Is it even worth the development effort, to use a small fraction of the 50D sensor (effectively turning 50D into a tiny sensor, consumer level camcorder), RAW or not? (I am talking about zoom RAW recording.) With a tiny sensor, you lose most of the advantages of the DSLR-based video-camera - shallow DoF and great light sensitivity.
of course it is!
in crop mode you get a 1:1 pixel representation. yeah, it is just a pretty tiny fraction of the sensor size. but it´s fine for establishing shots or high detail images.
from a technical point of view there should be no difference in dynamic range or sensitivity. just noise could be a little bit more noticeable.
for your shallow dop needs you can use the non crop resolutions. faces or close ups of organic things without sharp edges work perfectly.
to avoid aliasing in sound recorded with low sampling frequency ( to record a frequency range between 20 hz and 20 khz the sampling frequency has to be 40 khz according to shannon-nyquist)
there´s something called noise shaping. maybe turning the iso levels up to generate a strong random noise pattern could help prevent edge aliasing.
could be a reason why some people report moire reduction after applying noise reduction.
this method can also be used to avoid bending in your web videos: just ad a little amount of noise to your videos.
the encoder (mostly h264) detects the noise as detail and lowers its macroblocking giving smoother gradients in large flat surfaces (skies etc.)
the magical images from the 50D are in my opinion a result of the "low" resolution chip. a 50D with a 10 or 12 MP sensor would give you even better images. the 50D was called a pixel over kill when it was released. if they built in a lower res sensor with bigger sensel size dynamic and low light performance could even be better. but the 50Ds sensor was quite unique at that time. the first one without real spaces between the sensels.
and if i understand the live view down scale line skipping right, there would be less line aliasing as a result of less lines needed to be skipped.
is this correct?
so theoretically the 40D would give better line-aliasing results.
moire can only be solved in cam with an optical low pass filter since the visible moire effect is just a result of the nyquist-shannon theoreme. every image sensor has its high frequency limits (high frequencies meaning small image details).
a lower res chip can be the solution for micro moire.
an ideal cam would have a full frame 35 mm faveon (or 3 chip) style sensor with the exact pixel count you want to shoot and a specially designed LPF.
this is the reason why the 5DIII is so good for video: good LPF and a sensor pixel count that´s ideal for downscaling.
did some tests with the moire reduction in ACR. works pretty good in special cases!
