High ISO noise CCD's vs CMOS

Discussion in 'Digital SLR' started by mswlogo, Aug 14, 2006.

  1. mswlogo

    Frank B Guest

    Frank B, Aug 16, 2006
    1. Advertisements

  2. mswlogo

    Bill Guest

    That's great, but the problem is that the discussion was relating to the
    performance of the 10mp sensors from Nikon and Sony, such as the D200
    and A100 where high ISO performance is not as good as the D50.

    In fact, the D50 has some of the best noise levels of any current DSLR.
    Bill, Aug 16, 2006
    1. Advertisements

  3. mswlogo

    Isaiah Beard Guest

    I used to also be a big believer in Nikon until I saw similar results.
    True, the Nikon has 2 more megapixels, but this is a textbook example of
    how, at 8 megapixels and above, the resolution alone isn't everything.
    I would easily conclude that despite having a lower pixel count, I'm
    getting *more* usable image data from the 30D than I would be from the D200.
    Isaiah Beard, Aug 16, 2006
  4. mswlogo

    Frank B Guest

    Frank B, Aug 16, 2006
  5. mswlogo

    RichA Guest

    Looking at those two images, the little $599 Nikon kicked the $1200
    Canon's a--.
    RichA, Aug 17, 2006
  6. Yes, read noise is random. CCDs commonly have the 7-15 electron read
    noise, even in professional cooled CCDs used in astronomy.
    It was amazing to see Canon improve CMOS so much, which 5+
    years ago many said CMOS would never match CCD performance.
    I do not know the specifics of how they achieved that.
    No, I have not seen that either.
    It could be an interesting experiment.

    Roger N. Clark (change username to rnclark), Aug 17, 2006
  7. mswlogo

    Socrates Guest

    I'm reading ISO 3200 with Panda exif reader on the Nikon file, this on top
    of the photo being underexposed, so it is not a fair comparison.

    Socrates, Aug 17, 2006
  8. mswlogo

    ColinD Guest

    You're talking out of yours ...

    Colin D.
    ColinD, Aug 17, 2006
  9. SNIP

    All it would take is a white noise (all frequencies equally
    represented) target (could be made with Photoshop). Outputting it as
    an inkjet image or on an LCD might introduce it's own noise spike, but
    printing it on photopaper could come a long way. One could also buy
    such a target:
    Some of he math required for quantification could be derived from:
    <http://www.sinepatterns.com/docs/Random Target MTF Engineering.pdf>

    One would have to compare the actual result to that of theoretically
    perfect area sampling with a given fill-factor and sampling density,
    or to that of another camera for a relative (instead of an absolute)

    A program like Imatest also allows an analysis of the noise spectrum,
    e.g. by using a shot of a grayscale steptablet;
    <http://www.imatest.com/docs/tour_q13.html> , or as part of an SFR/MTF

    And for some more background info:
    Bart van der Wolf, Aug 18, 2006
  10. Thanks for the links.

    White noise would be sufficient to evaluate what happens assuming the
    processing is "dumb" (ie doesn't react differently to the signal). But
    what if it's not? If I had means of producing reliably targets and
    analyzing the results (well, I suppose Mathematica will be ok for the
    analysis) I'd give it a try (with both white noise and noise with
    various spectra, to see how it behaves, and from what I find, think
    about what to do next).

    Do you have any suggestions for how I can reliably produce a target? I
    can produce it easily in the computer, my question refers to getting it
    to print or displaying it. Sources of extraneous noise? I have zero
    experience in practical matters here, but I'm on vacation for two weeks
    and this sounds an interesting thing to try.
    How does the fill factor come into it? Are you referring to the effect
    of the sampling density/fill factor combination on the sampled signal,
    or something else? I supose also that the actual camera noise will have
    to be taken into account. Maybe at low ISOs it can be ignored. And how
    about the CFA? This is starting to sound like a serious project...
    achilleaslazarides, Aug 19, 2006
  11. This image looks out of focus. Noise level is much lower than
    d200,though. Here is an in focus 3200 ISO 30D shot from Steves Digicam. Compare it
    to the 1600 ISO D200 image

    Just Plain Bill, Aug 20, 2006
  12. But ... sometimes there is not enough light to expose
    "properly". Sometimes even with the f/2.8 super lens, wide
    open, there is not enough light at a shutter speed short
    enough to get rid of subject movement.

    Then sensor noise matters. You get a seriously underexsposed
    picture, but you can of course post-process it to get the
    shading "right" ... and you get noise. Effectively, this is
    like setting the camera at ASA 30000 or more. Both CCD and
    CMOS are linear at low light levels, so it actually works
    well ... noise excepted, of course.

    Doug McDonald
    Doug McDonald, Aug 21, 2006
  13. That's what the
    24mm f/1.4L
    28mm f/1.8
    35mm f/1.4L or f/2
    50mm f/1.2L[1] or f/1.4 or f/1.8
    (or f/1.0, if you find it, have the budget and want to go crazy)
    85mm f/1.2L of f/1.8
    100mm f/2
    135mm f/2
    are for ...

    I imagine Nikon has a similar product palette.


    [1] Yep, it's a new one, but you'll have to wait a few more months.
    (and the 70-200mm f/4 IS is to be found there too.)
    Wolfgang Weisselberg, Aug 25, 2006
  14. But ... sometimes there is not enough light to expose
    True, but:

    a) not everyone has the resources to own these lenses
    b) the DOF at apertures below 2 is so shallow that this might not yield
    acceptable pictures of some scenes

    Not that you were claiming otherwise, and I've lost track of exactly
    what the point was suppsoe to be here, but since in another context,
    there was argument over the value of high ISO settings compared to
    faster lens, I thought this is worth mentioning. I can understand some
    poeople not caring about high ISO for themselves. But to argue that
    others should have reason to care, well, that's just foolish.

    Marc Sabatella

    Music, art, & educational materials
    Featuring "A Jazz Improvisation Primer"
    Marc Sabatella, Aug 25, 2006
  15. mswlogo

    AaronW Guest

    He did mention 35/2, 50/1.8, and 85/1.8, which are very affordable,
    very good quality, adequate DoF, and still 1 stop faster than much more
    expensive f/2.8 zooms.
    Both ISO 1600 and f/2 primes are very useful.

    AaronW, Aug 25, 2006
  16. $90 for the 50mm f/1.8 is certainly not to high for
    multi-hundred $ DSLR owners. And you _can_ rent lenses (and
    bodies and flashes), too.
    True, but it WILL help focussing (manually and AF) even in the
    dark. And "What did you expect? Sshooting in available darkness
    usually means a shallow DOF and low contrast and sharpness
    wide open!"

    And if high ISO and fast lens won't be acceptable, pushing doesn't
    help and flash or more light is impossible, tripods not being an
    option due to moving objects ... well, then you are out of luck;
    you'll have to re-create the setting in a studio or where more
    light is available.

    Wolfgang Weisselberg, Aug 25, 2006
  17. Probably not. But a whole collection of them at different focal lengths
    may well be.
    I never said high ISO wouldn't be acceptable; on the contrary, I'm
    trying to argue it is necesary and indeed often preferable. As I said,
    I kind of lost track of the various points being made in this thread,
    but it *appeared* your lens listing was trying to demonstrate that high
    ISO wasn't necessary - just spend a few thousand dollars and carry
    around huge bag of lenses. That's probably not what your point really
    was, but it kind of came off that way to me.

    Marc Sabatella

    Music, art, & educational materials
    Featuring "A Jazz Improvisation Primer"
    Marc Sabatella, Aug 25, 2006
  18. mswlogo

    AaronW Guest

    I have a lot of AF error with 50/1.4, while 50/1.8 AF much better.
    Under good conditions, the 50/1.4 AF fine, so the lens is not
    defective. I think it is because AF is done at wide open, and 50/1.4
    wide open at f/1.4 is softer than 50/1.8 at f/1.8. The softer image
    makes it more difficult for AF sensors to work correctly.

    AaronW, Aug 25, 2006
  19. True, but still, I don't see the existence of these as a substitute for
    good high ISO performance. That's a limited range of focal lengths.
    Once again, I'm not sure that Wolfgang *intended* to suggest the
    existence of these lenses made high ISO performance irrelevant, but
    taken somewhat out of the context of the rest of thread, it appeared
    that way.

    Marc Sabatella

    Music, art, & educational materials
    Featuring "A Jazz Improvisation Primer"
    Marc Sabatella, Aug 25, 2006
  20. True, but start with a cheap, optically good lens and see what
    you need and if it appeals to you.
    High ISO is always preferrable to digital pushing (unless the
    camera does aught than digital pushing itself and you don't
    need any off-the-camera JPEG).

    I was responding to Doug McDonald: "Sometimes even with the f/2.8
    super lens, wide open, there is not enough light at a shutter
    speed short enough to get rid of subject movement."
    (Message ID: <ecdam3$m9e$>)

    I was trying to show that f/2.8 was _not_ the lower end at all.
    And by extension: if you find yourself often in places where a
    f/2.8 is too slow, consider a faster lens of matching focal
    length! (there are many that are cheap compared to your
    average zoom lens.)

    Wolfgang Weisselberg, Aug 28, 2006
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.