ONR Forensic Memory System (FMS) Phase II

The Forensic Memory System (FMS) addresses a critical need for wideband sensor systems. Wideband digitizers produce copious data samples that need considerable interface bandwidth to transport and distribute real-time data to heterogeneous compute resources performing signal analysis. The initial signal analysis component, signal detection for example, will indicate a signal of interest (SOI) is present, but further actions are required to produce detailed signal analysis results. The need to perform subsequent processing necessitates the recovery of wideband data samples associated with the detected signal before, during and after its presence. The task of managing and distributing the high data volume of wideband systems is addressed by the focused FMS subsystem that time indexes and stores the incoming data samples and later recovers data segments corresponding to SOI activity. An embedded circular buffer approach enables a wideband detection system to cue on specific events and direct data to additional processing specific to the signals and events of interest. An efficiently managed FPGA based buffering subsystem can be tailored to the targeted signal parameters, scenarios, and environments while having reasonable cost, hardware, and power demands. The FMS system utilizes High Bandwidth Memory (HBM) to provide extremely high data rates, and the FPGA packaging allows the insertion into existing hardware servers.  FMS is unique in that it can offer data recovery as a service to multiple downstream consumers and it separates the detection and data recovery functions. Multiple detection functions can operate on a wideband or derived data stream, but they don’t need to duplicate the buffering and data management functions. In the past capabilities like FMS existed as an element of stove-piped systems where a rigid service is not exposed to third party DSP utilities.  In addition, FMS adds the flexibility of supporting a range of data rates, data structures, and dimensions.  For example, FMS could store raw samples or samples that have been produced by a Polyphase Channelizer.  FMS has a significant impact in reducing the downstream processing requirements in that only data samples corresponding to signal activity are processed.  In a sparse wideband signal environment there can be significant reductions in the required hardware capacity for the mission. FMS will be a service that can be incorporated into an existing sensor server-based systems by adding an FPGA card and using the services APIs provided. ???????