Comments on: Demystifying Digital Cinema Camera Specifications

By: Graeme Nattress

Graeme Nattress — Sun, 01 Jun 2008 15:28:31 +0000

What gets me riled up, so to speak, Dylan, is that 4:2:0 is bad terminology. It’s bad for it’s intended use for YCbCr chroma subsampling. It’s doubly bad to apply it to a colour filter array sensor. Canon do some superb DSLRs that all use a Bayer Pattern Color Filter Array, and that there is no high end stills camera that uses an RGB stripe array speaks volumes. Given that, I can’t understand why some of the Bayer pattern explanation was plain wrong. It’s spoils the wonderful explanations of MTF and sampling theory that form the core of the rest of the video.

I’d also have been pleased if similar “scrutiny” had been applied to the issues of RGB Stripe sensors.

You can, obviously, from the Bayer Pattern in the RED get a full 2k RGB at full resolution. But in addition, you can get a measured luma resolution of 3.2k. Indeed, due to the way that a Bayer pattern works, the reconstruction does not produce 2×2 pixel chroma blocks like you’d get in a 4:2:0 video codec. You’ll find that the edges key very well indeed. And of course, because you can key at the higher 4k resolution, there’s a lot more room for better results.

Graeme

By: Dylan Pank

Dylan Pank — Fri, 23 May 2008 23:24:27 +0000

Thanks for posting these, and Kudos to Panavision for letting you host them. If you stick with the numbers there is a lot of very useful and interesting info here, though it gets SUPER technical.

The thing that’s got the reduser.net pack riled up is one comment by Larry Thorpe that Bayer pattern sensors (and the Red is not unique in this regard) are like 4:2:0 video codecs, in that the blue and red colour res is much lower than the green (I think or is it one of the other colours). Larry Thorpe does actually admit that in the real world, i.e. when you’re not shooting charts, this is not such a severe problem, though I can see it might give the SFX crowd pause for thought.

The Genesis is a single CCD camera but it doesn’t use Bayer, it actually has 3 times the pixels horizontally in RBG stripes so it records genuine full res RGB, albeit at only 1920*1080. the F23 of course uses 3CCDs.

Frankly though, with some notable exceptions, most of the reduser gang are a bunch of AICN style fan boys. It’s rather embarrassing the way Jim Jannard plays to them.

By: Bill

Bill — Tue, 20 May 2008 00:18:58 +0000

Matthew,

I see how the selections are being shown with your player. Makes more sense now. Good series, I’ve really enjoyed these.

By: Matthew Jeppsen

Matthew Jeppsen — Tue, 20 May 2008 00:10:35 +0000

Bill, if you’ll look in each video’s post, you’ll see those links back to Panavision as well as the mention of 720p and 1080p download options. With Panavision’s permission we are providing them here in our flash player for the convenience of FreshDV readers.

It is an eye-opening series of videos and I commend the presenters for their thoroughness.

-Matt Jeppsen

By: Bill

Bill — Tue, 20 May 2008 00:03:51 +0000

All 7 of these videos have been up on Panavisions website for the last month or so:

http://media.panavision.com/ScreeningRoom/Screening_Room/Box_Office.html

The people at Red seem unimpressed with their discussion of Bayer sensors, specifically Video 6 in the series:

http://www.reduser.net/forum/showthread.php?t=12811

By: Joel Smith

Joel Smith — Fri, 16 May 2008 20:26:09 +0000

Great explanations and surprisingly easy to understand given the technical detail he goes through.

Now if only manufacturers would stop fudging their numbers and tell the truth… One can wish.

I’m interested to see what he has to say about Bayer patterns and color resolution. Maybe he’ll put to rest some of the debate over the RED ONE resolution.