Neuropixels is an application specific integrated circuit (ASIC) with two “parts”.  The shank, 70 um wide, 24 um thick, and 10 mm long contains 12 x 12 um TiN sites (recording electrodes) shown on the left drawing. The 10 mm shank has 960 sites, each with a switch to connect the site to an output line leading to the electronics in the base. Of these 960 sites, 384 can be addressed at any one time. The base has filter, a reference input multiplexer, and amplifier for each active channel. From the initial amplifier output, the signal is split into a low frequency (local field potential) and high frequency (action potential) path. Again, each active channel has a filter, and programmable amplifier for both the LFP and AP paths. The output of these 768 analog signals are fed to 32 multiplexers and 32 analog to digital converts, one A/D for each multiplexer. The digital output of the A/D’s is combined into 4 95 channel data stream and sent over the flex cable to the head stage. The probe chip is cemented to a rigid PC board with extra components. Mostly capacitors to isolate AC and DC as well as a eeprom chip that contains the probe identity. This eeprom is read by the software. The base integrated circuit must be protected from an light since light creates enormous noise for all Si electronics. The “shiny” rectangular surface visible on the left is a thin Si wafer cemented in place with opaque black epoxy. 

Neuropixels 2.0 is currently undergoing a pre-production engineering prototype ramp up. This process allows optimization of the fabrication, assembly, and testing processes. Neuropixels 2.0 4-shank probe shanks are 10 mm long, 70 um wide, and 24 um thick with a pitch between shanks of 250 um; the edge-to-edge span of the shanks is 820 um. Each shank has 1280 addressable sites in 26 groups of 48 sites. Geometry and dimension details can be seen in the brochure on page 1 of this tab. We expect these devices to be available for sale in early Q2 2022 in both single shank and 4 shank versions. 

Neuropixels UHD (ultra high density) is a boutique project to investigate the impact of small dense sites on spike sorting, cell type identification and recording efficiency for very small cells. Initially the fabrication is sponsored for a limit probe count by UCL-Carandini Lab, Allen Institute for Brain Science, and HHMI Janelia Research Campus.  Probes will be made in 2 geometries both with 5x5 um sites on 6x6 um centers.  The first, a non-programmable probe, has a site array of 8x48 sites at the distal end of a 10 mm shank. The very first of these probes is now being tested. The second geometry is 816 rows of 8 sites each, for a total site count of 6528, covering 4.9 mm of the 10 mm shanks, programmable in groups of 16. The first programmable geometries will deliver in Q3 2021. A follow-on grant to Nick Steinmetz, U. Washington, Shawn Olsen, AIBS, and Tim Harris, Johns Hopkins BME, funded by the NIH BRAIN Initiative, 1U01NS113252-01, will fabricate a second generation and test function. Commercial release is not planned.   

Neuropixels NHP is a project sponsored by HHMI to produce probes for their investigator labs. Two probe types will be produced, both in two geometries. These probes are variants of Neuropixels 1.0, with much longer larger shanks. The final probe, 384 channels with shanks of 25 mm and 45 mm length having  2496 or 4416 recording sites, respectively (there are 6.5 or 11.5 virtual arrays of 384 pixels in the shank). As for Neuropixels 1.0, only 384 sites may be addressed at a time. Delivery of this probes will occur in August 2020.  If tests prove these probes to have the expected performance, they will be made available for sale to the community by imec, probably beginning in Q1, 2021. HHMI’s role will be complete. Check back here for details of performance, fixtures, and availability.   

Neuropixels NXT is a new project (Starting Q3 2020), funded by the NIH BRAIN initiative (1U01NS115587-01), located at Johns Hopkins University, Department of Biomedical Engineering. Tim Harris is Principal Investigator with Dan O’Connor of the Department of Neuroscience as the Co-Investigator. As for prior Neuropixels projects, imec is our design and manufacturing partner. The proposal summary is available. Our aim is to reduce the base area of the probe so that the limit of recording capacity is the tissue damage imposed by inserted shank count. This includes eliminating the head stage. We anticipate probes with ~650 channels with base area <6 mm2. Arrays of shanks arranged as a “comb” or a bed of nails, with shanks 6-10 mm should allow 10,000+ channels in freely moving, tethered mice. Data rates exceeding 1 Gb/s will be generated, so at this capacity, a head stage like “data hub” will be needed. The first generation of these very small devices will not be available for test by the PI’s until Q3 2023, with commercial deployment likely in 2025. A variety of translational applications for chronic geometries and acute surgical devices (an 800 channel recording stylet) are enabled by the very small base. We will work with NIH BRAIN program staff and other research groups to optimize the final geometries.