Search for: All records

Abstract Solving linear systems, often accomplished by iterative algorithms, is a ubiquitous task in science and engineering. To accommodate the dynamic range and precision requirements, these iterative solvers are carried out on floating-point processing units, which are not efficient in handling large-scale matrix multiplications and inversions. Low-precision, fixed-point digital or analog processors consume only a fraction of the energy per operation than their floating-point counterparts, yet their current usages exclude iterative solvers due to the cumulative computational errors arising from fixed-point arithmetic. In this work, we show that for a simple iterative algorithm, such as Richardson iteration, using a fixed-point processor can provide the same convergence rate and achieve solutions beyond its native precision when combined with residual iteration. These results indicate that power-efficient computing platforms consisting of analog computing devices can be used to solve a broad range of problems without compromising the speed or precision. 

Photonic computing has potential advantages in speed and energy consumption yet is subject to inaccuracy due to the limited equivalent bitwidth of the analog signal. In this Letter, we demonstrate a configurable, fixed-point coherent photonic iterative solver for numerical eigenvalue problems using shifted inverse iteration. The photonic primitive can accommodate arbitrarily sized sparse matrix–vector multiplication and is deployed to solve eigenmodes in a photonic waveguide structure. The photonic iterative eigensolver does not accumulate errors from each iteration, providing a path toward implementing scientific computing applications on photonic primitives.

 

Light propagation in random media is a subject of interest to the optics community at large, with applications ranging from imaging to communication and sensing. However, real-time characterization of wavefront distortion in random media remains a major challenge. Compounding the difficulties, for many applications such as imaging (e.g., endoscopy) and focusing through random media, we only have single-ended access. In this work, we propose to represent wavefronts as superpositions of spatial modes. Within this framework, random media can be represented as a coupled multimode transmission channel. Once the distributed coherent transfer matrix of the channel is characterized, wavefront distortions along the path can be obtained. Fortunately, backreflections almost always accompany mode coupling and wavefront distortions. Therefore, we further propose to utilize backreflections to perform single-ended characterization of the coherent transfer matrix. We first develop the general framework for single-ended characterization of the coherent transfer matrix of coupled multimode transmission channels. Then, we apply this framework to the case of a two-mode channel, a single-mode fiber, which supports two randomly coupled polarization modes, to provide a proof-of-concept demonstration. Furthermore, as one of the main applications of coherent channel estimation, a polarization imaging system through single-mode fibers is implemented. We envision that the proposed method can be applied to both guided and free-space channels with a multitude of applications.

 

Total-variation regularization is applied at each iteration of an iterative framework for optical diffraction tomography. Numerical and experimental tests are performed using various highly scattering objects, and significant improvement in reconstruction SNR are demonstrated.

 

We propose a coherent multi-dimensional (wavelength, spatial mode, polarization, etc.) photonic tensor accelerator capable of matrix-vector, matrix-matrix, and batch matrix multiplications in a single clock cycle. A proof-of-concept 2x2 matrix-matrix multiplication at 25GBd with 4.67 bit precision was experimentally demonstrated. 

Mode-group multiplexing (MGM) can increase the capacity of short-reach few-mode optical fiber communication links while avoiding complex digital signal processing. In this paper, we present the design and experimental demonstration of a novel mode-group demultiplexer (MG DeMux) using Fabry-Perot (FP) thin-film filters (TFFs). The MG DeMux supports low-crosstalk mode-group demultiplexing, with degeneracies commensurate with those of graded-index (GRIN) multimode fibers. We experimentally demonstrate this functionality by using a commercial six-cavity TFF that was intended for 100 GHz channel spaced wavelength-division multiplexing (WDM) system.

 

Optical diffraction tomography (ODT) is a label-free and noninvasive technique for biological imaging. However, ODT is only applicable to weakly scattering objects. To extend ODT to the multiple-scattering regime, more advanced inversion algorithms have been developed, including optimization-based ODT (Opti-ODT) and iterative ODT (iODT). In this paper, we propose a combined strategy, namely, an iODT initialization for Opti-ODT, based on the observed complementarity of their individual advantages. This study numerically demonstrates that under this combined strategy, the reconstruction can accurately converge to a better local minimum, especially in the case of multiply scattering objects with large optical path differences.

 

Optical diffraction tomography (ODT) is used to reconstruct refractive-index distributions from multiple measurements in the object rotating configuration (ORC) or the illumination scanning configuration (ISC). Because of its fast data acquisition and stability, ISC-based ODT has been widely used for biological imaging. ODT typically fails to reconstruct multiply-scattering samples. The previously developed iterative ODT (iODT) was for the multiply-scattering objects in ORC, and could not be directly applied to ISC. To resolve this mismatch, we developed an ISC update and numerically demonstrated its accuracy. With the same prior knowledge, iODT-ISC outperforms conventional ODT in resolving the missing-angle problem.