Photonic next-generation reservoir computer based on distributed feedback in optical fiber.

Reservoir computing (RC) is a machine learning paradigm that excels at dynamical systems analysis. Photonic RCs, which perform implicit computation through optical interactions, have attracted increasing attention due to their potential for low latency predictions. However, most existing photonic RCs rely on a nonlinear physical cavity to implement system memory, limiting control over the memory structure and requiring long warm-up times to eliminate transients. In this work, we resolve these issues by demonstrating a photonic next-generation reservoir computer (NG-RC) using a fiber optic platform. Our photonic NG-RC eliminates the need for a cavity by generating feature vectors directly from nonlinear combinations of the input data with varying delays. Our approach uses Rayleigh backscattering to produce output feature vectors by an unconventional nonlinearity resulting from coherent, interferometric mixing followed by a quadratic readout. Performing linear optimization on these feature vectors, our photonic NG-RC demonstrates state-of-the-art performance for the observer (cross-prediction) task applied to the Rössler, Lorenz, and Kuramoto-Sivashinsky systems. In contrast to digital NG-RC implementations, we show that it is possible to scale to high-dimensional systems while maintaining low latency and low power consumption.

Integrated photonic encoder for low power and high-speed image processing.

Modern lens designs are capable of resolving greater than 10 gigapixels, while advances in camera frame-rate and hyperspectral imaging have made data acquisition rates of Terapixel/second a real possibility. The main bottlenecks preventing such high data-rate systems are power consumption and data storage. In this work, we show that analog photonic encoders could address this challenge, enabling high-speed image compression using orders-of-magnitude lower power than digital electronics. Our approach relies on a silicon-photonics front-end to compress raw image data, foregoing energy-intensive image conditioning and reducing data storage requirements. The compression scheme uses a passive disordered photonic structure to perform kernel-type random projections of the raw image data with minimal power consumption and low latency. A back-end neural network can then reconstruct the original images with structural similarity exceeding 90%. This scheme has the potential to process data streams exceeding Terapixel/second using less than 100 fJ/pixel, providing a path to ultra-high-resolution data and image acquisition systems.

High-speed RF spectral analysis using a Rayleigh backscattering speckle spectrometer.

Persistent wideband radio frequency (RF) surveillance and spectral analysis is increasingly important, driven by the proliferation of wireless communication and RADAR technology. However, conventional electronic approaches are limited by the ∼1 GHz bandwidth of real-time analog-to-digital converters (ADCs). While faster ADCs exist, high data rates prohibit continuous operation, limiting these approaches to acquiring short snapshots of the RF spectrum. In this work, we introduce an optical RF spectrum analyzer designed for continuous, wideband operation. Our approach encodes the RF spectrum as sidebands on an optical carrier and relies on a speckle spectrometer to measure these sidebands. To achieve the resolution and update rate required for RF analysis, we use Rayleigh backscattering in single-mode fiber to rapidly generate wavelength-dependent speckle patterns with MHz-level spectral correlation. We also introduce a dual-resolution scheme to mitigate the trade-off between resolution, bandwidth, and measurement rate. This optimized spectrometer design enables continuous, wideband (15 GHz) RF spectral analysis with MHz-level resolution and a fast update rate of 385 kHz. The entire system is constructed using fiber-coupled off-the-shelf-components, providing a powerful new approach for wideband RF detection and monitoring.

High-speed broadband absorption spectroscopy enabled by cascaded frequency shifting loops.

Frequency shifting loops, consisting of a fiber optic ring cavity, a frequency modulator, and an amplifier to compensate for loss, enable high-speed frequency scanning with precise and easily controlled frequency steps. This platform is particularly attractive for applications in spectroscopy and optical ranging. However, amplified spontaneous emission noise accumulates due to the repeated amplification of light circulating in the cavity, limiting the frequency scanning range of existing frequency shifting loops (FSLs). Here, we introduce a cascaded approach which addresses this basic limitation. By cascading multiple FSLs in series with different frequency shifts we are able to dramatically increase the accessible scanning range. We present modeling showing the potential for this approach to enable scanning over ranges up to 1 THz-a tenfold increase compared with the state-of-the-art. Experimentally, we constructed a pair of cascaded FSLs capable of scanning a 200 GHz range with 100 MHz steps in 10 ms and used this platform to perform absorption spectroscopy measurements of an H13C14N cell. By increasing the operating bandwidth of FSLs, the cascaded approach introduced in this work could enable new applications requiring precise and high-speed frequency scanning.

Dynamic temperature-strain discrimination using a hybrid distributed fiber sensor based on Brillouin and Rayleigh scattering.

We present a distributed fiber sensor capable of discriminating between temperature and strain while performing low-noise, dynamic measurements. This was achieved by leveraging recent advances in Brillouin and Rayleigh based fiber sensors. In particular, we designed a hybrid sensor that combines a slope-assisted Brillouin optical time domain analysis system with a Rayleigh-scattering-based frequency scanning optical time domain reflectometry system. These sub-systems combine state-of-the-art sensitivity with the ability to perform both dynamic and quasi-static measurements. This enabled a hybrid system capable of temperature/strain discrimination with a quasi-static temperature resolution of 16 m°C and a strain resolution of 140 nɛ along 500 m of single mode fiber with 5 m spatial resolution. In contrast to previously reported techniques, this approach also enabled dynamic measurements with a bandwidth of 1.7 kHz and temperature (strain) noise spectral density of 0.54 m°C/√Hz (4.5 nɛ/√Hz) while temperature/strain cross-sensitivity was suppressed by at least 25 dB. This represents a dramatic improvement in measurement speed and sensitivity compared with existing techniques capable of temperature/strain discrimination in standard single mode fiber.

Massively distributed fiber strain sensing using Brillouin lasing.

Brillouin based distributed fiber sensors present a unique set of characteristics amongst fiber sensing architectures. They are able to measure absolute strain and temperature over long distances, with high spatial resolution, and very large dynamic range in off-the-shelf fiber. However, Brillouin sensors traditionally provide only modest sensitivity due to the weak dependence of the Brillouin frequency on strain and the high signal to noise ratio required to identify the resonance's peak frequency to within a small fraction of its linewidth. Recently, we introduced a technique which substantially improves the precision of Brillouin fiber sensors by exciting a series of lasing modes in a fiber loop cavity that experience Brillouin amplification at discrete locations in the fiber. The narrow-linewidth and high intensity of the lasing modes enabled ultra-low noise Brillouin sensors with large dynamic range. However, our initial demonstration was only modestly distributed: measuring strain at 40, non-contiguous positions along a 400 m fiber. In this work, we greatly extend this methodology to enable fully distributed sensing at 1000 contiguous locations along 3.5 km of fiber-an order of magnitude increase in sensor count and range. This highly-multiplexed Brillouin fiber laser sensor provides a strain noise as low as 34 nɛ/√Hz and we analyze the limiting factors in this approach.

High-resolution wide-band optical frequency comb control using stimulated Brillouin scattering.

We introduce a technique to manipulate an optical frequency comb on a line-by-line basis using stimulated Brillouin scattering (SBS). The narrow-linewidth SBS process has been used to address individual lines in optical frequency combs, but previous demonstrations required a dedicated laser to modulate each comb tooth, prohibiting complete comb control. Here, we use a pair of frequency shifting fiber optic loops to generate both an optical frequency comb and a train of frequency-locked pulses that can be used to manipulate the comb via SBS. This approach enables control of the entire frequency comb using a single seed laser without active frequency locking. To demonstrate the versatility of this technique, we generate and manipulate a comb consisting of 50 lines with 200 MHz spacing. By using polarization pulling assisted SBS, we achieve a modulation depth of 30 dB. This represents a scalable approach to control large numbers of comb teeth with high resolution using standard fiber-optic components.

Enhanced bandwidth distributed acoustic sensing using a frequency multiplexed pulse train and micro-machined point reflector fiber.

In this Letter, we present an enhanced bandwidth distributed acoustic sensor (DAS) that uses a frequency multiplexed interrogation system to probe a micro-machined point reflector fiber. The fiber contains a series of discrete point reflectors with reflectance as high as -48 dB, while the frequency multiplexed interrogator allows us to increase the effective pulse repetition rate by a factor of 10. Together, this enables a phase noise as low as -101 dB (re rad2/Hz) for a 2.5 km fiber with 10 m spatial resolution, corresponding to a strain noise of 0.095p ε/Hz. This scheme also enables a 10-fold increase in the sensor bandwidth without introducing noise due to interference fading. Finally, we demonstrate sensing at ranges up to 10 km using a fiber containing 1000 point reflectors, illustrating the scalability of this approach.

Frequency multiplexed coherent φ-OTDR.

We present a comprehensive analysis of a frequency multiplexed phase-measuring φ-OTDR sensor platform. The system uses a train of frequency-shifted pulses to increase the average power injected into the fiber and provide a diversity of uncorrelated Rayleigh backscattering measurements. Through a combination of simulations, numerical analysis, and experimental measurements, we show that this approach not only enables lower noise and mitigates interference fading, but also improves the sensor linearity. We investigate the sensor dependence on the length of the pulse train and characterize the sensor performance as a function of range, demonstrating operation from 1 to 50 km. Despite its relative simplicity, this platform enables state-of-the-art performance, including low crosstalk, high linearity, and a minimum detectable strain of only 0.6 p[Formula: see text] in a 10 km fiber with 10 m spatial resolution and a bandwidth of 5 kHz.

Quantitative amplitude-measuring Φ-OTDR with pε/√Hz sensitivity using a multi-frequency pulse train.

We report an amplitude-measuring Rayleigh-based sensor that uses a series of frequency-shifted pulses to extract quantitative distributed strain measurements. By using frequency multiplexing, we are able to inject a train of 10 pulses into the fiber at once. This allows us to use a higher average input power than standard phase-sensitive optical time domain reflectometry systems, improving the sensitivity. The sensor recovers the strain by tracking the time-dependent amplitude of the Rayleigh backscattered light from all 10 pulses. This approach enables a sensor with a noise floor of 1.5pε/√Hz over 10 km of fiber with 12 m spatial resolution, a 5 kHz bandwidth, and a dynamic range of 80 dB at 1 kHz. The sensor exhibits a high degree of linearity and is immune to interference fading.

Low-noise distributed acoustic sensing using enhanced backscattering fiber with ultra-low-loss point reflectors.

We present a low-noise distributed acoustic sensor using enhanced backscattering fiber with a series of localized reflectors. The point reflectors were inscribed in a standard telecom fiber in a fully automated system by focusing an ultra-fast laser through the fiber cladding. The inscribed reflectors provided a reflectance of -53 dB, significantly higher than the Rayleigh backscattering level of -70 dB/m, despite adding only 0.01 dB of loss per 100 reflection points. We constructed a coherent φ-OTDR system using a double-pulse architecture to probe the enhanced backscattering fiber. Using this system, we found that the point reflectors enabled an average phase noise of -91 dB (re rad2/Hz), 20 dB lower than sensors formed using Rayleigh backscattering in the same fiber. The sensors are immune to interference fading, exhibit a high degree of linearity, and demonstrate excellent non-local signal suppression (>50 dB). This work illustrates the potential for low-cost enhanced backscattering fiber to enable low-noise, long-range distributed acoustic sensing.

Suppressing non-local effects due to Doppler frequency shifts in dynamic Brillouin fiber sensors.

Brillouin fiber sensors have traditionally been limited to low-speed or static strain measurements due to the time-consuming frequency scans required. In the past decade, a number of novel high-speed measurement techniques have been proposed to enable Brillouin-based dynamic strain sensors. In this work, we present a new mechanism, which can limit the performance of high-speed dynamic Brillouin sensors. Specifically, we show that dynamic strain induced Doppler shifts can corrupt a distributed Brillouin strain measurement by introducing non-local signals throughout the fiber. We present a model showing that these non-local signals can introduce unacceptable levels of cross-talk or even exceed the local signal strength in reasonable operating conditions and experimentally observe these signals in a standard slope-assisted BOTDA sensor. Finally, we present a modified sensor architecture to address this issue and experimentally demonstrate low-noise (2.6 nε/Hz1/2), dynamic strain measurements with significantly reduced sensitivity to this type of non-local signal.

Quantitative strain sensing in a multimode fiber using dual frequency speckle pattern tracking.

We report an amplitude-measuring multimode fiber sensor capable of making quantitative strain measurements and extracting the algebraic sign of the strain. The Rayleigh-based sensor probes the fiber with pulses of alternating optical frequency and records the backscattered speckle patterns on a high-speed camera. We show that measuring the change in the speckle pattern induced by a change in optical frequency provides a form of in situ calibration, enabling the sensor to recover the magnitude and algebraic sign of the strain. The sensor, which can be positioned anywhere along 2 km of fiber, has a linear strain response, a 10 kHz bandwidth, and a strain noise of ${10.2}\;{\rm p}\unicode{x03B5} /\surd {{\rm Hz}} $10.2pε/√Hz.

Quantitative amplitude measuring φ-OTDR using multiple uncorrelated Rayleigh backscattering realizations.

We propose and demonstrate a technique to perform quantitative strain sensing using the amplitude of the Rayleigh backscattered light in a modified φ-OTDR system. While standard amplitude measuring φ-OTDR sensors can identify the presence of strain, they cannot perform quantitative measurements because the amplitude of the Rayleigh backscattered light exhibits a non-linear and unpredictable strain response. Here, we demonstrate a technique to computationally recover a linear strain response from a set of uncorrelated Rayleigh backscattering measurements. Using a combination of frequency and polarization multiplexing, we constructed a φ-OTDR system capable of recording 18 uncorrelated Rayleigh backscattering measurements in parallel. By combining information from these 18 measurements, the sensor achieves a linear strain response with total harmonic distortion below -35 dB. The sensor is immune to signal fading, has a minimum detectable strain of 5 pɛ/√Hz and a bandwidth of 500 kHz.

Speckle-based strain sensing in multimode fiber.

The diversity of spatial modes present within a multimode fiber has been exploited for a wide variety of imaging and sensing applications. Here, we show that this diversity of modes can also be used to perform quantitative strain sensing by measuring the amplitude of the Rayleigh backscattered speckle pattern in a multimode fiber. While most Rayleigh based fiber sensors use single mode fiber, multimode fiber has the potential to provide lower noise due to the higher capture fraction of Rayleigh scattered light, higher non-linear thresholds, and the ability to avoid signal fading by measuring many spatial modes simultaneously. Moreover, while amplitude measuring single mode fiber based Rayleigh sensors cannot provide quantitative strain information, the backscattered speckle pattern formed in a multimode fiber contains enough information to extract a linear strain response. Here, we show that by tracking the evolution of the backscattered speckle pattern, the sensor provides a linear strain response and is immune to signal fading. The sensor has a noise floor of 2.9 pɛ/√Hz, a dynamic range of 74 dB at 1 kHz, and bandwidth of 20 kHz. This work paves the way for a new class of fiber optic sensors with a simplified design and enhanced performance.

Multimode fiber Φ-OTDR with holographic demodulation.

We propose and demonstrate a method to perform quantitative phase-sensitive optical time domain reflectometry (Φ-OTDR) using multimode fiber. While most Φ-OTDR sensors use single-mode fiber, multimode fiber exhibits higher thresholds for non-linear effects, a larger capture fraction of Rayleigh backscattered light, and the potential to avoid signal fading by detecting many spatial modes in parallel. Previous multimode fiber based OTDR sensors discarded most of the backscattered light and thus failed to take advantage of these noise-reducing factors. Here, we show that by performing off-axis holography with a high-speed camera, we can record the entire Rayleigh backscattered field, maximizing the detected light level and making the sensor immune to fading. The sensor exhibits a high degree of linearity, a minimum phase noise of -80 dB [rel. rad2/Hz], and 20 kHz bandwidth.

Measuring vibrational motion from a moving platform using speckle field detection.

We investigate the ability of a holographic laser vibrometer to mitigate noise introduced when operating on a moving platform or when measuring a moving target. This motion introduces a fundamental limitation on the measurement sensitivity due to the time-varying speckle pattern produced as the illumination beam scans across the target surface. In addition, since existing systems record the phase of only a single speckle grain, speckle fading imposes a limit on the coherent processing interval and thus the frequency resolution of these measurements. In this work, we show that by measuring N speckle grains in parallel using holographic detection, we are able to provide a N1/2 improvement in the system sensitivity while simultaneously overcoming the limitations on the coherent processing interval imposed by speckle fading. The ability to perform sensitive vibrational measurements of a moving target or from a moving platform could dramatically increase the applications available to laser vibrometry.

Enabling time resolved microscopy with random Raman lasing.

Optical imaging of fast events and processes is essential for understanding dynamics of complex systems. A bright flash of illuminating light is required to acquire sufficient number of photons for superior image quality. Laser pulses can provide extreme brightness and are typically employed to achieve high temporal resolution; however, the high degree of coherence associated with the lasing process degrades the image quality with speckle formation. Random lasers are low-coherence sources of stimulated emission and do not suffer from speckle, but are rather broadband and have a relatively low output power limiting the scope of their potential applications. In this report, we demonstrate the use of random Raman lasing as a novel imaging light source with unprecedented brightness for a speckle-free and narrowband light source. We showcase the advantages of a random Raman laser to image the nanosecond scale dynamics of cavitation formation in water and quantitatively compare these images to those taken with incoherent fluorescent emission and coherent laser light as illumination source.

Intracavity frequency-doubled degenerate laser.

We develop a green light source with low spatial coherence via intracavity frequency doubling of a solid-state degenerate laser. The second-harmonic emission supports many more transverse modes than the fundamental emission, and exhibits lower spatial coherence. A strong suppression of speckle formation is demonstrated for both fundamental and second-harmonic beams. Using the green emission for fluorescence excitation, we show the coherent artifacts are removed from the full-field fluorescence images. The high power, low spatial coherence, and good directionality make the green degenerate laser an attractive illumination source for parallel imaging and projection display.

Principal modes in multimode fibers: exploring the crossover from weak to strong mode coupling.

We present experimental and numerical studies on principal modes in a multimode fiber with mode coupling. By applying external stress to the fiber and gradually adjusting the stress, we have realized a transition from weak to strong mode coupling, which corresponds to the transition from single scattering to multiple scattering in mode space. Our experiments show that principal modes have distinct spatial and spectral characteristic in the weak and strong mode coupling regimes. We also investigate the bandwidth of the principal modes, in particular, the dependence of the bandwidth on the delay time, and the effects of the mode-dependent loss. By analyzing the path-length distributions, we discover two distinct mechanisms that are responsible for the bandwidth of principal modes in weak and strong mode coupling regimes. Their interplay leads to a non-monotonic transition of the average principal mode bandwidth from weak to strong mode coupling. Taking into account the mode-dependent loss in the fiber, our numerical results are in qualitative agreement with our experimental observations. Our study paves the way for exploring potential applications of principal modes in communication, imaging and spectroscopy.