Formats

OpenReadout reads the 96 formats below. Each links to our notes on its layout and metadata and to the provenance log of its reader. Confidence says how thoroughly a reader has been checked against independent readers on public files (44 high, 38 medium, 14 low); known gaps are what it does not handle yet. To see the list for the version you have installed, run openreadout self formats.

Search formats

All 96 Light microscopy 16Electron microscopy 4Mass spectrometry 8Chromatography 9IR, Raman and UV-Vis 11NMR 4Flow cytometry 1Electrophysiology 12Plate readers and qPCR 6Bench biophysics 8Materials and electrochemistry 16Containers 1

96 formats

Format	Extensions	Field	Confidence	Known gaps
3DHISTECH MIRAX slide 3DHISTECH	.mrxs	Light microscopy	medium	6 gaps No specification: the layout comes from OpenSlide's public format documentation and the corpus files; placement of camera photos was derived by comparison with OpenSlide run as a black box Levels above 0 place camera photos at fractional pixels: values are bilinear resamplings as OpenSlide renders them (within a few grey levels of OpenSlide), not stored samples Slides without a camera position table (exported with overlaps removed) are listed but their pixels are refused (exit 6): no such development file yet Fluorescence slides with more than three filters, or filters sharing a colour component, are refused (exit 6) Level 0 is too large to read whole: read a region (--region) or a coarser level (--level) The scan-information records (focus, motor positions) and stitching records are listed as attachments, not decoded
Hamamatsu DCIMG Hamamatsu Photonics	.dcimg	Light microscopy	medium	5 gaps Derived without a specification from 15 corpus files (versions 7 and 0x1000000; ORCA-Flash4.0 and ORCA-Fusion; 16-bit); 8-bit frames are inferred Each file is one image: Bio-Formats groups sibling files named _NNN_NNN.dcimg into a Z stack, this reader does not (nothing in the file says the siblings belong together) Planes are returned in stored row order; Bio-Formats returns them bottom row first Only the first session of a file is read (no corpus file has more than one) No physical pixel size: the file records none
Imaris IMS (HDF5) Oxford Instruments / Bitplane (Imaris)	.ims	Light microscopy	high	4 gaps Imaris 3 (.ims TIFF-based files) is not this format; those files go to the TIFF reader Scene8 objects: their statistics and numeric record datasets are tables; surface meshes (SurfaceModel, BlockData), text fields and the legacy Scene group are not decoded Only the first data set is read when a file declares several (NumberOfDataSets > 1) Chunks compressed with filters other than deflate, shuffle, fletcher32 and LZ4 exit 6
Leica LIF Leica Microsystems	.lif.lifext.lof.xlif.xlef.xlcf.xllf	Light microscopy	high	7 gaps FLIM/TCSPC raw photon data (FALCON) is detected and described but not decoded; reading those planes exits 6 XLIF frames stored as single-page TIFF files are read (validated on two LAS X exports); JPEG and PNG frames follow the same rule unvalidated (a note says so); BMP and multi-page OME/Aivia TIFF frames are not read (exit 6) LOF, XLEF and XLCF are implemented from liffile's documentation and synthetic test files only; XLLF folder lists and XLIF files with TIFF frames are validated on real LAS X 3.7 exports (two depositors) Tile scans are stitched from stage positions without registration or blending; LAS X 'Merged' images may differ where it registered overlapping tiles Rotation (DimID 6) and XT/T-slice (DimID 7/8) addressing is validated on synthetic files only LIFEXT sidecar images (pyramid previews, histograms) are listed from the .lif but read by opening the .lifext itself Version-1 LIF containers (32-bit block lengths) are untested
Molecular Devices ImageXpress / MetaXpress plate (HTD + TIFF) Molecular Devices	.htd	Light microscopy	medium	4 gaps Z series (ZStep_ folders) and time series (TimePoint_ folders) follow the folder names; Z projections and stitched montages are not handled specially A plate folder without its HTD is read from the file names: missing planes cannot be detected there Excitation wavelengths and per-channel colours are not recorded in the files read; channel names come from the HTD (WaveName) or the plane files' illumination setting Cellomics/CellWorX variants of the HTD (.pnl) are not read
Nikon ND2 Nikon	.nd2	Light microscopy	high	6 gaps Lossy-compressed frames (eCompression 1) are not decoded: no public sample exists to derive the codec from Custom loops and unrecognized loop types are folded into T; NE-time sub-loops (pSubLoops) are not expanded Legacy (JPEG 2000) files: events and per-frame times are read, ROIs and custom-data columns do not exist in that generation ROI geometry is normalized from the documented tree layout but no corpus file contains ROIs; units are as stored RGB planes carry no excitation/emission wavelengths (the nd2 package assigns pseudo-wavelengths; we do not) RGB planes are returned R, G, B (modern files store B, G, R; validated on dye-coded planes); legacy JPEG 2000 RGB planes keep the codestream order their JP2 colour box declares as sRGB (inferred, no reference export)
Olympus FluoView OIB Evident (Olympus)	.oib	Light microscopy	medium	5 gaps Derived without a specification from 12 corpus data sets (FV1000/FV1200, FluoView 4.2): 16-bit grey planes, uncompressed or LZW; 8-bit and colour planes follow the TIFF reader but no corpus file has them Plane file names with axis letters other than C, Z, T and L become separate images (inferred, no corpus file) Line scans (XT) are exposed as planes whose height is the number of lines (Y is time); point scans and multi-area time lapse are not reinterpreted ROI (.roi) and LUT (.lut) files are listed, not decoded; channel colours are not reported Multi-page plane TIFFs are read from their first page only
Olympus FluoView OIF Evident (Olympus)	.oif	Light microscopy	low	5 gaps Derived without a specification from 12 corpus data sets (FV1000/FV1200, FluoView 4.2): 16-bit grey planes, uncompressed or LZW; 8-bit and colour planes follow the TIFF reader but no corpus file has them Plane file names with axis letters other than C, Z, T and L become separate images (inferred, no corpus file) Line scans (XT) are exposed as planes whose height is the number of lines (Y is time); point scans and multi-area time lapse are not reinterpreted ROI (.roi) and LUT (.lut) files are listed, not decoded; channel colours are not reported Multi-page plane TIFFs are read from their first page only
Olympus/Evident cellSens VSI Evident (Olympus)	.vsi.ets	Light microscopy	high	8 gaps No specification: the .vsi record tree and ETS layout are derived from 27 corpus files (11 with pixel data: cellSens Dimension 1.18/3.2/3.x, VS200 ASW, VS120 dotSlide 2.5, Stream Essentials 1.7; ETS versions 0x00030003/5/6); names, calibration, channels, exposure, objective, camera, times and dimension kinds come from records identified by comparison with Bio-Formats; everything else is only in vendor under numeric tags Dimension kinds other than Z (1), T (2) and channel (4) never occurred in the corpus; one would be exposed as T with a note and a check warning ETS sample types other than 8-bit (2) and 16-bit (4) unsigned, and compressions other than raw (0), JPEG (2) and JPEG 2000 (3), are not decoded (exit 6) Unstored whole-slide tiles are filled with the ETS background value (as Bio-Formats does); JPEG tiles are decoded with jpeg-decoder and may differ by a few levels from other decoders Where the tile grid is offset from the image corner (record 2410), coarser pyramid levels place it at the nearest pixel of offset/2^L; Bio-Formats truncates, so such levels can differ from it by one pixel Planes larger than 4 GiB (full-resolution whole slides) are read by region (--region) or at a pyramid level (--level), not whole Stacks without an ETS file (focus maps, sample masks, measurement/ROI layers) and blob_*.ets focus data are listed, not decoded; a stack directory holding several frame_t*.ets files is read from the first only The excitation wavelength (record 2474) is inferred from plausibility across 9 channels (Bio-Formats reports only emission)
Olympus/Evident OIR Evident (Olympus)	.oir	Light microscopy	high	6 gaps Compressed pixel blocks are not known to exist and are not handled; 4-byte samples are read as float32 on the strength of oirfile's documentation only (no corpus file) RGB (colour camera) acquisitions are exposed as three planar channels, not as interleaved RGB; no corpus file Line scans are exposed as XY planes whose height is the number of stored lines (no reinterpretation of Y as time); no corpus file POIR/MPOIR archives (zip collections of OIR files, mosaic layouts in matl.omp2info) are not read; open the .oir files inside Chunk names with axis letters other than t, l and z are listed by check but not exposed Excitation wavelength is reported only when exactly one enabled laser line is linked to the channel
OME-Zarr (OME-NGFF) open standard (OME-NGFF, Zarr) · also written by export	.zarr.ome.zarr.zip	Light microscopy	high	6 gaps Label images are exposed as images after the others (label ids as stored); their colours and properties are copied, not interpreted Codecs other than blosc (blosclz/lz4/snappy/zlib/zstd, byte and bit shuffle)/zstd/gzip/zlib/lz4/crc32c/sharding/transpose are not decoded (exit 4 on read, reported by check) String, structured and datetime data types are not read (exit 6); float16 is widened to float32, bool is uint8 0/1 Scale and translation transformations are composed (physical sizes, extra.origin_um); NGFF 0.6 coordinate systems and affine/rotation transformations are reported, not applied Axes other than t/c/z/y/x are read at index 0 Remote (HTTP/S3) stores are not opened: the binary has no network access
Revvity/PerkinElmer Harmony and Columbus exports (Opera Phenix, Operetta, Opera) Revvity (PerkinElmer)	.idx.xml.xml.flex	Light microscopy	high	5 gaps Harmony 4/5 Index.idx.xml with TIFF planes, Columbus ImageIndex.ColumbusIDX.xml with multi-page TIFF or .flex planes, and standalone Opera .flex files or measurement folders are read; the older Index.ref.xml layout with gzip-compressed planes (.tiff.gz) is not Standalone .flex: time series (Kinetic) and compressed pages are not validated on a public file; one well per file, pages as the Flex XML's Image@BufferNo say FLIM planes (several FlimID values) and sequence/fk/fl variants in file names are not told apart Flat-field profiles and skew-crop parameters are listed by info --view full, not applied The Z step is the PositionZ difference of the first field's first two planes
TIFF family (TIFF, BigTIFF, OME-TIFF, ImageJ, SVS, NDPI, LSM, QPTIFF, Leica SCN, Ventana BIF, Micro-Manager, MetaMorph STK/.nd, Thermo Fisher EER) open standard (TIFF 6.0 / OME); Leica Aperio, Hamamatsu, Zeiss, PerkinElmer, Molecular Devices MetaMorph, Thermo Fisher (Falcon EER) conventions · also written by export	.tif.tiff.ome.tif.ome.tiff.svs.ndpi.lsm.qptiff.btf.stk.nd.eer.scn.bif.gel	Light microscopy	high	12 gaps Old-style JPEG (6) is decoded only in its tables form (not lossless, not a single JPEGInterchangeFormat stream); LERC tiles exit 6 in default builds (the opt-in lerc feature decodes them with lerc-rs, which panics on some malformed input; SECURITY.md). JPEG 2000 tiles (Aperio 33003/33005, 34712) are decoded, irreversible (9/7) ones within 1 grey level of OpenJPEG (the reconstruction is not bit-exact) Packed 1-31-bit, half and 24-bit float and complex-integer samples are decoded only uncompressed or with LZW, deflate, PackBits or zstd (not inside JPEG, JPEG 2000, WebP, JPEG XL or LERC chunks) Ventana BIF: tiles are returned on their stored grid; the overlaps recorded in EncodeInfo/TileJointInfo are reported, not stitched (such files are partially decoded for --strict); volumetric BIF (ImageDepth > 1) exits 6 EER (Falcon electron-event movies): frames are decoded to counts on the sensor grid; the sub-pixel (super-resolution) positions of events are not used, the gain reference is not applied, and the TIFF Orientation is reported, not applied Planes larger than 4 GiB (whole-slide level 0) are read by region (--region), not whole; NDPI level 0 is read by its JPEG restart intervals (baseline JPEG only) NDPI files larger than 4 GiB (offset high bytes in tag 65324) and LSM files larger than 4 GiB (wrapped strip offsets) are not supported OME Modulo annotations (FLIM/lambda sub-dimensions along C/Z/T) are not expanded; the OME sizes are reported as stored Micro-Manager multi-file datasets without OME-XML and other TIFF dialects fall back to plain-TIFF behaviour (pages as Z) MetaMorph STK: only uncompressed stacks; .nd series: file names are built from the .nd keys (_w_s_t), stage positions become separate images, no time increment is derived from the member files Nikon NIS-Elements TIFF exports: files are grouped by the xy/c tokens of their names (validated on one public export; t/z tokens inferred); the LV and XML metadata blocks (objective, channel names, text information) are not decoded; the pixel size tag's meaning is inferred Palette (indexed-colour) pages return the stored indices; photometric WhiteIsZero pages return stored values (not inverted) JPEG: YCbCr stored as separate sample planes (PlanarConfiguration 2, e.g. Bio-Formats 8.x from planar input) and three-component streams in grey/palette/CMYK pages are refused (exit 6) rather than decoded to wrong colours; chroma upsampling differs from libjpeg-turbo by up to 3 counts
Yokogawa CellVoyager measurement (CV7000, CV8000, CQ1) Yokogawa	.mlf.mrf.wpi.mes	Light microscopy	medium	4 gaps Tiled fields (several PartialTileIndex values) are not stitched The objective's numerical aperture is not recorded in the files read Excitation is reported only for channels with a single laser; the emission band comes from the filter name (BP/) Correction files (shading, geometry, crosstalk) are listed, not applied
Zeiss AxioVision ZVI Carl Zeiss (AxioVision)	.zvi	Light microscopy	high	4 gaps Derived without a specification from 16 corpus files (12 depositors, 8 microscope stands, 2003-2025): 16-bit grey (pixel format 4) and 3 x 16-bit colour (format 8) are validated bit for bit; 8-bit grey and 3 x 8-bit colour are inferred (a note says so); other formats exit 6 Time series (tag 2821), multi-position and mosaic files are handled by inferred index tags that no public file exercises (a note says so); the acquisition-setup folder (Image/RootFolder) is listed, not interpreted Scale units other than micrometres (code 76) are not interpreted; uncalibrated axes have no physical size Layers, shapes (annotations) and the document summary property sets are listed, not decoded
Zeiss CZI Carl Zeiss Microscopy	.czi	Light microscopy	medium	8 gaps JPEG (id 1) subblocks decode with jpeg-decoder, which differs from libjpeg-turbo by up to a few grey levels on lossy streams (lossless JPEG is bit-exact); 12-bit DCT JPEG is not decoded JPEG-lossless (id 3), chunked (id 7) and camera/system raw (id >= 100) subblocks are not decoded (no public sample) Multi-file documents: part discovery ( ().czi) is inferred from synthetic fixtures; no public multi-file CZI was available Dimensions H, I, R, V, B that vary are one image per combination of coordinates; validated on H only (SIM phases; Airyscan-era multi-track files where some channels exist at H = 0 only, the others read as 0 and listed in extra.absent_channels); no public file varies along I, R, V or B Scenes above 4 GiB are read by region (--region), not whole; pyramid levels and regions are read with --level/--region, and export copies the pyramid Subblocks stored below their logical size without a pyramid flag (Airyscan fast-scan) are not upsampled Complex-valued pixel types are not decoded Super-resolved renderings (subblocks stored above their logical size, e.g. PALM) are exposed as their own image at the stored size, pixel size divided by the stored-to-logical ratio; they are not resampled onto the other channels' grid
EMD (Velox, HDF5) Thermo Fisher Scientific (Velox); HDF5-based	.emd	Electron microscopy	high	4 gaps Velox images, EDS spectra (traces), EDS spectrum images assembled from uint16 event streams and STEM-EELS spectrum images (Data/EelsSpectrumImage, validated on one dual-EELS file) are read; line profiles, EELS point spectra and the Data/SpectrumImage blob are listed, not decoded; other stream encodings are not read EDS spectrum images are summed over detector segments and frames (per-segment and per-frame cubes are not exposed); the first plane read decodes every event into memory (at most 400 million events) Berkeley EMD (0.2 and 1.0) data groups are read as images: X is each array's last dimension, Y the one before, leading dimensions are T; metadata groups are copied, not normalised beyond the microscope model and voltage Per-frame metadata of multi-frame Velox images is not exposed
FEI TIA / ES Vision series (SER + EMI) FEI / Thermo Fisher Scientific (TIA)	.ser.emi	Electron microscopy	medium	5 gaps Complex element data (types 9, 10) is described but not decoded Series whose elements differ in size are listed but not read Pixel-size units are not stored in the .ser; deltas below 1 mm are taken as metres Only the XML of the .emi is read (display layout and data copies are not) Line and area scans are exposed as T (elements in file order), not as scan axes
Gatan Digital Micrograph (DM3/DM4/DM5) Gatan	.dm3.dm4.dm5	Electron microscopy	high	5 gaps Old-style packed complex (DataType 5), RGB (8) and the other RGB/RGBA variants (15-22, 24-26) are described but not decoded (exit 6): no public file shows their byte layout; complex (3, 13), RGBA (23), binary (14) and 64-bit integers are decoded Packed Fourier transforms (DataType 27, 28) are returned as the stored half plane (X/2 + 1 columns), not mirrored into the full plane Thumbnails are exposed as raw-pixel attachments, not images Data with more than three dimensions (4D-STEM) is returned as frames of the first two dimensions, the scan positions flattened into T (extra.frame_grid gives the grid) Line plots overlaying several images (ImageSourceList) are returned image by image; display settings and annotations are listed by info --view full, not interpreted
MRC / CCP4 (MRC2014) CCP-EM open standard (cryo-EM, tomography, crystallography software)	.mrc.mrcs.map.ccp4.rec.st.ali.preali	Electron microscopy	high	4 gaps Complex modes 3 and 4 (Fourier transforms) are returned as stored (NX complex values per row), not expanded to the full transform MAPC/MAPR/MAPS are reported, not applied: planes are returned in file order FEI1/FEI2 extended-header fields are decoded from mrcfile's published layout; presence bitmasks are reported raw, not interpreted HDF5 extended headers (EXTTYP HDF5), .mrcz and bzip2-compressed files are not read; gzip-compressed files (.map.gz, .mrc.gz) are decompressed once at open
Agilent MassHunter (.d) Agilent Technologies	.d	Mass spectrometry	high	7 gaps Validated on MSScan layouts 5 (6400-series triple quadrupole 2006-2014: MRM, dynamic MRM, full, product-ion, neutral-loss and SIM scans, quadrupole profiles) and 6 (6200 TOF and 6500-series Q-TOF, acquisition software B.05 to B.10), and on the 6560 ion-mobility layout (drift-bin profiles, all-ions frames); other layouts are read through the file's own MSScan.xsd but are unverified Q-TOF profile spectra (MSProfile.bin, LZF) are checked against the scan index only (counts sum to the recorded TIC; the base-peak bin's m/z equals the recorded base-peak m/z): no export holds them; profiles are returned without the zero runs between peaks Precursor m/z is each MS/MS scan's own recorded value; ProteoWizard exports report a value averaged over the scans that share a precursor (differences up to 1e-5 relative) Ion-mobility data: drift-bin spectra only (the frame-summed spectra are not listed); all-ions high-energy spectra carry no precursor or collision energy (the frame method's energy ramp is shown by dump); compressed drift-bin profiles and CCS values are not supported Precursor-ion scans have not been seen; neutral-loss scans report the loss as extra.neutral_loss_mz, not as a precursor MSPeriodicActuals.bin, MSActualDefs.xml, the acquisition method's device settings and MSTree2.bin are listed but not decoded; ion-mode codes are reported as numbers DAD spectra (DAD1.sd/.sp) are listed but not decoded; DAD and LC-module signals (*.cd/*.cg) are exposed as traces
Bruker timsTOF (TDF/TSF .d) Bruker	.d.tdf.tsf	Mass spectrometry	high	4 gaps m/z and 1/K0 apply the file's MzCalibration (types 1, 2) and TimsCalibration (type 2) models, validated to <= 0.065 ppm and 5e-13 V*s/cm2 against vendor-library conversions of five files; other model types fall back to the acquisition-range approximation (tens of ppm off) and say so in notes and check prm-PASEF frames (MsMsType 10) are summed whole, not split per target window (no public prm-PASEF file to validate against) Spectra are raw sums of TOF bins over scans (no smoothing or centroiding); per-scan mobility arrays are not returned by spectra (library: TimsDataset::read_frame) TimsCompressionType 1 frames hold more scans than Frames.NumScans; all are read (NumPeaks and SummedIntensities count them; the reference conversion lists only the first NumScans)
imzML (mass spectrometry imaging) imzML open standard (converted from any vendor)	.imzml.ibd	Mass spectrometry	low	2 gaps Pixels are exposed as spectra with extra.position; no ion images are rendered The .ibd MD5 (IMS:1000090) is not verified; its SHA-1 (IMS:1000091) and UUID are
mzML (HUPO-PSI) HUPO-PSI open standard (converted from any vendor) · also written by export	.mzml.mzml.gz	Mass spectrometry	high	4 gaps Arrays other than m/z, intensity and chromatogram time/intensity (ion mobility, charge, noise, ...) are listed but not returned Byte-shuffled and dictionary zstd, and the truncation+prediction encodings (MS:1003089/1003090) are recognised but not decoded (exit 6) gzip-compressed files (.mzML.gz) are decompressed once at every open (restart points are kept in memory, not saved): info costs about two decompressions of the file; mzMLb is its own format id (mzmlb) Only the first scan of a spectrum and the last precursor are normalized; the rest stay in the element tree
mzMLb (mzML in HDF5) HUPO-PSI open standard (converted from any vendor)	.mzmlb	Mass spectrometry	medium	3 gaps The truncation + prediction encodings (MS:1003089/1003090) are recognised but not decoded (exit 6) HDF5 filters other than deflate, shuffle, Fletcher-32, LZ4 and Blosc (e.g. SZIP) are not decoded Everything the mzML reader does not do applies here too (see mzml)
mzXML (ISB) ISB/SPC open format (converted from any vendor)	.mzxml.mzxml.gz	Mass spectrometry	high	2 gaps peaks content types other than m/z-int pairs (e.g. m/z ruler) are not decoded (exit 6) No chromatograms: mzXML stores none
Sciex WIFF (.wiff + .wiff.scan) SCIEX	.wiff.scan.wiff2	Mass spectrometry	medium	7 gaps .wiff2 (SCIEX OS) files are encrypted and are refused (exit 6); convert them with the vendor's software Validated on MRM files (QTRAP 6500+ scheduled MRM, Analyst 1.7.2: 165 SRM chromatograms bit-exact; QTRAP 6500; a 2007 ten-sample Analyst file with one value per transition per cycle) and LC device traces, and on TripleTOF 6600 and 5600+ data-dependent files (Analyst TF; scan times, MS levels, polarity, precursors, TIC and base-peak m/z against the export); other scan types (Q1/Q3 scans, enhanced product ion, MRM³, SWATH variable windows) are listed but not decoded TOF spectra are the stored time-to-digital histograms (counts per bin, calibrated per scan); the vendor library's peak picking, which depositor exports hold, is not reproduced ZenoTOF (SCIEX OS) .wiff + .wiff.scan files open, but their precursor charges and rolling collision energies are not read (reported absent); not validated against an export TripleTOF 5600 files: the index TIC is not the sum of the stored counts (median +0.6 %, from -64 % on the most intense scans to +38 % on the weakest); every scan's largest count equals the index's base-peak intensity at the index's base-peak m/z, so the counts are returned as stored and total_ion_current is the index value The first half of every 2N-value MRM cycle (zero in the corpus files) is not interpreted; transitions are assigned to cycles by their scheduling windows, and a boundary cycle whose values are all zero is left out of the transitions that start or end there (inferred from one file) Multi-period methods: only period 0's experiments are described; multi-sample files expose one run per sample but only sample 1's acquisition time
Thermo RAW Thermo Fisher Scientific	.raw	Mass spectrometry	medium	9 gaps Validated file versions: 63, 64, 66; other versions open with a note and may misparse Detector controllers (UV/DAD channels, PDA field, analog pressure/temperature/A-D) are traces. An Accela PDA file (v63) is validated against an independent conversion (GNPS/MassIVE msconvert mzML of MTBLS773): all 10,501 PDA spectra × 401 wavelengths and their times exact, channel A equal to its UV 1 chromatogram (that conversion times it one 0.1 s sample later; our grid lines up with the PDA field). The Vanquish DAD/CAD (v66) and the analog channels are validated by internal consistency only (DAD channels equal the PDA field at their wavelength; stored per-scan values equal the samples). Units are inferred (mAU; no unit is stored) or taken from the method text; file versions before 63 expose only the mass spectrometer SRM scans (TSQ window records) are validated against the depositor's chromatograms, not against spectra; their per-peak flag bytes are not interpreted, and full-scan quadrupole (Q1MS/Q3MS) scans have no corpus file Scan filter text covers analyzer, polarity, data kind, ionization, in-source CID (sid=), turbo scan rate (t), dependent flag, multiplexing (msx), scan type, MS level, precursors with activation (CID, HCD, ETD) and energy, SPS MS3, and scan ranges (SRM product windows included); multi-range SIM scans list every range; rarer filter tokens (supplemental activation, FAIMS) are not composed Ion-trap (m/z-grid) profiles are validated against vendor-library conversions (LTQ Velos, Orbitrap ion-trap MS2/MS3 of ProteoWizard's test data); four-coefficient ion-cyclotron profiles are decoded from prior-art notes but have no corpus file to validate against The embedded instrument-method container is read for its stream list and the per-device method text; binary method data and the tune block are not decoded Centroid lists keep the peaks the file flags as reference/background ions, as conversions with the current vendor library do; conversions made before October 2020 (ProteoWizard up to 3.0.20239) left them out, so their peak counts are lower Orbitrap Exploris packets end in a per-peak annotation block (isotope-cluster marks, inferred) that is skipped; Orbitrap Astral and Ascend files have no corpus file Precursor charge is what the file records (Charge State); ion-trap (LTQ) and triple-quadrupole SRM scans record none, so none is reported
Agilent ChemStation data directory (.D: .ch, .uv, .ms) Agilent	.d.ch.uv.ms	Chromatography	high	7 gaps Signal-file versions other than 2 (.ms), 30, 81, 130, 131 (.uv), 179 and 181 are listed but not decoded Signal files are read whole; .ch files above 512 MiB are refused (exit 6) .reg register files, RUN.LOG and method (.M) files are listed with sizes, not decoded Vendor peak reports are tables[0] vendor_peaks: Result.xml (limits, baselines, named compounds and amounts), else Report.TXT (the printed report), and RESULTS.CSV (MSD ChemStation TIC integration); check finds every reported peak on our decoded signal, Result.xml areas recomputed between the vendor's limits (median ratio 0.9999 on 66 peaks) (2 + 4 GC runs, 144 peaks; 12 MSD runs, 46 peaks). Binary reports (REPORT.REG, .rpt) and calibration curves/levels are not read MS scans are reported as stored (thresholded m/z–intensity pairs at 0.05 m/z resolution), MS level 1, polarity unknown Agilent MassHunter (.d with AcqData/) and OpenLab CDS (.dx) are different formats: agilent-masshunter and openlab-cds read them .ch headers hold no point count: a version 81/179/181 body cut exactly at a value boundary cannot be told from a complete one (check reports the interval the header time range implies)
Agilent OpenLab CDS injection (.dx, with its .rx results) Agilent	.dx.rx	Chromatography	medium	5 gaps Detector signals (.CH, Signal179), instrument curves (.IT: pressure, flow, solvent ratios, temperatures) and DAD spectra parts (.UV, Spectra131: one trace, one channel per wavelength; bit-identical to rainbow-api, unit checked against the DAD's own channels) are traces; spectra directories (.UVD) and parts of other content types are listed, not decoded Vendor peaks (.rx) are a table with the vendor's retention times, areas, heights, area %, limits and baseline codes; compound amounts, calibration and custom fields (e.g. GPC results) are not read, and every corpus file has unnamed compounds Without the result set's .acaml sequence file, vendor peaks are matched to a signal by their baseline values (the vendor's baseline at a BB peak limit is the signal sample there); tables[0].extra.signal_mapping says how A result-set folder (.rslt) is not one data set: open its .dx files (openreadout batch 'set.rslt/*.dx') Validated on GC-FID (54 injections, OpenLab CDS 2.6 and a YADG GC injection), one LC-DAD/FLD result set whose detector signals were cut to 4 values by its publisher and two LC-DAD injections with spectra; no LC-MS .dx in the corpus
AIA/ANDI netCDF (chromatography and mass spectrometry) ASTM E1947 / E2077 open standard (exported by most chromatography data systems)	.cdf.nc	Chromatography	high	4 gaps netCDF CDF-5 (64-bit data) files are recognised but not read (ANDI files are CDF-1) Non-uniform chromatography sampling (raw_data_retention) is reported, but traces assume the regular actual_sampling_interval grid ANDI/MS: MS level is reported as 1 (the template has no MS level); library/SIM extensions (group variables) are listed, not mapped ANDI/MS: the total-ion chromatogram is a trace only when scan times are evenly spaced (within 1 %)
Cytiva ÄKTA UNICORN 6/7 result export (.zip) Cytiva	.zip	Chromatography	medium	5 gaps Validated on UNICORN 7.1, 7.3 and 7.7 exports (ÄKTA pure 25, pure 150L, avant); UNICORN 6 exports have no public sample Curves are read at full resolution; reduced-resolution point files, if an export has them, are not used Evaluated curves on a volume grid (IsoChroneType Volume) have volumes but no times Pool tables, method phases and the audit trail are kept in the vendor tree (info --view full), not as tables A UNICORN database backup (.bak) or a .result file from the UNICORN database is not an export and is not read
Cytiva ÄKTA UNICORN result (.res) Cytiva (GE Healthcare, Amersham Biosciences)	.res	Chromatography	low	4 gaps Validated on two files (ÄKTAprime 2009, Ettan LC 2017), both with the UNICORN 3.10 layout text; other UNICORN 3-5 systems (ÄKTA explorer, purifier, FPLC) are expected to share the layout but have no public sample Times are the sample index × the curve's interval from the method start; a sub-sample start offset some curves carry is read as milliseconds (unvalidated, < 1 sample) Binary blocks (CONFIG, CALIB, ColumnsChoosen, Template, fraction-collector configuration, the per-run block) are listed by info --view structure, not decoded The run start is derived from the header's end time minus the curves' duration (the logbook's own start text is local time without a zone)
Shimadzu LabSolutions data file (.lcd, .gcd) Shimadzu	.lcd.gcd	Chromatography	medium	5 gaps Read: chromatograms (older layout in mV; newer layout and .gcd in the unit of their Chromatogram Status record, e.g. µV), status traces (pump pressures, oven/cooler temperatures in the record's unit), the PDA field (stored µAU, shown in mAU) and LabSolutions' own peak tables (tables[0] vendor_peaks). Validated point for point against LabSolutions ASCII exports of an older-layout (LCsolution 1.25) and a newer-layout (5.109) file (25,038 trace points, 8 peaks in every column), bit-exact against chromConverter on 2 files, and against the vendor's own peak areas on 4 files (LC-UV, PDA, GC) Newer-layout LC chromatograms (LSS Raw Data/Chromatogram ChN) are named by their 2D Data Item title (DT 52 or 48) and scaled by their status record; the unit rule is the one validated on older-layout exports, a 5.109 export and a GC file's vendor peak areas, but no export exists of a file-version-5.01 (DT 48) chromatogram PDA channel chromatograms (a wavelength and bandwidth) are not extracted: their vendor peak tables are read and labelled with the channel's wavelength (channel 0 has none recorded); newer-layout peak records are read in their first 64 bytes only (no k', plates, tailing, resolution) Older-layout status traces are named by module model and quantity (LC-20AD pressure 1): the file stores no names; LabSolutions' export names (Pump A Pressure, Oven Temp.) are its own LC-MS (TLM Raw Data): MRM and SIM records are spectra (run 0), equal to msconvert's on 610 spectra of 3 CC0 files; full-scan and product-ion-scan spectra are refused (exit 6: the vendor's reader returns other values), their retention time, TIC and precursor listed; polarity unknown. MS quantitative results, compound and calibration tables and method streams are listed, not decoded
Thermo Scientific Chromeleon 7 archive (.cmbx) Thermo Fisher Scientific	.cmbx	Chromatography	medium	5 gaps Read: every archived 2D signal (detector channels, PDA-extracted channels, MS extracted-ion chromatograms, pump pressure and flow; encodings PtsLDiff, PtsLL2Df, PtsDDCmp) and every 3D field (PDA spectra: one trace, one channel per wavelength) of every injection, and the calibration standards a processing method holds, as traces with their unit, device and recorded range from the sequence file. Validated point for point against Chromeleon's own ASCII exports (6 HPAEC-PAD injections, 55,446 points, Chromeleon 7.3.1) and a Chromeleon 7.2.10 PDF report (57 GC-FID injections); every signal and 3D field of 10 archives equals the sequence file's recorded point count, times, minimum and maximum; Chromeleon's extracted channels equal our 3D fields bit for bit Chromeleon's own integration results, as it last saved them, are tables[0] vendor_peaks (retention time, limits, baseline, area, height, component, 50/10/5 % points; widths, asymmetry, tailing, plates and resolution computed from those points): equal to the PDF report's printed values, and every one of 5,360 stored peaks of 9 archives reproduced by our integration on our decoded signal within 1e-6. Injection details (position, volume, inject time, type, status, level, methods) are tables[1] injections and traces[].extra, equal to the vendor's exports and report. Amounts are not stored with the peaks and are not computed; injections Chromeleon saved no results for have none Audit trails, instrument methods (the script: stages, times, property settings such as gradients and oven programs) and processing-method settings are decompressed (LZMA2) and offered as attachments and in dump; report definitions, layouts and custom raw items are listed only Mass-spectrometry data (MSRawItem) is an embedded Thermo .raw file: listed with a hint to extract it and open it as thermo-raw; MS channels Chromeleon derives on demand have no stored points and are not traces. Unknown section tags, block layouts or result values never seen are refused (exit 6) Chromeleon 6 backups (.cmb) are a different format and are not read
Waters Empower ASCII raw-data export (.arw) Waters	.arw	Chromatography	low	3 gaps Read: 2D exports (a row of quoted field names, a row of their values, then time/value rows) as one trace with every exported field; equal to Appia's reading of the same files. Empower keeps its native raw data in its database: exports (this ASCII export and AIA/netCDF, read by andi-chrom) are how data leaves it Multi-column (3D PDA) exports are refused (exit 6): no public example with a licence; export PDA data as AIA/netCDF or as single-wavelength channels The export does not state the value's unit, and times are assumed to be minutes (Empower's export unit)
Waters MassLynx .raw directory Waters	.raw	Chromatography	high	5 gaps Spectra are validated point for point (m/z within 0.06 ppm, intensities equal) against vendor-library conversions of 13 acquisitions (SYNAPT G2/G2-Si/XS, Synapt MS, Xevo TQ-MS/G2-XS, ACQUITY SQD; 12-, 8- and 6-byte layouts, centroid and continuum, DDA, MSe, product-ion scans); the index flag 0x100 means values stored calibrated, otherwise the Cal Function n polynomial (T0, T1) is applied Lock-mass (lock-spray) correction is not applied: each reference scan's flagged lock-mass peak is reported (lock_mass_peak_mz); files MassLynx already corrected are read as stored Ion-mobility (HDMS/HDMSe/HDDDA/HD-MRM) and SONAR functions: the drift-resolved bins (_funcNNN.cdt) are not decoded; their stored drift-summed spectra are returned and say so MRM intensities (4-byte layout) are scaled so each scan sums to the TIC stored in _FUNCnnn.IDX (rainbow reports twice these values) and exposed as tables/traces, not spectra Photodiode-array functions are a trace of stored absorbance counts (unscaled); binary .EE/.CMP files and the third word of 12-byte values are listed, not decoded
Agilent Cary UV-Vis (.dsw, .bsw, .bsk) Agilent (Varian)	.dsw.bsw.bsk.csw	IR, Raman and UV-Vis	medium	4 gaps Validated against the Cary software's CSV exports on Cary 4000 Scan 3.00 files (absorbance over wavelength); Cary 60 files of Scan and Scanning Kinetics 5.0 are read the same way but have no export to check them against Only the spectrum and baseline stores are decoded; graph, report, database and baseline-info stores are listed by info --view structure Y modes other than absorbance and X modes other than nanometers have no corpus file: their spectra are returned with a finding Collection times are a local clock (no time zone is stored), read month/day/year as every corpus file writes them
Agilent FT-IR imaging (Resolutions Pro) Agilent	.dat.seq.dms.dmd.drd.bsp.dmt	IR, Raman and UV-Vis	medium	4 gaps A mosaic is read from its whole-image .dms file (or one tile at a time); tiles are not assembled when the .dms is missing Rows run top to bottom: the files store them bottom to top (inferred from how a mosaic's tiles are stacked in its .dms) The pixel size is the detector pixel size times the aggregation, read from the header; stage positions are not in the files Settings are read from the header's text settings (English Resolutions Pro); the selected-pixel spectrum stored in the header is not returned
Bruker OPUS Bruker	.0.1.2	IR, Raman and UV-Vis	high	4 gaps Any numbered extension (.0, .1, … .999) is an OPUS file when it starts with 0A 0A FE FE; the extensions listed are the common ones 3D series blocks (rapid scan, time-resolved) follow brukeropus's layout; no public series file was available to validate them Report blocks (quant and search results), embedded bitmaps and blocks of unknown type are listed by info --view structure, not decoded Data blocks without a data-status block are listed, not decoded
Galactic / Thermo GRAMS SPC (.spc) Thermo Fisher Scientific (Galactic Industries) GRAMS	.spc	IR, Raman and UV-Vis	high	4 gaps Big-endian (0x4C) files, per-subfile x arrays (TXYXYS, mass spectra) and old-format (0x4D) multifiles are refused 16-bit fixed-point values follow the documented rule but no development file has them Multifile maps are returned as spectrum series with a subfile table (z, W); the map grid is not rebuilt Other programs' .spc files (Bruker EPR, Becker & Hickl, EDAX, Shimadzu UV) are not Galactic SPC and are not read by this reader
JASCO Spectra Manager (.jws) JASCO	.jws.jrs	IR, Raman and UV-Vis	high	4 gaps Validated on FT/IR-4600/4700, NRS-5100 Raman, V-630/V-730 UV-Vis, J-1500 circular dichroism (1-3 channels) and FP-8300 fluorescence files; interval/kinetics files (.jwb) and channel or axis codes not seen are refused or returned unnamed with a finding Time-course files carry their x axis as time without a unit (the unit is not in a field that has been validated) Acquisition parameters are named for FT-IR (module 9) only; the other instruments' parameter records are kept raw in the vendor tree Text fields of flat files are decoded as UTF-8 or Latin-1; Japanese (Shift-JIS) titles may be garbled
JCAMP-DX IUPAC open standard (NMR, IR, UV-Vis, MS spectra) · also written by export	.jdx.dx.jcamp.jcm	IR, Raman and UV-Vis	high	3 gaps ##PEAK ASSIGNMENTS= tables and JCAMP-CS structure blocks are listed, not decoded ASDF compression inside (XY..XY) peak tables is not decoded (AFFN only) ##RADATA= (interferograms) is listed, not decoded
PerkinElmer Spectrum (.sp) PerkinElmer	.sp	IR, Raman and UV-Vis	medium	3 gaps One spectrum per file (2D constant-interval data sets); Spotlight image files (.fsm) are read by the perkinelmer-fsm reader Instrument settings are named from their values in the corpus (scans, resolution, detector, source, beamsplitter, apodization, laser wavenumber); other settings are in the vendor tree by member id UV-Vis (.sp from Lambda instruments, x in nm) follows the same layout by inference; no public UV-Vis .sp file was available
PerkinElmer Spotlight image (.fsm) PerkinElmer	.fsm	IR, Raman and UV-Vis	low	3 gaps Image rows are in acquisition order (stage y increasing by one step per row); which way up Spectrum IMAGE draws them is not in the file Only the 4-D constant-interval layout (one float32 spectrum per pixel block) has been seen Instrument settings are named as in the .sp reader (the same member ids); other settings are in the vendor tree
Renishaw WiRE Renishaw	.wdf	IR, Raman and UV-Vis	high	5 gaps x-list and y-unit codes are named only where the corpus confirms them (Raman shift, counts); other codes are reported as numbers with an x axis Property sets are decoded to a tree; keys WiRE does not name in the file appear as # and only those whose meaning the corpus shows are normalized (objective, laser name, serial numbers) Map pixels are placed from the stage coordinates on the WMAP grid; maps without coordinates fall back to storage order Line scans and depth or time series are traces with a positions table, not images Pre-WiRE-3 files, processed data blocks (baseline, cosmic-ray removal results) and embedded Z-stacks are not decoded
Thermo Fisher OMNIC Thermo Fisher Scientific (Nicolet)	.spa.spg.srs	IR, Raman and UV-Vis	medium	4 gaps Series files (.srs: rapid scan, high-speed, GC-IR, TGA-IR) are read from their key table: the series spectra, their times and the backgrounds; the per-series profiles (Gram-Schmidt, chemigrams) and processing records are listed, not decoded; no licensed .srs file was available, so the series layout rests on held test files OMNIC map files (.map) are not recognised Resolution and the instrument serial number come from the processing-history text (English or French OMNIC); files without that text have neither y-unit codes other than absorbance, transmittance, reflectance, log(1/R), single beam, Kubelka-Munk, volts, photoacoustic and Raman intensity are reported as intensity
WITec Project / WITec Data WITec (Oxford Instruments)	.wip.wid	IR, Raman and UV-Vis	high	4 gaps Format versions 0-7 (WITec Control 1.6 to WITec Suite SIX); spectral calibrations of the grating and polynomial kinds are validated, linear and lookup-table x calibrations follow the published format description without a development file Filter, cursor, colour-profile and mask-definition records (analysis settings) are listed, not decoded; the images they produced are read Acquisition settings come from each measurement's information text (English WITec software); the version-7 Suite SIX parameter records are kept in the vendor tree only The acquisition time is local time (the file holds no time zone)
Bruker TopSpin NMR Bruker		NMR	high	5 gaps Input is an experiment directory (or its acqus/fid/ser, or a pdata/ directory), not a single file Processed data: 1D, 2D and 3D components are read; 4D+ processing is not decoded Time-domain values are scaled by 2^NC by analogy with nmrglue's NC_proc rule (inferred); the digital-filter group delay is reported, and removed only when the FID is processed (analyze nmr-peaks, trace --process; 1-D rows only) Non-uniformly sampled ser files are returned as acquired rows; nuslist indices are available as a table; NUS reconstruction is out of scope DTYPA/DTYPP values other than 0 (int32) and 2 (float64) are not decoded
JEOL Delta NMR JEOL	.jdf	NMR	high	5 gaps 1D and 2D (32-point submatrix) data with real/complex axes are decoded; 3D+ and small-submatrix layouts, TPPI and envelope axes are unsupported (exit 6) Values are returned as stored; nmrglue returns the complex conjugate String parameters hold at most 16 characters (longer values are cut in the file) Acquisition start (ACTUAL_START_TIME, seconds since 1990-01-01 UTC) and header dates are inferred from corpus files Context, annotation and history sections are not decoded
Magritek Spinsolve benchtop NMR Magritek	.1d.2d	NMR	medium	5 gaps Input is an experiment directory (acqu.par and data.1d/data.2d), or any file in it; a folder of experiments lists them Prospa data types 501 (complex), 503 (x + real) and 504 (x + complex) are decoded; other type codes and multi-row files with an x block are listed, not decoded Plot files (.pt1/.pt2), MestReNova documents and nmr_fid.dx are listed, not decoded (open nmr_fid.dx with the JCAMP-DX reader) The x block's unit is not stored: it is named (ms or ppm) only when its spacing matches the dwell time or the spectral width Sub-experiment folders (e.g. 1D_T1IRT2/0000/) are opened one by one
Varian/Agilent VnmrJ NMR Agilent (Varian)		NMR	high	5 gaps Input is a data directory (.fid/ with fid and procpar), or its fid/procpar/text/log Sweeps are the fid's traces in disk order: arrayed and multidimensional data are not reordered by phase or array parameters (nmrglue's as_2d order) Block scale factors and drift corrections are reported, not applied (as nmrglue) Processed data (datdir/phasefile, datdir/data) are listed, not decoded: no public file or permissive prior art int16 samples (dp='n') are decoded from the header flags but covered by synthetic tests only
FCS (Flow Cytometry Standard) ISAC open standard (all cytometer vendors)	.fcs.lmd	Flow cytometry	high	6 gaps Histogram data sets ($MODE C/U, deprecated in FCS 3.1) are described but not decoded Integer fields that are not a whole number of bytes (bit-packed $PnB) are not decoded $BYTEORD permutations other than 1,2,3,4 and 4,3,2,1 are not decoded read_table/export return raw values; table --compensate/--transform and analyze gate apply scale conversion, spillover, transforms and gates FlowJo workspaces: transforms other than linear, log, logicle, biex and ArcSinh, derived parameters and non-OLS spectral unmixing are reported as unsupported FCS 1.0 and FCS 4.0 files are detected but not read
Axon Binary Format (ABF) Molecular Devices (Axon Instruments) pCLAMP	.abf	Electrophysiology	high	4 gaps Command (DAC) waveforms are synthesized (trace 1) for ABF 2 episodic files with step, ramp and pulse-train epochs only; stimulus files, user lists, alternating outputs, conditioning trains, triangle/cosine/biphasic epochs and ABF 1 files keep the epoch table only User lists, statistics, math, scope and voice-tag sections are listed but not decoded Pre-1.6 ABF 1 files: telegraph, protocol path and 20-entry epoch tables are not in the header and are not reported Old pCLAMP/Axotape files without an ABF signature (CLPX, FTCX) are not read
Axon Text File (ATF) Molecular Devices (Axon Instruments) pCLAMP	.atf	Electrophysiology	medium	2 gaps Only time-series ATF (first column time) is read; GenePix results files (also ATF) are read by genepix-gpr, GenePix array lists exit 6 Values are the decimal text as written by the exporting program (already rounded)
Blackrock NSx/NEV Blackrock Neurotech (Cerebus, NeuroPort)	.ns1.ns2.ns3.ns4.ns5.ns6.nev	Electrophysiology	high	3 gaps A recording directory is one session (each NSx file a trace, each NEV a table, clocks kept); NEV rows are not assigned to NSx sweeps NEV video-sync, tracking, button, configuration and log packets are counted (kind 3), not decoded; comment text is in info --view full Spec 2.1 NSx scaling needs the .nev with the same name next to it
CED Spike2 data file (.smr, .smrx) Cambridge Electronic Design (CED) Spike2	.smr.smrx	Electrophysiology	high	4 gaps 64-bit .smrx: the layout is derived from public files and Spike2's own exports (no documentation); Adc, rising events and empty marker channels are validated, RealWave, AdcMark/RealMark/TextMark and level events follow the same rules unvalidated Level-event channels are listed as event times; the level after each edge is not reported Interleaved AdcMark waveforms are returned in stored (interleaved) order Spike2 memory and virtual channels are not reconstructed; only stored channels are read
HEKA PatchMaster bundle (.dat) HEKA Elektronik (Harvard Bioscience) PatchMaster	.dat	Electrophysiology	high	4 gaps Stimulus (PGF) templates, amplifier (.amp), solution and marker trees are not decoded; the command waveform is not synthesized Files with separate .pul/.pgf files (DAT1), PULSE-era files, big-endian (PowerPC) files and PatchMaster Next v2000 (64-bit) bundles are refused Non-zero Y offsets and scaled float traces are refused (never seen in a development file) Values are not zero-subtracted; PatchMaster's zero level is reported per channel
Intan RHD2000/RHS2000 Intan Technologies (RHD/RHS recording and stimulation systems)	.rhd.rhs	Electrophysiology	high	3 gaps "One file per signal type" and "one file per channel" directories (info.rhd/info.rhs + .dat) are read as one recording; temperature sensors are not saved in those layouts Stimulation words return the signed current; the amp-settle, charge-recovery and compliance bits are not exposed Split recordings (one file per N minutes) are read file by file
Neuralynx (NCS/NEV/NSE/NTT) Neuralynx (Cheetah, Pegasus)	.ncs.nev.nse.nst.ntt.nvt	Electrophysiology	high	3 gaps A recording directory is one session: channels on different sample grids (rate, segment boundaries) are separate traces, never resampled; event and spike files are separate tables Raw data (.nrd) files are not read; video-tracker (.nvt) records give the extracted position, angle and target count, not the colour-transition points or target bitfields Sampling rate is the header value; when timestamps imply another rate (pre-Cheetah-5 1 MHz clock) it is reported in extra.timestamp_rate_hz, not applied
NWB (Neurodata Without Borders 2.x, HDF5) open standard (NWB) · also written by export	.nwb	Electrophysiology	high	4 gaps Traces are TimeSeries, ElectricalSeries and SpatialSeries under acquisition/ and processing/, and patch-clamp series (current/voltage clamp, their stimuli) there and under stimulus/presentation/, one trace per sweep (intracellular recording tables are tables, not used to group sweeps); ImageSeries, optical-physiology and other series types are listed by info --view structure, not decoded Tables hold numbers: text columns become category codes (extra.categories), ragged columns a per-row count (units' spike times also a table of their own), object-reference columns are left out NWB 1.x files and the Zarr backend are not read Timestamps are compared for uniform sampling; irregular series are returned with a time channel
Open Ephys recording (binary or Open Ephys format) Open Ephys GUI	.oebin.continuous	Electrophysiology	high	4 gaps Input is a recording directory (or its structure.oebin / a .continuous file); NWB-format Open Ephys recordings are read by the NWB reader Binary format: spike groups and binary (tracking) event streams are listed, not decoded; synchronized timestamps are reported per sweep, samples are not resampled Channel units come from structure.oebin as recorded; when empty, µV for neural and V for ADC channels (Open Ephys docs) Legacy format: records that are not full 1024-sample records on a regular clock are refused per channel; gaps and recording numbers split sweeps (Neo zero-fills gaps)
Plexon PLX/PL2 Plexon (MAP, OmniPlex)	.plx.pl2	Electrophysiology	medium	5 gaps PLX has no index: info walks every data-block header (fast, but proportional to the file size) PLX values are in mV by the gain formulas Neo documents; version 100/101 files assume a preamplifier gain of 1000 PL2 is derived from hex dumps of public files (Plexon's reader is a DLL); spike waveforms and events are confirmed against Neo on a PLX of the same recording, PL2 continuous records only for internal consistency PLX version 107 (and continuous channels of unequal length) are read but have no oracle-confirmed continuous data PL2 footer index and device settings are not read; info walks the data-record headers
SpikeGLX (.bin + .meta) SpikeGLX (HHMI Janelia), Neuropixels / NI-DAQ / OneBox	.bin.meta	Electrophysiology	high	3 gaps A run directory is one session with one trace per stream (AP, LF, NI, OneBox), each on its own clock; streams are not aligned on their sync pulses Probe sites come from ~snsGeomMap (µm) or ~snsShankMap (grid column/row, not converted to µm); imro bank/reference settings are reported in info --view full, not normalized imMaxInt defaults for probe types other than NP 1.0 and 2.0 (types 21/24) are inferred from the ProbeTable when the metadata omits it
WinWCP data file (.wcp) Strathclyde Electrophysiology Software (WinWCP)	.wcp	Electrophysiology	low	3 gaps The channel zero level (YZ) is reported, not subtracted (as Neo does); both public files have 0 Record status (accepted/rejected), type and group are reported per sweep; rejected records are not dropped Stimulus protocols (.vpr) and analysis results are not read
Applied Biosystems experiment document (.eds) Thermo Fisher Scientific / Applied Biosystems (QuantStudio 1-7 Pro, 12K Flex, ViiA 7, StepOne, StepOnePlus, 7500)	.eds	Plate readers and qPCR	high	4 gaps Raw optical images (apldbio/sds/images/*.tiff, quant/*.quant) are listed, not decoded; files that keep only images have no curves Genotyping (allelic discrimination): the vendor's calls are the table genotypes (SDS layout; codes mapped from 1000 Genomes genotypes of 4 reference samples at 2 markers of one QuantStudio 7 run); JSON-layout (Design & Analysis 2) genotyping results are not read Relative-quantification results (RQ, ΔΔCt) computed by the vendor are not parsed; analyze qpcr --ddcq computes them SDS/7500 layouts: the threshold comes from the analysis protocol (equal to the exported Ct Threshold in 8 StepOnePlus runs); an automatic baseline window is recovered per well from Rn − ΔRn (572 of 574 wells equal to the export) or left empty, never reported as the setting
Bio-Rad CFX data file (.pcrd) Bio-Rad Laboratories (CFX96, CFX384, CFX Opus; CFX Maestro / CFX Manager)	.pcrd	Plate readers and qPCR	low	1 gap Encrypted container (a zip whose single member is encrypted): detected and refused with exit 6; export RDML from CFX Maestro instead
Plate-reader exports Agilent BioTek, Molecular Devices, BMG LABTECH, Revvity/PerkinElmer, Tecan, Thermo Fisher Scientific	.txt.csv.tsv.tab.asc.xlsx.xlsm.xls.xlsb.ods.pda.sda.xpt.prt	Plate readers and qPCR	medium	5 gaps SoftMax Pro binary documents: SoftMax Pro 5 .pda plate sections (absorbance, endpoint or kinetic, one wavelength, 96 wells; 4 documents from 2 labs validated value-for-value against their text exports) and SoftMax Pro 6/7 .sda endpoint plates (absorbance, fluorescence, luminescence; one wavelength; 6 documents validated) are decoded; kinetic, spectrum and well-scan sections of 6/7 documents, several wavelengths, other plate sizes and cuvette sets are refused (exit 6: export them as text) Gen5 experiment files (.xpt): endpoint and kinetic reads decoded from the plate archives (validated on Gen5 2.08, 3.04, 3.11 and 3.15 files from three labs against their exports); the detection mode comes from the read name (number, ex,em, Lum), other names are unknown; flagged cells (OVRFLW etc.) are NaN without their text; area scans, spectra, well scans and protocol files (.prt) are refused Gen5 area scans and multi-plate kinetic exports, SoftMax Pro well scans and fluorescence spectra are not decoded Values calculated by the vendor software are kept as calculated reads; curve fits and group tables are kept verbatim only (openreadout analyze assay recomputes standard curves, IC50s and kinetics from the reads). SoftMax Pro 5 document templates give the layout (samples, groups, Blank group as blanks); SoftMax Pro text-export group tables and SoftMax Pro 6/7 group sections are not used as layouts, and document-stored reduced values are not read Day/month order of slash dates is assumed month-first when ambiguous (flagged)
Qiagen Rotor-Gene run file (.rex) Qiagen (Rotor-Gene Q, Rotor-Gene 6000; Rotor-Gene Q Series Software)	.rex	Plate readers and qPCR	low	2 gaps No Cq values: Rotor-Gene files keep raw channel readings and the run setup; the analysis lives in the vendor software (analyze qpcr --cq computes Cq) Melt channels are read when the file records them as readings of a melt step; gain optimisation and other auxiliary readings are listed only
RDML (Real-time PCR Data Markup Language) RDML consortium (open standard); written by Bio-Rad CFX Maestro, Roche LightCycler 96/480, Applied Biosystems StepOne, qPCR analysis tools · also written by export	.rdml.rdm.lc96p.xml	Plate readers and qPCR	high	3 gaps Digital PCR partitions (RDML 1.3 partitions, partition tables) are counted, not read Documentation, xRef, cDNA synthesis and commercial-assay elements are kept only in the vendor tree LightCycler 96 (.lc96p, or RDML with its app_data.xml/calculated_data.xml): the software's calls, replicate Cq mean/SD and exclusions are attached; an RDML cq under a call other than Positive is not a Cq, even when it is a plausible number (kept as cq_stored; the software's analyses list none for such graphs, and rdmlpython and other RDML readers report the stored number). Validated against the software's results table for 3 files (108 Cqs) and its absolute-quantification analysis in a fourth (56 Cqs, 8 Negative graphs without one); melting-peak Tm of LightCycler 96 analyses is not read
Roche LightCycler 480 experiment (.ixo) Roche Diagnostics (LightCycler 480, LightCycler 480 II; LightCycler 480 software 1.5)	.ixo	Plate readers and qPCR	low	5 gaps Read: run and instrument, thermal protocol, detection channels, sample names and per-channel target/type/concentration, raw fluorescence per cycle and melt readings (instrument units, no color compensation), and the vendor's absolute-quantification results (Cp and positive/negative call). Validated on 4 LightCycler 480 QC runs (software 1.5.0 and 1.5.1): every vendor-positive well rises and every negative stays flat in the analysed channel Only Legacy Absolute Quantification Analysis results are read (one channel, no ratio); Tm calling, genotyping, relative quantification and other analyses are listed in the vendor tree, not read; a call other than 0 or 2 gives no Cq Vendor melt smoothing and derivative arrays (DARZ/FORM binary properties) and the temperature log are not decoded; -dF/dT is computed from the raw melt readings Our own Cq (analyze qpcr --cq) is not the LightCycler 480's Cp algorithm (not public); use the vendor Cp in cq The closing checksum line is kept (vendor.trailer), not verified
Agilent Seahorse XF assay result (.asyr) Agilent (Seahorse)	.asyr	Bench biophysics	low	3 gaps OCR, ECAR and PER are not stored; they are computed (published compartment model, the file's constants) only for standard 96-well plates with Wave's AKOS settings, where they agree with Wave's own rates (OCR within 0.2 pmol/min, ECAR within 0.001 mpH/min); other plates (24-well, spheroid) return the emissions and a note saying why PER uses kVol 1.6 (not stored; what Wave prints for the same plate) and is withheld by --strict Raw LED on/off emission and reference arrays and the calibration arrays are not exposed; templates (.asyt) are not read
Bio-Rad Image Lab (.scn) Bio-Rad	.scn	Bench biophysics	high	5 gaps Validated on 13 development files from Image Lab 3.0.1 to 6.1.0 (ChemiDoc MP, ChemiDoc XRS+, Gel Doc XR+, a merged image); all hold one 16-bit scan. Scans imported from Typhoon/Molecular Dynamics .gel files carry a square-root encoding (scaler/moldyn) that is reported, not applied, and no corpus file exercises it Lane and band analyses, molecular-weight calibrations and volume tools are not stored in any file seen; other parts are listed by info --view structure and kept in the vendor tree Multi-channel (fluorescent multiplex) files have no public sample; each ScanImageTag would be its own image Scans imported from Molecular Dynamics .gel keep their stored square-root-encoded counts (the scale is reported, not applied) Leica SCN whole-slide files share the extension: they are TIFF and go to the TIFF reader
Cytiva Biacore result file (.blr) Cytiva (Biacore)	.blr	Bench biophysics	medium	4 gaps Validated on Biacore T200 Control Software 2.0.1-2.0.2 result files (kinetics/affinity and immobilization wizards, manual runs); other Biacore systems (3000, X100, S200, 8K, Insight) and evaluation files (.bme, .bie) are not read Sensorgrams and the reference-subtracted curves are returned as stored; kinetic or affinity fits (ka, kd, KD) are the evaluation software's and are not computed Report points are the control software's; at 1 Hz storage they average data the file does not hold Event-log codes, the quality and time-correction streams and the wizard template are kept or listed, not interpreted
Cytiva Biacore T200 evaluation file (.bme) Cytiva (Biacore)	.bme	Bench biophysics	medium	4 gaps Fits (kinetics, affinity) are the evaluation software's, returned as stored with their parameters, standard errors and Chi²; they are not recomputed The evaluation's own processed curves (aligned, blank-subtracted, in EvaluationItemNBinary), plots and report-point tables are not decoded; the sensorgrams are the result files' as the evaluation copied them Concentration-analysis items (calibration curves) are listed without their results Biacore 8K, S200 and Insight evaluation files are not read
GenePix Results (.gpr) Molecular Devices (Axon GenePix)	.gpr	Bench biophysics	medium	2 gaps GenePix array lists (.gal), settings (.gps) and the scan TIFF images are not read by this reader (the TIFF reader opens the images) Values are the text GenePix wrote; no normalization, background correction or flag filtering is applied
Malvern Zetasizer measurement file (.dts) Malvern Panalytical	.dts	Bench biophysics	low	4 gaps Size and zeta results are returned as stored; the size distributions, correlation functions and phase plots are not decoded Number and volume means and the diffusion coefficient the software exports are computed at export and are not returned Molecular-weight, protein-mobility and other record kinds are listed, not decoded ZS Xplorer .zmes files are not read
MicroCal ITC (.itc) Malvern Panalytical (MicroCal)	.itc	Bench biophysics	medium	4 gaps Validated on VP-ITC (VPViewer2000 1.4.8-1.30), iTC200 (1.25-1.26) and MicroCal ITC software 1.29 files; PEAQ-ITC .apitc and analysis projects (.apj) are not read Time and differential power are validated against Origin's raw exports; the cell temperature column is inferred; further data columns (7- and 9-column files) are exposed unnamed Integrated injection heats are the analysis software's (baseline and integration) and are not stored in the file; they are not computed Calibration lines (%) are kept verbatim in the vendor tree, not interpreted
Sartorius Octet BLI result (.frd) Sartorius (ForteBio)	.frd	Bench biophysics	medium	3 gaps One file is one biosensor; the experiment's other sensors are separate .frd files (read them together with batch) Kinetic fits (kon, koff, KD) are made by the Octet analysis software and are not in .frd files Only kinetics experiments (KineticsData) are read
Arbin MITS Pro result (.res) Arbin Instruments	.res	Materials and electrochemistry	medium	4 gaps Only the Jet 4 / ACE database of MITS Pro 4-8 is read; the MITS 10+ SQL Server data and the .xlsx/.csv exports are not Smart-battery, CAN-BMS and multi-cell ACI tables are listed (ls) but not returned The schedule (.sdu) is named, not read; step types are not in the file date_time is the tester's local time; the file stores no time zone
BioLogic EC-Lab (.mpr) BioLogic	.mpr	Materials and electrochemistry	high	4 gaps A file with a column id not validated against an EC-Lab export is refused (exit 6) with the advice to read its .mpt export; about 70 ids are known Quantities EC-Lab computes only when exporting text (energies, capacities, efficiency, P/W where not stored) are not returned Technique parameters (the settings module) are not decoded; the technique, channel and acquisition start are Instrument model and software version are not in a decoded field of the .mpr (the .mpt export names them)
BioLogic EC-Lab text export (.mpt) BioLogic	.mpt	Materials and electrochemistry	medium	2 gaps Every column EC-Lab exported is returned, including those it computed at export; columns not in the vocabulary keep a name made from EC-Lab's label Header facts are read from the key : value lines EC-Lab writes in English
Bruker BES3T EPR (.DSC/.DTA) Bruker	.dsc.dta	Materials and electrochemistry	medium	4 gaps ASCII data sets (IRFMT A), N-tuple axes (NTUP) and points whose values have different number formats are refused (exit 6) Values are returned as stored: no normalization by scans, gain, power or conversion time (Xepr's and EasySpin's optional scalings) The manipulation history layer (#MHL) is counted, not interpreted; processed data sets are read like acquired ones Only the standard parameters are normalized; device parameters (#DSL) are in the vendor tree as text
Bruker DIFFRAC RAW Bruker	.raw	Materials and electrochemistry	high	4 gaps RAW1.01 (DIFFRACplus) and RAW4.00 (DIFFRAC.SUITE) are read; the older RAW and RAW2 versions are refused (exit 6) Ranges whose records carry per-point parameters other than the measured 2θ, and RAW4 records wider than one float32, are refused (exit 6) The scanned axis is the drive the range header marks as moving, else 2θ for coupled and detector scans; other scan types are refused Detector, slit and goniometer settings are kept in the vendor tree; only the tube, wavelengths, step time, user and sample are normalized
Bruker DIFFRAC.SUITE BRML Bruker	.brml	Materials and electrochemistry	medium	3 gaps 2D detector frames (recorded views wider than one value) are refused (exit 6); integrated 1D counts are read Counts are returned as stored; absorber factors are a separate channel and are not applied Evaluation, template and instruction containers are not decoded
Bruker ESP/WinEPR EPR (.par/.spc) Bruker	.par.spc	Materials and electrochemistry	low	2 gaps The .spc file has no header: its layout (little-endian float32 for WinEPR, big-endian int32 for ESP) and point count come from the .par file; a data set without its .par file is not recognised Values are returned as stored: no normalization by scans, gain or conversion time
Gamry Framework (.DTA) Gamry Instruments	.dta	Materials and electrochemistry	medium	2 gaps Text columns (the Over overload flags) are kept out of the traces Header parameters are kept in the vendor tree; only the date, time, potentiostat, notes, area and scan rate are normalized
NETZSCH Proteus (.ngb-*) NETZSCH	.ngb-ss3.ngb-bs3.ngb-ds3.ngb-sd7.ngb-bd7.ngb-dla.ngb-cla	Materials and electrochemistry	medium	4 gaps The DSC signal is returned in µV as stored; Proteus's sensitivity calibration (to mW, mW/mg) is not applied The temperature program, PID settings and calibration tables are not decoded Analysis (.ngb-taa) and state (.ngb-od7) files are not measurement files and are refused Channels without a known meaning are returned as channel_ without a unit
Neware BTS (.nda) Neware	.nda	Materials and electrochemistry	medium	4 gaps Versions 29 (BTS 7) and 130 (BTS 8/9, both record layouts) are read; other versions are refused (exit 6) Auxiliary records of version-29 and BTS 9.0 files (0x65) are not decoded; BTS 9.1 records carry their temperature A current range missing from the known scale table is refused Cycle numbers are as stored (1-based); BTSDA's charge-first cycle renumbering is not applied
Neware BTS (.ndax) Neware	.ndax	Materials and electrochemistry	medium	4 gaps .ndc versions 5, 11, 14, 16 and 17 are read; other versions are refused (exit 6) In versions 11-17 only logged records carry capacities and energies (NaN between them); the step time between logged records follows from the logging interval (reported as assumed) Auxiliary channels of versions 5, 11 and 16 are not decoded The step program (Step.xml) is kept out; the executed steps are the steps table
PANalytical XRDML Malvern Panalytical	.xrdml	Materials and electrochemistry	high	3 gaps Area-detector frames (2D XRDML) are refused (exit 6); 1D scans of 1D and 2D detectors are read Intensities are returned as stored (counts, attenuation applied, as the schema defines them); count rates are intensities divided by the counting time in extra.counting_time_s Optics are kept in the vendor tree; only the tube, wavelength and detector name are normalized
Rigaku RAS Rigaku	.ras	Materials and electrochemistry	medium	1 gap Intensities are returned as stored; the attenuation column is a separate channel and is not applied (a file with factors other than 1 is flagged)
Rigaku RASX Rigaku	.rasx	Materials and electrochemistry	low	2 gaps Intensities are returned as stored; the attenuation column is a separate channel and is not applied (a file with factors other than 1 is flagged) Area-detector images stored in a .rasx are not read
TA Instruments data file (Universal Analysis, .001) TA Instruments	.001	Materials and electrochemistry	medium	3 gaps Values are the stored signals; Universal Analysis's analyses (onsets, peak areas, normalisation by size) are not computed Only the header key/value lines are decoded; calibration lines are kept in the vendor tree TRIOS (.tri) files are a different format (ta-trios)
TA Instruments TRIOS (.tri) TA Instruments	.tri	Materials and electrochemistry	medium	4 gaps Values are the stored signals; variables TRIOS computes (normalised heat flow, derivatives, viscosities of flow steps, DMA moduli) are not, except the oscillation moduli of parallel-plate rheometer steps Signals of unknown meaning are returned as signal_ without a unit Files of the 2019 generation (TRIOS 3/4, byte 2 = 12) store their data differently and are refused Per-point oscillation waveforms and analysis results stored in the file are not decoded
HDF5 (generic) open standard (The HDF Group)	.h5.hdf5.he5.hdf	Containers	low	2 gaps Structure only: groups, datasets (shape, type) and attributes are listed; no dataset is decoded as an image, table or trace HDF4 (.hdf without the HDF5 signature) is not read

What a format page contains

Each page shows our notes from docs/formats/: the byte layout, the metadata OpenReadout normalizes, and a vocabulary table of every public name in the reader’s code. It links to the reader’s provenance log, which records the files we examined, the prior art we consulted and what we inferred from each. Together they show how each reader follows the clean-room policy. OpenReadout doesn't write vendor formats; it exports to open ones instead (see export).

Requesting a format

Open a format request. The most useful thing to attach is a link to a public file under an open license, with the name and version of the acquisition software. Instruments that have no open reader anywhere yet are surveyed in Bench instruments without open readers.