I have been informed, in direct response to my question posed to DxOMark “How many individual cameras of a given make and model are tested for Dynamic range?” that the answer is “one camera”. DxOMark tests raw capture DR at different ISOs and reports the highest DR observed, with no mention of the associated ISO.
Several observations/questions come to mind.
First and foremost is that independent testing of the amount of extra raw-accesible dynamic range (ERADR) that exists beyond the point at which the JPEG-adjusted histogram trips the overexposure warning “blinkies” or “zebras” reveals considerable quantitative differences among cameras of the same make and model. In one popular full-frame DSLR the ERADR varies from ⅔ stop to Two and two-thirds stops (And possibly more than three stops according to unsubstantiated rumor).
Obviously, there is no assurance that the camera makers attempt to standardize to any reliable degree the DR of camera's of a single model. Not being au fait with camera production and marketing I cannot presume to offer a definitive reason for this state of affairs re: unreliable DR ratings of any particular camera model.
However, one likely cause does come to mind.
The semiconductor industry has, since Shockley co-invented the transistor up to the present day, been plagued with the phenomenon of “process variance” , “process variability”, or “process excursion”, the detection and limiting of which has been inadequate to prevent the obvious variance in dynamic range provided by our cameras’ light sensors. Literature from the semiconductor industry and researchers on causes of and proposed remedies for “process variance” continue, at pace, to the present.
It appears, then, that the camera manufacturers accept sensors guaranteed to provide a minimum required DR, then claim that there is a bit more DR “overhead” adequate to help “recover some blown highlights’ ...and the sweep the rest of the ERADR Under the rug, hoping no one will notice.
Well, those of us concerned with capturing the highest possible quality image data have noticed!
The reason we notice is that failure to fully utilize the brightest available stop of DR results in sacrifice of 50% or more of the total tonal spectrum and chromatic available to the image file (absolute numbers of lost tones depends on the bit-depth of the captured file). Failure to use the next brightest stop sacrifices 75% of tones, and failure to use the third brightest stop would leave the image file with only 12.5% of its tonal “birthright”. Additionally, use of progressively less of the available dynamic range at any ISO results in capture of progressively more noise. The photographer ignores his camera’s extra raw-accessible dynamic range (ERADR) at considerable risk to print quality.
The camera manufacturers ought explicitly notify potential buyers and current owners of
this state of affair relative to the indeterminate dynamic range of their camera model, and
the means by which to test their camera to determine its amount (in useable stops or parts thereof) of extra raw-accessible dynamic range (ERADR).
Those parties purporting to report accurate and useful tests on new camera models ought test an adequate number of camera's of each model. Results on dynamic range should be reported in terms of number of sensors tested, ISO’s at which tested, and the means, medians, and standard deviations from the means (assuming parametric distribution of data which, though likely, must be characterized as such).
I suggest that at the moment, except for the photographers who have independently tested their camera’s allotment of ERADR and actually use that extra DR , no photographers are utilizing their camera’s available dynamic range to the maximum possible extent.
It would be wise to recall and consider the late Bruce Fraser’s caution that proper exposure of raw image data for maximum image data quality is accomplished by exposing as closely as possible to blowing highlights, without actually doing so!
Sage advice, indeed!
Dave (uuglypher)