Originally posted by stevebrot I think he speaks without knowledge, except for the part that ISO between camera makes is not consistent. What piqued my curiosity is what he meant by not complying with ISO spec.
Much is made about ISO standards for sensor sensitivity, compliance, "cheating" and evaluation. The Wikipedia includes a good discussion of the current standard (dated 2006, confirmed as a standard, 2015) and how it relates to ISO film speed. Here are a few talking points:
- The standard is ISO 12232:2006. It may downloaded from HERE, but is not free (88 CHF)
- The standard(s) are based on full system output to an image file (not RAW)
- Several different techniques may be used by manufacturers and need not be the same technique for each point in the ISO range. These may be summarized as follows (the first two are new as of 2006):
- Recommended Exposure Index (REI): Arbitrary standard based on the manufacturer's best guess of what constitutes acceptable output.
- Standard Output Sensitivity (SOS): Based on pixel values for 18% gray on sRGB output.* This one is similar to the method for film speed.
- Saturation-based (SAT or Ssat): Similar to SOS, but with the basis being 100% white. Again the measured output is limited to sRGB.
- Noise-based: Subjective evaluation of image quality at two S/N points (40:1 and 10:1)
- In practice, manufacturers may use a mix of the above techniques. The standard defines weighting and reporting for those cases. The Wikipedia article gives a few examples.
- We call it ISO, but CIPA procedures bind implementations by Japanese makers
https://en.wikipedia.org/wiki/Film_speed#Digital_camera_ISO_speed_and_exposure_index
How any of the above relates to camera reviews and comparison between brands is questionable.** Two camera models may be fully standards compliant and yet exhibit quite different image characteristics at a particular "speed" setting. Near as I can tell, the best we might expect is a stop-wise linear progression in sensitivity along with reasonable results for a metered exposure. Expectation of equivalent assignment of "speed" to histogram results and expected quality between cameras is a pipe dream.
* My K-3 consistently reports this method for its full ISO range.
** FWIW, DxO Mark uses none of the ISO standard methods in their evaluation, though there is merit to their approach.
Steve