The Joy of Pixel Density

Discussion in 'Digital Cameras' started by John P Sheehy, Jul 15, 2008.

  1. I've made a direct comparison of RAW data per unit of area in the deep
    shadows of ISO 100 between the FZ50 (1.97 micron) and the 400D (5.7
    micron). Exposure is the same, same Av (f/4.5), same Tv (1/100), same real
    focal length (22mm), both shot at "ISO 100" pushed to ISO 13,500. Large
    crop is 100% for FZ50 (0.4MP), and small crop is 100% for 400D (0.05MP),
    and the other two are the other camera scaled to the 100% crop of each. As
    I already knew, the bigger pixels of the DSLR are inferior compared to the
    higher pixel density of the small sensor camera:
    John P Sheehy, Jul 15, 2008
    1. Advertisements

  2. John P Sheehy

    Dev/Null Guest

    Another Measurebator!
    Dev/Null, Jul 15, 2008
    1. Advertisements

  3. John P Sheehy

    Jufí Guest

    I suggest you call the NY Times or Scientific American since your expirement
    completely contradicts what every tester has shown to date. Do you think
    there might be a problem with your methodology? How about just shooting the
    same scene, same angle of view, and enlarging each image to 100%? Too

    Sorry, but larger pixels always win the day.
    Jufí, Jul 15, 2008
  4. John P Sheehy

    John Sheehy Guest

    Thank you for making sure you fully understood what was going on before
    commenting! You can go back to sleep now!

    John Sheehy, Jul 15, 2008
  5. John P Sheehy

    John Sheehy Guest

    Of course, but 2.89x, not 3x+. That's the point, or at least part of it.

    John Sheehy, Jul 15, 2008
  6. John P Sheehy

    John Sheehy Guest

    What were they testing, *exactly*? Did they have access to the RAW data,
    or just JPEGs and conversions?
    Not at all; this is as close to "all other things being equal" as we can
    expect to get, for comparing the IQ effects of pixel density in light-
    starved situations. In all probability, the lens on the 400D has nowhere
    near the MTF of the lens in the FZ50, but even a B&W checkerboard pattern
    with 1-pixel tiles would be relatively soft blown up 289%.
    No; too *irrelevant* to the issue of PIXEL DENSITY.
    Sorry, but that's wrong. With a few exceptional situations (specifically
    read noise, but not shot noise, at ISOs 1600 and above, in a very small
    number of DSLRs), smaller pixels filling the same sensor area give better
    imaging. Larger pixels are only universally better when they are doing
    the same job; IOW, when they have the same imaging responsibility.

    John Sheehy, Jul 15, 2008
  7. John P Sheehy

    Jufí Guest

    John, nothing you've posted here is correct. If you want, I'll shoot pix
    using my G9 and 5D at ISO 1600, and post the 100% crops. Resolution of the
    sensors is very close, only the size of the pixels is very different. It
    will shot that smaller pixels simply cannot compete with larger pixels.
    Unless I'm totally missing your point here...
    Jufí, Jul 16, 2008
  8. John P Sheehy

    John Sheehy Guest

    The Canon 180mm Macro, the Tamron 90mm macro, the Canon super-teles, and
    many other lenses are fully capable of using 200MP FF, now. Even for
    lenses that are more than marginally oversampled at 2 microns, the gains in
    DR and lower read noise at low ISOs is worthwhile. You just are going to
    need a lot more storage space, and give up the 11fps burst, at least at the

    John Sheehy, Jul 16, 2008
  9. John P Sheehy

    John Sheehy Guest

    I've used lenses I own on 8 and 10 MP APS DSLRs with stacks of TCs of 2.8x,
    and 4x, and gotten detail that was fragile enough to get lost downsampling
    to 70%. This means that these lenses should enjoy 8x to 16x the pixel
    density, without the TCs.

    John Sheehy, Jul 16, 2008
  10. John P Sheehy

    John Sheehy Guest

    Then the 400D would clearly be better, noise-wise, with about 60% the read
    noise of the FZ50, and about 4x as many photons collected. Optically, the
    comparison would depend on what lens was on the 400D.

    Of course, such a test would give a result already expected by almost
    anyone who knows anything at all about digital imaging, but would tell us
    absolutely nothing about the effects of pixel *density*; only about sensor

    John Sheehy, Jul 16, 2008
  11. John P Sheehy

    John Sheehy Guest

    I generally don't mind relevant cross-posting, but that comes from using
    Agent. I am using XNews now, and the method of cross-post management in
    XNews seems a bit inferior and dangerous, so I wind up seeing a lot of
    things twice.

    John Sheehy, Jul 16, 2008
  12. John P Sheehy

    John Sheehy Guest

    (Floyd L. Davidson) wrote in
    Pay attention, Floyd. I clearly stated that the optics were unequal.
    However, I also stated why it doesn't affect the comparison very much.
    Even an aliased image of B&W edges from a 5.7 micron pixel pitch is going
    to be softer than the 1.97 micron pixel-pitch, with 289% upsampling. How
    much clearer about that could I have been?
    What is "that"?
    Tell me how a 5.7 micron PP is going to competehere, with even the
    sharpest lens? Anyway, my main point here is the noise. The higher
    pixel density sensor clearly has far less noise per unit of area.

    John Sheehy, Jul 16, 2008
  13. John P Sheehy

    John Sheehy Guest

    Yes, you are totally missing the point here. The demonstration is of PIXEL

    John Sheehy, Jul 16, 2008
  14. John P Sheehy

    John Sheehy Guest

    Sorry; that last one should have been "PIXEL-LEVEL PERFORMANCE", not "PIXEL

    John Sheehy, Jul 16, 2008
  15. John P Sheehy

    Ray Fischer Guest

    "per unit area"

    In other words, if you know what results you want to get then you can
    adjust your measurement to get those results.
    Ray Fischer, Jul 16, 2008
  16. John P Sheehy

    Jufi Guest

    Uh, OK. So why not compare cameras with the same size sensors but different
    numbers of pixels?
    Jufi, Jul 16, 2008
  17. John P Sheehy

    ASAAR Guest

    That's also too easy since a G9 and 5D were mentioned. Switch a
    '0' and you can have the D30 while I'll take the D700 that's left.
    ASAAR, Jul 16, 2008
  18. John P Sheehy

    Jufì Guest

    I guess I just don't get the point of his comparison. Why bother to compare
    the results of two cameras that have virtually nothing in common when you
    can compare sensors of the same size with different pixel counts? And how do
    you account for the diminishing returns as you approach very small pixel
    sizes? I just don't see the point of this excercise at all, regardless of
    what someone's intrepertation of the results may be. Unless you're designing
    and building your own sensors, then you need to select a commercial product,
    and why not just compare available products?

    maybe it's me...
    Jufì, Jul 17, 2008
  19. John P Sheehy

    Jufì Guest

    What has been shown is that for the same number of pixels larger
    pixels win. What has not been shown, at least I have not seen it, is
    that for the same sensor area larger pixels win, in fact for many
    cases they clearly do not.

    Take a FF sensor, 10 MP pixels will win over 1 MP, even though the 1
    MP sesnor will have larger pixels. At some point however fewer pixels
    will win, for example 100 MP will win on a FF sesnor over 1000 MP, by
    the time the pixels are down to the size needed to fit 100 MP there
    will be little gain in going smaller.

    So somewhere there is the sweat spot, the 1Ds III has I believe 21MP,
    I doubt this will be the max number we see. BTW Most people would
    prefer the 21MP on the 1Ds III vs. the 11 MP on the 1Ds

    What about diminishing returns? At what point are the pixels so small that
    they become virtually useless? Very odd excercise.
    Jufì, Jul 17, 2008
  20. John P Sheehy

    John Sheehy Guest

    I'm not sure what you're trying to say here. What is "that"? My "that"
    was the upscaling factor. If you're talking about stacking images with
    sub-original-pixel alignment, yes, that is useful. You get higher
    resolution detail, albeit at a low contrast, and if your SNR is low
    enough, and you have the bit depth, you can sharpen it to good effect.
    However, that is not why I upscaled the 400D here. It was to show how
    the two different pixel densities compared, resolving the same area of a
    focal plane, not only in resolution, but in noise. The 400D has one of
    the lowest read noises in the industry at ISO 100, just behind the D3 and
    the 1-series Canons by about 1/3 to 1/2 stop; much better than anything
    fom Nikon/Sony CCD, the D2X, or any other MFR's DSLR except the
    exceptional Pentax K10D. Yet, it can't touch the tiny 2 micron pixels of
    the FZ50 here, which have almost exactly the same read noise relative to
    absolute signal at ISO 100 as the 400D, at the pixel level. With 8.35x
    as many pixels per unit of area, the effective read noise factors down by
    the square root of that, to about 35% the read noise of the 400D.

    John Sheehy, Jul 17, 2008
    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.