ARW File Documentation

Summary

An ARW (Sony Alpha Raw) file is the unprocessed raw image captured by a Sony Alpha or Cyber-shot camera’s sensor: a digital negative that stores the full tonal data before any in-camera demosaicing, white balance or compression. It is built on the TIFF/EP container and begins with the bytes 49 49 2A 00; its MIME type is image/x-sony-arw. To view thumbnails on Windows, install the free Raw Image Extension; to develop and edit an .arw, use Sony Imaging Edge, darktable or RawTherapee, then export to .jpg or the open DNG format.

Technical details

FeatureValue
Full nameSony Alpha Raw
File extension.arw
MIME typeimage/x-sony-arw
Format typeTIFF/EP-based camera raw image (binary)
DeveloperSony
Introduced2005 (Sony Alpha line; succeeded Konica Minolta SRF/SR2)
Base containerTIFF/EP (ISO 12234-2), extended with Sony tags
Magic number49 49 2A 00 (“II*\0”, little-endian TIFF)
Byte orderLittle-endian (Intel, “II”)
Raw located inSubIFD of the first IFD
Bit depthTypically 12–14 bits per sample
Compression modesUncompressed, lossy compressed, lossless compressed (code 32767 for Sony raw)
Colour filterBayer mosaic (undemosaiced sensor data)
Key Sony tagsSonyRawFileType (0x7000), SR2SubIFDOffset (0x7200)
Embedded previewFull-size and thumbnail JPEG(s)
MetadataEXIF + Sony MakerNote (camera, lens, exposure, sometimes GPS)
Open standardNo — proprietary; DNG is the open alternative
Related extensions.sr2, .srf, .dng, .nef, .cr2, .tiff
File signature (magic bytes)
49 49 2A 00

Offset 0, 4 bytes. In ASCII this reads I I * \0: the little-endian TIFF header. 49 49 (“II”, for Intel) sets the byte order to little-endian, 2A 00 is the magic number 42, and the next 4 bytes are a little-endian offset to the first Image File Directory (IFD). An ARW is a TIFF/EP file, so it shares this exact signature with an ordinary TIFF and with Nikon’s NEF. What makes it a Sony raw is not the header but the Sony-specific MakerNote tags and the SONY string inside the EXIF block; a plain TIFF viewer sees the structure but cannot render the undemosaiced sensor data.

What is an ARW file?

ARW stands for Sony Alpha Raw. It is the raw image format Sony has written since 2005, when it entered the interchangeable-lens market by acquiring Konica Minolta’s camera division; ARW succeeded Minolta’s earlier SRF and SR2 raw formats and is still produced by current Sony mirrorless bodies. A raw file is the data read almost directly off the camera’s image sensor, a digital negative. Unlike a JPEG, an ARW has not been demosaiced, white-balanced, sharpened or compressed into a finished picture inside the camera, so it keeps the full sensor bit depth and far more latitude to recover highlights and shadows or change white balance afterwards, at the cost of large files and the need for software to “develop” it.

Technically an ARW is not a bespoke format but an extension of an existing one: it is a TIFF/EP file (the Tag Image File Format for Electronic Photography, ISO 12234-2) with Sony-specific additions. Everything below is about how that TIFF structure stores the sensor data, the previews and the metadata, and why a generic TIFF viewer still cannot show you the picture.

The TIFF/EP header and the IFD chain

An ARW opens with the 8-byte TIFF header. The first two bytes, 49 49 (“II”), set little-endian byte order; the value 42 (2A 00) confirms it is TIFF; and the final 4 bytes are the file offset of the first Image File Directory.

Bytes 0-1:  49 49            byte order = little-endian ("II", Intel)
Bytes 2-3:  2A 00            magic number 42 (TIFF)
Bytes 4-7:  offset (uint32)  file position of IFD0

A TIFF is organised as a chain of Image File Directories (IFDs). Each IFD is a table of tagged entries followed by a pointer to the next IFD (zero if it is the last). Every entry is a fixed 12 bytes:

IFD entry (12 bytes):
  uint16  tag          which field this is (e.g. 0x0100 ImageWidth)
  uint16  type         data type: 1=BYTE 3=SHORT 4=LONG 5=RATIONAL ...
  uint32  count        number of values of that type
  uint32  value/offset the value itself, or an offset to it if it doesn't fit in 4 bytes

This is the key to the whole format: a reader walks the IFD entries, and any value larger than four bytes (an array, a string, the pixel data) is stored elsewhere in the file and reached through the offset in the last field. Because the format is offset-driven rather than positional, the raw pixels, the previews and the metadata can sit anywhere in the file, wired together by these pointers.

IFD0, the SubIFD, and where the raw actually lives

In an ARW the first IFD, IFD0, does not hold the full-resolution sensor data. It holds a smaller embedded preview and the pointers that lead everywhere else, in the TIFF/EP arrangement where the primary image is demoted to a sub-directory. IFD0 carries a SubIFDs tag (0x014A) pointing to a SubIFD, and that SubIFD describes the real raw frame: its ImageWidth and ImageLength (for example 6048×4024 on a 24-megapixel body), a BitsPerSample of typically 12 or 14, and the strip or tile offsets that locate the sensor data in the file.

The SubIFD’s Compression tag tells the developer how the raw bytes are packed. Sony’s raw compression is signalled by the value 32767 (a private code in the vendor range), while uncompressed raw uses the standard TIFF value 1. Recent bodies also offer a lossless-compressed mode; on some newer cameras the raw data is even double-referenced as both strips and tiles in the same SubIFD. Whatever the mode, the samples are a single-channel Bayer mosaic, not RGB pixels: each photosite recorded light through one red, green or blue filter, and turning that mosaic into a colour image (demosaicing) is the developer’s job, not something stored in the file. That is precisely why a plain TIFF viewer fails on an ARW: it can parse the IFDs but has no demosaicing engine and does not understand Sony’s compression code, so it shows nothing or only the embedded preview.

Embedded JPEG previews and fast thumbnails

An ARW carries one or more baseline JPEG images so that cameras, operating systems and browsers can show a picture instantly without running a raw pipeline. IFD0 typically references a full-size or reduced JPEG preview through its strip offsets, and a small thumbnail is stored for directory listings. This is what a file manager displays when it shows an ARW thumbnail, and it is why installing a raw thumbnail handler on Windows (the Raw Image Extension) makes previews appear in Explorer: the handler is reading that embedded JPEG, not developing the raw. The preview is a convenience only; it is a processed, camera-cooked image and does not carry the editing latitude of the sensor data beneath it.

EXIF and the Sony MakerNote: how ARW is recognised

IFD0 also points, via the ExifIFD tag (0x8769), to an Exif sub-directory holding the standard shot metadata: exposure time, aperture, ISO, focal length, lens model, timestamps and sometimes GPS coordinates. Inside the Exif block sits the MakerNote tag (0x927C), a vendor-private container that Sony fills with its own tags. This MakerNote, together with the SONY maker string, is what actually identifies the file as a Sony raw, since the TIFF header alone is identical to any other TIFF/EP file including Nikon’s NEF.

Two Sony-private tags matter for parsing. SonyRawFileType (tag 0x7000) records which raw variant the file is (for example uncompressed, lossy or lossless compressed). SR2SubIFDOffset (tag 0x7200), a survival from the older SR2 lineage, points to a further private sub-directory of decoding parameters. Some of that private data is lightly obfuscated, which is one reason each new Sony body needs an updated camera profile in raw software before it decodes correctly: the container is stable, but the exact tag contents and any compression tweaks change model to model.

Developing an ARW and exporting it

Because ARW holds a mosaic of sensor values rather than finished pixels, you do not edit it in place the way you retouch a JPEG. You open it in a raw developer, which demosaics the Bayer data, applies white balance and a tone curve, and lets you adjust exposure and colour non-destructively before exporting a rendered image. Sony’s own free application is Imaging Edge Desktop, which reads ARW with model-accurate colour; the free, open-source darktable and RawTherapee also decode Sony raw well, and macOS supports it natively in Photos and Preview.

Exporting produces a normal image, and the useful targets are a few. A JPEG is the universal, compressed result for sharing and printing, so keep the ARW as the master. A 16-bit TIFF is a lossless, fully-processed file for further work in an editor. And Adobe DNG is worth knowing as the open, documented raw format: converting an ARW to DNG keeps the raw sensor data while shedding Sony’s proprietary wrapper, which is why photographers use it for long-term archiving. What you cannot do is convert an ARW into another manufacturer’s raw, such as Canon’s CR2 or Nikon’s NEF: each vendor’s raw is tied to its own sensor data and private MakerNotes, so the open DNG is the only sensible cross-vendor raw target. One privacy note follows from the metadata above: the EXIF and MakerNote (including any GPS) travel inside the file, so strip them before publishing if location matters.

References