Results 1 to 9 of 9

Thread: DxOMark Sensor For Benchmarking Cameras

  1. #1
    BPN Viewer
    Join Date
    May 2008
    Posts
    2,940
    Threads
    288
    Thank You Posts

    Default DxOMark Sensor For Benchmarking Cameras

    It contains something similar to what Roger talks about:

    http://www.luminous-landscape.com/es..._cameras.shtml

    From footnote 27:

    "Arguably the Canon S90 is the best low-light camera in the database – at least when we take its limited size into account. In fact, creating an array of about 20 identical S90 sensors would result in a full-frame sensor which would, at least in theory, slightly outperform the reigning Nikon D3s!"
    Last edited by Desmond Chan; 02-11-2011 at 02:49 AM.

  2. #2
    BPN Viewer
    Join Date
    Jan 2011
    Posts
    21
    Thank You Posts

    Default

    However, dxomark.com site itself says otherwise.

    http://www.dxomark.com/index.php/en/...(brand2)/Nikon

    By far, the ISO quality of a 2.03micron pixel pitch can hardly get close to that of a 8.4 micron one. Perhaps until when Canon can make a 48MP sensor capable of outperforming a D3s.
    Last edited by Kenny Wong; 02-11-2011 at 04:36 AM.

  3. #3
    Banned
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,949
    Threads
    254
    Thank You Posts

    Default

    In my opinion, Dx0Mark ratings are like rating tools by a single metric, like the weight of a hammer. That may be fine for some hammers, but don't account for when you need a claw hammer and the rating said the ball-peen hammer was better, or when you need a screw driver (which, with it's low
    weight, scores low on the tool weight rating).

    DxOMark ratings may be fine if your application falls into the limited range of their benchmarks, but many types of photography do not. Where their benchmarks most apply is photography where you tune lens focal length to fill the frame with your subject (e.g. landscapes, portraits).

    Where their analyses fall apart is in wildlife, sports, astro and other types of photography where one is often focal length limited and the subject is small in the frame. They seem to go to great lengths to argue this type of photography is not relevant, but ask anyone who is trying to photograph a distant bird and they want the most detail in the subject, and it becomes highly relevant.

    For example, look at figure 2 at:
    http://www.clarkvision.com/articles/telephoto_reach/

    DXoMark rates the Nikon D3s, with its 8.46 micron pixels, as the top performer for sports/wildlife type photography. OK, then look at Figure 2 on the above page, where the D3s would produce about the same image as the Canon 1D Mark II image on the left. Compare that to the other 3 cameras in the figure. Which camera (pixel size) would you choose to get the most detail in your subject given the same telephoto lens? (This is not a Nikon versus Canon thin--I am taking about pixel size, not a specific camera).

    To produce an equivalent image as the Canon 7D did in Figure 2, the Nikon D3s would need a lens of double the focal length. Sure, the result would be a lower noise image, but does anyone think the 7D image has a noise problem? I certainly don't. The thing about noise is that people are sensitive to it at different levels. Once noise is low enough, it doesn't matter if it gets better as people won't perceive the noise as a problem in the first place. And who has the money to buy a double the focal length lens (if it existed), and the muscle to carry it around? The example image was shot with a 300 mm f/2.8 lens; 600 mm f/2.8 lenses do not exist, and would cost a lot of money. And in the case shown here (the Moon, Figure 2) detail is perceived as more important than any noise. This will also be true of all photography, and especially true for wildlife, sports, and other photography where you don't have the lens focal length and end up cropping the image.

    The bottom line is one model of performance is clearly inadequate and DxO is misleading people by arguing their model is more complete and relevant than it really is.

    For new people here, here is my web page on sensor performance:
    http://www.clarkvision.com/articles/...mance.summary/

    Also relevant:
    http://www.clarkvision.com/articles/...l.size.matter/

    and the above telephoto reach article.

    Roger

  4. #4
    BPN Viewer
    Join Date
    May 2008
    Posts
    2,940
    Threads
    288
    Thank You Posts

    Default

    Hi Roger, have you read the article linked? It's not from DxO Mark; it was written by another physicist in Netherland on DxO Mark ratings . I just thought it was an easier read for the mortals like me out there.

    So, what do you think about that physicist's conclusion on DxO Mark rating?

  5. #5
    Banned
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,949
    Threads
    254
    Thank You Posts

    Default

    Quote Originally Posted by Desmond Chan View Post
    Hi Roger, have you read the article linked? It's not from DxO Mark; it was written by another physicist in Netherland on DxO Mark ratings . I just thought it was an easier read for the mortals like me out there.

    So, what do you think about that physicist's conclusion on DxO Mark rating?
    Desmond,
    Yes, I have read the article. My post was modified from a post to another group a couple of weeks ago. Those who like the DXO benchmark and think they understand all conditions do not do much wildlife photography with big lenses. It is a markedly different ballgame, and the author of the article does not appear to be a wildlife photographer either. While he raises some good points in his conclusions, there is still too much emphasis on large pixels delivering high signal-to-noise ratio and not enough balance with subject detail. It is an extreme position. Similarly, some argue that many small pixels would be best. Their models for performance are incomplete. Like many things in life, there are multiple compromises and a compromise in one situation might not be the best solution for another situation. We have that condition in focal length limited telephoto situations versus other types of photography where you can choose your distance/focal length to fill the frame.

    For example, If you have a Nikon D3s or Canon 1DII with 8.4 micron pixels, why not get a camera with 100 micron pixels? That would give higher signal to noise ratios and much more dynamic range. 500 micron pixels? The ultimate on the DXO Mark rating would be a single pixel camera with a 24x36 mm pixel! Obviously these cameras would not produce the image quality of the D3s or 1DII. More resolution (pixels) generally improves perceived image quality up to the point where noise overwhelms detail. It is a compromise and people choose different levels that is best for their tastes, just like in many other things in life. There is no single camera that is best for all conditions, thus no single benchmark is appropriate for all conditions.

    Roger

  6. #6
    BPN Viewer
    Join Date
    Jan 2011
    Posts
    21
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    For example, If you have a Nikon D3s or Canon 1DII with 8.4 micron pixels, why not get a camera with 100 micron pixels?
    I think that you can't get a 100 micron pixel for 2 reasons,

    1) teh CMOS technology perhaps can't fabricate a large pixel of the size of 100 microns

    2) the sensor will have to have at lease 6MP pixels or else human eys may notice the "digital" effect. 12MP pixels could be a good compromise of both the number of pixels and the size of pixels. 16-18MP could be another good compromise of lens resulotion with a little bit tradeoff in pixel size and thus pixel quality.

    However, technology may allow higher SNR reserved in higher mega pixel sensor. At this moment and under the current technology, 12~18MP is already of the best compromise. Higher than this is just a demonstration of technology or marketing strategy in the sacrifice of the pixel quality. The only advantage a normal user can get from higher mega pixel sensors is that either he is used to cropping photo more than 100% crop or he'd like to make printouts larger than or up to A2 size.

  7. #7
    Banned
    Join Date
    Feb 2008
    Location
    Colorado
    Posts
    3,949
    Threads
    254
    Thank You Posts

    Default

    Quote Originally Posted by Kenny Wong View Post
    I think that you can't get a 100 micron pixel for 2 reasons,

    1) teh CMOS technology perhaps can't fabricate a large pixel of the size of 100 microns
    Kenny,

    A google search turns up CMOS sensors being sold with 80 micron pixels (just not in DSLRs), so hundred microns should not be a limit. I have used silicon photodiods in the 250, 500 and 1,000 micron sizes, and cmos diodes should be possible too.

    Quote Originally Posted by Kenny Wong View Post
    2) the sensor will have to have at lease 6MP pixels or else human eys may notice the "digital" effect. 12MP pixels could be a good compromise of both the number of pixels and the size of pixels. 16-18MP could be another good compromise of lens resulotion with a little bit tradeoff in pixel size and thus pixel quality.
    It wasn't that long ago that we didn't even have 6 megapixels. I have a full page image printed in Natures Best: a 3 megapixel image, and I've sold 16x18 inch prints in galleries.
    http://www.clarkvision.com/galleries...ght.f-600.html

    Quote Originally Posted by Kenny Wong View Post
    However, technology may allow higher SNR reserved in higher mega pixel sensor. At this moment and under the current technology, 12~18MP is already of the best compromise. Higher than this is just a demonstration of technology or marketing strategy in the sacrifice of the pixel quality. The only advantage a normal user can get from higher mega pixel sensors is that either he is used to cropping photo more than 100% crop or he'd like to make printouts larger than or up to A2 size.
    But all this does not address that fact that the DX0Mark benchmark does not appear to include sensor resolution, especially for focal length limited situations laikt that often encountered in bird photography. But I do agree that 12-18 megapixels in APS-C sensors is the optimum in getting high enough signal-to-noise ratios and enough megapixels.

    Roger

  8. #8
    Forum Participant
    Join Date
    Oct 2010
    Location
    Green Lane, PA
    Posts
    744
    Threads
    42
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    To produce an equivalent image as the Canon 7D did in Figure 2, the Nikon D3s would need a lens of double the focal length. Sure, the result would be a lower noise image, but does anyone think the 7D image has a noise problem? I certainly don't.
    Actually, I think the 7D has a noise problem, at least among modern DSLRs. I first noticed it when I shot side by side with one using my Nikon D90. With very similar settings, the results weren't even close. Since then I've looked at many photos on this and other sites and it seems to me that Canons below the Mark series often show noise above ISO 400.

    It appears to me that Canon went for megapixels and Nikon went for higher ISO performance, two different strategic decisions but each giving up one thing to excel in the other. Just my opinion, but a number of other people seem to share the same view over here.

  9. #9
    BPN Viewer
    Join Date
    Jan 2011
    Posts
    21
    Thank You Posts

    Default

    Quote Originally Posted by Roger Clark View Post
    Kenny,

    A google search turns up CMOS sensors being sold with 80 micron pixels (just not in DSLRs), so hundred microns should not be a limit. I have used silicon photodiods in the 250, 500 and 1,000 micron sizes, and cmos diodes should be possible too.



    It wasn't that long ago that we didn't even have 6 megapixels. I have a full page image printed in Natures Best: a 3 megapixel image, and I've sold 16x18 inch prints in galleries.
    http://www.clarkvision.com/galleries...ght.f-600.html


    But all this does not address that fact that the DX0Mark benchmark does not appear to include sensor resolution, especially for focal length limited situations laikt that often encountered in bird photography. But I do agree that 12-18 megapixels in APS-C sensors is the optimum in getting high enough signal-to-noise ratios and enough megapixels.

    Roger
    Even if it is possible that they can make larger pixels. But as long as 12-18 megapixels can hold the best details for cropping needs (including birding), then it is reasonable for vendors like Nikon and Canon to drive the pixel resolution and thus pixel size to suit for the 12-18 megapixel sensor.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
Web Analytics