Wednesday, December 16, 2015

CCD vs CMOS: Smear

Adimec publishes a nice demo of smear in CCDs and virtual absence of it in CMOS sensors. The smear manifests itself in:
  • Contrast reduction of the image
  • Questionable patterns depending on image contents
  • Higher noise level in the dark areas of the image
  • The three effects above combined: degradation of DRI (detection, recognition, identification)
  • For color: a shift of white balance, depending on the smear level

9 comments:

  1. Smear do exist in cmos and affects the darkest region of image, as it comes from the readout, when a pixel row contains saturated pixels and dark pixels, the supply or ground could be shifted locally creating a different response for the darkest (for example black could become grey).
    Then the pattern for cmos is horizontal io vertical for ccd, and to put it in evidence for cmos, an inverted target must be used (white rectangle region and dark background)

    ReplyDelete
    Replies
    1. True, and this effect is called banding or streaking in CMOS sensors.

      Delete
    2. But for a well designed CMOS sensor, you should observe such phenomena.

      Delete
    3. do you mean 'you should not'?

      Delete
  2. « for a well designed CMOS sensor » - seems a bit more complicated :
    -- Sony Exmor sensors from Digital Cameras does Not show Any Banding, neither at Low or High ISO setting : samples from NiKon D600, Pentax K5 & Q10, Samsung NX mini : http://www.dpreview.com/galleries/7467909648
    -- Sony Exmor sensors from Astro Cameras, here hundreds of subs stacked from a Zwo Asi174MM (Sony IMX174) : http://www.cloudynights.com/topic/503868-ngc-5676-and-ngc-5660-with-the-asi174mm-again-4-sec-subs/ = still shows huge amount of Banding in Astro. Low Light conditions, despite the sensor used to take NGC 5676 & 5660 is also a " Sony Exmor " ?!?!?

    ReplyDelete
    Replies
    1. Cooling limits dark current and that allows those horizontal line to emerge from noise floor. The same with Micron CMOS in cooled miniCAM5s - one of solutions is to use higher gain to dump those line below noise floor.

      Delete
    2. « solution is to use higher gain to dump those lines below noise floor » - as with Canon CMos sensors where the Huge low ISO Banding totally disappears at ISO 1600 : http://en.astronomike.net/deep-sky/nebulae/162986.html
      Why didn't MvZ pushed the gain of his Asi174MM to get rid of the Banding = still visible despite 550 / 380 subs stacked ?!

      Delete
    3. Higher gains gives lower dynamic range so it won't always be a good choice. Also it may not be doable if the noise floor is to low even with high gain. Those Sony sensors are really low on noise. DSLR isn't cooled so that bit more of noise may help clean the image.

      Delete
  3. " Those Sony sensors are really low on noise " - Not so low = they didn't implement the Digital CDS, despite their Exmor label ! Now we know : the Sony imx178 is the very first Astro sensor to include these Digital CDS = No FPN, No Banding = same as the first Exmor in DSLR camera : Nikon D7k, Pentax K5 !

    ReplyDelete

All comments are moderated to avoid spam.