Paddy Eason

Update - there's a chance the issues visible in these images were one-offs. Similar tests with another D20 didn't exhibit them. Further testing in progress, so for now, pls don't assume that all D20's will have these problems... of course if you are planning on using this new-ish camera, you would be doing your own tests.

The Genesis doesn't use a Bayer Array, it uses exactly the layout I showed you.

I know perfectly well how a Bayer Array works, and the shortcomings it has.

Ah - forgive me, I thought we were discussing the D20 here. Obviously the discussion switched to a different camera 5 pages ago, and I missed it.

For anyone who's interested, here are some small patches (actual pixels, no resizing) cut out of some D20 HD images from tests we did a few weeks ago.

Arri D20, shooting 4:4:4 uncompressed into the s.two. Camera shooting in "User 6" mode, which, if memory serves, is one of the higher gain (1 or 2 stops) video-ish colourspace settings, with highlight compression on.

Images taken into Shake, and saved out as highest quality jpegs. Yes, yes, yes, I know that conversion to jpeg potentially brings up a load of issues, but I think that you will still see some things of interest in these images. Most specifically the high level of noise in the red and blue relative to the green, and also the low resolution and "blockiness" of the red and blue relative to green. Also some obvious and colourful moire. Quite worrying. YMMV.

In the camera's defence, I should add that we shot a load of material of more "naturalistic" subjects, projected it all in a high-end grading suite with a high spec projector, and sat for a couple of hours commenting on how wonderful the material looked.

So it's probable that Arri have made the necessary compromises in the right places. Trouble is - I have a load of chromakey shots to worry about! ;)

Huh? Well, 1,920 x 1080 x 3 = 6,220,800 nicht wahr?

 

Anyway I've seen an e-mail from John Galt about this. He said the 12,440,800 pixels are arranged in square "macropixels" of two red, two green and two blue each. I would have thought they would have been arranged like this:

 

RGB|RGB|RGB

BGR|BGR|BGR

------------------

RGB|RGB|RGB

BGR|BGR|BGR

 

to reduce alaising effects, which is what they do with single-chip standard definition cameras that use this technique, but he said no, they're arranged like this:

 

RGB|RGB|RGB

RGB|RGB|RGB

------------------

RGB|RGB|RGB

RGB|RGB|RGB

 

and he didn't seem to understand why you'd want to do it any other way. He didn't seem to have any answer to the question: "why do they simply average adjacent pairs of pixels, instead of having just one taller pixel (which would give a better signal-to-noise ratio)?" either. He stopped replying after that!

I'm certainly no expert, but I think you are misunderstanding the way Bayer arrays work:

http://www.arri.com/news/newsletter/articl...9211103/d20.htm

I've always wondered why Arri haven't made a version of the D-20 specifically for stop-motion work.

It's a nice idea, but there's a couple of things that spring to mind -

- How big IS that market for stop motion cameras?

- If final delivery is 35mm print, and one is doing a DI, how much does one gain from shooting at a higher res (eg full 2880 wide as per the Arri's full RAW Bayer output)? Can you get 2880 images through any known post pipeline?

- Digital SLRs are getting pretty damn good, and shoot at considerably more that 1920 or 2880 wide. Usually with a smaller chip than 35mm too, which has got to be helpful when it comes to stop motion (DOF probs). I couldn't comment on the difference between 35mm film lenses and pro 35mm/digital stills lenses though.

But I can see why it seems a neat idea, technically. I like the idea of shooting extended dymamic range by autobraketing. Very nice.

How does anyone feel Corpse Bride looked? Shot on Nikon prosumer DSLRs, I believe (or was it Canon?).

whereas I consider HD to be a single stage process. Both require color correction in post.

After you take your SR tape (or s.two DMag or whatever) it seems to me you still have a few hoops to jump through before you get to the finish line...

Bill Bennett told me that the asa speed is fixed at 400.

Not true. As I understand it, they are working on (or have just released) an upgrade that takes the ASA to 320 (?). Just before that, through use of different user settings with varying amount of gain, you had a choice up to 200.

Of course, you can underexpose and "push" process in the grade any amount you want, with concurrent noise issues. From what I have seen, noise/grain is considerably less than what you would see with 35mm film of a similar ASA rating.

Really dont understand your complications with this setup.

ah, now, you've made me feel reaally stupid already... ;)

The problem is, yes, it's tape. Most of us already involved on this feature have spent decades avoiding anything with the word "video" in, and consequently our brains lock up when someone says "tape" or "timecode" or "deck". And this stuff ain't obvious, in terms of detailed workflow - you can ask the people sweeping up the mess on several features that are around in post right now... But your tech hints are actually quite revealing and reassuring (some of the same lines we have been thinking on). I may well email you - thanks for the offer!

P

We are in the middle of some testing of just this.

We ran an Arri D20, shooting 4:4:4 into both the SRW-1 deck at HQ quality (2:1 compression) and also the s.two disk array (no compression) simultaneously. Very shortly I will be having a look to see whether the compresson of the SR deck has any visible implications for grading, keying etc.

We are trying to work out a complele pipeline for our production (med budget vfx-heavy feature) using either the SRW1 or the s.two. Or both. It's certainly not obvious how to run all this through with complete safety, redundancy, convenience etc. Actually, it's a complete headache. I am really hankering after 35mm again!

Can anyone show me a tried-and-tested pipeline for SR > final print? :)

P.