Viper's 9Mp

[Landon D. Parks]

OK, I have a stupid question here. What is it with Vipers 9mp CCD's? They advertise it as 9mp, but is it true? Do they really have 9 million pixels on the ccd's or is this some kind of trick?

 

P.S) Another Viper question:

What if you wanted to shoot in 2:2.35 mode, and watch the live feed on a 16:9 monitor. Is there someway to letter box the 16:9 monitor so that when you view the 2:2.35 on the 16:9 screen, it wont appear streatched?

 

Thanks guys! (And gales)

[David Mullen ASC]

The Viper CCD uses 4 subpixels for every picture pixel. This allows them to map different aspect ratio images to 1920 x 1080 pixels in recording. I'm sure there is a box or method of converting the display so that 2.35 is letterboxed rather than stretched on the HD monitor.

[John Sprung]

The Viper's setup is that it divides the pixels up into subpixels vertically but not horizontally. Every output pixel comes from three or four subpixels, 3 or 4 high by one wide. That part I remember for sure.

 

I'm less sure of this, but I think that in the four subpixel mode, you get square pixels and a 1.78:1 aspect ratio, and conventional 1920 x 1080 HDTV. The three subpixel mode discards part of the top and bottom of the chip to give you a 2.37:1 aspect ratio squeezed into the same 1920 x 1080. It should work like having a 1.33:1 anamorphic lens on a conventional HDTV camera. If anyone wants, we can probably get Mark Chiolis to visit this forum and give us the details.

 

 

 

-- J.S.

[Mike Brennan]

An interesting question is why they haven't made all 9 million pixels (per ccd!) useable to create a 2kx4k (HxV) output....

 

 

Suggested this 3 years ago to them.

They seem content with leaving it in the marketplace as is, without any further development or upgrades.

 

 

Mike Brennan

[Mitch Gross]

An interesting question is why they haven't made all 9 million pixels  (per ccd!) useable  to create a  2kx4k (HxV) output....

Suggested this 3 years ago to them.

They seem content with leaving it in the marketplace as is, without any further development or upgrades.

Mike Brennan

<{POST_SNAPBACK}>

 

 

And record it to what?

[John Sprung]

An interesting question is why they haven't made all 9 million pixels  (per ccd!) useable  to create a  2kx4k (HxV) output....

<{POST_SNAPBACK}>

 

Because the subpixels are severely out of square. You'd need a 4:1 anamorphic lens to get to square pixels, and you'd have to rotate the camera body 90 degrees.

 

That and of course there's no way to record the output.

 

 

 

 

 

-- J.S.

[Mike Brennan]

"Because the subpixels are severely out of square. You'd need a 4:1 anamorphic lens to get to square pixels, and you'd have to rotate the camera body 90 degrees."

 

Huh?

As I understand it Viper has 3 chips each with 1920 horizontal pixels x 4320 vertical lines

 

I'm suggesting *all* 1920 pixels x 4320 lines are used rather than selecting or grouping 1080 or 720 of the lines to conform to HDSDI TV output.

 

Three 16x9 chips with 9 million pixels... lets use all of em!

 

 

"That and of course there's no way to record the output."

 

"Of course" there (potentially) is! The srw1 can cope with two 1920x1080 4:2:2 streams today!

 

Call it 1920x2160 to tape, sampled from a 1920x4320 chip.

 

Remember we are more sensitive to vertical res than horizontal res so the picture would be stunning. (this is why HDCAM samples horizontal res from 1920 down to 1440 but keeps verticle res at 1080, the vertical res is more important)

 

This could be a walk in the park for the company that makes million dollar 4k telecine machines!

 

 

Mike Brennan

[John Sprung]

I'm suggesting *all* 1920 pixels x 4320 lines  are used rather than selecting or grouping 1080 or 720 of the lines to conform to HDSDI TV output.

<{POST_SNAPBACK}>

 

OK, take a piece of paper and draw a square 4" x 4". Then draw three horizontal lines across the square, 1", 2" and 3" down from the top. You should have four rectangles, 1" x 4", stacked up to form a 4"x4" square. This is a picture of how the Viper's four subpixels combine to form one pixel. In four-subpixel mode, the Viper's output pixels are square. At a 4:1 ratio, its subpixels are severly out of square.

 

If you shoot with spherical lenses and you could take the full 1920 x 4320 output of the chips, you'd get a 16:9 image with vertical resolution four times the horizontal resolution.

 

So how does the human visual system react to a mismatch between horizontal and vertical resolution? The research shows that up to about a 2:1 ratio of resolutions, the image is accepted as having the lower of the two resolutions. Much beyond that, and the casual observer becomes aware that the picture is sharper on one axis than the other. It doesn't matter whether vertical is sharper than horizontal or vice versa.

 

 

 

-- J.S.

[John Sprung]

I just sent an e-mail to Mark Chiolis at Thompson asking him to have a look at this thread. The Viper is his baby. If he says anything different than what I said about it, he's right and I'm wrong. ;-)

 

 

 

-- J.S.

[Mark Chiolis]

I just sent an e-mail to Mark Chiolis at Thompson asking him to have a look at this thread.  The Viper is his baby.  If he says anything different than what I said about it, he's right and I'm wrong.  ;-)

-- J.S.

<{POST_SNAPBACK}>

 

Thanks for the heads up John:

 

I can give you the "basics" but for more detailed information, Jan Van Rooy or Peter Centen from our factory should ring in on this. I've cc'd them.

 

The FT25 Sensors (used in the Viper and the LDK6000 series of cameras) have a horizontal resolution of 1920 pixels and 4320 in the vertical domain. If you divide 4320 into 6, you get 720, which gives us a native 720 vertical resolution, perfect for HD sports at 720p60, given the bandwidths available in today's world. If you divide 4320 into 4 you get 1080, which gives us the 1080i native format resolution for the LDK6000 in "normal" HD broadcast productions and the 1080p23.98, 24 or 25 for the Viper in RGB 444. If you divide 4320 into 3 you get 1440, which when we take the center extraction gives us a native 1080 lines of vertical resolution for use in 2.37:1 widescreen capture mode. We are the only digital camera on the market today that provides you full resolution in the widescreen mode. All others just "chop" lines of resolution from the top and bottom, which will show up in projects designed for theatrical release.

 

Unfortunately due to physical limitations in components and technology it isn't possible for us to output independently 4320 "lines" of resolution, or even make full use of the 1440 lines for oversampling that way. There is also the little matter of what do you record all this "stuff" on in realtime. With recorders now available such (but not limited to) as the S.two, CineRAM and Sony's SRW series (slightly compressed to tape) we're finding that the RGB 444 output of today's Viper is now managable through the acquisition and post process, but now is not the time to try expanding that amount of data by a factor of 4, it will put us back into an "unusable" mode for the marketplace.

 

While we are always working on increasing the technology and driving things forward, if we as a company don't have income from our products, then we can't afford to drive additional advancements to our client partners, and if we're always in a state of "development" that doesn't make sense from an economic standpoint.

 

That's the short version. I'm sure that either Jan or Peter can add to this in a more technical fashion if you wish. I also have some diagrams, but they aren't easily postable. Here's a link to the brochure, which has a simplified diagram about halfway into it.

 

http://www.thomsongrassvalley.com/docs/Bro...er/viper_br.pdf

 

Mark Chiolis

Marketing Manager

Thomson Broadcast & Media Solutions

[Mike Brennan]

Thanks John,

 

If you shoot with spherical lenses and you could take the full 1920 x 4320 output of the chips, you'd get a 16:9 image with vertical resolution four times the horizontal resolution. 

 

So how does the human visual system react to a mismatch between horizontal and vertical resolution?  The research shows that up to about a 2:1 ratio of resolutions, the image is accepted as having the lower of the two resolutions.  Much beyond that, and the casual observer becomes aware that the picture is sharper on one axis than the other.  It doesn't matter whether vertical is sharper than horizontal or vice versa.

-- J.S.

<{POST_SNAPBACK}>

 

This is what I was getting at:), keeping the aspect ratio the same but increasing (disporportionatly) the horizontal res. So the test you quote states that at a 2:1 ratio of resolution it becomes noticable to the audience that something is wrong.

 

That would make a H1920 x V2020 worth a shot on the Viper? (maintaining a 16 x 9 aspect ratio ie rectangular pixels)

Could we relate this to HDCAM sampling? H 1440 x V1080 (disporportionate V res)

 

 

Mike Brennan

[John Sprung]

That would make a H1920 x V2020 worth a shot on the Viper?

<{POST_SNAPBACK}>

No, when you have more resolution in one direction than the other, up to a ratio of about 2:1, the human visual system sees the image as having the *lower* of the two resolutions. So, a 16:9 picture with 1920 pixels horizontally by 2160 pixels vertically would look the same to the audience as conventional 1920 x 1080. You gain no resolution by going to the effort of handling all that extra data. Square pixels are the most efficient pixels in that sense.

 

If you were to go farther than 2:1, say to the full 1920 x 4320, the audience would notice that the tops and bottoms of objects were obviously sharper than the sides.

 

What the Viper gains by using sub pixels is the ability to support multiple aspect ratios using spherical lenses.

 

 

 

-- J.S.

[John Sprung]

Could we relate this to HDCAM sampling? H 1440 x V1080 (disporportionate V res)

Mike Brennan

<{POST_SNAPBACK}>

 

Sony did that because of interlace. In progressive mode, 1920 x 1080 would be full square pixel resolution, and 1440 x 1080 would be lower in horizontal resolution. But for interlace to not flicker, the vertical resolution has to be filtered down to about 0.60 to 0.65 times full progressive resolution. So 1080i would be equivalent to about 1248 x 702 progressive pixels.

 

Sony chose to split the difference and go with less H-res than you'd really want for progressive, but more than you really need for interlace. By reducing the data in this way before compression, they could use less compression and therefore get less severe compression artifacts. In the early days of a new technology, you make trade-offs to just barely get there. Later on, it gets to be easy to do without any compromises.

 

-- J.S.

[Mike Brennan]

Sony did that because of interlace.  In progressive mode, 1920 x 1080 would be full square pixel resolution, and 1440 x 1080 would be lower in horizontal resolution.  But for interlace to not flicker, the vertical resolution has to be filtered down to about 0.60 to 0.65 times full progressive resolution. 

 

<{POST_SNAPBACK}>

 

Why doesn't output from HD cams in interlace mode, which is 1920x1080, jitter?

 

 

Mike Brennan

[John Sprung]

Why doesn't output from HD cams in interlace mode, which is 1920x1080, jitter?

Mike Brennan

<{POST_SNAPBACK}>

It does, but just not as often as in SD. It looks to me like they might be using a lower "Kell" factor in HD.

 

If you set up HD and SD cameras side by side, and shoot the same test pattern with the zooms set so it takes up the same number of lines in both systems, then any difference in the severity of interlace artifacts will be due to the "Kell" factors of the two cameras.

 

 

 

-- J.S.

[Guest Daniel J. Ashley-Smith]

Interpolation?

[Mike Brennan]

Thanks to John and Mark for making this most concise answers to a wackey post on the net!

 

 

Mike Brennan

[Landon D. Parks]

Interpolation?

Interpolation: The Method of creating more resolution from something with less resolution. See this link:

 

http://www.dpreview.com/learn/?/Glossary/D...polation_01.htm

[Landon D. Parks]

Well, the weird thing is Thompson is advertising the camera as 9 Milliion pixels for CCD.... And not subpixels.

 

I understand from reading into the tech sheets and stuff that it is infact Sub-Pixels. But they are still advertising it as if it had a total of 27.6 million pixels.

 

This is quote from an artical @ www.filmandvideomagazine.com:

The camera's optical block contains three unique 9.2 megapixel CCDs for a total of 27.6 million pixels.

 

It's just weird to me how they always seem to "Leave out" the part about there being sub-pixels instead of 27.6 ACTUAL pixels.

 

Does anyone agree with me?

 

P.S) I have heard that shooting the Viper in 2.39:1 mode does something to the dynamic range. Will this hurt the image any if your shooting mostly blue/green screen?

Edited January 19, 2005 by Landon D. Parks

[John Sprung]

Camera pixels are not the same as post production pixels. It's kinda like shop vac horsepower. They put more effort into finding ways to claim a large number than they put into making the thing work.

 

The camera guys count every light sensitive region on every chip. In post, a pixel means all the color and brightness data for one set of coordinates in the grid.

 

I have a supposedly 6.5 HP shop vac that's smaller than a real 6.5 HP industrial motor would be. It works nicely on a 15 amp circuit, whereas a true 6.5 HP motor would need over 40 amps just to run, and pull a starting surge over 120 amps.

 

So, file camera pixels along with shop vac horsepower and amplifier watts in the round file of advertising numbers.

 

 

 

-- J.S.