Friday, June 30, 2017

Sony Fast Readout May Create Banding Artifacts in LED Light

DPReview publishes an analysis of banding artifacts in Sony full-frame A9 camera with electronic rolling shutter. Apparently, its stacked stacked sensor reads 12 rows in parallel to progress through the entire 24MP sensor in 6.25ms time. This creates noticeable banding artifacts on LED advertisement panels and nearby objects in the examples shot at a soccer game:

DPReview concludes: "The a9's electronic shutter proceeds in 12-row chunks; a mechanical shutter is more analogous to an electronic shutter proceeding line by line, which would yield smoother bands. Not the more hard-edged, lower frequency (and therefore more readily identifiable) 12-row bands that are visually more distracting than if they had been gradual, single-pixel rows transitioning from one color or brightness to another."


  1. Seems to be more a problem of post-processing rather than the image sensor itself...

  2. Nope, it fully technical stuff. Like in Pentax K1 in pixel shift mode (electronic shutter) you will get banding with pulsing light source.

  3. Ahh... I hate this, when people start making assumptions when they do not know the tech. This is absolutely not a result of the new fast readout / conversion speed / scan rate, but merely an issue with the flashing light of LED and short exposure time.

    The same happens when I point my old A7S (mark I) in video mode* towards a dimmed laptop screen (vaio z1). Shortening the exposure time I get various sized bands and at extreme cases (under 1/200) I start seeing black sections in the image. Shooting a television set produces more random artifacts, as the TV has a "rolling backlight" opposed to the laptops global backlight.

    *video mode does true readout, while photo mode uses different scan rate for preview. You would need to take a still to judge the artifacts, with video mode its WYSIWYG

    With 1/200 or shorter time, I am not able to shot the dimmed laptop screen no matter if I use electronic shutter, electronic first curtain or just mechanical shutter.

    When the exposure time is non-integer multiple of the flashing period, some bands will appear.

    When the exposure time is shorter than the period, black/unlit sections will appear - shorter the time, more severe the black bands are.

    Judging the pictures, the guy runs on crazy short expo time with high ISO to avoid motion blur (I do take the same approach for shooting dancers)

    1. This is what DPReview is saying, essentially. The only difference is that they say that 12-row readout makes the band edges more abrupt and, hence, more objectionable.


All comments are moderated to avoid spam and personal attacks.