File I/O Supported Formats Guide

A comprehensive guide to all file formats supported by gwexpy, including how to read and write each format. This page covers only the end-user API (.read() / .write() class methods) and does not expose internal implementation details.

Supported Formats Overview

Format	Extension	Read	Write	Recommended Class / Method	Dependencies	Notes
GWF	`.gwf`	Y	Y	`TimeSeriesDict.read()` / `.write()`	— (gwpy built-in)	gwpy standard format
HDF5	`.h5`, `.hdf5`	Y	Y	`TimeSeriesDict.read(format="hdf5")`	— (gwpy built-in)	gwpy standard
LIGO_LW XML (DTTXML)	`.xml`, `.xml.gz`	Y	N	`TimeSeriesDict.read(format="dttxml")` / `FrequencySeriesDict.read(format="dttxml")`	—	`products` argument required
CSV / TXT	`.csv`, `.txt`	Y	Y	`TimeSeriesDict.read(format="csv")`	— (gwpy built-in)	ASCII format. Supports directory loading
Pickle	`.pkl`	Y	Y	`TimeSeries.read(format="pickle")`	—	Python serialization
WAV	`.wav`	Y	Y	`TimeSeriesDict.read("file.wav")`	scipy	`t0` always 0 (no absolute time)
MiniSEED	`.mseed`	Y	Y	`TimeSeriesDict.read(format="miniseed")`	ObsPy	Seismic waveform format
SAC	`.sac`	Y	Y	`TimeSeriesDict.read(format="sac")`	ObsPy	Seismic waveform format
GSE2	`.gse2`	Y	Y	`TimeSeriesDict.read(format="gse2")`	ObsPy	Seismic waveform format
KNET	`.knet`	Y	N	`TimeSeriesDict.read(format="knet")`	ObsPy	K-NET strong motion records
GBD	`.gbd`	Y	N	`TimeSeriesDict.read("file.gbd", timezone=...)`	—	`timezone` required
WIN / WIN32	`.win`, `.cnt`	Y	N	`TimeSeriesDict.read(format="win")`	ObsPy	NIED WIN format (improved parser)
MTH5	`.h5`	Y	Y	`TimeSeries.read(format="ats.mth5")`	mth5	Magnetometer data (design-level support)
ATS	`.ats`	Y	N	`TimeSeriesDict.read("file.ats")`	—	Metronix binary parser
ROOT	`.root`	Y	Y	`EventTable.read(format="root")`	— (via gwpy)	CERN ROOT tables
SQLite / SDB	`.sdb`, `.sqlite`, `.db`	Y	N	`TimeSeriesDict.read("file.sdb")`	—	WeeWX / Davis weather data
NetCDF4	`.nc`	Y	Y	`TimeSeriesDict.read(format="netcdf4")`	xarray, netcdf4	Auto-detects time dimension
Zarr	`.zarr`	Y	Y	`TimeSeriesDict.read(format="zarr")`	zarr	Cloud-optimized chunked arrays
Audio (MP3, FLAC, etc.)	`.mp3`, `.flac`, `.ogg`, `.m4a`	Y	Y	`TimeSeriesDict.read("file.mp3")`	pydub (+ffmpeg)	`t0` always 0 (same as WAV)
NDS2	(network)	Y	N	`TimeSeries.fetch(...)`	nds2-client	Network data server
TDMS	`.tdms`	Y	N	`TimeSeriesDict.read("file.tdms")`	npTDMS	National Instruments
ORF	`.orf`	N	N	—	—	Not implemented (stub)

Note: Formats marked “gwpy built-in” (GWF, HDF5, CSV/TXT, Pickle) are handled through gwpy’s built-in IO pathways. Since gwexpy extends gwpy, these are available out of the box.

Format Details

GBD — GRAPHTEC Data Logger `.gbd`

Extension: .gbd Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/logger.gbd", timezone="Asia/Tokyo")

Required arguments:

timezone (str or tzinfo) — Logger’s local timezone (IANA name, e.g. "Asia/Tokyo" or UTC offset). Required. Omitting raises ValueError.

Key optional arguments:

channels (iterable[str], optional) — List of channel names to load. Defaults to all channels.
digital_channels (iterable[str], optional) — Channel names to treat as digital. Defaults to auto-detecting ALARM, ALARMOUT, PULSE*, LOGIC*.
unit (str or Unit, optional) — Physical unit override. Default 'V'.
epoch (float or datetime, optional) — Epoch (GPS seconds) override. Datetimes are converted to GPS.
pad (float, optional) — Padding value. Default NaN.

Dependencies: None (native implementation)

Notes:

Digital channels (ALARM, PULSE*, etc.) are binarized to 0/1.
If the HeaderSiz field is missing from the header, a ValueError is raised.
Scale factors are automatically extracted from the AMP section.

Implementation: gwexpy/timeseries/io/gbd.py

ATS — Metronix `.ats`

Extension: .ats Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.ats")

Reading via the mth5 library (for .atss files):

from gwexpy.timeseries import TimeSeries

ts = TimeSeries.read("path/to/data.atss", format="ats.mth5")

Required arguments: None

Key optional arguments: None (metadata is automatically extracted from the binary header)

Dependencies:

Standard reader: None (native binary parser)
ats.mth5 format: Requires the mth5 library. Raises ImportError if not installed.

Notes:

Supports ATS header versions 80/81. CEA/sliced headers (version 1080) are not supported (NotImplementedError).
LSB values (mV/count) are automatically converted to Volts (V).
Channel names are auto-generated from header info in the format Metronix_{system}_{serial}_{type}_{sensor}_{serial}.
The ats.mth5 format requires mth5’s filename conventions. Use the default binary parser if file names do not conform.

Implementation: gwexpy/timeseries/io/ats.py

SDB — WeeWX / Davis Weather Station `.sdb`

Extension: .sdb, .sqlite, .sqlite3 Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/weewx.sdb")

Required arguments: None

Key optional arguments:

table (str, optional) — Table name to read. Default 'archive'.
columns (list[str], optional) — Column names to read. Defaults to known weather columns (barometer, outTemp, windSpeed, etc.).

Dependencies: None (uses stdlib sqlite3 + pandas)

Notes:

Automatic conversion from Imperial to SI units (e.g. degF to degC, inHg to hPa, mph to m/s, inch to mm).
The table must have a dateTime column (UNIX timestamp).
Sample rate is auto-estimated from the median dateTime interval.

Implementation: gwexpy/timeseries/io/sdb.py

TDMS — National Instruments `.tdms`

Extension: .tdms Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.tdms")

Required arguments: None

Key optional arguments:

channels (list[str], optional) — Channels to read. Channel names use the format "GroupName/ChannelName". Defaults to all channels.
unit (str, optional) — Physical unit override.

Dependencies: npTDMS — Raises ImportError if not installed, with a message suggesting pip install nptdms.

Notes:

Channel names are stored in "GroupName/ChannelName" format.
wf_increment (sample interval) and wf_start_time (start time) are read from TDMS properties.
Start times as numpy.datetime64 or datetime are automatically converted to GPS time.

Implementation: gwexpy/timeseries/io/tdms.py

WAV — Audio File `.wav`

Extension: .wav Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/audio.wav")

Required arguments: None

Key optional arguments: None

Dependencies: scipy (scipy.io.wavfile) — included in gwexpy’s recommended dependencies.

Notes:

t0 is always set to 0.0 (GPS seconds). WAV files do not carry absolute timestamps.
Multi-channel files have channel names channel_0, channel_1, etc.
Mono files are automatically loaded as a single channel.
Writing uses the gwpy standard WAV writer pathway.

Implementation: gwexpy/timeseries/io/wav.py

MiniSEED — Seismic Waveform `.mseed`

Extension: .mseed Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

# Read
tsd = TimeSeriesDict.read("path/to/data.mseed", format="miniseed")

# Write
tsd.write("output.mseed", format="miniseed")

Required arguments: None

Key optional arguments:

channels (list[str], optional) — Channels to read. Specify by trace ID (NET.STA.LOC.CHA) or channel code.
unit (str, optional) — Physical unit override.
epoch (float or datetime, optional) — Epoch override.
timezone (str, optional) — Timezone specification.
pad (float, optional) — Gap fill value. Default NaN.
gap (str, optional) — Gap handling method. "pad" (default) or "raise".

Dependencies: ObsPy — Raises ImportError if not installed, with a message suggesting pip install obspy.

Notes:

Data is read via ObsPy’s read() function.
Gaps are padded with NaN by default. Use gap="raise" to raise an error instead.
Automatic trace merging (merge(method=1)) is applied.

Implementation: gwexpy/timeseries/io/seismic.py

SAC — Seismic Waveform `.sac`

Extension: .sac Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.sac", format="sac")
tsd.write("output.sac", format="sac")

Required arguments: None

Key optional arguments: Same as MiniSEED (channels, unit, epoch, timezone, pad, gap).

Dependencies: ObsPy

Notes:

SAC is typically one trace per file. Multi-trace writing behavior depends on ObsPy.

Implementation: gwexpy/timeseries/io/seismic.py

GSE2 — Seismic Waveform `.gse2`

Extension: .gse2 Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.gse2", format="gse2")
tsd.write("output.gse2", format="gse2")

Required arguments: None

Key optional arguments: Same as MiniSEED (channels, unit, epoch, timezone, pad, gap).

Dependencies: ObsPy

Implementation: gwexpy/timeseries/io/seismic.py

KNET — K-NET Strong Motion Records `.knet`

Extension: .knet Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.knet", format="knet")

Required arguments: None

Key optional arguments: Same as MiniSEED (channels, unit, epoch, timezone, pad, gap).

Dependencies: ObsPy

Notes:

Read-only. No writer is registered.

Implementation: gwexpy/timeseries/io/seismic.py

WIN / WIN32 — NIED WIN Format `.win` / `.cnt`

Extension: .win, .cnt Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.win", format="win")
# or
tsd = TimeSeriesDict.read("path/to/data.cnt", format="win32")

Required arguments: None

Key optional arguments:

century (str, optional) — Century part of the year. Default "20".

Dependencies: ObsPy — If ObsPy is not installed, the reader is not registered and an ImportError is raised.

Notes:

Uses gwexpy’s improved parser over ObsPy’s standard WIN reader, with the following fixes:
- 0.5-byte (4-bit) delta decode: fixed lower nibble sign handling; skip unused nibble on odd delta counts.
- 3-byte (24-bit) delta decode: fixed operator precedence and sign-preserving unpack/shift.
Gaps are merged with NaN.

Implementation: gwexpy/timeseries/io/win.py

DTTXML — Diag DTT XML (Time Series & Frequency Series)

Extension: .xml, .xml.gz Read/Write: Read Y / Write N Recommended API (time series):

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/dtt_output.xml", format="dttxml", products="TS")

Recommended API (frequency series):

from gwexpy.frequencyseries.collections import FrequencySeriesDict

fsd = FrequencySeriesDict.read("path/to/dtt_output.xml", format="dttxml", products="PSD")

from gwexpy.frequencyseries.matrix import FrequencySeriesMatrix

fsm = FrequencySeriesMatrix.read("path/to/dtt_output.xml", format="dttxml", products="TF")

Required arguments:

products (str) — Product type to extract. Required. Raises ValueError if omitted.
- Time series: "TS"
- Frequency series: "PSD", "ASD", "FFT"
- Matrix: "TF", "STF", "CSD", "COH"

Key optional arguments:

channels (iterable[str], optional) — List of channels to read.
unit (str, optional) — Physical unit override.
epoch (float or datetime, optional) — Epoch override.
timezone (str, optional) — Timezone specification.
native (bool, optional) — If True, uses gwexpy’s native XML parser. Recommended for complex TF data (subtype 6 phase loss fix). Default False. (FrequencySeriesDict / FrequencySeriesMatrix only)
rows, cols, pairs — Matrix filtering (FrequencySeriesMatrix only).

Dependencies: None (native implementation)

Notes:

Supports both time-series and frequency-domain data. The output type is determined by the products value.
The .xml extension is auto-identified (format="dttxml" can be omitted).

Implementation: gwexpy/timeseries/io/dttxml.py, gwexpy/frequencyseries/io/dttxml.py

GWF — Gravitational Wave Frame `.gwf`

Extension: .gwf Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.gwf", format="gwf")
tsd.write("output.gwf", format="gwf")

Required arguments: None (follows gwpy standard arguments)

Dependencies: — (gwpy standard. Internally uses python-ldas-tools-framecpp, etc.)

Notes:

Handled through gwpy’s standard IO pathway. No custom reader/writer in gwexpy.

Implementation: gwpy standard (gwpy/timeseries/io/gwf.py)

HDF5 — General Scientific Data `.h5` / `.hdf5`

Extension: .h5, .hdf5 Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.h5", format="hdf5")
tsd.write("output.h5", format="hdf5")

Required arguments: None

Dependencies: h5py (included in gwexpy’s required dependencies)

Notes:

gwexpy’s TimeSeriesDict.read() has extended HDF5 loading logic.
Supports automatic layout detection (LAYOUT_DATASET / LAYOUT_GROUP).
Supports keymap and ordering restoration.

Implementation: gwpy standard (gwpy/timeseries/io/hdf5.py) + gwexpy extension (gwexpy/timeseries/collections.py)

CSV / TXT — ASCII Text `.csv` / `.txt`

Extension: .csv, .txt Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

# Single file
tsd = TimeSeriesDict.read("path/to/data.csv", format="csv")

# Load all files from a directory
tsd = TimeSeriesDict.read("path/to/data_dir/")

Required arguments: None

Dependencies: None (gwpy standard)

Notes:

gwexpy supports loading all CSV/TXT files from a directory as a single TimeSeriesDict.

Implementation: gwpy standard (gwpy/timeseries/io/ascii.py) + gwexpy extension (gwexpy/timeseries/collections.py)

Pickle — Python Serialization `.pkl`

Extension: .pkl Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries import TimeSeries

ts = TimeSeries.read("path/to/data.pkl", format="pickle")
ts.write("output.pkl", format="pickle")

Required arguments: None

Dependencies: None

Notes:

Uses gwpy’s standard serialization pathway. Only use with files from trusted sources (standard pickle security considerations apply).

Implementation: gwpy standard

ROOT — CERN ROOT `.root`

Extension: .root Read/Write: Read Y / Write Y Recommended API:

from gwexpy.table import EventTable

table = EventTable.read("path/to/data.root", format="root")

Dependencies: — (via gwpy’s table/root pathway)

Notes:

gwexpy’s table/io/root.py is a re-export of gwpy’s module of the same name.
Primarily used for event table data.

Implementation: gwpy standard (gwpy/table/io/root.py) — via gwexpy: gwexpy/table/io/root.py

NDS2 — Network Data Server

Extension: None (network protocol) Read/Write: Read Y / Write N Recommended API:

from gwexpy.timeseries import TimeSeries

ts = TimeSeries.fetch("channel_name", start, end)

Dependencies: nds2-client (Python bindings)

Notes:

Network-based data retrieval, not file I/O.
Uses gwpy’s standard fetch() method.

Implementation: gwpy standard (gwpy/timeseries/io/nds2.py)

Audio — MP3 / FLAC / OGG / M4A (via pydub)

Extension: .mp3, .flac, .ogg, .m4a Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

# Read (auto-detected by extension)
tsd = TimeSeriesDict.read("path/to/audio.mp3")

# Explicit format
tsd = TimeSeriesDict.read("path/to/audio.dat", format="flac")

# Write
tsd.write("output.flac", format="flac")

Required arguments: None

Key optional arguments:

channels (iterable[str], optional) — Channel names to read (e.g. "channel_0").
unit (str, optional) — Physical unit override.

Dependencies: pydub — Raises ImportError if not installed, with a message suggesting pip install pydub. MP3/M4A encoding also requires ffmpeg (apt install ffmpeg or equivalent). FLAC may work without ffmpeg.

Notes:

t0 is always set to 0.0 (GPS seconds). Audio files carry no absolute timestamps (same behavior as WAV).
On read, sample values are normalized to the [-1.0, 1.0] range.
Multi-channel files use channel names channel_0, channel_1, etc.
On write, data is rescaled to peak-normalized 16-bit PCM.

Implementation: gwexpy/timeseries/io/audio.py

NetCDF4 — Scientific Data `.nc` (via xarray)

Extension: .nc Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.nc", format="netcdf4")
tsd.write("output.nc", format="netcdf4")

Required arguments: None

Key optional arguments:

channels (list[str], optional) — Variable names to read. Defaults to all variables with a time dimension.
unit (str, optional) — Physical unit override. Defaults to the file’s units attribute.
time_coord (str, optional) — Name of the time coordinate. Auto-detected if omitted (looks for "time").

Dependencies: xarray + netcdf4 — Raises ImportError if not installed, with a message suggesting pip install xarray netcdf4.

Notes:

Only variables with a time dimension (time) are converted to TimeSeries. Variables without a time dimension are skipped.
Time coordinates are automatically converted from datetime64 to GPS seconds.
Multi-dimensional variables (time + spatial, etc.) are flattened along non-time dimensions, resulting in varname_0, varname_1, etc.
On write, the datetime64 time coordinate is reconstructed from t0 and dt.

Implementation: gwexpy/timeseries/io/netcdf4_.py

Zarr — Cloud-Optimized Chunked Arrays `.zarr`

Extension: .zarr (directory store) Read/Write: Read Y / Write Y Recommended API:

from gwexpy.timeseries.collections import TimeSeriesDict

tsd = TimeSeriesDict.read("path/to/data.zarr", format="zarr")
tsd.write("output.zarr", format="zarr")

Required arguments: None

Key optional arguments:

channels (list[str], optional) — Array names to read. Defaults to all arrays in the root group.
unit (str, optional) — Physical unit override.

Dependencies: zarr — Raises ImportError if not installed, with a message suggesting pip install zarr.

Notes:

gwexpy Zarr convention: each array in the root group corresponds to one channel. Array attrs store sample_rate (Hz) and t0 (GPS seconds).
If sample_rate is not set, the inverse of dt is used; if neither is present, defaults to 1 Hz. t0 defaults to 0.0.
Supports all store types supported by the zarr library (directory stores, zip stores, etc.).
On write, sample_rate, t0, dt, and unit are saved as array attributes.

Implementation: gwexpy/timeseries/io/zarr_.py

Design-Level Support (No Dedicated Implementation in gwexpy)

The following formats are listed in the design specification (io_support.csv) but do not have dedicated reader/writer implementations in the gwexpy repository at this time.

Format	Extension	Designed Read/Write	Notes
MTH5	`.h5`	Read Y / Write Y	Via `mth5` library. Partially supported through `ats.mth5`

Unimplemented (Stub) Formats

The following formats are registered as placeholders (stubs) in the IO registry. Calling .read() raises an UnimplementedIOError or NotImplementedError. These are reserved for future implementation pending specification or sample files.

Time Series Stubs (`gwexpy/timeseries/io/stubs.py`)

Format Name	Registered Classes	Notes
`orf`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	ORF format
`mem`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	MEM format
`wvf`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	WVF format
`wdf`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	WDF format
`taffmat`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	TAFFMAT format
`lsf`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	LSF format
`li`	TimeSeries, TimeSeriesDict, TimeSeriesMatrix	LI format

Frequency Series Stubs (`gwexpy/frequencyseries/io/stubs.py`)

Format Name	Registered Classes	Notes
`win`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Implemented for time series but not frequency domain
`win32`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Same as above
`sdb`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Implemented for time series but not frequency domain
`orf`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`mem`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`wvf`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`wdf`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`taffmat`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`lsf`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented
`li`	FrequencySeries, FrequencySeriesDict, FrequencySeriesMatrix	Not implemented

Basic Usage of `.read()` / `.write()`

All gwexpy data classes use gwpy’s IO registry. Here is the basic usage pattern.

Reading

from gwexpy.timeseries.collections import TimeSeriesDict
from gwexpy.timeseries import TimeSeries

# Auto-detect format from extension
tsd = TimeSeriesDict.read("path/to/file.gbd", timezone="Asia/Tokyo")

# Explicitly specify format
tsd = TimeSeriesDict.read("path/to/file.dat", format="miniseed")

# Read a single channel
ts = TimeSeries.read("path/to/file.gbd", timezone="Asia/Tokyo")

Writing

# Write to a supported format
tsd.write("output.mseed", format="miniseed")
tsd.write("output.h5", format="hdf5")

Frequency Series

from gwexpy.frequencyseries.collections import FrequencySeriesDict
from gwexpy.frequencyseries.frequencyseries import FrequencySeries

fsd = FrequencySeriesDict.read("path/to/dtt.xml", format="dttxml", products="PSD")

How the `format` Argument Works

Omitted: The format is auto-detected from the file extension using identifier functions registered via io_registry.register_identifier(...).
Specified: The reader/writer registered for that format name is called directly.
When auto-detection fails (e.g. .xml is used by formats other than DTTXML): specify format explicitly.

Reference: Source Implementation Files

Implementation files referenced in the creation of this document.

Module Path	Summary
`gwexpy/timeseries/io/__init__.py`	Time series IO module registration entry point
`gwexpy/timeseries/io/gbd.py`	GBD reader. `timezone` is required
`gwexpy/timeseries/io/ats.py`	ATS reader (binary parser). `ats.mth5` variant available
`gwexpy/timeseries/io/sdb.py`	SDB reader (WeeWX SQLite). Registered under `sdb`, `sqlite`, `sqlite3`
`gwexpy/timeseries/io/tdms.py`	TDMS reader. Depends on `npTDMS`
`gwexpy/timeseries/io/wav.py`	WAV reader. `scipy.io.wavfile` wrapper. `t0=0` fixed
`gwexpy/timeseries/io/seismic.py`	MiniSEED / SAC / GSE2 / KNET reader/writer. Depends on ObsPy
`gwexpy/timeseries/io/win.py`	WIN/WIN32 reader. Requires ObsPy. Improved 4bit/24bit delta decode
`gwexpy/timeseries/io/dttxml.py`	DTTXML time series reader. `products` is required
`gwexpy/timeseries/io/audio.py`	Audio reader/writer (MP3/FLAC/OGG/M4A). Depends on pydub. `t0=0` fixed
`gwexpy/timeseries/io/netcdf4_.py`	NetCDF4 reader/writer. Depends on xarray. Auto-detects time dimension
`gwexpy/timeseries/io/zarr_.py`	Zarr reader/writer. Depends on zarr. gwexpy store convention
`gwexpy/timeseries/io/stubs.py`	Time series stubs (`orf`, `mem`, `wvf`, `wdf`, `taffmat`, `lsf`, `li`)
`gwexpy/timeseries/io/hdf5.py`	HDF5 IO (gwpy re-export)
`gwexpy/timeseries/io/ascii.py`	ASCII IO (gwpy re-export)
`gwexpy/timeseries/io/nds2.py`	NDS2 IO (gwpy re-export)
`gwexpy/timeseries/io/cache.py`	Cache IO (gwpy re-export)
`gwexpy/timeseries/io/losc.py`	GWOSC IO (gwpy re-export)
`gwexpy/timeseries/io/core.py`	Core IO (gwpy re-export)
`gwexpy/frequencyseries/io/__init__.py`	Frequency series IO module registration entry point
`gwexpy/frequencyseries/io/dttxml.py`	DTTXML frequency series reader. `products` required. `native` option available
`gwexpy/frequencyseries/io/stubs.py`	Frequency series stubs (`win`, `win32`, `sdb`, `orf`, `mem`, `wvf`, `wdf`, `taffmat`, `lsf`, `li`)
`gwexpy/frequencyseries/io/hdf5.py`	HDF5 IO (gwpy re-export)
`gwexpy/frequencyseries/io/ascii.py`	ASCII IO (gwpy re-export)
`gwexpy/frequencyseries/io/ligolw.py`	LIGO_LW IO (gwpy re-export)
`gwexpy/table/io/root.py`	ROOT IO (gwpy re-export)
`gwexpy/timeseries/collections.py`	`TimeSeriesDict.read()` extensions (HDF5 layout detection, directory loading)
`docs/developers/design/design_data/io_support.csv`	Design data: supported format list

File I/O Supported Formats Guide

Supported Formats Overview

Format Details

GBD — GRAPHTEC Data Logger .gbd

ATS — Metronix .ats

SDB — WeeWX / Davis Weather Station .sdb

TDMS — National Instruments .tdms

WAV — Audio File .wav

MiniSEED — Seismic Waveform .mseed

SAC — Seismic Waveform .sac

GSE2 — Seismic Waveform .gse2

KNET — K-NET Strong Motion Records .knet

WIN / WIN32 — NIED WIN Format .win / .cnt

DTTXML — Diag DTT XML (Time Series & Frequency Series)

GWF — Gravitational Wave Frame .gwf

HDF5 — General Scientific Data .h5 / .hdf5

CSV / TXT — ASCII Text .csv / .txt

Pickle — Python Serialization .pkl

ROOT — CERN ROOT .root

NDS2 — Network Data Server

Audio — MP3 / FLAC / OGG / M4A (via pydub)

NetCDF4 — Scientific Data .nc (via xarray)

Zarr — Cloud-Optimized Chunked Arrays .zarr

Design-Level Support (No Dedicated Implementation in gwexpy)

Unimplemented (Stub) Formats

Time Series Stubs (gwexpy/timeseries/io/stubs.py)

Frequency Series Stubs (gwexpy/frequencyseries/io/stubs.py)

Basic Usage of .read() / .write()

Reading

Writing

Frequency Series

How the format Argument Works

Reference: Source Implementation Files

GBD — GRAPHTEC Data Logger `.gbd`

ATS — Metronix `.ats`

SDB — WeeWX / Davis Weather Station `.sdb`

TDMS — National Instruments `.tdms`

WAV — Audio File `.wav`

MiniSEED — Seismic Waveform `.mseed`

SAC — Seismic Waveform `.sac`

GSE2 — Seismic Waveform `.gse2`

KNET — K-NET Strong Motion Records `.knet`

WIN / WIN32 — NIED WIN Format `.win` / `.cnt`

GWF — Gravitational Wave Frame `.gwf`

HDF5 — General Scientific Data `.h5` / `.hdf5`

CSV / TXT — ASCII Text `.csv` / `.txt`

Pickle — Python Serialization `.pkl`

ROOT — CERN ROOT `.root`

NetCDF4 — Scientific Data `.nc` (via xarray)

Zarr — Cloud-Optimized Chunked Arrays `.zarr`

Time Series Stubs (`gwexpy/timeseries/io/stubs.py`)

Frequency Series Stubs (`gwexpy/frequencyseries/io/stubs.py`)

Basic Usage of `.read()` / `.write()`

How the `format` Argument Works