The image sensor voltage is proportional to linear light. Sensor acquisition is usually 12-14 bit. 13 actual stops of linear light cannot be contained within the 12 bit word even as it easily is more than enough to accommodate all 13 stops as non-linear light. This is the opto electrical transfer function. Yet the premise of raw to the video maker is that you bypass OETF processing and work directly through the metadata on the underlying data set of the imager.
Working directly with linear RAW data is practical for still photos, they aren't excessively large. With video, linear RAW occupies too much storage space, restricting your maximum capture duration. Fortunately, there's no perceptible loss in converting linear RAW into log RAW when the internal processing is done at the native bit-depth of the sensor. That's because log RAW matches the luminance response of the human eye far more closely than linear RAW. With lossless data compression, 12-bit encoding is more than wide enough to encode 14+ stops of log RAW data.
Last edited:



