ASC vid manual... let's figure some stuff out

[Tony Hall]

I'm going to post a message that I posted over at DVXUser, but didn't really get a good answer to anything:

As I approach the center of the American Cinematograher Video Manual, a few confusing things haunt me and I was wondering if anyone could shed some light on them.



Topic #1 -

Is DV 480i SDTV? Why do I ask? Because 480i resoltion is defined as 704x480. DV is 720 pixels wide and it uses them all. We know that it uses them all because you can see all 720 in our NLE programs.

Here's what I think: If you take 8 pixels from each side of the 720x480 image captured by DV, you get the 704 "active" pixels that the video manual states as the width of 480i video. This must be what is broadcast. Now, if you multiply 704 by .9091 (which is the pixel's vertical aspect ratio) you get 640, which makes a perfect 4:3 aspect ratio with the 480 lines.

Here are some resolutions from the American Cinematography video manual. It says that DV's resolution is 500 lines... I'm assuming it means TV lines, since there's only 480 vertical lines used in DV. It's kinda strange though because the XL1S claims 530 lines.

http://img61.imageshack.us/img61/3103/video-chart.gif

I did have someone say that it looks like they're rounding the numbers off, but if that's the case, what are they talking about: horizontal resolution or vertical resolution?



Topic #2 -

Quote from the manual:

The dynamic range of a high end CCD may exceed 70dbs - the ability to reproduce a contrast range of 600:1, which is a range from absolute black to two and a half stops beyond the whitest white displayable on a video monitor. It should be self evident that compression of the highlight information is necessary in order to reproduce the signal.

It sounds like he's saying that a high end CCD can capture 9.2 stops of contrast, but in order for it to be displayed as a video signal, it must be compressed to about 7 stops... is this correct... that there can be no more than 7 stops of contrast in video? This would be consistant with what John Jackman says in his book "Lighting for Digital Video and Television". He says

some film manufacturers today are claiming 500:1 on their latest, best film stock. But remember that this is an idea on the camera original; you'll never get a distribution print with that range. Video cameras haven't reached that far yet. A good prosumer camera like the Sony VX200 or the popular Canon XL1s might expose well through six stops, or 64:1. Pro camcorders like the Sony DSR-500 can consistently hit seven stops, or 128:1.

Barry Green and I were debating this point because in this: http://www.theasc.com/magazine/product.htm article by Jon Fauer, it is claimed that the DVX has 8 stops of contrast. I think this is wrong... not only is it one stop more than is even mentioned in the lighting book, but it's also only two stops less than human vision. The quote above from the AC video manual suggests that a video monitor can't even display more than 7 stops of contrast.

Here's another question that arises with this discussion... do regular photographers and film users measure contrast ratio the same as video people? Because in regualar photography we count the "usable" contrast ratio... it sounds like they count pure white and pure black in video. Here's a quote by Barry Green:

Michael tested it and found almost 8 stops of latitude. 16mm showed about 11 stops, and 35mm film showed 13 stops. The pictures are posted on his site at http://24p.com/ASC_Pictures.htm

and here's my response:

We've got a conflict of facts. One of two things are possible:

a) Michael doesn't know how to measure contrast range

or

b) Michael is using different standards than are used in still photography and every book I've read on the subject... which is still doing it wrong.

The pictures he has posted on his website are meaningless (for this discussion) unless someone was there to measure the actual values of the subject.

Human vision is said to be about 10 stops and no film on earth has topped that yet. B&W negative film is 7 stops, color negative film is 6 stops, color slide film is 5 stops, digital SLR's about about 5 stops. There isn't a film stock out there that claims to produce a ratio of more than 500:1, which is 9 stops.

Topic #3 -

I found this a little odd:

From the AC Manual:

[Because of diffraction] HDTV lenses should be operated at wide apertures and never beyond 5.6

The manual is very vague and seems to suggest that HD's resolution is the reason for the diffraction issue. If that's the case then why isn't it an issue in still photography past f5.6?



Topic #4 -

From the AC Manual:

HDTV cameras require lenses with much higher performance characteristics. The wide screen aspect ratio lowers the CCD's sensitivity by ten percent because the area of active pixels is smaller.

Hmm... I know that larger pixels are more sensitive to light because each pixel has a larger aperture (so to speak). I don't understand how not using part of the CCD would affect the pixels that you are using... do you? Also, because of the way the XL2 does this, it won't have the same affect will it (actually the opposite)?

[David Kennett]

Tony,

I agree - some things don't seem to make sense!

TOPIC 1
Years ago - before digital video and pixels, there were scan lines. And since there was no other logical way to measure horizontal resolution, it was measured in LINES. And the number of horizontal "lines" was relative to the height of the pic. Thus 480 lines Vert-res by 480 lines Hor-res would be equal. But since the picture was wider (by 4/3) the TOTAL number of H-lines was 640. By using more than 640 Hor-lines (or shall we say pixels today?), horizontal resolution could be improved. In a nutshell - Hor-res can be about anything, and still be standard NTSC video. VHS is about 250, S-VHS and laser disc are 400, and DV is 540 (720 times 3/4). These are theoretical maximums - your mileage may vary!

TOPIC 2
Don't know anything about this!

TOPIC 3 and 4
I'm with you - a lens is a lens. If the imager is better, the lens must be better, and we've had high resolution lenses for years. I have JVC HD10. 1280 by 720 HD video. Switch to 1280 by 1020 still, and the top and bottom of the picture appear! The lens doesn't know the difference. The center part of the CCD doesn't know the difference.

Not everything in print is correct - not even in a big expensive official book! Not even here!

[Jeff Donald]

Just about all lenses show effects of diffraction. Very high resolution lenses can be diffraction limited. This means that all the other defects are so well corrected that only diffraction remains. This is what the ASC is implying and may very well be the case.

In 8 bit (DV) with gamma 2.2 you only have about 6.6 zones of dynamic range. The CCD's RAW files are capable of about 9 zones with 12 bit A to D convertors. The number of usable zones is also limited by noise, which is a big factor. The use of curves in the conversion of the RAW data can also extract more usable zones by compressing the highlights and expanding the shadows. Larger pixels generally have lower noise levels. The differences in the number of zones of dynamic range may be between theoretical limits and the real limits of human vision. RAW files are linear, but human vision is not. We are not as sensitive to lower light levels and the last zone or two (darkest) may be lost.

The trade off is between dynamic range and contrast. Dynamic range can only be increased by lowering contrast. We tend to prefer images with greater contrast, hence lower dynamic range. This is a trade off the plagues all camera lens manufactures.