Intan RHD2000 (.rhd) and RHS2000 (.rhs)

Intan recording controllers and their software write .rhd (RHD2000) and .rhs (RHS2000, with stimulation) files. OpenReadout returns the recorded signals as traces, together with the header settings: sample rate, filter bandwidths and channel names. Derived from Intan’s public application notes “RHD Data File Formats” and “RHS Data File Formats” and checked on public files (RHD 1.5 and 3.3, RHS 1.0 and 3.3, including stimulation), with Neo (BSD-3) as the reference reader. Provenance: docs/provenance/intan.md.

Crate: openreadout-intan. Three layouts are read: the traditional single file (a header followed by data blocks, IntanDataset) and the two split layouts that keep the same header alone in info.rhd / info.rhs next to raw .dat files, “one file per signal type” and “one file per channel” (SplitDataset, section below). Every stored signal kind is one trace with one sweep, in data-block order.

Detection

Bytes 0–3 as a little-endian u32: RHD_MAGIC (0xC6912702) or RHS_MAGIC (0xD69127AC) → definite (Family::Rhd / Family::Rhs). The .rhd/.rhs extension without the magic number → extension only.

Header (IntanHeader, parse_header, Cursor)

All values little-endian; strings are Qt QStrings: u32 byte length (0xFFFFFFFF = null), then UTF-16LE.

field	RHD	RHS	our name
magic, version (i16, i16)	✓	✓	family, version (→ format_version major.minor)
sample rate (f32)	✓	✓	sample_rate_hz
DSP enabled (i16), DSP cutoff (f32)	✓	✓	dsp_enabled, dsp_cutoff_hz
lower bandwidth, [lower settle bandwidth], upper bandwidth (f32)	✓	✓ (settle)	lower_bandwidth_hz, lower_settle_bandwidth_hz, upper_bandwidth_hz
the same three/four “desired” values	✓	✓	skipped
notch mode (i16: 0 off, 1 50 Hz, 2 60 Hz)	✓	✓	notch_mode
desired and actual impedance-test frequency (f32)	✓	✓	impedance_test_hz (actual)
amp settle mode, charge recovery mode (i16)		✓	amp_settle_mode, charge_recovery_mode
stim step, charge recovery limit, target voltage (f32)		✓	stim_step_a, charge_recovery_limit_a, charge_recovery_target_v
notes 1–3 (QString)	✓	✓	notes
temperature sensors (i16, v1.1+)	✓		temperature_sensors
DC amplifier data saved (i16)		✓	dc_saved
board mode (i16; RHD v1.3+)	✓	✓	board_mode
reference channel (QString; RHD v2.0+)	✓	✓	reference
number of signal groups (i16)	✓	✓

Each signal group: name, prefix (QString), enabled (i16), channel count (i16), amplifier count (i16); when enabled and non-empty, one record per channel (IntanChannel): native_name, custom_name (QString), native_order, custom_order, signal type (signal_code), enabled, chip_channel, [RHS: command_stream], board_stream, four Spike Scope values (skipped), impedance_ohm, impedance_phase_deg (f32); group records the group name. header_len is where the data begins.

Signal types (SignalKind::from_code): RHD 0 amplifier, 1 auxiliary, 2 supply, 3 board ADC, 4 digital in, 5 digital out; RHS 0 amplifier, 3 board ADC, 4 board DAC, 5 digital in, 6 digital out. Only enabled channels are stored, in header order (enabled).

Data blocks (block_layout, BlockPart)

block_samples() = N: 128 for RHS and for RHD version 2.0 or later, 60 for older RHD files (the notes tie N to the hardware; the version rule is Neo’s and holds for every corpus file). A block holds N time indices (i32; u32 before RHD 1.2, signed_timestamps()), then the parts below in this order; each BlockPart has kind, offset (within the block), channels and samples (per channel per block). Within a part, channel-major: all samples of channel 0, then channel 1, …

SignalKind	family	samples per block	stored	value (scaling)	unit
Amplifier	both	N	u16	raw × 0.195 − 6389.76 (= (raw − 32768) × 0.195)	µV
DcAmplifier	RHS, if dc_saved	N	u16	raw × 19.23 − 9845.76 (= (raw − 512) × 19.23)	mV
Stimulation	RHS	N	u16	stim_steps: ±(bits 0–7), bit 8 = negative, × stim_step_a	A
Auxiliary	RHD	N / 4	u16	raw × 37.4 × 10⁻⁶	V
Supply	RHD	1	u16	raw × 74.8 × 10⁻⁶	V
Temperature	RHD	1 per sensor	i16	raw × 0.01	°C
BoardAdc	both	N	u16	RHD mode 0: raw × 50.354 × 10⁻⁶; mode 1: (raw − 32768) × 152.59 × 10⁻⁶; mode 13 and RHS: raw × 312.5 × 10⁻⁶ − 10.24	V
BoardDac	RHS	N	u16	raw × 312.5 × 10⁻⁶ − 10.24	V
DigitalIn / DigitalOut	both	N	one u16 word	bit native_order of the word, 0 or 1, one channel per enabled line	–

A trace’s rate is sample rate × samples per block / N (auxiliary at a quarter, supply and temperature once per block). The stimulation part holds one word per amplifier channel; its amp-settle (bit 13), charge-recovery (bit 14) and compliance (bit 15) flags are not returned. The RHD digital-output word is inferred from Neo (the RHD note lists only digital inputs); no corpus RHD file has digital outputs. A zero stimulation magnitude with the sign bit set decodes to +0.0.

Trace extra: signal, samples_per_block, first_time_index; the first trace also carries family, board_mode, dsp_enabled, dsp_cutoff_hz, bandwidth_hz, lower_settle_bandwidth_hz, notch, impedance_test_hz, reference, notes, stim_step_a, charge_recovery_limit_a, charge_recovery_target_v, amp_settle_mode, charge_recovery_mode. Channel extra: native_name, custom_name, native_order, chip_channel, board_stream, command_stream, impedance_ohm, impedance_phase_deg, bit, decoding, group. Channel name is the custom name, else the native name. start_s = first time index / sample rate (time indices start above zero in split recordings and can be negative before a trigger).

Split layouts (SplitDataset, SplitLayout { PerSignalType, PerChannel }, name())

Derived from Intan’s public “RHD/RHS Data File Formats” notes and Neo’s IntanRawIO (BSD-3, read as documentation), checked against four corpus directories. The header file is the traditional header with no data blocks. open takes the directory or its info.rhd/info.rhs (is_split_header: a file named info.* with .dat files or time.dat next to it); a directory is claimed (session_files) only when it holds info.rhd/info.rhs and nothing but .dat/.rhd/.rhs files and companions (SESSION_COMPANIONS: settings.xml, notes; info_file finds the header). time.dat holds one time index per sample (i32; u32 before RHD 1.2) → extra.first_time_index, start_s.

Layout (SplitStream: kind, channels, files, samples; StreamFiles { Interleaved { path, words, len }, PerChannel }): when any per-type file exists the directory is one file per signal type, else one file per channel.

SignalKind	per signal type (signal_file)	per channel (channel_file: prefix + native name)	stored (split_scaling)
Amplifier	amplifier.dat	amp-A-000.dat	int16, × 0.195 µV (no offset)
DcAmplifier (RHS)	dcamplifier.dat	dc-A-000.dat	u16, as in the traditional file
Stimulation (RHS)	stim.dat	stim-A-000.dat	u16 stimulation word, as in the traditional file
Auxiliary (RHD)	auxiliary.dat	aux-A-AUX1.dat	u16, × 37.4 µV
Supply (RHD)	supply.dat	vdd-A-VDD1.dat	u16, × 74.8 µV
BoardAdc	analogin.dat	board-ANALOG-IN-1.dat	u16, per board mode
BoardDac (RHS)	analogout.dat	board-ANALOG-OUT-1.dat	u16
DigitalIn / DigitalOut	digitalin.dat / digitalout.dat: one u16 word per sample, bit native_order per line	board-DIGITAL-IN-01.dat: one u16 (0 or 1) per sample	–

Per-type files interleave channels sample by sample (channel 0 sample 0, channel 1 sample 0, …) for the enabled channels in header order. Per-channel directories list a channel only when its file exists (RHS software may omit amplifier or stimulation files); header channels without a file are reported (missing_file). Every stream holds one sample per time index — auxiliary and supply inputs are written at the full rate in these layouts (the corpus files hold as many auxiliary samples as time indices; Neo labels them with the traditional quarter/block rates) — so a trace’s rate is sample rate × samples ÷ time indices, which is the sample rate in every corpus directory. Temperature sensors are not saved in the split layouts. Trace extra adds layout, data_file (per-type) and channel extra.data_file (per-channel).

check finding codes

Split layouts: truncated (a .dat or time.dat length that is not a whole number of samples), length_mismatch (a stream with a different number of samples than time.dat) (errors); missing_file, no_time_file, data_after_header, time_index_gap (warnings).

Traditional files: truncated (a partial data block), bad_rate (errors); time_index_gap (time indices that do not advance by one sample), no_samples, unknown_board_mode (warnings). A header that cannot be parsed to its end, or that runs past the end of the file, fails to open as corrupt (exit 4).

Observed corpus values

file	family, version	N	blocks	signals
intan-rhd-test-1.rhd	RHD 1.5	60	500	192 amplifier, 6 supply, 8 board ADC (mode 0), 8 digital in
intan-time-split-121054.rhd	RHD 3.3	128	47	32 amplifier, 3 auxiliary, 3 digital in; first time index 138880
intan-test-tetrode-163225.rhd	RHD 3.3	128	353	4 amplifier; board mode 13; first time index 5400192
intan-rhs-test-1.rhs	RHS 1.0	128	500	32 amplifier + stimulation, 3 ADC, 2 DAC, 2 digital in
intan-rhs-stim-intantestfile.rhs	RHS 1.0	128	144	128 amplifier + stimulation (pulses), 16 digital out in one word
intan-rhs-fpc-multistim-240514-082243.rhs	RHS 3.3	128	726	8 amplifier + stimulation, 2 ADC, 2 DAC, 2 digital in, 2 digital out
intan-fps-rhd-231117/	RHD 3.3, one file per signal type	–	–	64 amplifier, 6 auxiliary (full rate), 1 ADC, 1 digital in; 24320 samples
intan-fps-rhs-240329/	RHS 3.3, one file per signal type	–	–	64 amplifier + DC + stimulation, 2 ADC, 1 DAC, 1 digital in, 1 digital out; 57088 samples
intan-fpc-rhd-multistim-240514/	RHD 3.3, one file per channel	–	–	8 amplifier, 6 auxiliary, 2 ADC, 2 digital in, 2 digital out; 98048 samples
intan-fpc-rhs-stim-250327/	RHS 3.3, one file per channel	–	–	no amplifier files; 5 DC, 4 stimulation, 1 digital in; 68352 samples

Vocabulary (every public identifier in openreadout-intan must appear here)

identifier	meaning
IntanReader, IntanDataset, FORMAT_ID, open, header, MAX_HEADER_LEN, RHD_MAGIC, RHS_MAGIC	entry points
SplitDataset, SplitLayout { PerSignalType, PerChannel }, layout, SplitStream, kind, files, samples, StreamFiles { Interleaved, PerChannel }, path, words, len, signal_file, channel_file, info_file, split_scaling, session_files, is_split_header, SESSION_COMPANIONS	split layouts
Family { Rhd, Rhs }	chip family
IntanHeader, family, version, sample_rate_hz, dsp_enabled, dsp_cutoff_hz, lower_bandwidth_hz, lower_settle_bandwidth_hz, upper_bandwidth_hz, notch_mode, impedance_test_hz, amp_settle_mode, charge_recovery_mode, stim_step_a, charge_recovery_limit_a, charge_recovery_target_v, notes, temperature_sensors, board_mode, dc_saved, reference, channels, header_len, block_samples, enabled, signed_timestamps, parse_header	header
Cursor, block, at	header reader
IntanChannel, native_name, custom_name, native_order, custom_order, signal_code, chip_channel, command_stream, board_stream, impedance_ohm, impedance_phase_deg, group	channel records
SignalKind { Amplifier, DcAmplifier, Stimulation, Auxiliary, Supply, Temperature, BoardAdc, BoardDac, DigitalIn, DigitalOut }, name, from_code	signal kinds
BlockPart, kind, offset, samples, block_layout, scaling, stim_steps	data blocks and scaling

Byte-source integration

The existing layouts are read through Input/Fs/SourceFile, including directory sessions and companions where applicable. Path, memory and callback namespaces use the same parser and scaling. See byte sources.

---

How this reader was derived, file by file: provenance log.