> How would you implement an 1.743956737 pixels to one
> mapping? It might not be possible at all because there simply
> isn't any usable x-to-y mapping.

Interesting that you mention it, a pretty sophisticated resampling must be going on if the CCD chips have square pixels, since DV has non-square pixels, both in 4:3 and in 16:9 modes.

Well, maybe the CCD pixels are non-square then.

Does the CCD even have pixels?

Doesn't it just have monochromatic light detectors than have some unknown form factor. The Fuji SuperCCD SR chip, for example, has two different types of light sensors - larger and smaller. With any CCD at least three images must be taken so that color values can be recorded. Some camcorders use three CCDs while others use some sort of filtering / pixel shifting etc.

Now when there are lots of light detectors then the data they give can be processed in such a way that the image looks good on the _output_device_of_choise_ (a TV set in the case of DV tape).

A TV screen and a computer display generate the picture in slightly different ways (various forms and arrangements of those light emitting phosphor dots than make pixels). So there must be some kind of adaptation taking place somewhere...

1) in this ccd I suppose that the distance between pixels is the same in vertical and orizzontal, so is like square sensor detection.

2) the usable ccd area is 1152 x 864 =995328

3) in NTSC 4/3 are used 960 x 720 = 691200 remapped in

960>720 (4>3) and 720 > 480 (3>2)

4) in PAL 4/3 are used 960 x 720 = 691200 remapped in

960>720 (4>3) and 720 > 576 (5>4)


Francesco

> How would you implement an 1.743956737 pixels to one
> mapping? It might not be possible at all because there simply
> isn't any usable x-to-y mapping.

may be 1,75 to one similar to 7 to 4

----------------------------------------------------
Ciao

in 16/9 i think that the dvcam use

1152 x 648 =746496

1152 > 720 is like 8>5 (anamorfic)

PAL 648>576 10> 9
NTSC 648>486 4> 3


Francesco