To realise novel, low-cost, photonic technologies that can support 100Gb/s Ethernet in next-generation dense wavelength-division-multiplexed metro transport networks, we are developing arrayed photonic integrated circuits that leverage colourless reflective modulators. Here, we demonstrate a single-channel, hybrid reflective electroabsorption modulator-based device, showing error-free 25.3Gb/s duobinary transmission with bit-error rates less than 1 × 10(-12) over 35km of standard single-mode fibre. We further confirm the modulator's colourless operation over the ITU C-band, with a 1.2dB variation in required optical signal-to-noise ratio over this wavelength range.

Exponentially-increasing demands on the current telecommunication infrastructure are driving the development of next-generation ultra-high-bandwidth network architectures with sufficiently low energy consumptions. Within the scope of the EU FP7 C3PO project, we are developing novel, energy-efficient, colourless photonic technologies for metro applications. The colourless transmitters will leverage reflective photonic integrated circuits, specifically reflective electroabsorption modulator-based phase and amplitude modulators, in conjunction with multi-frequency lasers and low-loss piezoelectric beam-steering optical matrix switches, in order to achieve wavelength reconfigurability without the requirement for tuneable lasers. A specific target is a dynamically reconfigurable metro node which supports duobinary modulation for high dispersion tolerance and efficient spectral usage, to enable 100 Gb/s Ethernet dense wavelength-division-multiplexed transport networks. We report on recent progress towards these metro transport networks, providing the latest system test results obtained using novel hybrid photonic integrated devices.

The author gives a slide presentation of next generation access passive optical networks.

Two dimensional arrays of InGaAsIJnP based multiple quantum well surface modulators driven by standard high-speed CMOS have been demonstrated in experimental parallel optical interconnect and artificial neural processing systems. Transition times were fast enough for lOOMBit/s operation and the potential exists to increase array dimensions to include 100''s of devices and transmission rates to many GBit/s. Novel architectures employing computer generated holographic beam splitters and weight matrices outside the modulator and detector layers were employed.

A family of fixed and reconfigurable optical interconnection networks offering the potential of Terabit throughput are described. The designs are based on a bus architecture that combines space, wavelength and other multiplexing methods and have the following attractive features:- graceful growth; wide-sense non-blocking operation; low space switch crosspoint count; and simple control algorithm.

The classical optical matrix-vector multiplier, or crossbar, offers a number of obvious advantages over its electronic counterpart when the data transmission rate is very high. However, there are two principal limiting factors in its design, namely, insertion loss and crosstalk. Insertion loss is unavoidable and sets a loss-limited upper bound to the switch dimensions due to the finite optical power budget. More important, the low-contrast ratio of present-day spatial light modulators (SLMs) and wiring crosstalk caused by aberrations in the imaging optics also sets a limit to switch dimensions. Fortunately, this crosstalk limit is dominant in the lower transmission rate systems which are of less interest. This paper examines the relative magnitudes of the loss and crosstalk limit and their effect on switch dimensions and how the switch performance can be improved by tailoring the input and output optical field shapes. It is also shown how the crossbar can be made bidirectional without loss or crosstalk penalty. A bulk-optic 32 × 32 bidirectional crossbar with a transmission rate of 140 Mbit/s is proposed which uses multimode optical fiber inputs and outputs and a twisted-nematic liquid crystal SLM cross-point matrix.

A heteroassociative memory and the corresponding stable eigenmodes (states) of a ring resonator may be arranged to perform roles equivalent to the electronic logic and the latch memory of a classical finite state machine (FSM). Heteroassociative FSMs may be created by preloading (programming) an eigenstate memory with a set of associatively linked stable states. The links are completed by seeding the FSM with external stimuli. The underlying principle is similar to associative chain recall. However, a state transition can only occur in this system when the combined current FSM state and next input vector are identified as a new state stimulus. One-dimensional, open or closed, chains of stable states are quite straightforward to program; however, when branches are required, many unprogrammed states and links between states may appear. One solution to this problem is to use multiple-feedback loops, each loop corresponding to a higher level of decision. The lower level loops create keys which allow access to the higher levels only when prerequisite states have been reached.