http://arxiv.org/abs/1905.12528
Energy consumption and hardware cost of signal digitization together with the management of the resulting data volume form serious issues for high-rate measurement systems with multiple sensors. Switching to binary sensing front-ends results in resource-efficient systems but is commonly associated with significant distortion due to the nonlinear signal acquisition. In particular, for applications that require to solve high-resolution processing tasks under extreme conditions, it is a widely held belief that low-complexity 1-bit analog-to-digital conversion leads to unacceptable performance degradation. In the Big Science context of radio astronomy, we propose a telescope architecture based on simplistic binary sampling, precise hardware-aware probabilistic modeling, and advanced statistical data processing. We sketch the main principles, system blocks and advantages of such a radio telescope system which we refer to as The Massive Binary Radio Lenses. The open engineering science questions which have to be answered before building a physical prototype are outlined. We set sail for the academic technology study by deriving an algorithm for interferometric imaging from binary radio array measurements. Without bias, the method aims at extracting the full discriminative information about the spatial power distribution embedded in a binary sensor data stream. We use radio measurements obtained with the LOFAR telescope to test the developed imaging technique and present visual and quantitative results. These assessments shed light on the fact that binary radio telescopes are suited for surveying the universe.
M. Stein
Thu, 30 May 19
56/57
Comments: N/A