Another D800 theoretical question

Discussion in 'Digital Photography' started by kallisti, May 12, 2012.

  1. kallisti, May 12, 2012
    Last edited: May 12, 2012

    kallisti macrumors 65816


    Apr 22, 2003
    One theoretical advantage of the sensor on the D800 is the increased dynamic range compared to other sensors.

    From DxOMark data, it has a dynamic range of around 14 EV. When shooting RAW this should allow more leeway in recovering highlight and shadow detail in images with high contrast.

    I've read however that NEF files only allow for around 10 EV within an image. Has Nikon changed NEF files to accommodate the larger dynamic range the sensor on the D800 can capture or is the added dynamic range from the sensor effectively wasted due to software limitations? Or is my source for the dynamic range of NEF files in error (read it in a book on HDR photography and possible the limits on dynamic range related more to sensor technology at the time of publication and not limitations in the file format itself)?

    Again, this is a theoretical question. While it sounds awesome to have a sensor with the ability to capture a broader dynamic range, if the RAW files aren't capable of recording this data then it ends up being a waste from a practical perspective.
  2. sapporobaby, May 12, 2012
    Last edited: May 13, 2012

    sapporobaby macrumors 68000


    Sep 27, 2007
    3 earth minutes from your location....
    It could mean that you need Nikon's software or Nik software to fully exploit the captured data.
  3. mulo macrumors 68020


    Aug 22, 2010
    Behind you
    if your sources are credible, then you you would be wasting part of the sensors capabilities.

    But, like most things, I suspect there is more to it then meets the eye.
  4. charlieroberts macrumors 6502a


    Feb 5, 2007
    Well think about it this way, if the NEF file only allowed for 10... how did they check to see that the sensor captures 14? They must have gotten the data off the camera somehow, so i doubt your source is correct.
  5. sapporobaby macrumors 68000


    Sep 27, 2007
    3 earth minutes from your location....
    A quick post from a friend of mine who is a Dr of optical Physics wrote:

    "I've read however that NEF files only allow for around 10 EV within an image."
    that's nonsense
    it might be true of D70
    anyway, the DXOMark dynamic range is evaluated for 8MP images generated by lowpass filtering and downsampling from 36MP.
    so that would make it possible a DR over what 14 bits per pixel can record
    since the nef is always for the original resolution
    converting from 14 bit 36MP image to 8MP you get essentially 16 bits or more per pixel.

    In short, he said the data is wrong.... Move on to the next topic.
  6. OreoCookie macrumors 68030

    Apr 14, 2001
    Sendai, Japan
    The sensor scores for the DxOMarks are calculated from RAW files of the camera. So your claim does not make any sense: the claim that one gets 1x EVs of dynamic range is the result of a measurement, so of course Nikon's RAW files do not limit that in any way. As a matter of fact, there is no single .nef format, at least each set of cameras with distinct sensors have their own. Older .nef files have a color depth of 12 bits while newer ones store 14 bits. Your fear is unfounded.

    However, dynamic range measurements are done under lab conditions which means, in general you will not get 14 EVs out of your camera in practical applications since this may depend on the type of pictures, the RAW converter you use and that optimizing for dynamic range may negatively affect other aspects of image quality. But that's normal and has nothing to do with the manufacturer or the D800 in particular.
  7. kallisti thread starter macrumors 65816


    Apr 22, 2003
    Thanks for the replies. In hindsight it was a stupid question. The book I read was clearly in error. But it raised a question that made me think about the possible limits of the various points in the chain leading from image capture to final output and how it is possible for there to be a rate-limiting-step at one point that potentially negates advantages gained at another step.
  8. Ruahrc, May 13, 2012
    Last edited: May 13, 2012

    Ruahrc macrumors 65816

    Jun 9, 2009
    Actually, I would not be so fast to write off your question...

    If you assume a linear recording system (which I believe is the case for camera sensors and RAW files), a 14-bit NEF file should be able to record 14 stops of dynamic range. That is, the difference between the smallest and largest value recordable within a 14-bit system is 16384, or 14 stops: log2 of 16384 is 14, or conversely 2^14 is 16384. The calculations are easy (as long as you assume linearity) since data bits operate in the base2 system, and light stops are also base 2. Number of bits should be theoretical number of stops.

    Therefore, while the 14-stop DR of the D800 may theoretically fit within the 14-bit NEF files it produces, we seem to be up against the limit now? Is it possible that the D800 can actually outperform 14 stops, and that the near 14-stop value recorded by DxO was really a function of the 14-bit NEF file? (Less some real-world inefficiency- probably mainly due to thermal noise in the sensor creating a greater than zero noise floor)

    Will the next generation of high end sensors need to incorporate 16-bit ADCs and 16-bit NEF files in order to exceed the DR of the D800? Medium format cameras already use 16-bit ADCs and 16-bit files. In fact I believe the internal processing chain of the D800 and D4 are also 16-bit now, and this is quite possibly because of the dynamic range limitation of the 14-bit imaging chain compared to the sensor.

    Honestly, I had personally hoped that the D4 and D800 would have introduced 16-bit NEFs.

    But you are correct in your conclusion that somewhere along the imaging chain there is going to be the "weakest link", which will ultimately limit the performance of the system. I believe that in most real cases it is going to be the quality of the ADCs built into the sensor. However, in the most efficient and most well-engineered scenario, all components will be designed and implemented such that they are all performing at or near their peak theoretical performance, so that unnecessary capability is not "wasted" but also that no one particular step is crippling the rest of the chain.

Share This Page