Has sensor sharpness exceeded lens sharpness? If so, what is the point of more megapixels?

WestEndBoyWestEndBoy Posts: 1,456Member
edited January 2014 in General Discussions
Hello All, please read the following:

http://www.dxomark.com/en/Reviews/Looking-for-new-photo-gear-DxOMark-s-Perceptual-Megapixel-can-help-you

While I am not sure how much stock to put in this, it rings true.

For example, the sharpness on my D800 seems to be related more to which high quality professional lens I am using rather than the D800's 36 megapixel sensor.

Another example, what about the D7100's 24 megapixel sensor which would be equivalent to 50ish megapixels on my D800 when pixel density is considered? And the D7100 is a DX camera optimized for use with DX lens which have generally inferior sharpness to FX lenses.

While diffraction is only a serious limitation on lens sharpness at F/8ish or slower, current lenses do not come anywhere close to achieving the theoretical diffraction limited sharpness that can be achieved at faster F-stops.

I would appreciate hearing all of your views and thoughts?

Comments

  • AdeAde Posts: 1,071Member
    The short answer is no. E.g., if the Nikon D850 comes out tomorrow with a superior sensor than the D800, then the DxO P-Mpix scores for all the lenses will be higher when re-tested on the D850, which implies that the D800 doesn't actually capture "all" the sharpness transmitted by the lenses.

    The reason has to do with the fact that sharpness of a system (as described by an MTF curve) is equal to the sharpness of each of the system's individual components multiplied together.

    E.g., MTF of sensor+lens == MTF of sensor * MTF of lens.

    This means, given a specific sensor (e.g., D800), a better lens will give you more sharpness.

    But similarly, given a specific lens (e.g., 85/1.4G), a better sensor will also you give you more sharpness.
  • TaoTeJaredTaoTeJared Posts: 1,306Member
    The resolution limits of the sensor vs lens has been debated ad nauseum on multiple threads on sites all over the internte (note the article is from 2012.) (old site where it was just the topic) Simply stated, we are far away from sensors out-resolving good to great optics. This is shown with even older Leica and Zeiss lenses from 40yrs ago still resolve up with the best lenses today.

    DxO is doing quite a bit of verbal manipulating/massaging/dissimulating with it's descriptions of the results to convince people their self designed tests are not as bad as they really are. 45% of megapixels are not lost, nor is resolution lost with most modern lenses of the same build class. (Take a holga modified lens (with plastic glass) put it against a pro lens, yes there is a difference.)

    "DX lens which have generally inferior sharpness to FX lenses"
    This is inaccurate or probably better put, an incomplete statement. The variable in that is the "build quality" (i.e. consumer glass vs pro-type glass.)
    ...current lenses do not come anywhere close to achieving the theoretical diffraction limited sharpness that can be achieved at faster F-stops.
    I do not understand what that means. "theoretical diffraction limited sharpness that can be achieved at faster F-stops" I have never heard that phrase used "for faster f-stops".
    D800, D300, D50(ir converted), FujiX100, Canon G11, Olympus TG2. Nikon lenses - 24mm 2.8, 35mm 1.8, (5 in all)50mm, 60mm, 85mm 1.8, 105vr, 105 f2.5, 180mm 2.8, 70-200vr1, 24-120vr f4. Tokina 12-24mm, 16-28mm, 28-70mm (angenieux design), 300mm f2.8. Sigma 15mm fisheye. Voigtlander R2 (olive) & R2a, Voigt 35mm 2.5, Zeiss 50mm f/2, Leica 90mm f/4. I know I missed something...
  • snakebunksnakebunk Posts: 993Member
    edited January 2014
    People use Nikon 1 cameras on FX tele lenses with good result and therefore I guess the pixel count could reach at least 100 MP on FX cameras before it's unnecessary to add more pixels (just a rough guess though).

    Also, the color of a single pixel is calculated not only with the value of that pixel but also with the values of neighboring pixels (each pixel is only sensitive to one color). Therefore, "too many" pixels may help in post production algorithms to get the colors right when the image is downsampled.
    Post edited by snakebunk on
  • kenadamskenadams Posts: 222Member
    edited January 2014
    I said that a while ago on a similar thread: I recently re-read Thom Hogan's test of the D80 (which I have). "80", mind you, not "800". He said that with the new 10MP sensor, cameras might be getting in a range where they could start to outresolve a lens. That was back in 2007, on a - let me repeat that - 10MP sensor. I've never heard any complaints in that regard.

    All the while there's a thread on the Fred Miranda forum where people with old manual AI lenses keep producing the most amazing images.
    Post edited by kenadams on
  • sevencrossingsevencrossing Posts: 2,800Member
    edited January 2014
    If so, what is the point of more megapixels?

    well there is "My camera has got more mega whatnots then yours" :)

    which why people are getting hot flushes over a D4x


    Post edited by sevencrossing on
  • henrik1963henrik1963 Posts: 567Member
    edited January 2014
    As I understand it: If you have a not-so-good lens and use it on a D800 you will get better image quality than you would have, using that lens on a D700. This means that the quality of the sensor will have an impact on the resulting image quality regardless of the lens used.
    Post edited by henrik1963 on
  • henrik1963henrik1963 Posts: 567Member
    Following PitchBlack - in the real world - if you have a D700 + a dog of a lens you might be better of buying a better lens before moving to a D800. A bad lens is a bad lens regardless of minor gains in resolving power due to a better sensor.
  • MikeGunterMikeGunter Posts: 543Member
    Hi all,

    Take note that film is continuous tone. There's no pixels. The transition from pure black to white is step-less, sort of.

    That's part of the problem for me in comparing analog lenses with mega-pixels and analog film.

    My best,

    Mike

  • IronheartIronheart Posts: 3,017Moderator
    Folks, This question comes up every time a new high-watermark for megapixels is reached. The answer is an unequivocal "No". Lenses are analogue devices, some are sharper than others for a variety of reasons, but we are not even close to being limited "by the glass".

    Here's one proof point, the 41mp Nokia Lumina 1020. If that 41mp sensor were produced in a DX size with the same pixel pitch it would be a 283mp sensor. In an FX size that would be 666mp.

    Now do we think the Zeiss lenses in front of that 41mp sensor are somehow limiting the ability to capture images? No, they are not. Is Nikon worried about a 54mp sensor somehow "out resolving" their high-end glass? No. Would they worry about a 600mp sensor in the same vein? Probably not. At this point though, pixel binning start to make sense as it will increase ISO sensitivity, DR, and reduce noise. You probably don't want the 8.5GB raw file such a sensor would produce.

    Something else to think about, that 41mp sensor has a pixel size of 1.1 micrometers or microns, the D7100 has more than 3.5 times greater at 3.9microns. Modern chip FAB techniques can produce feature sizes measured in tens of nanometers or 1000x smaller than the pixel on that Nokia. The real limiter may be the size of the wavelength of visible light 390-700 nanometers. If an FX sized sensor were produced with 700 nanometer pixels, it would be a 1700mp sensor. My prediction is this is the real limit of sensor megapixels, it has nothing to do with glass, which is really just focusing and transmitting the light to the sensor in the analogue realm.
  • TaoTeJaredTaoTeJared Posts: 1,306Member
    As I understand it: If you have a not-so-good lens and use it on a D800 you will get better image quality than you would have, using that lens on a D700. This means that the quality of the sensor will have an impact on the resulting image quality regardless of the lens used.
    I have an old 24-85 that no one will take off my hands so it just sits there. It looked fine on my D300 but on the D800 it looks worse - much worse. That is pixel peeping though - printed image I'm sure it would look fine.

    D800, D300, D50(ir converted), FujiX100, Canon G11, Olympus TG2. Nikon lenses - 24mm 2.8, 35mm 1.8, (5 in all)50mm, 60mm, 85mm 1.8, 105vr, 105 f2.5, 180mm 2.8, 70-200vr1, 24-120vr f4. Tokina 12-24mm, 16-28mm, 28-70mm (angenieux design), 300mm f2.8. Sigma 15mm fisheye. Voigtlander R2 (olive) & R2a, Voigt 35mm 2.5, Zeiss 50mm f/2, Leica 90mm f/4. I know I missed something...
  • haroldpharoldp Posts: 984Member
    Hello All, please read the following:

    And the D7100 is a DX camera optimized for use with DX lens which have generally inferior sharpness to FX lenses.

    While diffraction is only a serious limitation on lens sharpness at F/8ish or slower, current lenses do not come anywhere close to achieving the theoretical diffraction limited sharpness that can be achieved at faster F-stops.

    I would appreciate hearing all of your views and thoughts?
    As a type, DX lenses have a smaller light circle than FX lenses. Since all design parameters influence trade-offs, at the same build cost and engineering competence, a DX lens might out resolve an FX lens of the same focal length since it has to correct for a smaller image circle.

    I have most of the best Nikon and Leica lenses, and my best Olympus 4/3 lenses out resolve them slightly but corrected for 1/4 of the image circle.

    As was mentioned elsewhere in this thread, most Nikon DX lenses are marketed and priced as 'kit' type lenses and built to a lesser standard.

    Nikon's 17-55mm f2.8 is a DX lens, and is as good as a fast zoom gets, but it is priced like a professional lens.

    I have Fuji X series DX lenses which are easily the equal of my Leica and Nikon pro optics.

    Nikon's (and probably Canon's) super tele's (like my 400 2.8 G) are sharpest wide open which indicates that they are already diffraction limited even at f4 (possibly 2.8).

    The new generation of lenses (Leica APO Summicron, Zeiss Otus etc.) are designed to challenge sensors of much higher resolution than are available today commercially. Military and industrial optics and sensors are in a separate class entirely.

    Sorry for rambling.


    .... H




    D810, D3x, 14-24/2.8, 50/1.4D, 24-70/2.8, 24-120/4 VR, 70-200/2.8 VR1, 80-400 G, 200-400/4 VR1, 400/2.8 ED VR G, 105/2 DC, 17-55/2.8.
    Nikon N90s, F100, F, lots of Leica M digital and film stuff.

  • Golf007sdGolf007sd Posts: 2,840Moderator
    @Ironheart: +1 Well said buddy :)
    D4 & D7000 | Nikon Holy Trinity Set + 105 2.8 Mico + 200 F2 VR II | 300 2.8G VR II, 10.5 Fish-eye, 24 & 50 1.4G, 35 & 85 1.8G, 18-200 3.5-5.6 VR I SB-400 & 700 | TC 1.4E III, 1.7 & 2.0E III, 1.7 | Sigma 35 & 50 1.4 DG HSM | RRS Ballhead & Tripods Gear | Gitzo Monopod | Lowepro Gear | HDR via Promote Control System |
  • michael66michael66 Posts: 231Member

    Take note that film is continuous tone. There's no pixels.
    Hmmm... I thought the grains of film were directly analagous to pixels. In other words, a 35mm ( 24x36 ) frame of fine grained 100 ISO film is about 15MP.
  • SymphoticSymphotic Posts: 711Member
    I've thought about this a lot. My vendor calculates MTF curves for our video cameras and we are not yet limited by the lenses, but by the sensors. We don't make still cameras and I don't know how to access the pixel data on the D800 or any other still camera to calculate MTF using our system, but my colleague and I have gone around with this quite a bit. We have found that you use point and shoot technique with your D800, you don't get great results, but you don't get great results with the D700 either. Once you have focusing and stabilizing technique down, the differences are significant, and more megapixels allow for resolving useful fine detail not available with lower MP cameras. Better lenses yield better results, but there are no lenses we have tried where we can't see a difference.

    This is important in his work as his photos end up in court cases. For this reason he has pretty much stopped using his D700.

    Your mileage may vary: we both like having lots of megapixels. (I am not a sports, fashion, or wedding photographer.) If I could justify buying a D4, I would. I almost did, but the D800 is very useful with all those megapixels.

    I scratch my head trying to figure out the DXO mark scores: It is possible to calculate an MTF curve for a lens by oversampling the imager results, allowing one to cancel out the MTF of the imager, which would otherwise be limited by the partially sampled pixels where the knife edge image falls across them. (That's why those bars the measure MTF with are long and slanted.) But I never really figured out what DXO mark is doing.
    Jack Roberts
    "Discovery consists in seeing what everyone else has seen and thinking what nobody else has thought"--Albert Szent-Gyorgy
  • AdeAde Posts: 1,071Member
    For sure DXO also oversamples to calculate MTF. It's typical in photography to have 4x oversampling by using a binning algorithm, on slant-edges as you say.

    But what they are reporting is their own "perceptual megapixel" metric. I speculate -- but don't know for sure -- that they are calculating their own model based on SQF (Subjective Quality Factor). The vanilla SQF is described here:

    http://www.bobatkins.com/photography/technical/mtf/mtf4.html

    Today we say that SQF is "acutance" but you may have a different definition for acutance in your field. SQF is basically the integral of the MTF over some defined range corresponding to human eye perception.

    Tools like Imatest can be used to measure both MTF and SQF from your D800.
  • SymphoticSymphotic Posts: 711Member
    Thanks, Ade.
    Jack Roberts
    "Discovery consists in seeing what everyone else has seen and thinking what nobody else has thought"--Albert Szent-Gyorgy
  • PistnbrokePistnbroke Posts: 2,443Member
    I made this point to tom hogan and this is what he replied to me ...

    lenses already exceed the ability of the camera; the implied diffraction
    of very high pixel counts means that a lens that can produce an MTF of .9+
    is already getting only .7 to .8 recorded on the 24mp DX cameras and the
    D800.

    there you go God has spoken.
  • WestEndBoyWestEndBoy Posts: 1,456Member
    Here is an interesting article relevant to many of the points debated in this thread:

    http://photographylife.com/dx-or-fx-for-sports-and-wildlife-photography
  • donaldejosedonaldejose Posts: 3,675Member
    jshickele: just read that article; makes me think I should not buy a D400 if one comes out and just concentrate on shooting FX. considering all the things Thom says if the top of the DX line and the bottom of the FX line meet at the same price point, the better buy will be the bottom of the FX line.
  • roombarobotroombarobot Posts: 201Member

    #4 on this article also states well the often misunderstood FX and DX depth of field vs field of view relationship.

    But, what Nassim ends with, is the truth, particularly for me. "At the end of the day, however, keep something else in mind – any camera, whether DX or FX is capable of producing excellent results. It is not the gear, it is the guy behind the camera :)"

  • WestEndBoyWestEndBoy Posts: 1,456Member

    #4 on this article also states well the often misunderstood FX and DX depth of field vs field of view relationship.

    But, what Nassim ends with, is the truth, particularly for me. "At the end of the day, however, keep something else in mind – any camera, whether DX or FX is capable of producing excellent results. It is not the gear, it is the guy behind the camera :)"

    Well said. A point that is often overlooked in the quest for gear. I love my D800, but often find myself impressed with the results of my Coolpix A with a DX sensor.
  • WestEndBoyWestEndBoy Posts: 1,456Member
    jshickele: just read that article; makes me think I should not buy a D400 if one comes out and just concentrate on shooting FX. considering all the things Thom says if the top of the DX line and the bottom of the FX line meet at the same price point, the better buy will be the bottom of the FX line.
    I tend to think you are correct. It seems to me that unless we are comparing cameras from more then two generations apart, FX always trumps DX. It is why I have no faith in Micro 4/3rds or CX.

    However, if the D400 ever does come and it has the specs and features that people are asking for, I might understand why somebody might choose a D400 over a D610, though I would still choose a D610 myself.
Sign In or Register to comment.