No abstract

Over the past few years, 3M has been contracted to develop 14" Magneto-Optic (MM) disk media for advanced optical data storage systems. The high capacity and high data transfer rate potential for this media format appeals to users who require higher performance than the standard 5.25" format. The present article is a status report covering a brief history of 14" (356mm) M-O disk development, initial spaceflight feasibility testing and ongoing prototype development work. BACKGROUNDSeveral high performance optical recording systems have been proposed to take advantage of the recording performance advantages of the 14" (356mm) erasable formatThe increased recording surface area allows for storage capacities of greater than 5 Gbytes per side and the increased media surface velocity allows data transfer rates in the range of 20 Mbps per channel. The multi-channel Spaceflight Optical Disk Recorder (SODR), presently under development for NASA, is a two-disk system with target performance levels of 20 Gbytes capacity and 600 Mbps data rate These levels are achieved by using a 9 element laser diode array ,a 0.68 Numerical Aperture (NA) objective lens and a disk speed of 1000 RPM. The center diode is used for focus and tracking on a preformat pilot track of 13.3 l.Un track pitch and the other 8 diodes are used for read/write ( fig. 1).Previously contracted media development work for the Optical Disk Buffer application was targeted for a media construction consisting of a Winchester-sized aluminum substrate with a dust-defocussing coating. This media required a non-bifringement defocussing layer with thickness of 0.007" .0003" and vacuum deposition uniformity extended to the 14" disk size. The process-related hurdles of fabricating the defocussing layer and coating uniformity were addressed in earlier programs; several samples were provided to a hardware developer and used for technology demonstrations. At this stage, the provided media was characterized by a Carrier-to-Noise-Ratio (CNR) >50dB with 1 micron features and 1 dB uniformity across the disk radius.The cunent program (Advanced Erasable Optical Disk fAEOD)) has the goal of developing 14" media for the SODR space-based recording system mentioned above. This AEOD program is divided into 3 phases of media development: feasibility study, prototype development and manufacturing. The objective of the recently completed feasibility study was to compare two disk constructions for the spaceflight application. One disk construction was the previously developed aluminum construction; the other was a construction based on a 1.2mm chemically hardened glass substrate ( fig.2). The objectives of the prototype development phase are to improve media performance and to deliver prototype disks that will survive the spaceflight environments. The manufacturing phase is expected to involve scale-up considerations and certification testing. The following sections detail some results from the feasibility study and current efforts of the prototype development phase. AEOD FEASIBILITY STUDYT...

Analog computing paradigms are promising solutions to the growing computational demands of machine learning applications. Despite being susceptible to errors, analog and mixed-signal platforms have the potential to achieve higher speed and power efficiency for artificial neural network (ANN) applications than digital computers. Driven by the development of digital fixed-point ANN accelerators, low-precision ANN models have proven to be successful in compressing the size of ANNs and conforming the models to the data format of digital accelerators. While the inputs and weights of these digital, fixed-point ANN models can have low bit widths, the intermediate results (e.g., activations) must be preserved in high precision. As a result, these digital fixed-point models and training algorithms cannot be migrated easily to analog accelerators, because the analog intermediate results typically suffer from reduced precision due to noises and device imperfections. Here, we report on a training method for mixed-signal ANNs that considers two types of analog impairments, namely, random noise and distortion (deterministic in nature). The results show that mixed-signal ANN trained with our method can achieve the same classification accuracy as the digital fixed-point model with noise levels up to 50% of the ideal quantization step size. We demonstrate our training method on a mixed-signal, convolutional neural network based on diffractive optics.