Video Quality on AF100/GH2

[Ben_B]

I agree people will continue to use dslr's, but as camera makers follow panisonic and start making video cameras with large sensors, the dslr's will dwindle. Not completely, but the sacrifices are rediculous. This is why,I never got rid of my XHA1, I use it with the t2i, and hv40. These are just different tools, for different kinds of shooting.

I shot a pageant a few weeks back, It was an hr long with no intermission or breaks in between, If I tried to use a dslr for that gig I would have been sued for ruining the shoot. So I can see the need for a proper video camera.

GH2 can record until your card fills up without stopping...nearly 3 hours with a 32GB card. No 12-minute limit.

That has NOTHING to do with in-camera sharpening. That is aliasing/moire. The GH2, like all dSLRs, lacks an OLPF - optical low pass filter which is essential to have in a proper video camera but impossible to have in a still camera.

Read: Barry's informative article on aliasing.

Actually I believe still cameras do have OLPFs, just not nearly as strong as video cameras. They're there to prevent aliasing in still photos (which has caused some photographers to complain about a lack of sharpness in digital vs film) but not strong enough to do the job for video.

 

[Rick Burnett]

Correct, both still and video cameras have optical low pass filters. They are designed so they reduce the spacial frequency of the incoming light (which is basically diffusing it, but in today's cameras uses layers of birefringent material). The problem is in a still camera, they are designed with the best possible scenario which is using ALL pixels. They are typically designed between the density of green photosites (50%) and red/green photosites (25% each). If you designed them to just the red/blue, the image is typically too soft I've been told. This is also why you still get a tiny amount of aliasing in even still cameras on REALLY sharp spatial steps because they try to average out the benefit.

When you go to video mode, you are skipping and/or binning AND leaving larger areas not being used. That light information in those areas is not getting spread out to surrounding pixel sites, which is why you get steps and patterns.

To really understand what is going on, go into photoshop and create a diagonal line WITHOUT anti-aliasing turned on. Then zoom in. Then, use the 'blur' effect over and over. What you see is there is still stair stepping, but as the line gets larger from the blur, you cannot see the stepping because the gradient of the edge as it gets blurrier hides it.

 

[berylium]

While I agree editors want flat images, I guarantee you that what I can do in-camera on my HPX2700 32 bit DSP is much more precise and non-degrading than what can be done in Apple Color in 8-bit color depth with a sub-sampled codec. Take Film-Rec gamma, one of the features that put Varicam on the map--it has Dynamic Range(200-600%) and Black Stretch controls to really dial in the look or at least capture all the dynamic range possible for post. The very sophisticated matrix and color correction circuits in a P2 Varicam can't be duplicated with the same flexibility and precision in post, ditto detail circuits and knee circuits.

A good DIT can do anything on set a colorist can in post and more. It doesn't make much sense to do so in RAW, but in an 8-bit codec world, baking in a look can still make sense and not just in time savings in post. As a DP, I'd rather bake in the look than have an editor decide what look I was trying to achieve. I've seen my footage absolutely destroyed by over zealous editors/colorists, to the point where I didn't even recognize the project anymore.

Dino:

Your post is very informative and persuasive. I've been shooting flat with my Canon 550D, and will try baking in the look (color and contreast) in camera for my next project. Should I bake in the sharpness as well or leave that to post?