Neural Recording Chips: A Guide to DAQ ASICs

NeuroNexus · neuronexus.com · Resources

Every NeuroNexus headstage and DAQ system is built around one recording chip, a DAQ chip. It performs the first step of acquisition, and it sets the channel count, the noise floor and whether the system can stimulate. This guide compares the families and says how to choose between them.

Research use only. NeuroNexus products and the procedures described here are for non-clinical research use, and must not be used in human or veterinary medical procedures.

---

What Is a DAQ Chip?#

A DAQ chip (neural recording ASIC) is the integrated circuit inside a headstage or recording system. It amplifies, bandpass-filters and digitizes many channels on silicon, before the data leaves the implant site.

Why on-chip processing matters:

• Lower noise: Neural signals are 50–500 µV. Amplifying at the source prevents pickup from cables and connectors
• Reduced cable mass: One SPI or LVDS link replaces dozens of analog wires. That matters most in freely moving work
• Power efficiency: Modern recording ASICs consume 0.1–0.5 mW per channel, enabling battery-powered or implantable systems

Position in the recording chain:

Neural tissue → Silicon Probe → Package → Headstage [DAQ chip] → DAQ system → Host computer

The chip's analog front-end amplifies each electrode's signal, the on-chip ADC digitizes it, and the digital output is streamed to the data acquisition system.

---

Chip Families#

Intan RHD Series — Record-Only#

The Intan RHD family is the workhorse of extracellular recording. The three variants share one analog front-end and differ only in channel count.

Shared specifications:

• Input-referred noise: 2.4 µV RMS
• ADC resolution: 16-bit
• Sampling rate: 12–30 kHz (programmable)
• Bandwidth: 0.02 Hz – 20 kHz (programmable high and low cutoffs)
• Input signal range: ±5 mV
• Digital interface: SPI
• Supply voltage: 3.2–3.6 V

Best for: Acute and chronic extracellular recording in rodents and large animals. The 32-channel RHD2132 appears in more NeuroNexus headstages than any other chip. The RHD2164 doubles the channel density per headstage.

---

Intan RHS2116 — Record + Stimulate#

The RHS2116 adds stimulation to the same 16-channel front-end. Each channel independently records or delivers a constant-current pulse, which is what makes closed-loop protocols possible on one electrode.

Key specifications:

• Recording channels: 16 (same analog front-end as RHD series)
• Stimulation channels: 16 (constant-current, per-channel)
• Stimulation range: 0.01–2500 µA (programmable in 0.01 µA steps)
• Pulse width: 10–1000 µs
• Frequency: 0.1–5000 Hz
• Compliance voltage: ±10.7 V
• Input noise: 2.4 µV RMS (recording)
• Power: 18 mW

Best for: Closed-loop stimulation experiments, intracortical microstimulation (ICMS), deep brain stimulation (DBS) research, and any protocol requiring simultaneous stimulation and recording on the same electrodes.

---

Neuropixels — Active Probes from imec#

Neuropixels probes, made by imec, put the recording circuit in the probe base and a switch under every site, so one shank has hundreds of selectable sites and the probe records a chosen subset of them at once. NeuroNexus does not make Neuropixels; XDAQ acquires them, the Commutator PRO carries them, and Radiens reads their data. Figures below are imec's published data sheets.

Neuropixels 1.0:

• Sites: 960 on one 10 mm shank (70 × 24 µm); 384 recorded at once
• ADC resolution: 10-bit
• Sampling: 30 kHz AP band (0.3–10 kHz) and 2.5 kHz LFP band (0.5–500 Hz), recorded simultaneously
• Input-referred noise: 5.9 µV rms AP band, 9.2 µV rms LFP band (typical)
• Power: about 15 mW

Neuropixels 2.0:

• Sites: 1,280 on one shank or 5,120 on four (250 µm apart); 384 recorded at once (the Quad Base version records 1,536 at once from 5,120)
• ADC resolution: 12-bit
• Sampling: 30 kHz, one full band (0.5 Hz–10 kHz)
• Input-referred noise: about 6.8 µV rms AP band, about 5 µV rms LFP band (typical)
• Power: about 22 mW (Quad Base about 88 mW)

Best for: large-scale single-unit recording along a long, thin shank, with the site subset chosen per session. Record only.

---

SiNAPS-1 — Active Pixels on the Shank#

SiNAPS-1 is the circuit inside the SiNAPS probes NeuroNexus carries (Corticale's active-pixel technology). An active pixel under every site amplifies and multiplexes on the shank, so every site is recorded all the time: 256 on one shank, 1,024 on four or eight.

Key specifications:

• Sites: 256 (one shank) or 1,024 (four or eight shanks), all recorded at once
• ADC resolution: 12-bit
• Sampling: 20 kS/s per pixel, fixed
• Input-referred noise: 6.5 µV rms (300–7,500 Hz)
• Multiplexing: 32:1 on the shank; no headstage
• Interface: the SiNAPS interface box to Smartbox

Best for: dense recording in which the whole shank should be in the data every session, and multi-shank layouts up to eight shanks. Record only. How the two active probes compare, point by point, is set out in Neuropixels, SiNAPS and Silicon Probes: How to Choose.

---

Comparison Table#

Neuropixels figures are from imec's data sheets at neuropixels.org; Intan and SiNAPS figures from the chip records behind the NeuroNexus datasheets.

---

Choosing the Right Chip#

Step 1 — How many channels do you need?

• 16–64 channels, standard extracellular: → Intan RHD series (scale from 16 to 64 per headstage)
• 256 or 1,024 channels, every site recorded at once: → SiNAPS-1
• 384 channels chosen from 960 to 5,120 sites on a 10 mm shank: → Neuropixels

Step 2 — Do you need stimulation?

• Yes, on the same electrodes as recording: → Intan RHS2116 (16 ch, constant-current)
• No: → any record-only chip

Step 3 — What ADC resolution and noise floor do you need?

• Highest resolution and lowest noise: → RHD series (16-bit, 2.4 µV rms)
• The active probes sit close together: SiNAPS-1 and Neuropixels 2.0 at 12-bit, Neuropixels 1.0 at 10-bit, with AP-band noise of 5.9 to about 7 µV rms across the three

Step 4 — What is your weight / cable budget?

• Freely-moving rodent: → RHD2216 or RHD2132 headstages (lightest, SPI interface)
• Head-fixed or tethered: any chip is suitable

---

DAQ Chips in NeuroNexus Products#

Each headstage is built around one DAQ chip, and that chip sets the headstage's channel count, noise floor and stimulation capability. The chip model appears on every headstage and DAQ datasheet page.

• Browse Headstages — filter by channel count, stimulation capability, and weight
• Browse DAQ Systems — filter by total channels and throughput

---

Further Reading#

• DAQ Chip — glossary definition
• Headstage — glossary definition
• Data Acquisition System — glossary definition
• Silicon Probes — glossary definition
• NeuroNexus Data Acquisition Systems Overview — system-level overview