I was struck by the following added comment to the recent MFDB articlehttp://www.luminous-landscape.com/reviews/...ii.shtml#update
Unfortunately DxO Labs has complicated matters. Their DxO Mark camera tests have become widely read by photographers around the world, and are quoted as gospel by many. Regrettably there are concerns that I, and others have expressed about some aspects of DXO's tests, and I have been in communication with them over this in the past.
The only aspect that is relevant to this discussion is with regard to dynamic range. The standard definition is, as mentioned above, how many F stops above and below middle gray can be recorded while delivering full texture and detail. The DXO definition, according to their web site, is the range between zero signal to noise and full saturation of the sensor.
This approach is not inherently flawed, it's just that it does not take into account the linear nature of sensors. It is therefore not a particularly relevant way of measuring DR to a photographer, as opposed to an engineer.
This is a common complaint about DxO's tests, and is often fodder for uninformed complaints to the effect that "DxO is garbage, my camera doesn't have that much DR". It is good that Michael points out that what DxO tests for in their DR plots is an engineer's definition of DR Engineering DR
: max recordable signal (light intensity) divided by noise at zero signal; or nearly the same, the range of signal levels over which the signal/noise ratio is greater than one. (Slight quibble: Michael misstates this definition as "the range between zero signal to noise and full saturation of the sensor.
" which is incorrect; it is the range between SNR of 0dB
and saturation, and 0dB is S/N=1, not zero.)
This engineering definition is different from the photographer's working definition of dynamic range given above by Michael (and which he calls the "standard definition"; not for engineers, of course, but in photography circles, yes). For the purposes of matching to DxO, let me translate that definition into more technical terms:Photographer's DR
: The range of signal levels over which signal/noise ratio (SNR) exceeds the photographer's minimum acceptable SNR.
DxO's approach is not inherently flawed (and the difference has nothing to do with the linear nature of sensors). I suspect the reason that DxO doesn't present "photographer's DR" is that its very definition depends on a criterion that varies from person to person; what is minimal acceptable image quality to one may be unacceptable to another. The difference between these two definitions has been the source of much of the consternation over DxO's DR results. How then can the photographer then get useful information about "Photographer's DR" from DxO?
DxO has another graph, for signal/noise ratio (in dB; in more photographer-friendly terms, 6dB=1 stop) as a function of signal (in percentage of saturation). For instance, for the D3x the graph looks like this:
On this graph, base ISO SNR drops to 0dB (SNR=1) at .015% of saturation, or 12.7 stops down (log(.00015)/log(2)=-12.7). Indeed DxO quotes 12.8 stops DR at ISO 100 (the "screen" tab of the DR plot). But suppose you don't think S/N=1 is good enough for acceptable image quality; instead you need S/N of 16 or greater for your standards of IQ. Then looking at the graph, ISO 100 exceeds S/N=16 (4 stops=24dB) starting at about .6% of saturation, or about 7.4 stops down. Then for you, the D3x has 7.4 stops of "usable" DR at base ISO. Note that if we maintain the same criterion uniformly, the DR at ISO 1600 is about three and a half stops (a little less than 10% of saturation).
Looking at the same graphs for the P65+, one sees an ISO 100 engineering DR of a little over 11.5 stops per pixel
, and a S/N>16 dynamic range of about 7 stops (lower end at about .8% of saturation). However this is DR per pixel, and the Phase One has about 3x more pixels than the D3x. DR scales with the number of pixels in the image, so to compare equitably one should compare DR at a fixed scale in the image by aggregating together 3 pixels worth of DR of the Phase One back; this increases the DR by about the square root of three, adding another 0.6 stop to its DR when matched to the pixel scale of the D3x. According to the engineering standard the MFDB is still well short of the D3x, however according to a standard more relevant to photography the back comes out slightly better (but less than 1/3 stop), mostly because of the larger sensor area collecting more light over the frame. The difference is not however the many stops DR advantage that some MFDB proponents claim.
A final caveat: The useful DR to a photographer can be limited by more than indicated by the S/N figure of merit; for instance Canon DSLR's have a base ISO plagued by a lot of pattern noise in shadows, which can be visually much more objectionable than the random grain of unpatterned noise while not showing up in the noise standard deviation. On the other hand, pattern noise seems very well controlled on the D3x. The pattern noise can limit the useful DR -- how much one is willing to push shadows -- more than might be indicated by the S/N graphs.