Technology Issues and Data Communications Services

This note has been compiled as a form of report on work in progress . It has been prepared at this time because of the high level of interest and activity in issues associated with the availability of access to digital data communications services in the wider community.

The material is likely to be quite time sensitive as the technology and service availability issues are subject to rapid change as a consequence of business plan decisions by the various players in what is a very competitive industry.

This report has been compiled by Keith Malcolm with assistance from Maurie Daly.

Communications Laboratory

3 June 1998

This note is a brief outline of data communications technology issues that is derived from the results of work undertaken in the Communications Laboratory. For some time we have been undertaking desk based study of technologies suitable for universal data communications at a consumer level. These studies have been supported by formal and informal trials of various possible systems. These notes outline the information gathered through our study work. External activities such as RTIF funding of trial systems, suggest that it is timely to issue this note.

The issues of availability of, and access to, data communications services in urban, suburban and regional areas continues to attract attention at all levels.

For simplicity, this note will address issues associated with access to "the Internet" at a consumer level. Some of the identified issues will also be relevant to small business and to organisational users such as schools.

Over the last 10 years, "the Internet" has grown from being a closed network of links between educational and research institutions to something that approximates a consumer service in information access and delivery. To some extent, that amorphous thing called "the Internet" has developed via applications such as "the World Wide Web" (WWW) (the Web) into something like an electronic public library. Unfortunately, the organisation of the Web leaves a lot to be desired so access is more in the form of extended "browsing" rather than by means of concise and targetted extraction of information. This aspect of the Web has resulted in substantial changes in the nature of Internet traffic such that traditional methods of access are less attractive to users. Whereas traditional access demands were in the form of intermittent (often infrequent) and short exchanges at low net data rates over campus LAN’s, the growing usage is characterised by extended connection periods usually from home using the PSTN on a dial-up basis. This is causing problems for internet service providers, telecommunications carriers and users.

The emergence of electronic services such as electronic banking and bill paying (as an early part of the much vaunted "information economy") are precursors that suggest that it is time to examine a range of issues before the problems of universal data communications services become intractable.

The gathering volume of experience is showing up a number of issues that need to be addressed if the information economy is to become a reality and not just a glassy-eyed dream. The principal issues that need to be addressed include:

• performance of available access methods
• cost of access
• availability of access.

Practical electronic data communications systems can be traced back to computer to computer links established by means of mechanical teleprinter systems operating at speeds of around 50 to 110 baud. At this data rate, links can be readily established even using methods such as HF radio. In practice, links between systems would have been established generally by means of permanently connected circuits or on-demand switched circuits such as were provided by communications carriers for telex service. These low speed links were (and still are) suitable for applications such as e-mail and (to a lesser extent) ftp exchange of data files.

The next discernible step was to the development of voice-band modems for operation over the PSTN or on private voice networks. Early systems operated typically in an assymetrical mode such as 75/1200 or 300/1200 baud and continued to be used for applications such as e-mail, ftp and telnet.

In more recent times, we have seen a rapid development in voice-band modems for use on PSTN or permanently connected copper-pair cable networks. These systems initially offered fixed data rates such as 2400 bits/s or 4800 bits/s (note the change in definition of speed from baud to bits/second) but very quickly emerged adaptive modems that were able to evaluate line quality on a pseudo dynamic basis and thereby select the highest possible data-rate for each connection. Speeds moved rapidly through 9600, 14400, 19200, 28800, 33600 to 56600 bits/s and have now approached the speed achievable through dedicated (eg ISDN) data lines (64 kbits/s). The performance requirements of these modems are defined by the ITU-T in its’ "V-series" Recommendations.

Practical experience shows that while 9600 bps is acceptable for applications such as picture phones and "white-board conferencing", it is barely usable for WWW access. For Web access, data rates in excess of 19.2 kbps are required.

The PSTN network is the most widely used means of access for data delivery services but it was designed and implemented to carry voice rather than data traffic. Voice traffic is much more tolerant of impairments than is data because of the redundancy in human speech and the ability of a person to extract the speech message even among noise and distortion.

The performance of copper-pair cable systems is dependent on the length of the consumer local loop and the nature and quality of the cable connection between the user and the local exchange. Achievable data rates decline with length of the subscriber’s loop and decline rapidly if the subscriber connection involves anything (such as pair-gain systems or digital multiplexers) other than a straight physical copper connection. Active systems such as pair-gain systems are often used in congested locations to provide additional lines where it is not economical to install additional physical cable pairs.

The data performance of the subscriber loop is also adversely affected by noise and cross-talk levels. Since noise levels can vary over time, and cross-talk will vary with traffic loading in the cable, and both factors are distance dependent, these impairments will be time and location variable.

The outcome of these factors is that the data performance of the customer access network is not uniform and cannot be readily predicted to any reasonable degree of confidence.

The ISDN system has been designed on the basis of digital carriage of speech and data traffic. It is configured on the basis of 64 kbit/s channels so provides the promise of more reliable or consistent and higher speed data delivery that is possible via the PSTN. While ISDN uses the same physical pair cable network in the subscriber loop as the PSTN, ISDN requires careful selection of cable pairs and appropriate "conditioning" of the pairs if it is to work satisfactorily. At the present time, ISDN is not universally available.

There have been a range of proposals for digital subscriber line systems to operate over the copper pair cable local loop. These systems use complex modulation and equalisation techniques to use the "above voice-band" capacity of pair cable systems. There are a family of DSL systems that have been developed that provide a digital data capacity ranging from about 128 kbps up to about 50 Mbps. These systems are always asymmetrical, providing more capacity in the download direction. The capacity is distance dependent with typical performance being 1.5 Mbps over about 5 km range up to 50 Mbps over about 300 m range. DSL systems are not yet available as customer products in Australia so cost information is speculative. However, we can expect that costs will be moderately high because of the complexity of the hardware that is needed at both the customer premises and at the telephone exchange end of the local loop. Some US suppliers are offering user terminal cards at prices in the range $US500 to $US1000. Depending on the actual transmission technology adopted by the carrier, the usage cost might be relatively low and based only on data throughput rather than on connection time or connection occurrences if a packet switched, rather than a circuit switched, technology was to be employed.

Cable modem systems have been developed to use a part of the capacity of a TV cable distribution system for the delivery of digital data. A single TV channel block of spectrum can provide more than 10 Mbps of data capacity which permits the establishment of an "ethernet" style of shared connection. A typical cable modem would provide a connection at a peak data rate of around 500 kbps (connection speed) with the actual throughput being dependent on the number of sharing users active at any instant (just like the internet at large). This service is available to Telstra customers in locations where the Telstra/Foxtel cable is installed. Installation costs around $500 and monthly access/usage charges are $65 per month (for up to 100 Mbytes download).

It is, in principle, possible to use cellular telephone systems for data as well as speech traffic. The GSM system incorporates as a standard feature the ability to carry data at 9600 bps. As noted above, this is not really sufficient for WWW applications but is useful for e-mail and fax and similar purposes.

GSM modifications have been proposed that would permit a 64 kbit/s data rate but these proposals require the allocation of all time slots of each RF channel to a single 64 kbit/s data channel. In addition to any issues about availability of hardware, these proposed modifications may well be not compatible with the existing GSM channel allocations and consequent implementation arrangements in Australia.

The AMPS analogue telephone system inherently has a low data rate capacity. It is limited to about 2400 bit/s because of the sharp cut-off filtering that exists in the voice circuits of AMPS hardware. However, an "overlay" technique has been developed (called Cellular Digital Packet Data (CDPD)) that provides a data capacity of 19200 bit/s. This system has been implemented in the USA as a relatively low cost data service. The service has been provided on the basis of a single monthly access fee with no ongoing access or usage charges. It operates on a packet switched basis, using the network capacity "as required" to send and receive data and thus "mirrors" the "best endeavours" basis of the rest of the Internet rather than providing a guaranteed connection. As such, the system would appear to match the cost expectations of the domestic user but the capability is limited in terms of WWW access and is probably more appropriate to applications aimed at portable rather than desk-top PC’s. An interesting aside is that with the impending closure of analogue AMPS telephone service, the CDPD technique could provide a useful avenue for exploitation of the existing network of AMPS infrastructure.

A number of systems have been developed in the USA to provide wireless local loop connections for internet applications. These systems use frequency hopping or spread spectrum techniques in the 900 and 2400 MHz ISM bands. The permissible operating power levels are restricted by FCC regulations so the systems are limited in range by attenuation of the radio signal. The potential range depends on operating power and frequency. Systems exist that provide in-building or across-street coverage for wireless LAN applications extending to systems that provide up to 10 to 25 km or more in open rural areas. Systems exist that provide typically 19.2 kbit/s capacity up to about 128 kbit/s capacity in the 900 MHz band and up to 2 Mbit/s capacity in the 2.4 GHz band. These systems are of only limited application in Australia because of differences in the spectrum allocation arrangements (especially with regard to the 900 MHz ISM band devices - the USA band is 902 - 928 MHz whereas the Australian band is only 916 - 928 MHz). In addition, because of the strong dependency of performance on the radio path conditions, the systems need careful design and are of reduced reliability in other than clear path situations. Accordingly, the availability of these systems is restricted because of lack of adequate technical support from the suppliers.

There are two practical approaches to the use of satellite communications channels for internet access purposes not including the simplistic approach of use of a satellite voice telephone connection.

The existing satellite telephone services (Mobilesat etc) are simply too expensive in terms of capital costs and call charges and are too limited in data rate to be seriously considered as internet access methods. Even the much touted LEO systems will provide only a limited data rate capability and early indications are that equipment and call charges will be even higher than those associated with existing sat-phone services.

Satellite internet access systems can be provided on either a symmetrical basis or an asymmetrical basis. In this context, a symmetrical service is one that provides equal data rate in both forward and backward directions. Existing VSAT (Very Small Aperture Terminal) systems can provide a symmetrical connection that permit choice of data rate over a wide range. The penalty is that the cost of the hardware needed at each end of the link is very high ($10’s of thousands).

An asymmetrical link is one in which the forward link (from the ISP to the user) is wideband (relatively) while the backward link (the user to the ISP) can be narrow band. An asymmetrical circuit would usually be established by using the satellite to deliver data to the user while the user communicated to the ISP over a dial up telephone connection. In the case of TCP-IP data connections there is a link between forward and backward data rates because of the need for acknowledgment data to be sent. This sets a ratio of about 10:1 so that the satellite data channel speed is set by the speed obtained over the backward channel. An asymmetrical link can have relatively low equipment costs (the user needs a conventional telephone modem and a satellite receive-only terminal) but the equipment costs are likely to be in the order of $1500. Access costs include the dial-up line access charges plus the satellite transponder access charges. The satellite transponder costs can be quite low because it is possible to share the transponder capacity across a large number of users. DirecPC is one such system that is operational in the USA and it provides unlimited access at a monthly fee of about $US29. The system operates at Ku band (12 GHz) and requires that the satellite have coverage over the intended service area. The cost of the user terminal is related to the signal level provided from the satellite. DirecPC have provided a service in Auckland (NZ) by means of a terrestrial relay of the satellite signal (presumably because the satellite signal is receivable but of too low a level to operate on economical domestic receive-only systems).

An Australian company offers an asymmetrical service that currently uses a C-Band satellite feed from an off-shore based ISP up-link. Use of the service requires a terrestrial connection to a local ISP and the use of a complete satellite receive-only system (dish, LNC, receiver) plus a data decoder. Thus the setup cost for the satellite side of the link is additional to what would be incurred for a regular bi-directional modem link to an ISP. A short trial in the Laboratory confirms that the technology does work and is practical and delivers a data throughput in the download direction up to 10 times that achievable through a PSTN modem link.

The cost of access to, and use of, systems is a non-trivial issue. Even for a standard telephone service there are the multiple elements of initial installation cost and ongoing access and usage charges. At least for residential customers, the initial installation and monthly access charges for a standard telephone service would seem to be acceptable given the level of penetration of private telephone connections. However, usage charges are another issue. The availability of untimed local calls appears to be a significant factor in achieving an aceptable cost structure and this poses major problems in the case of rural customers who may be at significant distances from the nearest community. In these cases, the cost of usage is the highest and the system performance and reliability the lowest and we have the existence of special charging regimes to "equalise" the costs for such users in comparison with the charges imposed on urban users.

For data (internet) services, there is the additional cost element of the ISP charge for access to the internet. These charges are typically time or data volume charged. Some ISP’s provide "unlimited" access but these services are often characterised by low net data throughput and random disconnection or inability to connect at busy times. For data service, there is the additional element of the initial cost of the modem to connect to the subscriber loop.

Despite its quality, throughput and reliability limitations, the PSTN is the most common means of access to internet services.

For a regular user, internet connections will easily overwhelm the normal voice usage of a residential telephone service so even for the lowest cost case of an urban customer, the 25c cost of a local call every time the user logs on to the internet becomes a significant aggregate cost. For rural users who might not have an ISP within the local call zone, the access cost is even more significant as it will be a timed call. For limited Internet use such as for e-mail, the cost of a timed call is unlikely to be significant except for very long distance calls as the connect time is typically short and probably within a single unit time period. Web surfing, however, is a very time consuming activity and would be unrealistically expensive under a timed calls regime.

The cost of ISDN as an internet access medium is even more significant. The monthly rental charges are much higher than those for PSTN lines and connections are made on a timed basis. For the domestic user, even though the data performance of ISDN is much better (guaranteed 64 kbit/s) the cost is so high as to be effectively prohibitive.

Wireless Connection Technologies offer the prospect of no cost connections but this is offset by the capital cost of the wireless modems required. Low power units have been available at prices around $AU 500 but prices seem to have been rising and the units that offer longer ranges now seem to cost about $AU 5000 per modem. Note that two modems (one at each end of the link) are required for wireless interconnect. The major cost factor appears to be the unavoidable consequence of the use of technologies necessary to meet the FCC regulatory requirements for "no-licence" operation. If wireless techniques are to be considered seriously as viable for domestic internet connections then there would seem to be a need for a programme to develop low cost modems. This would need to be accompanied by the identification of suitable UHF spectrum at low cost either on an individually licensed or a class licensed basis. It is possible that new PCS services to be offered in the 1800 MHz band will provide viable data connection services. The operating band will impose a practical constraint on such services in terms of useful radio range. Also, a somewhat more "user friendly" approach to charging than that currently employed even for PSTN access will be needed if the services are to be attractive. As a particular issue, the current prices being bid for 800/1800 MHz spectrum would appear to be so high as to militate against any truly low cost service offering (prices appear to be based on an assumption of high value services).

Speech quality PSTN access is effectively universally available, but outside of urban areas, the data transmission quality of the PSTN is variable. For rural users, the quality is probably inadequate for effective WWW access purposes.

ISDN is, at present, only available on a limited basis and the cost of ISDN access imposes a practical barrier to its wider use even if availability was to improve.

The effective range limitations and capital costs of wireless systems preclude their use for general internet access. In any event, the limited spectrum available for such systems will constrain adoption of wireless for the foreseeable future.

There have been cases where wireless systems have been adopted on a wide scale for new local loop implementation. However, this has only occurred in locations where a copper pair local loop network was effectively non-existent and the high cost of wireless local implementation has been "offset" against the rapidity of roll-out in comparison with a wired system.

The preceding notes suggest that we have not yet found a connection technology that provides an economical and realistic match to the needs of universal internet connectivity and usage.

Cable modem systems using Pay-TV cabling networks have been touted as one solution, but Australia does not have (and probably never will have) a ubiquitous cable infrastructure.

ADSL systems that operate over the existing copper pair cable local loop offer the prospect of a range of digital transmission speeds. At the present time, such systems are relatively expensive in terms of the terminal unit costs and probably offer much higher data rate than is useful or necessary for internet connectivity. ADSL systems have been developed initially as a telephone company’s answer to video delivery in the local loop rather than simply as a practical digital data connection system.

Internet data is transmitted in the form of data packets and the effective data rate that is available to users is limited more by congestion in the upstream links of the internet rather than by the capacity of the local loop. This suggests that a low cost, narrow band DSL system might be a worthwhile avenue for study. Such a system could be conceived on the basis of multiplexing bi-directional data above the baseband speech channel in the subscriber loop. The data would be injected upstream of the users telephone and extracted at the MDF in the exchange (on the subscriber side of any switching plant) where it would be aggregated with other data streams to a progressively higher level. Since the data itself is structured for packet switching, there is no need for any switching task at the service provider end. Such an approach would provide a method for permanent connection of data services that did not "load" the remaining part of the PSTN. It is not apparent that any such system exists at present so substantial development work would be needed. However, if a data delivery service is being seriously considered for inclusion as a part of the USO then a system such as that outlined will need to be developed as all existing techniques seem to be too expensive. To be economically viable, it is assumed that any data delivery service would need to provide permanent connections at an access fee (rental) of about $10 per month, a modest installation fee (say around $250) and with traffic charged at (say) 20c per megabyte of incoming data.

Since wireless systems offer a number of attractive characteristics, particularly in the case of rural, regional and even outer suburban users, an alternative approach might be to undertake the development of a low cost wireless modem system that was tailored to the needs of Australian users. Such a system would offer a reliable data rate comparable to the best achievable via the PSTN and would be suitable for both short range and longer range applications. It would be necessary to identify a suitable block of spectrum that could be made available at no or low cost (class licensing of user terminals with frequency management undertaken by the ISP would appear to be a possibility). From a consideration of radio propagation issues, a spectrum block in the UHF band (say 300 to 1000 MHz) would be the best. Obvious choices include the (part) channel 27 (520-526 MHz), geographic sharing of TV channels or an allocation in the defence bands below 420 MHz. From our review of available radio systems we have not been able to identify any existing sources of equipment appropriate for a consumer quality (and priced) service so there would appear to be an opportunity for some development work to be undertaken.