arXiv subjects

Samiran Ganguly

Publications and source records attributed to Samiran Ganguly.

23 records · Page 2Linked to original sources

arXiv2018

A Multiscale Materials-to-Systems Modeling of Polycrystalline Pb-Salt Photodetectors

We present a physics based multiscale materials-to-systems model for polycrystalline $PbSe$ photodetectors that connects fundamental material properties to circuit level performance metrics. From experimentally observed film structures and electrical characterization, we first develop a bandstructure model that explains carrier-type inversion and large carrier lifetimes in sensitized films. The unique bandstructure of the photosensitive film causes separation of generated carriers with holes migrating to the inverted $PbSe|PbI_2$ interface, while electrons are trapped in the bulk of the film inter-grain regions. These flows together forms the 2-current theory of photoconduction that quantitatively captures the $I-V$ relationship in these films. To capture the effect of pixel scaling and minority carrier blocking, we develop a model for the metallic contacts with the detector films based on the relative workfunction differences. We also develop detailed models for various physical parameters such as mobility, lifetime, quantum efficiency, noise etc. that connect the detector performance metrics such as responsivity $\mathcal{R}$ and specific detectivity $\mathcal{D^*}$ intimately with material properties and operating conditions. A compact Verilog-A based SPICE model is developed which can be directly combined with advanced digital ROIC cell designs to simulate and optimize high performance FPAs which form a critical component in the rapidly expanding market of self-driven automotive, IoTs, security, and embedded applications.

cond-mat.mes-hall↗

arXiv2018

Hardware based Spatio-Temporal Neural Processing Backend for Imaging Sensors: Towards a Smart Camera

In this work we show how we can build a technology platform for cognitive imaging sensors using recent advances in recurrent neural network architectures and training methods inspired from biology. We demonstrate learning and processing tasks specific to imaging sensors, including enhancement of sensitivity and signal-to-noise ratio (SNR) purely through neural filtering beyond the fundamental limits sensor materials, and inferencing and spatio-temporal pattern recognition capabilities of these networks with applications in object detection, motion tracking and prediction. We then show designs of unit hardware cells built using complementary metal-oxide semiconductor (CMOS) and emerging materials technologies for ultra-compact and energy-efficient embedded neural processors for smart cameras.

cs.CV↗

arXiv2018

From materials to systems: a multiscale analysis of nanomagnetic switching

With the increasing demand for low-power electronics, nanomagnetic devices have emerged as strong potential candidates to complement present day transistor technology. A variety of novel switching effects such as spin torque and giant spin Hall offer scalable ways to manipulate nano-sized magnets. However, the low intrinsic energy cost of switching spins is often compromised by the energy consumed in the overhead circuitry in creating the necessary switching fields. Scaling brings in added concerns such as the ability to distinguish states (readability) and to write information without spontaneous backflips (reliability). A viable device must ultimately navigate a complex multi-dimensional material and design space defined by volume, energy budget, speed and a target read-write-retention error. In this paper, we review the major challenges facing nanomagnetic devices and present a multi-scale computational framework to explore possible innovations at different levels (material, device, or circuit), along with a holistic understanding of their overall energy-delay-reliability tradeoff.

cond-mat.mes-hall↗

arXiv2017

Reservoir Computing using Stochastic p-Bits

We present a general hardware framework for building networks that directly implement Reservoir Computing, a popular software method for implementing and training Recurrent Neural Networks and are particularly suited for temporal inferencing and pattern recognition. We provide a specific example of a candidate hardware unit based on a combination of soft-magnets, spin-orbit materials and CMOS transistors that can implement these networks. Efficient non von-Neumann hardware implementation of reservoir computers can open up a pathway for integration of temporal Neural Networks in a wide variety of emerging systems such as Internet of Things (IoTs), industrial controls, bio- and photo-sensors, and self-driving automotives.

cs.ET↗

arXiv2017

Evaluating Spintronic Devices Using The Modular Approach

Over the past decade a large family of spintronic devices have been proposed as candidates for replacing CMOS for future digital logic circuits. Using the recently developed Modular Approach framework, we investigate and identify the physical bottlenecks and engineering challenges facing current spintronic devices. We then evaluate how systematic advancements in material properties and device design innovations impact the performance of spintronic devices, as a possible continuation of Moore's Law, even though some of these projections are speculative and may require technological breakthroughs. Lastly, we illustrate the use of the Modular Approach as an exploratory tool for probabilistic networks, using superparamagnetic magnets as building blocks for such networks. These building blocks leverage the inherent physics of stochastic spin-torque switching and could provide ultra-compact and efficient hardware for beyond-Boolean computational paradigms.

cond-mat.mes-hall↗