Parametric devices are traditionally recognized as phase sensitive amplifiers and for their ability to perform noiseless signal regeneration. In its simplest form, the device uses a single pump wave in the anomalous dispersion regime that is coupled with a signal seed to a nonlinear waveguide. Efficient amplification and wavelength conversion is achieved by satisfying phase matching condition among pump, signal and idler waves. This process is described by degenerate (one-pump) four-photon mixing as two pump photons are annihilated to generate a signal and an idler photon such that 2wpump → wsignal + widler , as illustrated in Figure 1. Parametric devices are also capable of providing not only phase sensitive amplification but also, wavelength conversion, signal conjugation and regeneration. Their versatility and complex nature, in their single or multiple-pump forms, have thus established parametric devices as unique processing devices.

 Figure 1: Conventional (one-pump) parametric device: two pump photons are annihilated to generate a signal and an idler photon. l0 is the zero dispersion wavelength (ZDW).

With the increasing speed of data transmission, the need to close the gap between signal processing speed and transmission rate has become a necessity.  While the analysis of an arbitrary-rate, arbitrary-format data stream can be done off-line with a capture-and-dump approach, real time processing would be able to capture details and transient events in a realistic physical link. Furthermore, the terabit-scale capacity of current commercial links is supported by channel granularities ranging from 10 to 100 Gb/s. Driven by emerging high bandwidth applications, such as high-definition video, future links are expected to introduce both higher aggregate capacities and higher channel granularities in order to satisfy customer needs. Unfortunately, this trend will also widen the presently existing gap between the transmission rates and the processing layer speed. In fact, while it is relatively common to plan for 10 Tbps transmission over single fiber strands, data processing is still performed by low bandwidth electronics with rates in the GHz range.

Figure 2: Array of conventional demultiplexers for recovery of all tributaries from optical input with high data rate OR Multi channel processor for recovery of all tributaries from optical input with high data rate.

While electronic processing rate may never approach those of the optical transport, it is still possible to close this gap by all-optical preprocessing. In that regard, optical time division (OTD) demultiplexing was recognized early and used to decompose fast optical signal to lower, electronic-compatible rates. However, conventional demultiplexing schemes commonly produce a single tributary at a time. Device replication to many parallel data pipelines is therefore necessary to recover all the subrate tributaries. This type of parallel architecture (Fig. 2) significantly increases the complexity of the processor, challenges synchronization and hinders its scalability. Consequently, the scalability of classical OTD devices is fundamentally limited to applications requiring only a few subrate outputs. Motivated by this simple principle, we have proposed a new preprocessing architecture called Multicasting Parametric Synchronous Sampling (MPASS), where a single multichannel processor replaces the complex array of the conventional scheme (Fig. 3). Indeed, the femtosecond-scale parametric response in materials such as silica, establish parametric devices as exceptional candidates for this type of ultra-fat all-optical processing. The MPASS processor operation is illustrated in Fig. 4. The operation rests on 3 building blocks: parametric-multicasting (PM), synchronization (PS) and gating (PG). The parametric multicasting block replicates the input signal to M distinct spectral copies. The copies are then delayed with respect to each other within the delay block, shifting spectrally adjacent replicas by precise time intervals. The sub-rate parametric gate simultaneously samples all the multicast copies, resulting in spectrally-distinct, temporally-shifted slices of the original signal. Finally, the samples are separated using a coarse wavelength demultiplexer, resulting in M continuously-streaming, sub-rate data outputs. This topology can be varied to fit the specific application and the optimal system design will ultimately depend on the input signal rate, the required output granularity and the available spectral parametric bandwidth.

Figure 4: Multicast Parametric Synchronous Sampling (MPASS). Input channel at rate B is parametrically multicast to M spectrally distinct copies, synchronized and subsequently sampling by a single parametric gate operating at rate R in order to produce M subrate streaming outputs.

The infrared region beyond 2 μm is of interest for silicon four-wave mixing (FWM) since two-photon absorption (TPA) and the resulting free-carrier absorption are reduced as the photon energy becomes less than half the band-gap energy of silicon. Thus, silicon could potentially be an attractive platform for parametric nonlinear optics for applications that require a mid-infrared (mid-IR) source such as light detection and ranging (LIDAR) and free-space communication, remote chemical and bio-molecular sensing, and infrared spectroscopy.

We demonstrated wide band four-wave mixing in silicon waveguides over 630nm with a 2μm pump generated from telecom-compatible fiber-optic sources. The measurements indicate that the nonlinear parameter g is comparable to that in near-infrared band and, combined with a reduction in the TPA-induced nonlinear absorption, results in a t tenfold improvement of the nonlinear figure of merit. Consequently, two-stage mixer technology that combines NIR (silica) and mid-IR (silicon) waveguides holds considerable potential for application in a wide range of mid-infrared nonlinear optic devices.

Fiber optic communications represent the single most valuable enabler of the ubiquitous internet and multimedia presence in the modern society. Our group is continuously engaged in research and development of future generation optical transmission in terms of novel processing techniques availing a high capacity, low power consumption transport.

The group's interest particularly lies in leveraging off of the strength in information theory and all optical signal processing in designing the spectrally and energy efficient systems. The activity in this area is in close connection to the one of a kind in the world experimental testbed in the CALIT2 Photonics systems laboratory. The group particularly pursues large scale system experimental demonstrations whose quality and rigor are unmatched by any other University laboratory in the world.

The Photonics systems laboratory boast several records such as: the first ever experimental demonstration of the effectiveness of electronic dispersion compensation in high speed optical communications, first errorless transmission of 640 Gb/s, as well as the first errorless wavelength-transparent micro-second all optical delay. The current research interests include highly spectrally efficient transmission, novel modulation formats, advanced equalization techniques as well as the ultra-dense bi-directional transport.

Contact: Dr Nikola Alic (nalic@ucsd.edu)

Conventional LIDAR/LADAR device typically operates in fixed band (wavelength) and requires high-power transmitter. The present state of LIDAR technology is the consequence of historical development, and qualitatively better device can be developed by abandoning the conventional restraints. Firstly, fixed-wavelength operation is seen as a crippling disadvantage that eliminates features inherent to arbitrary spectral access. Secondly, high power transmitter requirement effectively eliminates any designs on device portability. Consequently, the LIDAR construction must be revisited by dispensing with conventional notions of high-power, noisy detectors and large collection optics.

Recent advances in three dissimilar disciplines have provided a basis for LIDAR reengineering. Firstly, high-gain parametric conversion has opened the continuous NIR/SWIR bands (1000-2500nm); secondly, high-speed signal processing developed for conventional communication band (1550nm) became applicable in the contiguous NIR-SWIR range; finally, new class of free-space to fiber interfaces have qualitatively raised the efficiency of fiberized receivers.  The combination of these basic advances makes the redefinition of the LIDAR transceiver, at least in principle, a practical consideration: new technology should lead to the concept of ultraportable LIDAR, capable of dual-mode (sensing/communication) operation in single or multiple (simultaneous) bands.

In simple illustration of the concept, the optical pulse in conventional (1550nm) band is encoded by data or correlation sequence and used as a parametric pump. A weak seed tone, positioned in NIR band is mixed with the encoded pulse to generate the launch (probe/carrier) pulse anywhere in continuous NIR/SWIR band. Parametric process guarantees not only high conversion efficiency but also that the encoded information (phase or amplitude) is imprinted on the probe pulse serving in either data carrier or sensing role. If operating in sensing mode, the launched pulse is reflected of the target, with return photons now serving as the seed entering the parametric receiver. Pulse centered at the original pump wavelength (1550nm) and captured SWIR light now mix, generating the photons at the original (seed) wavelength. With the efficiency of the conversion process exceeding 40dB, the receiver often does not need any preamplifiers. More importantly, the optical gain is, in principle, spectrally invariant, freeing the designer from constrains of operating in rare-earth ion band.  

While the introduction of parametric mixing effectively eliminates the need for separate preamplifier prior to LIDAR receiver, it also creates a much more important capability. The, parametric transceiver is a self-coherent construct. As long as the spectral linewidth of the pulse-forming laser is kept within practical limits, the received signal phase and amplitude will be processed in coherent mode.

The challenge of ADCs stem from the continual and growing need of digital signal processing that evolves towards higher rates and improved accuracy (dynamic range), for instance receive weak signals in the presence of strong ones or high noise. Existing ADCs need to improve operation to faster sampling rates and higher effective number of bits (ENOBs). The performance boundaries for an ADC are set by the signal to noise ratio (SNR), linearity, aperture jitter and width of the sampling gate. Sampling gate ambiguity of conventional electronic-based ADCs is limited by the finite transition speed of semiconductor circuits. To circumvent the rate barrier, photonic technology has been employed as photonic-based ADC or photonic-sampled ADC in order to improve performance compared to conventional electronic-based ADCs. Photonic architectures are used as a front-end to perform signal copying to several wavelengths (multicasting) and sampling. An example is illustrated in Fig.1 where a signal is optically sampled and subsequent electronics that are used for final digitization. The photonic platform of particular interest is based on the third order non-linearity of four-wave mixing (FWM) in optical silica fiber. The silica possesses femtosecond (10-15s) parametric response and the bandwidth and efficiency of highly non-linear silica fiber (HNLF) is compatible with any rate of interest. This offers a solid physical foundation for a parametric mixer that can perform multicasting and sampling. In a practical implementation, short optical pulses can be used to drive a fast, nonlinear gate and perform the sampling of the incoming signal.

Figure: Optical sampling of an analog waveform with subsequent electro-optic conversion and quantization.