Nothing physically changes, the scan rates change and the beam draws stuff longer or shorter and since pixels of a CRT can show content of multiple pixels rather (i.e you can have a red pixel spot be half lit) than just one as it is in digital panels you get "subpixel scaling" which lets lower or higher resolutions still look ok or good.

mhdbox385 wrote:

Also I noticed that the picture quality is good and sharp at resolutions 800x600 720x400 640x480 640x350 320x200, and "not so good" at other resolutions, why is this so?

Depends on quality of graphics card DAC, cable and some other factors.

Hmm ok, so nothing changes physically. Only image box changes, in size and shape, according to how the beam moves. In that case, it seems to me, that almost all the time we are not getting pixel-perfect image, because one would have to adjust the image box to such a setting that a pixel would fall just in between the grills. Still it is very interesting that the image quality on crt is better than on lcd(which can do pixel-perfect).

The other thing I noticed that, when I start pc, on post screen the monitor reports 640x350, on ms-dos command line the monitor reports 720x400, in a 320x200 game the monitor again reports 720x400, in a 640x480 game it reports 640x480. I always thought that ms-dos shell was 640x480? And why the crt in 720x400 mode while the game is in 320x200? I mean, why can't the crt be in 320x200 mode? is it the vga card that sends 720x400 while in 320x200 game, or is it the crt that makes the conversion(I don't think?) ? and why?

The monitor has absolutely no idea about the horizontal pixel count through a VGA cable. It only knows that the signal carries 400 (visible) scan lines per frame and synchronizes its horizontal scanning speed to this rate. Then it draws everything that comes through the cable when it comes through.

So it doesn't know or care if it gets 720x400 (text mode) or 320x400 (320x200 mode) because the pixels are just unknown time spans of a continuous analog signal. The 200 scanlines are doubled by the VGA card and sent to the monitor that way.

alvaro84 wrote:

The monitor has absolutely no idea about the horizontal pixel count through a VGA cable. It only knows that the signal carries 400 (visible) scan lines per frame and synchronizes its horizontal scanning speed to this rate. Then it draws everything that comes through the cable when it comes through.

So it doesn't know or care if it gets 720x400 (text mode) or 320x400 (320x200 mode) because the pixels are just unknown time spans of a continuous analog signal. The 200 scanlines are doubled by the VGA card and sent to the monitor that way.

Why does is it even report horizontal resolution then? Why not just report "400 lines" or something?

Here, I attached some photos. The 1280x1024 and 800x600 are fed from 2007 laptop, others are fed from p-200 system with s3 pci video card.

mhdbox385 wrote:

Why does is it even report horizontal resolution then? Why not just report "400 lines" or something?

I think it doesn't. These are the monitor's guesses and it just has no way to tell the horizontal resolution.

alvaro84 wrote:

mhdbox385 wrote:

Why does is it even report horizontal resolution then? Why not just report "400 lines" or something?

I think it doesn't. These are the monitor's guesses and it just has no way to tell the horizontal resolution.

Not just guesses. It knows the vertical and horizontal refresh and it knows the number of horizontal lines - so it can infer the horizontal resolution from them.

dionb wrote:

Not just guesses. It knows the vertical and horizontal refresh and it knows the number of horizontal lines - so it can infer the horizontal resolution from them.

Just guesses.
Indeed, HSYNC frequency divided by VSYNC frequency gives (more or less) the number of horizontal lines, ie. vertical resolution.
But there's no pixel clock signal on the VGA connector, so no way for a monitor to know the number of pixels per line.
Monitors can only assume the most common horizontal resolutions:
350 scanlines usually means 640x350
400 - 720x400
480 - 640x480
600 - 800x600
768 - 1024x768
1024 - 1280x1024
1200 - 1600x1200

Another question about changing the shape and position of image box.
crt starts by firing electrons for r then g then b, and repeats. rgb, rgb, rgb... When I move or shape the box, what if the electron that was supposed to hit red phosphor starts by hitting blue phosphor and the whole image is all wrong? and in fact even natural scaling can't happen this, it seems? I think need some good video or animation about how the beam makes it to right place at the right time.

mhdbox385 wrote:

The other thing I noticed that, when I start pc, on post screen the monitor reports 640x350

Yes, since around 1994, BIOSes use the 640x350 graphics mode to display that "Energy Star" logo, even though eg. CMOS Setup displays in 80x25 text mode.

I always thought that ms-dos shell was 640x480?

If you mean the command line, then it defaults to 80x25 text mode, and on VGA, characters are 9x16, therefore 720x400.

And why the crt in 720x400 mode while the game is in 320x200? I mean, why can't the crt be in 320x200 mode? is it the vga card that sends 720x400 while in 320x200 game, or is it the crt that makes the conversion(I don't think?) ? and why?

CGA monitors only support 200 scanlines.
EGA monitors - 200 or 350.
VGA monitors - 350, 400, or 480.

So, on a VGA, the card has to double-scan the 200-scanline modes, and when the monitor detects 400-scanline, it assumes 720x400.

mhdbox385 wrote:

Another question about changing the shape and position of image box.
crt starts by firing electrons for r then g then b, and repeats. rgb, rgb, rgb... When I move or shape the box, what if the electron that was supposed to hit red phosphor starts by hitting blue phosphor and the whole image is all wrong? and in fact even natural scaling can't happen this, it seems? I think need some good video or animation about how the beam makes it to right place at the right time.

Regardless of how the beam scans they almost always land in the right spot on the front of the screen. The keywords are convergence and purity. Convergence is that all the beams line up like they should, i.e when you get no misalignment of red or blue beams you don't get a colored edge on a white object and purity is making sure that a beam for say red won't land wrong color. Purity errors are seen as patches of discoloraton in solid colored objects. In white for example they appear as darkened parts, not unlike unevenly worn backlight on modern LCDs.
These two are adjusted mechanically with magnetic controls on the tube and deflection coils assembly and in general are set well enough that neither are a problem.

When you have high enough resolution and move a box as you describe then at some point some (color) detail of the box just lands between the phosphors and contributes to the blurring. Gradually you lose more and more info as the resolution increases or details shrink.