GLANCE
Open FPGA data acquisition for closed-loop neuroscience.
256 channels at 30 kS/s
What it is
GLANCE — Gigabit Low-latency Acquisition for Neuroscience and Closed-loop Experiments — is an open-source electrophysiology system. It records from Intan RHD2000-style neural chips on a MicroZed (Xilinx Zynq-7020) with a custom carrier board: the programmable logic talks to up to two cables of amplifiers over a DDR SPI protocol, and the processor streams the data to your host over gigabit Ethernet.
It is built for experiments that have to react. Every 30 kHz sample leaves the board as its own UDP datagram — never batched, never held for a bigger packet — so nothing on the acquisition side stands between a spike and your detector. Per-stream sequence numbers make packet loss provable rather than suspected.
The firmware, the FPGA design, the carrier PCB and the Open Ephys plugin are all open, and all developed in the open.
On the board
- LFP extraction in hardware — a two-stage FIR cascade decimates the 30 kHz broadband to a ~3 kHz band inside the FPGA, arriving as a second stream on the same port.
- Headstage motion — a BNO055 IMU on each cable, stamped with the same master clock as the neural samples, so movement and spikes align with no host-side clock matching.
- Stimulus playback — arbitrary waveforms out of two on-board SMA connectors, software- or hardware-triggered, driven by the FPGA so playback never depends on the network.
- Reconfigurable fabrics — the board carries several FPGA configurations on its SD card and loads whichever one matches the headstages actually plugged in.
The repositories
Acknowledgements
We are grateful for Federal support, including ARL, NIMH (R01MH139216), and NINDS (R01NS147469). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.