Page 1 of 3 123 LastLast
Results 1 to 10 of 26

Thread: There must be a limit somewhere

Hybrid View

Previous Post Previous Post   Next Post Next Post
  1. #1
    45-57-617
    Join Date
    Oct 2006
    Location
    Toowoomba, Queensland
    Posts
    644

    There must be a limit somewhere

    Hello to you all.

    I genuinely want to discuss an issue relating to the current state of play w.r.t digital sensor technology.

    For several months / years I have been watching the increase in the abilities of the digital camera scene and I wonder ...

    Is it true that we are pretty much at the smallest pixel size on a sensor now ? Is it not true that going too much smaller than the size of pixel on a say 16M full-frame DSLR leads to worsening ISO performance and increasing noise ?

    Without the figures in front of me, what is the Nikon D3 sampling each 'pixel' at ? Is it 12 bits per 'pixel' ? So is that 12 bits per red, per green and per blue ? A total of 36 bits per 'pixel' ?

    IF (note the capitals) one were to sample at 48 bits per pixel would one not need a larger voltage to avoid excessive noise ? To get a larger voltage one needs a larger pixel right ? So if one were to sample at 48 bits per pixel one would start to max out in noise and ISO at what ? 6M pixels for a full frame 35mm sensor ?

    Yes I do know my physics enough to know there is a limit. I just think we are pretty much at that limit. To get the colours of my scanner the camera needs to sample at 48 bits per pixel. To get this I need perhaps a 6M pixel full frame DSLR. 6M pixels would be pretty crappy right ?

    I know my 2.4GHz pc is nearly 10 years old and the new ones are around 3.2GHz. Moore's Law is breaking down is it not? So too I think in the digital photography realm.

    Nice replies only !

  2. #2
    Steve Smith's Avatar
    Join Date
    Jan 2011
    Location
    Isle of Wight, near England
    Posts
    707

    Re: There must be a limit somewhere

    Quote Originally Posted by Steve McLevie View Post
    Is it true that we are pretty much at the smallest pixel size on a sensor now ?
    It is my view that both film and digital are at the level where the laws of physics are the limiting factor. I also think that for the same format size, the resolution is now roughly equal.


    Steve.

  3. #3

    Join Date
    Oct 2003
    Location
    Westport Island, Maine
    Posts
    1,236

    Re: There must be a limit somewhere

    My son, pretty much an expert on this stuff, tells me that the Hot Area is sensitivity - high ISO without noise, rather than stretching the limits on pixels. That's what he loves about his D3s, which if memory serves is 12MP.

    He works for filmmaker Ken Burns (Dad gets to brag), and last night was talking about the Red video camera, which can run 250 frames per second, with each frame a 14 megapixel image. He thinks it will revolutionize things like fashion and journalism photography, since there apparently is a way to tag frames in the data stream before it leaves the camera. Others can then take that frame, immediately export it, and it's ready to go - anywhere. Faster than anything before, and there's full-motion video available, too. In any event, he's impressed.

    I couldn't muster the gumption to tell him about my development time tests for sheet film...

    I certainly would not count Mr. Moore out just yet. The path is littered with the bodies of those who did.
    Bruce Barlow
    author of "Finely Focused" and "Exercises in Photographic Composition"
    www.brucewbarlow.com

  4. #4

    Join Date
    Jan 2002
    Location
    Besançon, France
    Posts
    1,617

    Re: There must be a limit somewhere

    Is it true that we are pretty much at the smallest pixel size on a sensor now ?

    I'll refer to medium-format sensors which directly compete with (or, for some people, outperform) large format on film.
    If we consider a silicon sensor in the class of 22 Mpix, the pixel grid has a pitch of about 9 microns (example : Leaf Aptus 22)
    The smallest periodic object that can be detected with such a sensor has a pitch of 2x9 = 18 microns.
    This is still bigger than the actual capability of the best 'digital' view camera lenses. At the centre of the field, those lenses are close to be diffraction-limited when stopped down to f/11, f/8 being the recommended aperture for top-notche lenses designed for the 4.5x6cm format on silicon.
    At f/11 for visible light, the smallest periodic feature passing through the lens has a period of about 11x0.7 = 8 microns. Hence a 22 Mpix sensor with 9 microns of pixel pitch fails by a factor 2 to correctly sample the analog-optical image delivered by a top-notch lens @f/11.
    With a 80 Mpix sensor of about the same size, roughly 4.5x6 cm, the pixel pitch is divided by about 2, and now those sensors have a sampling capability that matches what a diffraction limited lens @f/11 can deliver.
    With a 22 Mpix sensor and top-notch lenses, some moiré (aliasing) patterns can be visible in the final image, in certain shooting situations, for certain subjects like fabric or other fine-pitched periodic objects.
    With a 80 Mpix sensor and state-of-the art lenses, the probabilty to get annoying moiré effects is very low. And there apaprently is no benefit to use a finer grid that what is necessary to sample a diffraction-limited image.

    Regarding noise, dynamic range and sensitivity, a quick look at the technical specs of a Kodak sensor, for example this one
    http://www.kodak.com/global/plugins/...00LongSpec.pdf will give us some ideas.
    We find that in this sensor, the pixel can store a maxium of 60,000 electrons. The natural statistical fluctuation of this figure is about 250 (the square root of 60,000)
    If we consider that the smallest gray level detectable is about 256 (this is a very rough approach, but we just want to have the proper order of magnitude), it means the the maximum number of distinct gray levels for 60,000 electrons is only 256, such a monochrome image can be coded on 1-byte only (8-bit).
    Of course you can accumulate images in your computer, there is no limit, but for a single-shot, the maximum number of electrons per pixel is a limiting factor. And you cannot get more electrons than incoming photons per pixels. This is not as obvious at it might seem actually.

    In terms of noise, it is very difficult to compare film with a digital sensor, but people involved in astronomical photography have done this kind of comparison. If we try to imagine that film behaves like a photon counter, in very weak ligh levels, we get for our beloved Tri-X flm an equivalent quantum efficiency of about 0.5%. it means that a photon counter that would miss 99.5% of the photons is as efficient as a Tri-X silver halilde film in terms of equivalent noise.

    The Kodak specs quoted above mentions a quantum efficient in the range of 20 to 30%.
    An efficiency above 80% has been achieved for monochrome silicon detectors in astrophysics, but the mere fact that we have to filter each pixel for RGB shots implies that each color pixel has to loose about 66% of the incoming white-light photons !
    So in terms of what can be expected for a monochrome image, Tri-X at 0.5% of efficiency versus silicon at 80%, we actually reached the limits for film at the end of last century, and we are close to the absolute limits for silicon.
    So the only solution is to use BIG pixels capable of counting zillions of electrons before saturating
    And big pixels imply big sensors, back to the 4x5" (silicon) format in the future ??

  5. #5

    Join Date
    Oct 2009
    Location
    San Mateo, California
    Posts
    742

    Re: There must be a limit somewhere

    I do not think we are at the limit yet. I know that the surface of the sensor is not 100% filled with photosites, and there is more sensitivity to be had. Probably have a few more generations of sensor before the physical limit of gathering photons is achieved, but there is more that can be gained by other things which may be possible with software and multiple exposures. I also think that the 35mm DSLR is a transitional camera and will go the way of TLRs sometime in the next decade.

  6. #6

    Join Date
    Dec 2007
    Location
    Pennsylvania
    Posts
    140

    Re: There must be a limit somewhere

    I don't think Moore's law is breaking down. As the technology advances the chips/sensors change so that 2.4GHZ doesn't necessarily equal 2.4GHZ. I mean that a 10MP sensor from 4 years ago and one from 1 year ago are two very different beasts, probably mostly in the high ISO limit, so they are both 10MP but their similarity ends at the name.

    It seems to me that as we approach the limit of the current chip/sensor technology there is usually a new design on the horizon that will allow for further expansion or as Bruce refered to above a new direction that innovation takes. The high ISO race is on now and that's great for consumers. I give it 2 years before we have compacts with ISO 6400 performance that the current full frame cameras can only dream of. This will have the added benefit of cheap FF DSLR's for the rest of us.

    It's the same trickle down you see in automotive circles. The current Ferrari Grand Prix race cars may have a braking system that cost 2.5 million dollars to develop but once it's developed and then a version is made from consumer grade parts, you wind up with a race designed Ferrari braking system on your Ford Taurus. Everyone wins.

  7. #7

    Join Date
    Jul 2007
    Location
    Austin TX
    Posts
    2,049

    Re: There must be a limit somewhere

    Emmanuel, those are nice comments on the technology at current state of the art. But I'm not clear on the point of quantum efficiency of film vs sensor.

    You are quoting efficiency of Tri-X at 0.5%. Is that during exposure or during development which results in a cascade (multiplication) of silver in the emulsion? And of course that multiplication is a form of noise since it is not directly related, spacially, to the original photon flux site.

    The quantum efficiency of silicon CMOS sensors is probably near the limit due to the purity of material and perfection of the junction parameters so little further gain can be achieved by reducing recombination rates. There is a possibility of increasing the number of electrons per incoming optical photons by employing avalanche multiplication within the high field junction depletion region. Such avalanche multiplication is analogous to PMT gain and in a way to the multiplication of silver in film (physics is different) but fraught with an increase in noise.

    As you alluded to, the easiest way to improvement in dynamic range is larger pixels with larger sensors. But that is not compatible with a high volume consumer market.

    Nate Potter, Austin TX.

  8. #8
    Abuser of God's Sunlight
    Join Date
    Aug 2004
    Location
    brooklyn, nyc
    Posts
    5,796

    Re: There must be a limit somewhere

    We're not anywhere near the limit of could theoretically be done with silicon technology, but I suspect we're close to the limits of what makes practical sense with pixel density. Sensors in better quality consumer DSLRs are down to around 4.8 micron pixel pitch. This allows 80 lp/mm to be resolved at a very high MTF. This is well in excess of what any large format lens can resolve, and is pushing the limits for any lens.

    I think we're going to see the quality of these sensors continue to improve, in terms of noise, color, and dynamic range, and we'll see the finer pixel pitches migrate to the larger sensors. But I don't think we're going to see pixel densities continue to increase that much past where it already is in the small sensors, unless marketing plays a big role.

    Ok, one reason I might be wrong: if the pixels are small enough to allow a high degree of oversampling, manufacturers could get rid of anti-aliasing filters and get a significant performance bump. They've already eliminated them on some of the MF sensors (in those cases I don't know how they didi it).

  9. #9
    Mike Anderson's Avatar
    Join Date
    Jan 2010
    Location
    San Diego
    Posts
    681

    Re: There must be a limit somewhere

    Here's one smart person's take on the laws of physics and the limits sensitivity:

    http://theonlinephotographer.typepad...-of-light.html

    To summarize, Ctein says there's room for a 10 times improvement in sensitivity:

    What the Nikon D3S can do at ISO 6400, these technologies would let it deliver at 64,000
    I'm no scientist but I always thought you could trade sensitivity for pixel density (common sense, right ), and if that's true there's still a lot of room for improvement in pixel density.

    ...Mike

  10. #10

    Join Date
    Oct 2009
    Location
    San Mateo, California
    Posts
    742

    Re: There must be a limit somewhere

    Quote Originally Posted by Mike Anderson View Post
    Here's one smart person's take on the laws of physics and the limits sensitivity:

    http://theonlinephotographer.typepad...-of-light.html

    To summarize, Ctein says there's room for a 10 times improvement in sensitivity:
    10 times is just a bit more than three stops.

Similar Threads

  1. Camera without swings? Does it limit too much?
    By jvuokko in forum Style & Technique
    Replies: 16
    Last Post: 25-Dec-2010, 18:30
  2. Beating the diffraction limit
    By dh003i in forum Lenses & Lens Accessories
    Replies: 18
    Last Post: 3-Sep-2010, 14:42
  3. Limit image circle to only usable area?
    By Darin Boville in forum Lenses & Lens Accessories
    Replies: 7
    Last Post: 25-Nov-2009, 12:58
  4. Diffraction Limit on Macro Lenses
    By DolphinDan in forum Lenses & Lens Accessories
    Replies: 36
    Last Post: 27-Oct-2009, 07:30
  5. Removing stop limit from 150 G-Claron barrel
    By john wilton in forum Lenses & Lens Accessories
    Replies: 0
    Last Post: 16-Feb-2009, 22:07

Bookmarks

Posting Permissions

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