Technical and commercial evolution
Time to market
One of the most obvious changes which is happening at the moment is that the very relaxed pace at which the traditional carrier consortia operated is being dramatically challenged. There used to be at least four years between the initial consortium discussions about a new cable and its introduction into service. FLAG, which is a very complex cable system, reduced this to 29 months. In their simpler technical and organisational envir¬onments, both Gemini South and the US-UK segment of Atlantic Crossing entered comm¬ercial service only 18 months after initial discussions and some 15 months after the con¬struction contracts were signed.
Transmission speed at a given wavelength
Until 1998 most cable systems used 2.5 Gbps (STM-16) as the “standard” SDH interface at the end stations, but several of the recently announced systems, including Project Oxygen, plan to use 10 Gbps transmission (STM-64). This speed now seems poised to gain general acceptance, even though it requires rather special and expens¬ive electronics. There seems to be a feeling that it will not be productive to move on to even higher speeds, such as 40 Gbps, in the near future. Conventional wisdom holds that it will be more interesting to increase overall capacity by even further increasing the number of wavelengths being used than by increasing the speed of transmission.
In 1997 already, experiments had been reported with soliton transmission at 20 Gbps over 200 km (KDD) and at 20 Gbps over many thousands of km (by looping) (Alcatel).
More wavelengths
The number of wavelengths which can be trans¬mitted on a single fibre (pair) is governed by several factors, of which the flatness of the optical gain in the EDFAs at different wavelengths, the dispersion characteristics of the fibre itself, and the signal-to-noise requirements of the signal processing (10 Gbps requires 6dB more than 2.5 Gbps) are the most critical. A good technical description of the issues involved can be found at http://www.ans.alcatel.com/telecom/transpt/optical/techpaps/. The ITU has defined various schemes for wavelength separation, in order that equipment from different vendors can inter-operate, and the most widely used of those has a separation of 100 Ghz (just under 0.8 nm) in the 1525-1560 nm passband.
For sub¬marine cables being installed today, 4- or 8-way WDM at 2.5 Gbps is the "obvious" choice, but the companies which specialise in this equipment have 40-way systems avail¬able, and those are already being deployed on terrestrial cable systems. Project Oxygen plans to use 16-way WDM at 10 Gbps. Bell Labs has demon¬strated 100-way WDM at 10 Gbps over 400 km on a single fibre, for a total capacity of 1 Tbps. Some people feel that it might be possible to transmit as many as 1000 wavelengths within the next 5-10 years.
The prospect of signalling at many more wavelengths is very attractive to owners of existing submarine (or land-based) cable systems, since it seems to offer them a way of providing additional band¬width which is quick, easy and cheap to install. However, early designs of EDFAs and the dispersion characteristics of some installed cables will mean that easy upgrades will not be possible everywhere. The ownership of well-adapted cable systems might concentrate very great economic power in the hands of a few "bandwidth barons".
More fibres per cable
“Normal” optical fibre costs about 0.1 $ per metre to fabricate, and retails in the region of 0.2-0.25 $ per fibre-metre when several fibres are installed in a cladding, unless there are few fibres and the cladding is very exotic. So, a 16-fibre indoor cable might cost 4 $ per metre. Double armoured submarine cable can cost in the region of 20 $ per metre, but, until very recently only contained two fibre pairs. The “standard” has recently moved to four pairs per long-distance cable.
The limit to the number of fibre pairs in a submarine cable does not come from the cost of the extra fibres, but from the economics of building repeaters. These are very expensive devices (because they have to be very reliable in an inhospitable environment), with a very limited production run. A significant part of their costing depends linearly on the number of fibres that has to be amplified. If and when the industry becomes convinced that building repeaters to handle 8 or even 16 fibre pairs is an investment which makes economic sense, then such an upgrade could happen quite quickly.
More pure optical switching
As far as I can judge we are close to the point where production fibre optic networks will start to operate in an (almost) "all-optical" mode. Until now network providers have always had to assume the presence of an electronic switching layer in their overall network architecture. SDH technology, which offers features for redund¬ancy, multi-plexing, and operational control, is a basic element of any modern wide-area network. But we are very close to the point where the optical switching layer will be able to provide many of these high-level functions. One way of thinking about this is that the optical layer, by using WDM to transmit signals at multiple wavelengths, will have divided each fibre into multiple independent paths.
Of course, the electronic switching layer, and SDH management tools in particular, are not going to disappear overnight, but it does look as though an important trans-formation is about to take place. The possibility of completely avoiding the need for an SDH layer when linking the local area networks of any widely distributed organisation via an "all-optical" network will surely provide a considerable financial incentive for many companies. It is possible that the speed of change will be a bit slower on submarine cables, in view of their strong need to integrate well with the operational needs of existing telecoms carriers, which are heavily dependent on SDH operations.
Summary
Gemini is operating at 15 Gbps per fibre pair today. Project Oxygen plans to start (at least partial) deployment, operating at 160 Gbps per fibre pair, by the end of 2000.
Submarine cable systems with total capacities of several (tens of?) Tbps, installable by the year 2005, look to be entirely plausible.