Robin Whittle - First Principles Consulting 25 June 1998
For the latest updates in reverse date order, click
here!
This article explores in detail the physical challenges an the technological methods of Wavelength Division Multiplexing - the sending of multiple closely spaced wavelengths of modulated infra-red light through a single fibre, amplifying it optically, and splitting the wavelengths up again so they can be demodulated into 2.5 or 10 Giga bit per second data streams. While WDM can also be used in local and campus sized networks, this article concentrates on the long-haul applications - intercontinental submarine systems and land-based intercapital fibre systems. In particular it looks at Telstra's involvement in the Perth - Indonesia - Singapore segment of SEA-ME-WE-3, and in the carrier's upgrade of is existing intercapital fibre network.
The topics covered include: SDH and Sonet transport data rates and physical link rates; 16 to 32 wavelengths in the 1530 to 1560 nm range; fundamentals of modulation, sidebands and therefore data carrying capacity; signal-to-noise and bit-error rate; effects of fibre: attenuation, non-linear distortion, chromatic dispersion and polarisation mode dispersion; problems with reflections in fibre splices and optical components; Erbium Doped Fibre Amplifiers (EDFAs); powering of undersea optical amplifiers; chromatic dispersion correction using either fibre Bragg gratings or lengths of oppositely dispersing fibre; wavelength demultiplexing using fibre Bragg gratings and multilayer optical filters; laser diodes and their stability problems; tunable laser diodes using movable micro-machined mirrors (see the WDM Links file above); "gain slope" issues with EDFAs - ensuring even gain for all wavelengths of cascades of dozens of EDFAs over thousands of km of fibre; 2.5 Gbps being more practical than 10 Gbps for systems greater than a few hundred km because of polarisation mode dispersion.
One paragraph of the article, concerning economics and how WDM
will
hopefully provide the great bandwidth that Australia needs, over the
thousands
of kilometres of ocean, as well as on land:
A number of engineers within Telstra helped me enormously with
this
article - but due to a policy of subjecting all quotations to the Nth
degree
of scrutiny by something resembling an "Investor Relations Thought
Police",
I am unable to acknowledge them personally. Thanks!
(Note in 2006: WDM is further developed now, of course, including
with Raman amplifiers. See the WDM report at Budde.com.au
. )
There is a great deal of information in this article regarding the practicalities and fundamentals of HFC Internet, and how Telstra has developed unique solutions to enable a limited number of Motorola Cable Routers to service an unusually large number of Fibre Serving Areas. This includes a dual conversion approach, where signals are converted from electrical, to optical and back to electrical twice in their path from the home to the Cable Router.
The many limitations of the service were examined, such as the lack of fixed IP addresses and subnet ranges of IP addresses, the proprietary connection protocol which precluded the connection of Unix machines and most importantly the very limited connectivity which was allowed - for instance no incoming TCP sessions.
Update March 1998: It is evident that Telstra is making significant progress on many of these fronts - or in some cases the problems are being overcome by users here and with similar "Roadrunner" systems in the USA.
The table on page 92 of the November issue paints a grim picture of the connectivity (in terms of incoming and outgoing TCP and UDP, for port numbers below and above 1024) which the service offered. Thankfully this table is now only of historical interest, since many or all of the restrictions were lifted late in February 1998. I will investigate this and Telstra's latest plans (such as for discounted rates and business services) in the coming months - and write it up as an article. Suffice to say that the very low user numbers of 1997 are now history - customer uptake is obviously increasing. The service is available to all homes and businesses in Melbourne and Sydney passed by Telstra's HFC cable - so the maps in the October issue are out of date now. There seems to be a new spirit of communication between Telstra and the keen users who have effectively been the paying guinea pigs for this adventurous, over-hyped and initially severely limited service. Check the latest details regarding this service at Eric Pircher's Big Pond Cable Internet FAQ - see the Resources - links to other sites section below.
The decorative illustration in the November issue was not my idea! This may have given the impression that I or Australian Communications considered the failings of Telstra's HFC Internet service to be the result of some deliberately mean decisions. I don't believe this was the case. The team struck me as totally dedicated to making the best Internet service anywhere - its just that they were inexperienced with the Internet and did not take enough notice of the many people who were trying to set them straight regarding issues such as openness and full-connectivity (ie. lack of firewalls). Now, it seems, the team is far better attuned to the divergent needs of customers, has a lot more technical experience with the Internet and a better understanding of the open methods of communication and development that the Internet facilitates.
Areas covered by this two part article include: Motorola Cybersurfr cable modems and the Hewlett Packard's "BIDS" server system; upstream and downstream modulation techniques and data rates; radio-frequency splitting and combining using two stages of electro-optical-electrical conversion; expansion of this architecture as demand increases; the necessity of volume based billing, when raw Internet traffic costs $0.19 a megabyte - due to the high cost of bandwidth to the USA; $0.35 a megabyte charge for both upstream and downstream; proprietary connection software; block on TCP port 80 outgoing - and a flaky proxy server . . .; blockages on traffic according to port and protocol; impossibility of explaining to ordinary users how such connectivity limits will restrict the operation of their Internet applications; dynamically assigned single IP addresses; impossibility of using the service for many high-value business applications, such as web, email and encrypted https servers; Telstra's goals in achieving full connectivity and volume-based billing; impossibility of broadband usage at $0.35 per megabyte; and difficulty in user's estimating their usage, at potentially $0.17 per second.
Read all about it at http://telstra.com.au/internetdirect/
. (Since July 97 I have been a happy user of this service.)
This is a welcome development - since ISDN has previously only been available from a small number of "overlay" exchanges and via expensive "BMUXs" at locations remote from those exchange sites. However, where is the promised price reduction? Telstra has previously justified its high rental and call costs for ISDN on the basis of the costs of the overlay network.
The article examines how Telstra seems to have got away with introducing Calling Number Display on an opt-out basis with the new OnRamp BR-ISDN service - without meeting the requirements the company agreed to in the AUSTEL Privacy Advisory Committee's report on Calling Number Display (http://www.ozemail.com.au/~firstpr/cnd/).
The article also probes Telstra's decision not to provide SPC tariffs - which are the basis on which many Australian ISPs use ISDN's 64 kbps channels to link permanently to each other and to their upstream providers. A report on this issue is available here: spc.txt .
Update, March 1998: as far as I know, the SPC replacement
is
yet to materialise.
Northgate's approach to HFC design is described. This could be described as a "star-of-stars" topology with lots of cable but relatively few amplifiers per km - aiming for the best possible upstream and downstream signal-to-noise ratios. Northgate are negotiating with the councils of other regional cities.
Here is a photo of Northgate cable technician Damien Chapman preparing coaxial cables for an amplifier site in the first Fibre Serving Area in Ballarat North. While the subtlety and sophistication of HFC signals and technology is the subject of my articles, it is worth noting that it is bloody hard yakka installing the cables! The hardline coax is based on an extruded aluminium tube, and is very stiff. All connectors need to be fitted carefully and tightened correctly to ensure water-tightness and proper electrical connection. The entire installation needs to resist expansion and contraction, corrosion, vibration and flexing from the wind - as well as other environmental threats such as possums and tree branches - for a few decades. Any loose connections, corrosion or water damage will lead to intermittent radio-frequency distortion and/or ingress of interfering signals and/or egress of HFC signals to interfere with radio reception.
Update 21 March 1998:
Northgate
are not making the progress they anticipated with HFC telephony or
Internet
services. They have become one of the few licensed carriers in
Australia
- bold stuff for a small team of people whose age is around half that
of
the average carrier CEO. Check their web site:
http://www.northgate.com.au (in 2003 its gone . . . )
for some extraordinarily unconventional marketing of a very unusual
service
- untimed, fixed fee, long distance and international calls from most
places
in Australia. They claim this is the first such service ever in the
world.
Update 25 june 1998:
actually,
northgate have closed their ballarat sales office, although the hfc
network
is still operating. they are concentrating on telephony services
instead, and are probably the only carrier in australia to have a
web-site
and invoices which eschew the use of capital letters. they are
building
their telephony business by word of mouth and via web - so far without
conventional advertising.
Some time later . . . Northgate ceased operations in
Australia.
A second generation service was introduced in May 1997 - based on
Motorola
Cybersurfr cable modems. See the articles mentioned above - October and
November 1997.
This was written as Optus Vision launched its HFC cable telephony service - the first such service in the world. It was pretty bold to go straight into service with equipment which was still being tested, but they did and by late 1996 the service was slowly (by comparison with the number of people who would like it) being expanded. Update 21 March 1998: In early 1998, the company has a few tens of thousands of telephone customers using its HFC system. Number portability of Telstra numbers to the Optus HFC service is promised in May. (I will believe it when I see it!) Until portability is implemented, customers choosing Optus HFC telephony will be assigned a number different from the Telstra service which they are presumably going to terminate.
This article is partly a discussion of Optus Vision's HFC network, and partly a discussion of the telephony and data equipment of OV's two suppliers: ADC and Motorola. Consequently it describes in detail the ADC and Motorola HFC telephony equipment, with comparisons for spectral efficiency and latency (in which ADC looks better in both respects). However OV's initial services used the Motorola equipment, due to a variety of factors - not all of which may be due to the equipment itself. Update March 1998: I understand that Optus is now preferring to install Motorola telephony equipment in preference to ADC's.
Click here: diag1.html to see two diagrams from this article, which explain ADC's OFDM and Motorola's DQPSK and compares the differing symbol times for the two approaches.
Then I look at how ADC's system can be extended to support data services - using interfaces developed by the company's Australian partner NetComm.
Motorola's approach to data over HFC is quite separate from its telephony approach - I look (on paper - rather than in the flesh) at the Motorola CyberSURFR cable modem.
There is a diagram depicting in detail the FSA, with its fibres, sub-systems inside the electro-optical node, bi-directional amplifiers, taps and home interface unit with switched cable TV port, high and low pass filters, modulator and demodulator and drivers for POTS, BR-ISDN or Internet/Ethernet interfaces.
There is extensive discussion of the noise, distortion and other difficulties encountered when passing signals along tens of kilometres of fibre and kilometres of coaxial cable. A breakout box (based largely on material published by Brian Bauer and John Holobinko at ADC) details the many things that can go wrong with HFC systems, especially ingress of interference into the upstream signal path - which funnels and amplifies many noise sources towards the node. Note that reading this may make you think that HFC could never work reliably - but you can get the same idea about childbirth when reading a book on that subject too.
There is a discussion of the need for encryption of telephony (and Internet traffic) on the shared media of HFC. Motorola's telephony engineers knew this from the start, and in the course of writing the article, ADC stated it would provide encryption should a customer request it.
There is a discussion of how the modulation and data multiplexing scheme of Motorola's telephony gear leads inevitably to latency - time delay within their system - which would best be avoided.
Note that Optus Vision had not committed itself (and still has not in late November 96) to any HFC data technology. Update March 1998: Optus Vision is now wholly owned by Optus, and they are apparently planning an HFC Internet service for introduction "some time in 1998".
There is a breakout box discussing the difficulties of how to charge for usage of HFC cable modems for Internet traffic. This is the Internet Bulk Billing Blues and you can read it here: int_bill.html
Mark Smeaton edited this article as he did all the previous ones. The editor's contribution in fine tuning an article should never be underestimated.
I don't like the cover. HFC is much more elegant and conceptually rigid than this graphic depicts - despite my notes on its susceptibility to ingress and other disturbances.
Errata: In the right column of the table on page 67, it should read "QAM 64: 30Mbps".
There is a discussion of CCS7 - the inter-exchange network protocol - which supports advanced network features and replaces a variety of older protocols used by the exchanges which are made obsolete by the FMO.
The second part of the article describes and compares the various methods of implementing a Full Service Network - whereby one cable (or whatever) system can carry all services from telephony, through Internet to digital video - and where there are high levels of automation in the service activation and maintenance areas of the carrier's operations. The technologies are HFC, Fibre to the Curb (FTTC) and Fibre to the Home (FTTH).
I give a critical analysis of these, and predict that the high electronic costs of FTTC will mean that it is not deployed for some time. HFC will do fine for the foreseeable future - and by then, the cost of optical-electronic components should be low enough to make the FTTH approach most viable. In Japan, there were plans to go straight to FTTH, but I have also heard of HFC projects there as well.
There is a discussion of the need for encryption on the shared media of HFC and FTTH. I report on Telstra CEO Frank Blount's speech at ISSLS in which he notes 7.5% monthly growth of Internet traffic and the apparently recent realisation in Telstra circles that the Internet is going to be bigger than they thought, sooner than they thought.
Update: Vince Pizzica, John Nacheff and other people at Telstra in Melbourne and at my local exchange in Heidelberg went out of their way to help me with this article - it was a pleasure to work with them all. The FMO is being implemented rapidly, but there are still many country areas waiting to be upgraded. The Communications Minister Richard Alston has put pressure on Telstra to speed things up, but my impression was that they were doing it about as fast as was humanly possible.
I understand that the new Euro ISDN software for the FMO exchanges is being tested and will be installed late in 1996 or early in 1997. This will provide one of the prime benefits of the FMO - normal phone exchanges (and their remote access units) will be able to do Basic and Primary Rate ISDN. This will free Telstra from its reliance to date on B-Mux units in local exchange buildings to drive BR-ISDN lines and 2 Mbps links from the B-Muxes back to a few dedicated ISDN exchanges - a costly and difficult system the company hoped to get rid of some years ago. Only when this happens, may ISDN prices be dropped significantly - to do so beforehand would result in demand which could not be satisfied.
Seven months after ISSLS, Telstra had an HFC Internet service! However I expect that the FSN work is a long way from being implemented en-masse. I wonder what became of the big contract with Bellcore to develop FSN management software - before the network elements to be managed had been defined. It seemed to me like a waste of millions of dollars to write software without a specification which would enable it to be tested realistically.
I did the diagrams for this and the August article. For the previous articles, Mark created them from my suggestions.
There is an overview of digital TV and cable modem standards and some discussion of analogue TV signals and the (now rather forlorn) concept of "Interactive TV" - which I always thought was a non-starter and a contradiction in terms.
There is a detailed diagram depicting how the Optus Vision HFC exchange would handle the upstream and downstream signals for all the various things the system might carry: analogue TV, signals to and from the analogue Set Top Units, digital video signals, signals for Internet cable modems and for ADC's telephony system.
Another diagram represents the allocation of upstream and downstream spectrum to the various signals. Methods of expanding an existing HFC system are examined. The article uses the word "node" to refer to the electro-optical conversion unit in the Fibre Serving Area, and "hub" for an intermediate site between the exchange and the FSA - where the fibres to the FSA are driven from. This is the terminology of Optus and the rest of the world - except for Telstra which uses the terms in the opposite sense.
There is a discussion of how Optus Vision's telephony system (which was actually launched in late June 1996) integrates with its Nortel exchanges and how they work with Telstra's.
After considering the typical spectral efficiencies for upstream and downstream data (1.15 and 4.28 bits per Hertz respectively were my estimates) it becomes possible to calculate the total data capacities of the HFC system with a 5 to 65 MHz upstream and an 85 to 750 MHz downstream path. This is 69 and 2,846 Mbps respectively - if the cables are being used for nothing else. There are likely to be reasons why the figures will be less than this - for instance below 30 MHz is relatively noisy due to ingress from various radio transmitters and electrical interference sources. However there are also ways of quadrupling the upstream capacity by sending the upstream signals of each of four cable "tributaries" to the node on the one fibre at four different sets of frequencies.
Since bandwidth for cable modems is shared dynamically, and since HFC FSA's can be split in two with relative ease when the traffic (and hence revenue) warrants it, I am quite relaxed about HFC's data carrying capacity. I don't foresee problems for quite a while.
Update: This is all relevant and accurate 18 months later. A manager from New Zealand Telecom, who was installing HFC in early 1996, told me "We photocopied the article and gave it to all our people, telling them 'This is all you need to know' ". See the August 96 article for a better diagram of what is inside the electro-optical node and the home-based equipment.
Asymmetrical Digital Subscriber Line is a means of getting up to 8 Megabits per second (Mbps) down several kilometres of ordinary twisted pair telephone line, whilst simultaneously receiving up to 0.8 Mbps upstream from the home and supporting a standard analogue POTS (Plain Old Telephone Service) phone.
This article looks at Telstra's ADSL trial, and in particular at Amati's DMT (Discrete Multi Tone) modulation technique - which has some similarities to OFDM. With diagrams, I explain multi-level amplitude modulation and then Quadrature Amplitude Modulation (QAM). AT&T's CAP is basically QAM, and then there is a detailed description of DMT, with special reference to how the modulation depths of the QAM carriers are adjusted to account for varying signal-to-noise ratios on the twisted pair.
There is extensive discussion of the noise, attenuation, distortion, echo and phase difficulties inherent in twisted pair lines and such things as bridge taps. Also some discussion of the massive DSP requirements of ADSL modems and the stringent requirements on the analogue circuitry they need - most particularly the Analogue to Digital Converter.
A table gives data rates for the various classes of operation in the emerging ANSI ADSL standard. 6 Mbps is seen to be the target, after MPEG tests revealed this was the minimum data rate for fast action sports programming with MPEG-2. There is a breakout box on MPEG-2 and the data rates for differing kinds of programs as determined by subjective testing.
Telstra's trial is described in text and a diagram, and there is some discussion of the Video On Demand (VOD) component.
There is a table of educated guesses (from Amati and Telstra Research Labs) on distance limits in kilometres for DMT ADSL at various upstream and downstream data rates on various gauges of wire. Amati estimate they can do 3.5 km over 0.4mm wire with 6 Mbps downstream plus 448 kbps duplex (channels upstream and downstream). TRL estimated that 2 Mbps ADSL with 176 kbps duplex (BR-ISDN!) could work to the distance limit for POTS on the three gauges of wire: 0.4mm, 0.64mm and 0.9mm. The distance limits are 4.2km, 7km and 11km respectively.
Detailed diagrams (Editor Mark Smeaton nearly went cross-eyed doing them to my specifications) show how noise and crosstalk determine the data carrying capacity of the hundreds of QAM carriers up to 1.1 MHz, including for situations where upstream and downstream signals share the same frequencies.
Update: I think this is all still reasonably relevant. The discussion at the start mentions Video On Demand as if it was not far off, with little attention to the Internet - but as time has moved on, real, industrial strength (good enough for entire suburbs) video servers are still some way off, and Internet traffic is growing at more than 15% per month.
Be sure to check for ADSL WWW sites starting with those listed below. I have not really kept up to date with developments in ADSL. Analog Devices has made an ADSL chipset and so no-doubt have others, but I have not checked for them recently.
ADSL is generally expected to play second fiddle to HFC and perhaps other broadband access technologies. I did not closely investigate Telstra's ADSL trial, but in June 96, at Telstra Research Labs, there was a 6 Mbps Amati unit receiving MPEG-2 video. Despite extensive research into mathematical modelling of ADSL modems and the behaviour of twisted pair lines, and especially the problem of impulse noise, one of the engineers at TRL was heard to say something like "Still no-one can predict which phone lines it (the 6 Mbps unit) will run on."
In 1994, it was hoped that a passive hybrid could be used to couple to the POTS phone and the ADSL modem in the home - to provide the lifeline POTS service when there is no power. (Active hybrids would be easier to design, but only work when power is supplied externally.) In early 1996, I heard people talking still about the need for active hybrids.
I don't think I suggested the use of the word "Superhighway".
In 1997, Telstra has signalled its continued interest in ADSL, and some local development is taking place at NEC in Melbourne. However it would only be widely deployed in areas beyond the reach of HFC cables, where twisted pair cables are less than three or four kilometres long. No announcements of deployment have been made - I think everyone is waiting for someone to start a big installation so the electronic production economies of scale can lead to cost-effective ADSL equipment for the home and the exchange site. Update March 1998: I heard that Telstra were very close to a serious ADSL deployment in major regional centres such as Ballarat - but that contractual problems regarding the pay TV component of the proposed service were a fatal stumbling block. This is a shame. ADSL has a lot to recommend it - the ability to bring broadband communications to many homes and businesses without laying new cables. Telstra's deployment would probably have been the first serious ADSL deployment anywhere.
Australian Communications was very well edited by Mark Smeaton and later by Richard Chirgwin.
The telecommunications section of the site is here .
The issues section of this site contains my January 1998 63 page submission to the Senate Select Committee on Information Technology. This submission contains a great deal of material on the various privacy and content regulation issues I have been working on since 1992.
My paper Music Marketing in the Age of Electronic Delivery .
Back to main directory of the First Principles web site.