Falcon Server

Falcon Server provides real-time data processing and closed-loop control for neuroscience experiments, enabling low-latency detection, decoding, and feedback to neural systems.

Key Features:

• High-Performance Real-Time Processing: Processes streaming experimental data with low latency suitable for closed-loop systems.
• Multi-Threading Architecture: Implemented in C++ with each processing-graph node mapped to an individual thread and communication via thread-safe buffers.
• Customizable Processing Graphs: Loads and executes arbitrary processing graphs and supports implementation of new processing nodes and data types.
• Low Latency Performance: Achieves round-trip latency well below 1 ms on Neuralynx systems, below 15 ms on Open Ephys hardware, and contributes intrinsic latency under 0.5 ms.
• Scalability: Maintains consistent performance across tested workstations including 32-core and 4-core configurations.

Scientific Applications:

• Real-time population burst detection: Detects population bursts recorded from the hippocampus of freely moving rats with an average latency of approximately 40 ms.
• Population neural decoding/encoding: Supports computationally intensive algorithms for decoding and encoding activity from large cell assemblies in closed-loop experiments.

Methodology:

Implemented in C++ with a multi-threaded processing graph (nodes mapped to threads) using thread-safe buffers; streams data from Neuralynx and Open Ephys; generates closed-loop TTL pulses via a USB digital-output module; tested with processing graphs up to 32 parallel pipelines and eight serial stages on 32-core and 4-core workstations.