Sensor dynamic range: measurements vs. field experience

Discussion in 'General Technical Discussion' started by Pa, Feb 7, 2018.

  1. I usually think that technical measurements of camera sensor dynamic range, particularly those of Bill Claff presented in his Photonstophotos web page, are the most reliable way to judge true dynamic range. My limited experience has born out his measurements.

    Yet many reviewers claim improved dynamic range between cameras that do not show a noticeable difference in the measurements. Case in point: D850 vs. D810. The measurements show a trivial difference at iso 64 (11.63 vs. 11.6), which is almost certainly within the uncertainty band for the measurement.

    Yet many reviewers claim significantly improved DR for the D850 at iso 64. Here I quote Glenn Nagel who did the research to find these comments:

    From Photography Life:

    "Even if you don’t need all the crazy fps and 4K video features, there are three most important features you do not want to miss – electronic front curtain shutter(EFCS) + silent shooting, ISO 64 for improved dynamic range and a much better autofocus system." This is in comparison to the D800.

    "...the D850 definitely has better dynamic range than its predecessor at base ISO, which is a remarkable achievement." In comparison to the D810.

    From Tony Northrup:

    "Where there is a clear improvement is in the dynamic range..." and this was in comparison to the D810.

    From F Stoppers:

    "At ISO 64, it's safe to say that the dynamic range of the D850 is a solid improvement over the Nikon D810."

    Can any of you technically-minded people here shed any light on this?
     
  2. I'll believe Bill Claff's measurements before I believe Tony Northrup et al's opinions with no data to back them up. DXOmark measured the identical dynamic range for both cameras.

    Larry
     
  3. Hey Jim, numbers don't lie. I would go with the facts from Bill rather than the opinions of those who have not measured the dynamic range of either camera and are just presenting opinions.

    That said, the D850 seems to have a lower noise floor (at iso 400 and above) which would affect the perception of dynamic range. I didn't buy the D850 for "better" dynamic range, but rather for more pixels and better AF....

    Cheers,
    alexis and Georgie Beagle

    "never let the facts get in the way of a click bait blog post" - Georgie Beagle
     
  4. I agree with both of you folks, but I didn't want to prejudice the comments in my original post. Remember that I am a retired engineering professor, so I trust objective measurements over subjective opinion.

    There may be a few subtleties we should not overlook. For example, downsizing a D850 image to the same pixel size as that of a D810 or D800 could provide some apparent DR improvement, but I again I doubt that it would be very noticeable.

    Can you think of any more?
     
    Last edited: Feb 7, 2018
  5. MNglass

    MNglass

    32
    Dec 7, 2005
    Let me preface this by saying that I'm a numbers guy and agree with you that Bill Claff has a better handle on objective truth. But I also believe that our vision is a far more complex (and subtle) thing than any single measurement suggests. For example, it has always looked to me that even if you have two sensors with near identical noise profiles, images from the higher resolution sensor will look better.

    So maybe the higher resolution, in capturing slightly more detail, subtly makes the files look better. It may be technically inaccurate to say that the dynamic range is improved, but not necessarily wrong to say that the files look like they have more DR. Think of the old grey square on a white background/ black background visual illusion.
    Checker shadow illusion - Wikipedia
     
    • Agree Agree x 1
  6. I think that is essentially what I said here:

     
  7. TonyW

    TonyW

    Jan 15, 2010
    UK
    I also trust Bill Claff's judgement on these things and have proved to my satisfaction at least that his figures can be trusted testing on a couple of my own cameras - the difference between his and my results insignificant (and his probably more accurate representation!)
    ISO Invariance

    The differences as you pointed out appear to be trivial and well within the "uncertainty band" (I like that and will use it elsewhere I am sure ;):))

    The bottom line for me is that it is not really possible to get more DR out of a sensor than the declared A/D converter figures. Therefore 14 bit capture and A/D conversion cannot yield more than 14 EV DR. You are then going to lose a couple of stops due to noise which means you will be lucky to achieve 12 EV DR

    It is not clear how the other reviewers reached their conclusion on clear, solid or better DR without any test data available. If they did choose to test then shooting a back-illuminated Stouffer wedge is not IMHO a great help in our world of image capture although you will likely see the magic 14 EV DR.

    However it would be hoped that if we use one comparison site to evaluate DR for a range of sensors then those comparisons may be useful as it would be assumed that the test methodology does not change (and that it is actually monitored correctly)
     
  8. MNglass

    MNglass

    32
    Dec 7, 2005
    Sort of. I suspect you are right. Except I'm not talking about downsizing. Again, this is all just speculation but I'm just thinking that as the extra resolution provides more detail, the perception of more tones might be perceived as 'greater range' but I have no way to test or prove that. And frankly I'm way outside my wheelhouse when talking about the fine points of visual perception :confused:

    But my experience is that higher resolution makes 'noise' look finer, less intrusive, with sharper edges so even though the per pixel noise may be similar, the higher resolution file looks better. Maybe something like that applies to DR. But again, the test engineer in me says 'you don't really know.'
     
  9. I enjoy technical discussions when everyone agrees with me. :D

    A further confirmation of my opinion comes from Thom Hogan, to whom I e-mailed a copy of my first post in this thread. He is currently working on his D850 review, so I thought he might have some useful input. He hasn't yet given me permission to quote him directly, but he said he doesn't think there is a clear DR improvement between the D810 and D850.
     
  10. Precisely. And that is the way I should have stated it in my original post
     
  11. As a D850 user who came from a D810 I agree with just about all that is said here. And as a social scientist I can attest to the fact that human mind is most fallible and humans are most capable of seeing what they want to see.
     
  12. I agree with that, but we were discussing DR at base iso.
     
  13. Ya, but I love throwing a spanner in the works....;)

    This discussion got me thinking and digging further into Bill's work.. (He really is a wonderful source of real information)

    From his Derived sensor characteristics we see:

    Nikon D810 well depth: 78,087 Read noise: 3.7 electrons
    Nikon D850 well depth: 60,912 Read noise 0.9 electrons

    From this data we see that the D810 pixels hold more electrons which is expected as they are larger, but the Read noise is smaller for the D850. What does this mean? For me, it seems that the D810 should have more DR but it is hampered by the read noise.

    Personally, I am totally impressed that the DN of the D850 is as good as the D810 considering the pixels are smaller. And I am really impressed they got the read noise down to 0.9 electrons. Ok as a chemist I am not happy with fractional electrons.

    Cheers,
    alexis and Georgie Beagle

    " mom, fractional electrons are statistical..." _Georgie Beagle. Shakes head, drops mike and walks off stage...
     
    • Like Like x 2
    • Agree Agree x 1
  14. I think it's time for an illustrated book of the sayings of Georgie Beagle!
     
    • Like Like x 2
  15. Are RAW files actually true RAW, or is there some smarts happening through the path from the sensor to the card which could be giving a perception of variation between sensors and processors?
     
  16. TonyW

    TonyW

    Jan 15, 2010
    UK
    Raw files by definition are simply unprocessed sensor data that are waiting for image manipulation to give a particular rendering. So no demosaicing, sharpening, white balance, TRC, etc applied

    Image rendering will be done on this raw data in post, the only exception would be camera JPEG you see on the camera LCD

    A manufacturer may decide to encode specific edit instructions within the raw data so that their software will interpret and render a particular look but this is done in post well after acquistion.

    Some third party raw editors come with presets that mimic the manufacturers own quite closely, but again this is in pp

    So raw must be truly raw to qualify for the name
     
    • Like Like x 1
  17. designdog

    designdog

    29
    Apr 25, 2013
    Richmond Va
    Where is the "vs field experience" from the title of this post? I have the D5 and the D850. The former is much maligned in the DR category, while the latter is much praised. Yet in my actual "field experience" with the two, I see absolutely no difference (in DR performance).

    Perhaps it is because I try to take photographs without the need for much exposure correction. +/- 1 stop at the very most, and rarely that. One other aspect never quite mentioned is burning and dodging - something we do all the time. This is actually a form of exposure correction, but, again, for me it is usually the equivalent of ½ stop or less, and it is mostly burning, at the edges of the frame. The DR issue does not come to play when darkening the image.

    Far more important, to me at least, is ISO performance, tone reproduction, and color fidelity. I do respect Thom Hogan and Bill Claff, and I do acknowledge that the lab tests seem to prove the DR issue out, I simply fail to recognize its importance in the real world of taking pictures - at least my real world...
     
  18. By "field experience" I was referring to the comments of the various reviewers mentioned in my first post, assuming they were based on field experience since they mention no measurements.

    In my own case, dynamic range is important because I do more travel photography than anything else. I am often faced with having to shoot in non-ideal light, so being able to expose for highlights and bring up shadows without introducing excessive noise is very helpful. That requires a camera sensor with wide dynamic range. I have a D750 and a D7200, and the extra stop of DR of the D750 is often noticeable.

    Of course what I am describing was called "burning and dodging" back in the film days and still is by some people.

    Perhaps I could have chosen a more apt title for the post.
     
  19. Pawl

    Pawl

    577
    Aug 10, 2013
    EC
    Nikon say the BSI image sensor design used in the D850 offers an "overall a one-stop image quality (image noise) improvement" - that is most likely how they acheive the D850 IQ being so good.
     
    Last edited: Feb 10, 2018
Loading...
Similar Threads Forum Date
Curved image sensors General Technical Discussion May 30, 2017
How to Practically Compare Dynamic Range Between Bodies? General Technical Discussion Apr 26, 2016
Sensor Scratched? General Technical Discussion Oct 1, 2015
Fijin D F-L001 Sensor cleaning device General Technical Discussion Sep 24, 2015
New Way To Clean Your Sensor General Technical Discussion Sep 9, 2015