Posted by Barry Watzman on 01/13/07 15:05

Do not confuse the phosphor triads on a CRT with "pixels". In a
properly used CRT, the diameter of the electron beam is several times
larger than the size of the phosphor triads. CRTs do not have fixed
pixels. Yes, the phosphor triads are fixed, but it's not proper to use
a CRT in an application where the pixel size and the phosphor triad size
are similar, and, indeed, due to the width of the electron beam and
other issues (electron beam focus, convergence, etc.) you normally
can't. A "pixel" should normally encompass at least several complete
phosphor triads (or their equivalent in a line-structure tube such as a
trinitron or linatron).



Tim Smith wrote:
> In article <Qe2dnZVs-ovroDXYnZ2dnUVZ_vKunZ2d@comcast.com>,
>>>> Right, and this was in the same fantasy world you live in where a
>>>> 720x576 DVD cannot be fully resolved except by a 1440x1152 display?
>>>> Did you happen to see a unicorn out your window while you were
>>>> playing it?
>>> Plonk.
>> Oh good, you've killfiled me. That saves me the effort of having to
>> correct your ridiculous stream of misinformation. Have fun living in
>> that fantasy bubble of yours.
>>
>> To anyone else reading this thread, know that Mr. "M.I.5¾" has
>> absolutely no idea what he's talking about.
>
> Actually, there *may* be something to his claim, in an obscure set of
> circumstances. Consider for a moment a CRT display. If you put up an
> image consisting of a grid of lines that are one pixel wide, and look at
> those lines with a magnifying glass, you'll find that the pixels aren't
> mapped one-to-one to phosphor triads. (Even without looking with a
> magnifying glass, a moment's thought and playing with the CRT controls
> will make this apparent. If you play with the image size controls, you
> can smoothly shrink or enlarge the image, but the phosphors are fixed,
> so it is the mapping from pixels to phosphors that is changing).
>
> For an LCD display, or a DLP display, or a plasma display, you'd expect
> there to be a one-to-one mapping from input pixels to device pixels.
> However, that is not always so. My DLP, for example, has two modes it
> can use when displaying a 1280x720 input. One is called TV mode and one
> is called PC mode. The explanation they give is that the system is
> designed so that 1280x720, at a one-to-one mapping, is slightly larger
> than the screen, to match the overscan of conventional televisions. In
> PC mode, it maps the 1280x720 input pixels to a slightly smaller set of
> pixels on the display, so that they will all be visible.
>
> It does a very good job of this. It's got good antialiasing software.
> But it is not perfect. You can tell that you've effectively lost some
> resolution.
>
> Now imagine that I had a 2560x1440 display instead, and it had TV and PC
> modes. It it accepted a 1280x720 signal in PC mode, and scaled it up to
> something slightly smaller than 2560x1440, I think it could do that in a
> way that would look a lot better than the 1280x720 display does when it
> shrinks the image slightly.
>
> So, if he is using a display technology where there is not a one-to-one
> mapping from input pixels to device pixels (such as a CRT, or a display
> in some kind of scaling mode), then a 1440x1152 display could actually
> do a better job than a 720x576 display at displaying 720x576 DVD images.
>
> For a display where there is a one-to-one mapping, it might be possible
> to make it so that a DVD looks subjectively better at 1440x1152 than at
> 720x576. Each pixel from the DVD would be mapped to 4 pixels on the
> display. If those are mapped in the simplest possible way--the 4 pixels
> all take the same value, then I don't think there will be much
> difference. You've just made bigger pixels. Although I suppose we
> should probably consider the arrangement of subpixels on an LCD.
> Representing each image pixel by 4 device pixels might make it so that
> artifacts due to subpixel structure might be less visible.
>
> However, suppose the device doesn't use the simplest possible mapping?
> Consider 3 2x2 pixel squares on the device, arranged in an L shape. I'll
> name these pixels A, B, C, ..., P in the drawing below. And consider 4
> image pixels, which I'll name 1, 2, 3, 4:
>
> A B C D
> 1 2
> E F G H
> I J K L
> 3 4
> M N O P
>
> In the simplest mapping, we'd set ABEF all to the value of 1, CDGH all
> to 2, and so on. But how about instead setting each device pixel to a
> blend of the nearest 4 image pixels? F, for example, is nearest 1,
> followed by 2 and 3, and then 4. So set it to a weighted mix of those
> four pixels, weighted by some function of distance.
>
> I could see this looking better, especially if you are viewing too close
> to the display, so that various artifacts are visible.
>
> The 1440x1152 display wouldn't be displaying MORE information than the
> 720x562 display, so his claim about fully resolving the DVD would not be
> correct, but it could look better in the cases I've mentioned above, and
> that could be what he is seeing.
>

[Back to original message]