A camera’s sensor, color encoding, and viewing transform answer different questions. Compare them separately.
Encoding is not the sensor
Log encodings distribute code values efficiently across exposure. They do not create sensor dynamic range or, by themselves, determine a highlight look.
- LogC3 EI800
- LogC4
0.00 stops: linear 0.18000; LogC3 0.39101; LogC4 0.27840.
The plot uses LogC3’s EI800 exposure-value formula, including its linear branch, and the LogC4 software formula. LogC3 parameters vary with exposure index; LogC4’s software encoding has a fixed definition. The latter must not be confused with its in-camera hardware encoding. ARRI LogC3 specification, ARRI LogC4 specification.
ARRI Wide Gamut 3 and 4 define RGB coordinates; LogC3 and LogC4 define transfer functions. A large primary triangle is not a measurement of a sensor’s spectral accuracy or the colors it captures.
- AWG3
- AWG4
Showing AWG3, AWG4. xy area does not measure perceptual gamut volume.
RED’s Log3G10 name refers to 18% gray near one third of the signal range and ten stops above it near 1.0. It does not mean three stops below gray. R3D is a RAW format; RWG/Log3G10 is one developed representation, not an inherent display interpretation of every R3D file. RED white paper.
A reference value makes these distinctions concrete: linear 0.18 maps to about 0.391007 in LogC3 EI800 and 0.278396 in the LogC4 software encoding. These are normalized encoded signals, not a measured screen brightness or a camera dynamic-range rating. See the reference table.
When mixing cameras, identify the exact RAW development settings or log/gamut pair. Use documented input transforms, then assess matching under the actual illumination. Brand-level claims about “natural,” “clinical,” or “film-like” color are not substitutes for those tests.
What RAW actually records
A conventional Bayer block has two green-filtered sites, one red, and one blue. RGGB, BGGR, GRBG, and GBRG are different orientations of that pattern. Each site records a filtered intensity; demosaicing estimates the missing channels using neighboring samples. Other sensor layouts also exist.
RGGB: 18 green, 9 red, 9 blue sites in this 6 × 6 sample.
The doubled green sampling supports spatial detail; it is not a direct copy of the human cone mosaic. Demosaicing, black-level subtraction, white balance, and color conversion all affect a developed image.
RAW tools and SDKs differ in supported cameras and output behavior. Retain their development settings with the image so a later application can reproduce the intended interpretation.
Separate decoding from the manufacturer’s look
An ARRI display rendering LUT and an ACES input transform perform different jobs. The first can include rendering for a particular display; the second brings a specified camera encoding into ACES, which still needs its chosen rendering. Comparing the two as if they were competing LogC decoders confounds input conversion with appearance. ARRI’s Look File 4 documentation explicitly separates the creative modification from display rendering. ARRI workflow guide
To compare an ARRI look, RED output rendering, or another managed pipeline, hold the source interpretation, exposure, white balance, output target, and viewing conditions fixed. Label the exact transform/version. Differences may be intended rendering choices; an attractive result does not establish a more accurate camera characterization. Avoid inventing a proprietary tone curve from an unrelated S-curve or attributing all highlight behavior to the log encoding.
RAW to ACES needs a characterized conversion
The name of a decoder is not enough to specify its behavior. Current LibRaw lists output color 6 as ACES, but that does not establish that every executable called dcraw implements the same option. Nor does a 16-bit integer TIFF command establish an ACES half-float EXR interchange workflow. Check the exact executable, version, output scaling, and metadata. The ACES white is not D65. LibRaw parameters
rawtoaces provides a dedicated conversion path using LibRaw, camera metadata, and spectral characterization when available. Its documented matrix methods include spectral data and metadata-based alternatives; the chosen method affects what can be claimed about the result. A metadata fallback is not a newly measured spectral characterization of that camera. rawtoaces documentation
For a still-RAW conversion, retain the original file and record the decoder version, black-level handling, white-balance choice, matrix method and data source, exposure scaling, output primaries, and encoding. Validate a neutral ramp, a chart under the intended illumination, and unclipped highlights in a downstream ACES application. Commands alone cannot validate color accuracy, and demosaiced output is already a processed estimate rather than untouched sensor samples.