Sony (Germany)

Structurally anisotropic materials are ubiquitous in several application fields, yet their accurate optical characterization remains challenging due our incomplete understanding of how anisotropic light transport properties arise from the microscopic scattering coefficients. In fact, even when the dynamics of light transport is directly measured, coarse simplifications are often introduced due to a lack of established theoretical models or numerical methods. Here, we apply a general Monte Carlo implementation capable of handling direction-dependent scattering to the analysis of light transport in a sample of polytetrafluoroethylene (PTFE) tape. Using only a set of transient transmittance intensity profiles, the analysis retrieves the tensor components of the diffusive rates and the scattering coefficients along all three directions, in excellent agreement with Monte Carlo simulations.

The Computed Tomography Imaging Spectrometer (CTIS) is a snapshot hyperspectral camera based on computational imaging. Typical designs use a Keplerian beam expander to limit the imaged scene and decrease the incident ray angles. We have found that the use of a Galilean beam expander instead can be beneficial. Often a smaller system with a better optical quality is achievable with the disadvantage of a vignetted image. Results of a comparison between both designs based on a prototype will be shown. Furthermore, we present our miniaturized CTIS system, which uses the Galilean design.

Operating deep neural networks (DNNs) on devices with limited resources requires the reduction of their memory as well as computational footprint. Popular reduction methods are network quantization or pruning, which either reduce the word length of the network parameters or remove weights from the network if they are not needed. In this article we discuss a general framework for network reduction which we call `Look-Up Table Quantization` (LUT-Q). For each layer, we learn a value dictionary and an assignment matrix to represent the network weights. We propose a special solver which combines gradient descent and a one-step k-means update to learn both the value dictionaries and assignment matrices iteratively. This method is very flexible: by constraining the value dictionary, many different reduction problems such as non-uniform network quantization, training of multiplierless networks, network pruning or simultaneous quantization and pruning can be implemented without changing the solver. This flexibility of the LUT-Q method allows us to use the same method to train networks for different hardware capabilities.

Nowadays more and more TV terminal devices for digital terrestrial broadcast are also equipped with broadband network interfaces. In view of this fact, we developed a novel redundancy on demand (RoD) system to improve the reception of terrestrial broadcast signals with the help of additional redundancy data transmitted via broadband networks. We propose various forms of redundancy, which can be provided on request by the receiving terminal in case of reception errors. RoD does not require any changes of the current broadcast network structure and is compatible with conventional receivers. Simulation results show that a small portion of additional redundancy data provided by the broadband network leads to significant performance improvements of the reception, making high data rate broadcasting feasible in a realistic environment. RoD introduces a new degree of freedom for network planning helping to reduce transmission cost and network energy consumption.

Nowadays more and more TV terminal devices for digital terrestrial broadcast are also equipped with broadband network interfaces. In view of this fact we developed a novel Redundancy on Demand (RoD) system to improve the reception of terrestrial broadcast signals with the help of additional redundancy data transmitted via broadband networks. We propose various forms of redundancy, which can be provided on request by the receiving terminal in case of reception errors. RoD does not require any changes of the current broadcast network structure and is compatible with conventional receivers. Simulation results show that a small portion of additional redundancy data provided by the broadband network leads to significant performance improvements of the reception, making high data rate broadcasting feasible in a realistic environment. RoD introduces a new degree of freedom for network planning helping to reduce transmission cost and network energy consumption.

A front-end MIMIC (millimeter wave monolithic integrated circuit) with a 4:2 MUX/DEMUX (multiplexer / demultiplexer) functionality for 60 GHz transceiver with multiple antennas has been developed. The chip has been realized using 100 nm gatelength metamorphic InAlAs / InGaAs HEMT (high electron mobility transistor) technology on GaAs substrate together with CPW (coplanar waveguide) technology. The chip has been designed for functionality over the frequency range of 55-65 GHz. The novel topology for the MIMIC will be described. The single building blocks and details for the DPDT (double pole double throw) switch design will be explained, simulated and measured results will be compared. Finally, the overall performance for the integrated chip will be presented.