ITU-T, ETSI, ANSI and EIA/TIA.


All IP Networks …special challenges

ISDN was probably introduced before the market was ready for it, and while it achieved reasonable penetration in Europe, it failed to become universally popular. However it did introduce the concept of integrated digital access for voice and data, and effectively achieved what transmission planners had been trying to achieve for about a century. With a lossless network and the 0dBr point could be put in the handset itself, so that a uniform OLR of 10dB could be practically realised for every call between ISDN customers.

VoIP over all IP network also allows this goal to be achieved. However while ISDN was very much an evolution of the telephone network, with the data being carried on voice circuits. In an IP world, the situation is the reverse with voice being carried on what was originally a network designed specifically for data. The biggest difference is that in an ISDN, when party A calls party B, a dedicated circuit is set up between the two parties for the duration of the call. So resources are allocated to that call, and barring faults or natural disasters, the call has the full use of that resource until it is completed.

In a VoIP call, the A party's speech is digitised and then every 20 or so milliseconds packetised and sent to Party B. In a simple IP network each packet makes its way through the network independently of any other packet. Depending on the network loading at the time, each packet may arrive by a different route, and may even arrive in a different order from which they are sent. This variable delay is a significant problem for VoIP, and is dealt with by buffering the packets at the receive end, so that the buffer length is the proportional to the maximum likely delay through the network. This of course adds further to the end to end delay which can degrade the perceived quality of the speech. The other degradation often occurs for economic reasons is due to low bit rate encoding. By using low bit rate encoding, transmission costs are reduced, but at the expense of increased distortion and usually additional delay.

The VoIP networks which network operators will use to replace the existing voice networks will minimise delay by using very fast switches and routers and using special protocols to reduce the variability of the delay. It is likely that the need for low bit rate coding will be unnecessary, as the difference in cost between 8kbps and 64kbps will be insignificant in networks designed to switch bandwidths in the Mbps.


Private VoIP networks

Where private VoIP networks are interconnected with public networks everything is simplified if the 0dBr reference point in the private network is the same as the public one, and standards compliant phones are used. If a non standard phone with say a high SLR e.g. 18dB is used in a private network, the easy fix would appear to be to add 10dB gain in the gateway to the public network, which would certainly fix any calls between that phone and the public network. However two undesirable situations result.

  1. No standard IP phone can be connected to that network, as it will sound too loud (by 10dB) to public network users.
  2. The non standard gateway setting corrects the send levels to Public network telephones but not to the other phones on the private IP network.

The second effect could in theory be corrected by introducing a 10dB loss at the gateway for the signal path from the public network to the private one. This would also necessitate the phones having an RLR of -8dB so that calls within the private network as well as beyond it would sound the same.

A further difficulty with running non standard levels in private networks is that the analogue to digital conversion processes will be forced to operate outside their normal dynamic range introducing noise and distortion.


Summary

VoIP allows the easy implementation of the optimal end to end loudness rating (OLR). SLR/RLR values of +8/+2 for handsets has been standardised by the major standards bodies. As far as speech levels are concerned, there are no compromises required if these loudness ratings are used, the user of any IP phone will be able to hear and be heard by any the user of any other IP phone. If IP phones do not have uniform LRs, then the IP world ends up no better than the analogue world. That is a public network surround by a collection of proprietary private networks with a limited range of proprietary (and expensive) handsets, and variable performance….i.e. back where we were 30 years ago!