[mythtv-users] Resolution

[James L. Paul]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Wednesday 03 December 2003 03:08, papenfuss at juneau.me.vt.edu wrote:
> > not pixel-based and do not scale NTSC or PAL video. The "resolution" of a
> > good consumer TV is approximately equivalent to 352x480, which is
> > considered the number of pixels needed to represent a SVHS quality image.
> > The signal quality of most analog cable TV sources in the US is about VHS
> > quality or slightly better, commonly about 280 lines. So, 352x480 is
> > actually overkill for my cable source, which is about 352x280. It's
> > certainly a waste to use more than half my bitrate to encode at 720x480!
>
> Not quite... your cable source should probably be captured at
> (280*4/3=)372x480... read on...

You are right, I was forgetting that the analog signal is vertical lines of 
resolution, not horizontal.

> 	I got off on this tangent yesterday, and from what I was able to dig up
> after a great deal of google research and critical reading is that an
> analog video signal's "lines of resolution" is almost always misunderstood.
>  This the distilled result of the research:
>
> - The most common misconception of a sources "lines of resolution" (e.g.
> VHS is 240 lines, cable TV is 330 lines, etc), is that it refers to the
> vertical resolution.  That would be the number of scan lines, and is
> *fixed* for a particular video format (NTSC, PAL, etc).  For NTSC, the
> number of lines is fixed at 525, about 480 of which are viewable picture
> lines. Since a VHS VCR has no temporal storage (a head swiping a scan on
> the tape is simply drawn as a line on the TV in real-time), it can't have
> 240 lines of vertical resolution. If this were the case, there would either
> be black bars on the top and bottom, or only every other line would be
> drawn... think about it and it makes sense.
>
> - The *actual* meaning of "lines of resolution" is (loosely) the number of
> lines that can be seen *horizonally*.  In other words, how many vertical
> lines can be put on the screen before they cannot be individually
> distinguished and blur together.  Thus, it is a measurement of the analog
> bandwidth capability of the recording medium.
>
> - More accurately, the horizontal lines is defined in terms of a 1:1 aspect
> ratio chunk of the screen.  It's weird to think about from a computer
> standpoint, but works like this:  Make the largest *square* you can on the
> screen, and put vertical lines closer together until you can't make them
> out. Then count these lines.... that's the lines of resolution as defined
> by analog video folks.  Note that not quite all the screen is used in this,
> since standard TV is 4:3.  So, in order to convert the number of "lines of
> resolution" into something useful for capturing purposes, the aspect ratio
> also needs to be taken into account.
>
> SO, with this in mind, take a few data points.  NTSC VHS has 240 lines.
> Formatted at 4:3, this translates into 480 vertical pixels (NTSC viewable
> scan lines), and horizonally (240 * 4/3) = 320.  So, to capture VHS-quality
> video, you need a capture resolution of 320x480... 1/2 D1 is close at
> 352x480.  To capture high-quality broadcast at 330 lines, (330 * 4/3) =
> 440... SVCD is close at 480x480.  Capturing any more than that wastes
> encoding time, space, encoding quality, etc.

This is a much better description than mine. As I remember it though there was 
some effect caused by interlacing, real-world signal loss/distortion and 
quality of video tape that made VHS generally considered 352x240, although I 
can see now the 352 is obviously inflated. (Perhaps that explains why all the 
VCDs I made from my DirecTV looked better than VHS from the same source.)
I think it's the way the interlaced frames are helically laid down on the 
tape, there is significant bleed on VHS, but SVHS used denser tape and 
probably smaller heads.

So the way I understood it, although all NTSC video has 525 lines, about 480 
of them viewable, the analog storage and reproduction of VHS tape doesn't 
accurately retain the resolution of all them, and results in about half. It's 
as if the interlaced video interpolates itself unnecessarily. This may be 
totally wrong. I've never questioned it though since seems to match my 
experience.

> 	Hope that helps... I know when I figured it out it made a lot more
> sense.  Makes better video with less CPU useage, too... :)

Your explanation is the clearest I've seen, thanks! It definitely put a finer 
edge on describing why resolution matters! And it's not just better due to 
lower CPU, it has more bits to work with per pixel which means less 
compression and fewer artifacts/higher quality.

> -Cory
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE/zj7zT8BYaKRUpkQRAgZwAJ44tGhhNQhlZYG7AKDcCE0wkJ8+/gCfZAgT
flshUpoBQYP0Huo0C7Gf2/I=
=DVM5
-----END PGP SIGNATURE-----