pycbsdk 9.11.0

Python bindings for CereLink SDK (Blackrock Neurotech Cerebus devices)

pycbsdk

Python bindings for the CereLink SDK, providing real-time access to Blackrock Neurotech Cerebus neural signal processors.

Built on cffi (ABI mode) — no compiler needed at install time.

Installation

pip install pycbsdk

# With numpy support (zero-copy array access for continuous data)
pip install pycbsdk[numpy]

Windows prerequisite: The bundled shared library requires the Microsoft Visual C++ Redistributable. Most Windows systems already have it installed. If you get a DLL load error, download it from Microsoft.

Quick Start

from pycbsdk import Session
import time

with Session("HUB1") as session:
    # Register a callback for spike events
    @session.on_event("FRONTEND")
    def on_spike(header, data):
        print(f"Spike on channel {header.chid} at t={header.time}")

    # Register a callback for 30kHz continuous data
    @session.on_group(5, as_array=True)
    def on_continuous(header, samples):
        # samples is a numpy int16 array of shape (n_channels,)
        print(f"Group packet: {len(samples)} channels")

    time.sleep(10)
    print(session.stats)

Features

• Callback-driven: decorator-based registration for event, group, config, and catch-all packet callbacks
• Context manager: automatic cleanup on exit
• numpy integration (optional): zero-copy arrays for continuous data, ring buffer accumulator, blocking read_continuous() collector
• Device support: LEGACY_NSP, NSP, HUB1, HUB2, HUB3, NPLAY

API Overview

Session

session = Session(device_type="HUB1", callback_queue_depth=16384)

Callbacks (decorator style):

Decorator	Description
@session.on_event("FRONTEND")	Spike / event packets for a channel type
@session.on_group(5)	Continuous sample group (1-6)
@session.on_group(5, as_array=True)	Same, but data as numpy array
@session.on_config(pkt_type)	Config / system packets
@session.on_packet()	All packets (catch-all)
session.on_error(fn)	Error messages

Configuration access:

• session.get_channel_label(chan_id) — channel label string
• session.get_channel_smpgroup(chan_id) — channel's sample group (0-6)
• session.get_group_channels(group_id) — list of channel IDs in a group
• session.runlevel — current device run level
• Session.max_chans(), Session.num_fe_chans(), Session.num_analog_chans()

Commands:

• session.send_comment("marker text", rgba=0xFF0000) — inject a comment
• session.set_digital_output(chan_id, value) — set digital output
• session.set_runlevel(level) — change system run level
• session.set_sample_group(n, "FRONTEND", group_id) — configure sampling
• session.set_spike_sorting(n, "FRONTEND", sort_options) — configure spike sorting

CCF Configuration Files:

• session.save_ccf("config.ccf") — save current device config to XML file
• session.load_ccf("config.ccf") — load config from file and apply to device

Recording Control (requires Central):

• session.start_central_recording("filename", comment="session 1") — start recording
• session.stop_central_recording() — stop recording

Clock Synchronization:

# Convert device timestamp to Python's time.monotonic()
@session.on_event("FRONTEND")
def on_spike(header, data):
    t = session.device_to_monotonic(header.time)
    latency_ms = (time.monotonic() - t) * 1000
    print(f"Spike latency: {latency_ms:.1f} ms")

• session.device_to_monotonic(device_time_ns) — convert device timestamp to time.monotonic() seconds
• session.clock_offset_ns — raw clock offset (device_ns - steady_clock_ns), or None
• session.clock_uncertainty_ns — uncertainty (half-RTT), or None
• session.send_clock_probe() — send a sync probe

Statistics:

stats = session.stats  # Stats dataclass
print(stats.packets_received, stats.packets_dropped)
session.reset_stats()

numpy Integration

Requires pip install pycbsdk[numpy].

# Blocking data collection
data = session.read_continuous(group_id=5, duration=2.0)
# data.shape == (n_channels, ~60000), dtype int16

# Ring buffer for ongoing collection
reader = session.continuous_reader(group_id=5, buffer_seconds=10)
import time; time.sleep(5)
data = reader.read()        # most recent samples
data = reader.read(1000)    # last 1000 samples
print(reader.total_samples, reader.dropped)
reader.close()

Supported Devices

Device Type	Description
LEGACY_NSP	Legacy NSP (default)
NSP	NSP
HUB1	Gemini Hub 1
HUB2	Gemini Hub 2
HUB3	Gemini Hub 3
NPLAY	nPlay

Development

# Build the shared library
cmake -S . -B build -DCBSDK_BUILD_SHARED=ON -DCMAKE_BUILD_TYPE=Release
cmake --build build --target cbsdk_shared --config Release

# Install pycbsdk in development mode
cd pycbsdk
pip install -e ".[dev,numpy]"

# Point to the shared library
export CBSDK_LIB_PATH=/path/to/libcbsdk.dll  # or .so / .dylib

License

BSD 2-Clause. See LICENSE.txt.