Ping JPS: Luminance noise v. RGB noise

Discussion in 'Digital SLR' started by Alan Browne, Jan 2, 2006.

  1. Alan Browne

    Alan Browne Guest


    Please look at the two graphs at the bottom of:

    Note how the luminance noise of the Max 7D (yellow line) is lower at 200
    and 400 than at 100.

    Then in the RGB graphs, the 200 noise is very slightly worse than at 100
    (except for green channel), and clearly worse for 400 (than 100) execept
    for the green again.

    Why is the luminance noise behaving one way, and the RGB noise differently?

    Alan Browne, Jan 2, 2006
    1. Advertisements

  2. Alan,

    The luminance is typically made up from 59% green, 30% red and 11% blue.
    As the green contribution is greatest, perhaps that accounts for some of
    what you are seeing? How accurate are the figures in any case? [Other
    questions, perhaps later....]

    Why does the green channel, and hence the luminance noise, increase at ISO
    100 in any case? It sounds as if the camera's sensor is being operated in
    some less sensitive, and higher noise mode, just to give ISO 100.
    Wouldn't an ND filter be better?!

    David J Taylor, Jan 2, 2006
    1. Advertisements

  3. Alan Browne

    Alan Browne Guest

    Hi David,

    That would explain a lot as at ISO 100, the green is close to R&B, but
    with those weightings it would shoot up considerably in the mix.
    How accurate? I have no idea. I assume that dpreview's method is at
    least good in relative comparisons.
    Yes: I believe that the "natural" unity gain, lowest noise point, of the
    sensor lies somewhere between 100 and 200. For the same sensor, Nikon
    and Pentax limit the lower ISO number to 200. Maxxum brought it down to
    100. By 'compressing' the output to get 100 they seem to cause noise to
    appear (which seems unusual, but may be an artifact from the processing).

    I've been shooting the Maxxum 7D predominantly at ISO 200 and 400 over
    the last few months. In looking at some of the underexposed areas of
    some recent images (eg: for the [SI]), I noticed a lot of speckles. No
    big deal, these were seriously underexposed areas ... but it caused me
    to re-look at the graphs and review my ISO 200 preference.

    Given the R.G.B/59.30.11 weights you provide above I now realize that
    the luminosity noise graph is misleading if not read with that in mind.

    This will probably cause me to shoot more at ISO 100 and 200 and less at
    400. (I was trying to limit flash power usage/recycle time).

    Alan Browne, Jan 2, 2006
  4. Alan Browne

    JPS Guest

    In message <dpbih9$3vm$>,
    My guess would be that for ISO 100, the green channel is posterized for
    some reason, increasing its noise.
    JPS, Jan 2, 2006
  5. Alan,
    It is hard to tell what is going on at ISO 100 without access to
    raw data. What I do find interesting is the "new trick"
    manufacturers seem to be playing as a function of ISO. At
    higher ISO, they are averaging pixels to smooth noise
    at the cost of spatial detail. That can be seen with the
    7D image in comparison to the 20D image at ISO 3200:
    the 20D shows more detail, or compare 7D ISO 100 to 7D ISO3200.
    The red and blue channels of the 7D are much noisier than
    the 20D, whereas the green channels of the two cameras are

    The 20D has been studied extensively by amateur astronomers.
    It is a photon noise limited system (at its given quantum
    efficiency) and has a read noise of only about 3 electrons.
    For any large signal, then the noise is pretty much determined
    by quantum efficiency between sensors and transmission
    of the optics. It's hard to imagine factors of 2 lower noise
    than the 20D, unless manufactures started using back-side
    illuminated CCDs.

    Roger N. Clark (change username to rnclark), Jan 2, 2006
  6. Alan Browne

    Alan Browne Guest

    I don't really care about noise above 800 as I don't photograph there
    often enough to care.

    Shows that the 7D compares favourably or quieter than the 20D up to ISO
    800. The 20D does better in the blue channel; the 7D does better in
    green and red.

    IAC, I was interested in the discrepency between the 7D's luminance
    noise v noise per RGB channel. David's answer regarding the channel
    weighting used to for the lumincance noise appears to be the best candidate.

    Alan Browne, Jan 3, 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.