Optical telephone cables

Modern optical fiber repeaters use a solid-state optical amplifier, usually an Erbium-doped fiber amplifier. Each repeater contains separate equipment for each fiber. These comprise signal reforming, error measurement and controls. A solid-state laser dispatches the signal into the next length of fiber. The solid-state laser excites a short length of doped fiber that itself acts as a laser amplifier. As the light passes through the fiber, it is amplified. This system also permits wavelength-division multiplexing, which dramatically increases the capacity of the fiber.

Repeaters are powered by a constant direct current passed down the conductor near the center of the cable, so all repeaters in a cable are in series. Power feed equipment is installed at the terminal stations. Typically both ends share the current generation with one end providing a positive voltage and the other a negative voltage. A virtual earth point exists roughly half way along the cable under normal operation. The amplifiers or repeaters derive their power from the potential difference across them.

The optic fiber used in undersea cables is chosen for its exceptional clarity, permitting runs of more than 100 kilometers between repeaters to minimize the number of amplifiers and the distortion they cause.

The rising demand for these fiber-optic cables outpaced the capacity of providers such as AT&T. Having to shift traffic to satellites resulted in poorer quality signals. To address this issue, AT&T had to improve its cable laying abilities. It invested $100 million in producing two specialized fiber-optic cable laying vessels. These included laboratories in the ships for splicing cable and testing its electrical properties. Such field monitoring is important because the glass of fiber-optic cable is less malleable than the copper cable that had been formerly used. The ships are equipped with additional propellers that increase maneuverability. This capability is important because fiber-optic cable must be laid straight from the stern (another factor copper cable laying ships did not have to contend with).

Originally, submarine cables were simple point-to-point connections. With the development of submarine branching units (SBUs), more than one destination could be served by a single cable system. Modern cable systems now usually have their fibers arranged in a self-healing ring to increase their redundancy, with the submarine sections following different paths on the ocean floor. One driver for this development was that the capacity of cable systems had become so large that it was not possible to completely back-up a cable system with satellite capacity, so it became necessary to provide sufficient terrestrial back-up capability. Not all telecommunications organizations wish to take advantage of this capability, so modern cable systems may have dual landing points in some countries (where back-up capability is required) and only single landing points in other countries where back-up capability is either not required, the capacity to the country is small enough to be backed up by other means, or having back-up is regarded as too expensive.

A further redundant-path development over and above the self-healing rings approach is the "Mesh Network" whereby fast switching equipment is used to transfer services between network paths with little to no effect on higher-level protocols if a path becomes inoperable. As more paths become available to use between two points, the less likely it is that one or two simultaneous failures will prevent end-to-end service.

As of 2012 operators had "successfully demonstrated long-term, error-free transmission at 100Gbps across Atlantic Ocean" routes of up to 6000km , meaning a typical cable can move tens of terabits per second overseas. Speeds improved rapidly in the last few years, with 40Gbit/s having been offered on that route only three years earlier in August 2009 .

Switching and all-by-sea routing commonly increases the distance and thus the round trip latency by more than 50%. For example, the round trip delay (RTD) or latency of the fastest transatlantic connections is under 60ms, close to the theoretical maximum for an all-sea route. While in theory a great circle route between London and New York City is only 5600km  this requires several land masses (Ireland, Newfoundland, Prince Edward Island and the isthmus connecting New Brunswick to Nova Scotia) to be traversed, as well as the extremely tidal Bay of Fundy and a land route along Massachusetts' north shore from Gloucester to Boston and through fairly built up areas to Manhattan itself. In theory using this partly land route could result in round trip times below 40ms, not counting switching. Along routes with less land in the way, speeds can approach speed of light minimums in the long term.