HDDSLR vs RED and vs 35mm

[Olivier Koos]

Hi there,

 

I did some crazy tests, like camparing the 5dmk2 against 3perd 35mm Kodak Vision 3 250D or the red, here it is;

 

http://olivierkoos.blogspot.com/2009/10/5d...35mm-3perf.html

http://olivierkoos.blogspot.com/2009/11/re...2-shootout.html

 

hope u like it ;)

 

O.

[Jason Hinkle RIP]

cool - very interesting. i like the way you shot the identical scene with each camera, i think that gives a much better comparison than other shoot-outs I've seen. The 5D holds up surprisingly well considering the massive price difference. I think it has it's own look which is very good in the right hands. thanks for sharing.

[John Sprung]

It would be nice to see these on a big screen. I can't really tell much difference in a computer window smaller than a post-it note.

 

 

 

 

 

-- J.S.

[Daniel Carruthers]

Did you do any color correction?

I noticed the 35mm shots have more saturation,and contrast

The canon 5D looks flat in comparison, but still very impressive for a camera under 5 grand

[David Mullen ASC]

Most of the problems of different digital cameras show up when you attempt any aggressive color-corrections in post. Just straight from the camera, many of them seem to shoot fine footage.

[Keith Walters]

Most of the problems of different digital cameras show up when you attempt any aggressive color-corrections in post. Just straight from the camera, many of them seem to shoot fine footage.

Exactly. The pictures look fine when they leave the camera, and God alone knows how many engineer-hours were spent getting them that way. OK it looks like video, but it's usually very good video. It's when people try to make it look like film that the trouble starts.

 

As far as I'm concerned, trying to achieve a "film look" from a video camera is like trying convince people reconstituted orange juice is freshly squeezed, by adding orange dye to it.

 

The only real benchmark is how much of a particular scene actually gets meaningfully captured by the pickup device. If certain pixels aren't captured in the first place, apart from painting them in by hand, there is simply no way of recovering them. Yet some people appear unshakeably convinced that there is.

 

Or occasionally, when a tiny tendril of what people have been saying about this for decades finally manages to seep through their inch-thick concrete skulls and take root, they then start denigrating the true performance of film.

[K Borowski]

Most of the problems of different digital cameras show up when you attempt any aggressive color-corrections in post. Just straight from the camera, many of them seem to shoot fine footage.

 

I have to disagree, somewhat, with what you and Keith are saying here.

 

 

A lot of digital files can look pretty decent when well exposed, but I think they are a lot less forgiving of exposure errors, not just blown highlights, but when you go under too. THe underlying lack of dynamic range comes into play; the image looks too flat if not tweaked, and looks artificially contrasty or noisy if it IS tweaked to try to bring out some vibrancy to dull-looking shots.

 

Maybe the notion that "Gee, with Photoshop you can fix ANY mistake," also lulls people into making more aggressive corrections than they would have with a negative and a contact printer. Most people would have just gone back and reshot a negative that was under two or more stops. People tend to get very comfortable with corrections that look good on 18" monitors but won't look good on anything larger. . .

 

 

I've also read several articles where cinematographers like to shoot digital as flat and low-contrast as possible just to be able to pull out as much information in post, even with some film shoots now with the DI. In fact, I may be thinking of something you yourself said on here about "Manure," David.

[John Sprung]

OK, there are three things going on here:

 

1. How much dynamic range do you have?

 

2. How many bits do you use to represent that range?

 

3. Do you use your bits in linear or log mode?

 

Karl's right that silicon chips have less dynamic range than silver halide grains. That's an important part of the problem. Cinematographers have long dealt with the limits of dynamic range through the venerable craft of adding fill light to the dark areas.

 

The other issue when you try really aggressive stretching of the curves has to do with bits. Each bit you have doubles the number of different levels of brightness into which you can divide your dynamic range. So, with 8 bits you have 256 levels, with 10 bits you have 1024 levels, and so forth. The smallest brightness difference that the human visual system can detect, over most of its range, is about 1%. This is called a JND, for Just Noticeable Difference. Do some really aggressive stretching of the curves, and you'll see that contour map banding effect, because the differences between original quantization levels get magnified to be bigger than a JND. Before you get that far, there's a more subtle breakdown of the reproduction of texture, things get an ugly greasy plastic look. Bottom line, the more bits you start with, the farther you can stretch things in color correction.

 

The linear/log thing is that log is more of a "fair" distribution of bits, while linear starves the shadows and wastes bits on the highlights.

 

 

 

 

-- J.S.