Excellent tests Sony A350

Discussion in 'Sony' started by Focus, Jun 8, 2008.

  1. Focus

    Alan Browne Guest

    My statement shows that I grasp it quite well.

    Despite that it is also clear that Sony have some work to do on their
    noise reduction algorithms (or better, preventing noise in the first
    place) v. Canon and Nikon.

    It's also germane to note that the sensor in question (-350) is a CCD
    whereas most of the other high density DSLR sensors are now CMOS.

    It is curious that Sony chose CCD. Is it cheaper to manufacture?
    Are you interested in the discussion or in what you've said elsewhere?

    A 16 Mpix back for my Hasselblad is still in the $10K range. Even
    used, they are extremely expensive.

    A higher Mpix 'blad (H3D) is even more prohibitive... at 22 Mpix: $25K.
    At 39 Mpix: $44K (with lens, VF, etc. etc.).

    Mamiya offer a tempting 22 Mpix kit at $10K with lens. But that doesn't
    leverage my Hasselblad lenses very well.

    No MF digital for me ... for now.
     
    Alan Browne, Jun 10, 2008
    #21
    1. Advertisements

  2. Binning happens before reading.
    Hence only one read and only once the read noise --- over a much
    stronger signal.
    Huh?

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 11, 2008
    #22
    1. Advertisements

  3. Wolfgang Weisselberg wrote:
    []
    Are you sure about binning being done on-chip? I would appreciate a
    reference.

    Thanks,
    David
     
    David J Taylor, Jun 11, 2008
    #23
  4. Of course. I didn't suggest otherwise. I said that it is not necessary
    to reduce the pixel count of an image to reduce noise. IOW, you can
    upsample both so that they're the same size, and view from a distance.
    Downsampling one to match the other may be fair in terms of noise, but it
    is not fair to the higher-MP image in terms of resolution, and the
    ability to distinguish signal from noise at the higher frequency.
    By "ruining an image" I am talking about losing resolution. It does not
    mean the the result looks bad at the pixel level, compared to nothing
    else. Look side by side at the original and the downsampled or binned
    version at the same image or subject size, and it is obvious how
    undetailed the latter is (unless you have oversampled the optics by a
    good margin).
     
    John P Sheehy, Jun 12, 2008
    #24
  5. "David J Taylor" <-this-bit.nor-this-
    bit.co.uk> wrote in
    I don't know if you're being sarcastic here or not, but if you oversample
    the lens and its defects, it is much easier to correct the defects in
    software, and you'll always have a better starting point, and an image
    that is relatively immune to artifacts caused by rotating, perspective
    correction, etc, and you can freely resample to any resolution you want
    with any algorithm with less artifact as well. Depending on neighboring
    pixels for maximum contrast only leads to artifact.

    Oversampling is good; the storage requirements aren't, though, with
    current RAW storage methods.
     
    John P Sheehy, Jun 12, 2008
    #25
  6. I agree with you about oversampling (I have mentioned it before on this
    newsgroup), but I suspect that significantly better lenses will be
    required if people are to take the maximum advantage of significantly
    higher resolutions.

    Whether there is any net benefit for the majority of photos which people
    take and display is another matter! <G>

    David
     
    David J Taylor, Jun 12, 2008
    #26
  7. Focus

    Mr.T Guest

    Was the part saying you can have one *OR* the other (low noise or high
    pixels) when you start with more pixels, but only one (less noise) if you
    start with less pixels, beyond you?

    If all you want is low noise, stick with a 2Mpixel sensor. Not that you'll
    find a new one anymore, it doesn't seem anyone else wants them.
    For the same reason you can probably get one cheap S/H :)

    MrT.
     
    Mr.T, Jun 12, 2008
    #27
  8. Focus

    Mr.T Guest

    So do you disagree (with the part you snipped) that more pixels to start
    with gives you a choice, higher res AND more noise, or same res and same
    noise.
    (ALL ELSE BEING EQUAL, AS I SAID ALREADY MANY TIMES)
    If not, what are you arguing about?
    If so, then please state your reasons, for a change.
    That is why I said "all else being equal, which it never is!"
    I do try to anticipate argumentative types, but they just ignore what is
    written for the sake of an argument anyway.

    Do you need lessons in how to use Google groups?
    Probably difficult for you by the looks.
    Are you just interested in arguing by ignoring what has actually been said
    then, even when someone agrees with you?
    Count me out.

    So what? The choice is yours.
    I can't afford a Ferrari either, but I don't expect a Corolla to perform as
    well.
    I wouldn't waste time arguing about it though.

    MrT.
     
    Mr.T, Jun 12, 2008
    #28
  9. Focus

    Mr.T Guest

    Once again, you are *NOT* losing resolution compared to the lower res
    camera, simply not GAINING any. (for noisy subjects)
    There is a difference!
    You WILL gain resolution though, when noise is not an issue.

    If you don't actually like choice, don't choose a camera that gives you any
    :)

    MrT.
     
    Mr.T, Jun 12, 2008
    #29
  10. .... you get all that noise crap back in.
    Now, if there was no noise, it'd be different.

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 12, 2008
    #30
  11. ["Followup-To:" header set to rec.photo.digital.slr-systems.]
    Please tell me what the point of "high pixels" (I think you tried
    to say "a higher pixel count") is, if you don't get additional
    image details out of them?

    Equating "high pixels" with "high resolution" is a fallacy,
    even if you have a perfect lens.

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 12, 2008
    #31
  12. Yes (otherwise it's just downsampling, which can produce
    arbitrary sizes).
    http://support.svi.nl/wiki/PixelBinning
    http://micro.magnet.fsu.edu/primer/digitalimaging/concepts/binning.html
    http://www.ccd.com/ccd103.html
    http://www.dpreview.com/articles/compactcamerahighiso/page2.asp

    A Summing Gate is provided in each of the 8
    output sections for pixel binning if required.
    http://www.fairchildimaging.com/main/library/pdf/ccds/ccd456.pdf

    On-chip binning capability
    (Yes, that's a 12kpix * 128 pixel time-delay-integration Chip)
    http://www.fairchildimaging.com/main/documents/CCD10121RevNRExport.pdf

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 12, 2008
    #32
  13. ["Followup-To:" header set to rec.photo.digital.slr-systems.]
    Do you propose equal chip size and equal pixel size, yet
    different pixel count? If not, what is equal?

    And how do more pixels give you higher resolution, e.g. when
    you use a pinhole camera and have already more pixels than
    the pinhole can produce in resolution?

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 12, 2008
    #33

  14. I think thisn is a clear case of you assuming that I am trying to make a
    point that I am not. You have brought too much baggage into this
    exchange.

    My point is that software binning does not improve an image in any way.
    It only reduces the storage requirements, and has the illusion of looking
    less noisy at 100% pixel view.

    If you take any noisy image, and either pixelate a copy by binning it,
    reducing the size of the copy and bringing it back up to the original
    size, the original always looks more detailed, and no noisier than the
    reduced version (unless a very strong small radius sharpening is applied
    to the original).
     
    John P Sheehy, Jun 12, 2008
    #34
  15. The more you oversample, the more easily the noise (which is concentrated
    at the nyquist) is distinguished from the signal.
     
    John P Sheehy, Jun 12, 2008
    #35
  16. Thanks, but none of these seem to apply to DSLRs.

    David
     
    David J Taylor, Jun 12, 2008
    #36
  17. Focus

    Mr.T Guest

    Well equal sensor size, and same lens for a start. Or else you are simply
    comparing apples and oranges.
    If you keep ignoring the fact I've already pointed that out, I sure can't
    help you.

    MrT.
     
    Mr.T, Jun 12, 2008
    #37
  18. Focus

    Mr.T Guest

    Which has nothing to do with what I said then.

    MrT.
     
    Mr.T, Jun 12, 2008
    #38
  19. Focus

    Alan Browne Guest

    Funny, your whole "discussion" style is argumentative. (when you're not
    reminding people about what you've already said).
     
    Alan Browne, Jun 12, 2008
    #39
  20. You assume that noise is only high frequency, when it's
    white noise (aka completely random).

    Your scheme starts falling apart once you realize that even with
    recombining the 'subpixels' you incur the read noise with _every_
    single subpixel. At some point you'll no longer be photon noise
    bound, and then your s/n ratio will really suffer, whatever you do.

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 13, 2008
    #40
    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.