@guentergunter
Maybe is because of the 1880 buffer and upscaling (then in-camera H264) that MkII has badder moiré than MkIII?
No. Aliasing (and thus moiré) always any only occurs, when you
skip lines.
In other words: Your 5D2 has 5616 columns and 3744 rows (→ 5616 x 3744 = 21026304 pixels total) of arranged silicon semiconductors which by light induce an electric voltage that is then interpreted by some mathematical logic behind to produce an image.
The very problem is, that semiconductors can only detect the intensity of light, but not it's color.
The solution: A
Bayer filter!
That means: Every one of those 21026304 pixels, your 5D2 sensor has, consists of four silicon semiconductors.
Let's now imagine a white item in from of a dark background that moves quite slowly in front of your camera. It will then first be detected by e.g. a semiconductor with a green filter, then one with a blue filter and so on. So, it always happens, that an item for a short period of time (and sadly often enough the very moment you shot) only covers half of the arranged four semiconductors which form a pixel.
So, only the green or blue or red semiconductor induces voltage and thus the mathematical logic behind means it's a green or blue or red item.
When the item moves ahead, it will cover more and more pixels completely, but at it's border, it will always just cover parts of pixels, as well.
The logical misinterpretation resulting in those wrong colored pixels at the edge of objects is what is called aliasing.
Moiré is an addition: It means that those pixels form patterns like e.g. colored streaks.
To prevent aliasing, you need to blur the light that reaches your sensor, so that it always fills one pixel completely.
In other words: Every digital camera has a blur filter in addition to the bayer filter in front of it's sensor.
And this only works for just the very resolution the sensor has.
So, when you now record a movie with 1920x1080 pixels from a sensor that has 5616x3744 pixels you have three options:
1) You only take the part in the middle of the sensor that matches your resolution.
→ no more aliasing; but higher crop factor
2a) You only read every second or third or even seldom pixel (= skip pixels/lines).
→ aliasing pops up, since the blur filter doesn't match your resolution anymore; but no higher crop
2b) You still read every second or third or even seldom pixel, but you implement a by voltage controllable variable blur filter.
→ no aliasing and no crop!
3) You take the full image from the sensor and downscale it by software on the fly in the camera before encoding.
→ no aliasing and no crop!
RED as an example uses method 1). That's why decreasing the recorded resolution increases the crop factor.
The 5D3 has one of those crazy by voltage controllable variable blur filters. That's why this camera uses the same sensor size for photos and video (no crop)
and has no aliasing.
The 5D2 has none of these solutions from stock, so there is aliasing (in video).
But there's hope: The
aliasing filter from mosaic engineeringAny questions left?
