The problem

Recently, received data from the talented Paolo Papale, working in Pieter Roelfsema’s lab. Electrophysiological data from Utah array recordings with massive parallel recordings (about 1000 recording sites in visual cortices), digitized via Blackrock Neurotech board, output in proprietary .nev format.

Rather than request proprietary tools, better to process data independently.

The solution

Simple solution, using the brilliant Python NEO package. Loading data with BlackRock I/O interface becomes trivial:

python
import numpy as np
import os
from neo.rawio import BlackrockRawIO
import time

recording_path = './path_to_recording/path_to_subfolder'
now = time.time()

print('> Loading array: %s/instance1_B001.nev . . .' % (recording_path))
reader = BlackrockRawIO(filename='%s/instance1_B001.nev' % (recording_path))
reader.parse_header()

print('Getting analog signals . . .')
signals = reader.get_analogsignal_chunk(block_index=0, seg_index=0, i_start=None, i_stop=None, channel_indexes=None)
signals = signals[:, 0:128]  # Here, instance has 144 channels, but last 16 are not signals

print('Writing to disk . . .')
if os.path.isfile('./raw_array_1.bin'): os.remove('./raw_array_1.bin')  # Better safe than sorry
with open('./raw_array_1.bin', 'wb') as FileToWrite:
    signals.tofile(FileToWrite)

print('Total time: %s' % (time.time() - now))

Produces a .bin file, easily loaded for further processing, e.g., using Kilosort for spike sorting.
Simple, efficient. For a 500mb .nev file, conversion takes less than a minute.