[Roland Wandinger]

Don't know if I'm way off here but here I go:


Instead of using scalers to generate more lines than 480/576 we get from a progressive DVD player could one not just resize the beam-spot on a CRT projector to an oval size so the lines get "stretched" vertically (no stretching horizontally) till they just touch each other?


A friend told me that there are in fact CRT projectors which are capable of this.


Of course there would have to be a least 4 different memories for the spot size for non-anamorpich and anamorphic PAL and NTSC.

 

[oferlaor]

Jeez Roland,


I don't know why we didn't think of that - I'll bet there are some pretty sore Faroudja owners out there that are kicking themselves after reading your posting


Seriously, is that the only thing you think scalers are capable of?


Please read home theater hi-fi's explanation of progressive signals.


Scaling has nothing to do with the spaces between scan lines...

 

[Michael Grant]

I think it's important to realize that the 720 x 480 pixels stored on a DVD really represent infinitesimally small sample points. It is the job of the display equipment (DVD player, scaler, display device) to "fill in the gaps" in between.


Representing each pixel as a big fat square is probably the worst choice you can make, visually speaking. The right thing to do is to smoothly interpolate between the individual 720x480 pixels. That's where line multipliers and scalers come in.


If scaling weren't helpful, there would be no market for higher-resolution projectors just to watch DVDs; but in fact, DVDs do indeed look better on high-resolution equipment. The finer resolution gives the scaler the opportunity to "fill in the gaps" in as controlled and visually pleasing a fashion as possible.


To be fair, some people do buy 480-line Seleco projectors on this forum, and I don't fully understand why pixel structure doesn't bother them. Maybe they sit far enough from the screen so their eyes do the smoothing for them.


CRT projectors provide a degree of natural smoothness in the horizontal direction due to their continuous display medium (although scaling even in that dimension helps). But the scan lines do force a fixed vertical resolution structure that line multipliers help to overcome. This is something that high-end video equipment dealers understood long before all-digital video equipment was common.

 

[Mark Rejhon]

Actually, you typically get better picture quality from a scaler than from resizing the beam spot size as you indicated.


Making the beam spot size larger makes the image softer, because the beam spot size is the equivalent of two pixels high (to fill in the black gaps during interlacing).


By scaling instead, you can preserve sharpness while eliminating the gaps between scanlines.


The sharpness of the two approaches are like night and day.

 

[Roland Wandinger]

Quote:

Seriously, is that the only thing you think scalers are capable of?

Of course not! But in the end isn't scaling/interpolating just a "guess" a scaler takes to increase lines/pixels and it does introduce artifacts (depending on the quality of the scaler)?

Someone on this forum once said something like: the more scaling the worse?


I personally use a BARCO Cine 7 which is a 7" CRT.

With anamorphic PAL DVDs I get 576P which is almost all the "scaling" I need. I can just barely see scan lines. Maybe around 650p would be ideal for my projector.


The only difference I see when I scale to say 1280x768 is that I don't see anymore scan lines.

Maybe with 8" and 9" CRT it's a whole different story?

Quote:

Making the beam spot size larger makes the image softer, because the beam spot size is the equivalent of two pixels high (to fill in the black gaps during interlacing).

Why would there be any sharpness lost by resizing the beamspot I mean you are not loosing focus, are you?

To make the beam-spot the equivalent of two pixels isn't that basically what linequadruplers do -> make 2 lines out of 1?


I have never seen a CRT where the beam-spot has been resized so I don't know. It was just an idea of mine and it seem like any easy way to safe the $$$ for a scaler.


As I said I might be way off here and it seems I am.

 

[Michael Grant]

If you can see scan lines or pixel structure, then you are not seeing what the original picture was intended to look like.


If you are familiar with digital audio reproduction (CDs), then know that the principles are nearly identical. On a CD, the music is represented as samples spaced 1/44100 second apart. There is no information on the CD about what the music is supposed to sound like in the "gaps" between the samples. So somehow those "gaps" have to be filled in.


If you were to try a simple approach and hold each sample out for a full 1/44100 of a second, you would sacrifice sound quality, particularly in the higher frequencies; and a significant amount of harmonic distortion (though all of it in the inaudbile range, I'll grant) would be introduced.


That is what you're proposing to do with a video signal by expanding the width of the CRT beam. The impact on the video image would be the same as well: a loss of high-frequency content and the addition of distortion in even higher frequencies. Since image "sharpness" is communicated by high-frequency content, you will have an image that is "softer" than it should be.


Back to digital audio. The proper thing to do is to interpolate the signal to fill in the blank spaces between samples. In other words, the sound to be played in the gaps is computed from the samples that surround it, and their neighbors, and their neighbors... Shannon theory tells us exactly how that interpolation should take place. Unfortunately, in practice, it is impossible to perform perfect interpolation, so CD players pull all sorts of tricks to come close.


One common technique is to oversample the digital signal by computing intermediate samples to place in between the existing ones. When that's done you have an audio signal whose samples are, say, 1/88200 or 1/176400 of a second part. If these intermediate samples are chosen carefully, then the subsequent interpolation or filtering that is needed to finish the job is a lot easier to accomplish. In fact, our ears probably would naturally do most of it.


Oversampling is precisely what video scalers and line multipliers do. As the amount of oversampling increases, the scan lines or pixel structure become less visible. The natural filtering that occurs in the CRT, the lenses, and our own eyeballs help us to complete the illusion of a great picture.


So yes, even if the DVD only has 480 lines of resolution, the image can seem sharper can help to display the image on a higher-resolution device, as long as the scaling is done well.

 

[TEW]

Ronald:

Your idea is actually not such a bad idea.

It's also not a new idea. I've read discussions years ago by knowledgable people who seemed to think this was a viable approach if you didn';t have $20,000 for a Faroudja. I remember a technican reporting that adjusting the beam spot to "an egg shaped spot " presented "suprisingly few" disadvantages.

The thing is, good video processors are now available for a lot less money than they were then. Nowadays you can get a good processor for $1000, so the idea of reshaping the beam spot to avoid investing in one is not as popular of an idea.

Still, if you dont' have a grand to invest in one, this is probably your best approach

 

[Roland Wandinger]

Quote:

That is what you're proposing to do with a video signal by expanding the width of the CRT beam. The impact on the video image would be the same as well: a loss of high-frequency content and the addition of distortion in even higher frequencies. Since image "sharpness" is communicated by high-frequency content, you will have an image that is "softer" than it should be.

Why would there be a loss of high-frequency by simply stretching the beam vertically? There is no impact on the videosignal itself?


What happens to a pixel when you adjust the H-Size of you CRT? Doesn't it just get stretch in the horizontal direction?


Quote:

Nowadays you can get a good processor for $1000, so the idea of reshaping the beam spot to avoid investing in one is not as popular of an idea.

You are of course right that today one can buy a videoprocessor rather cheap but I bought a SKYWORTH 1050P for just $260 and it does all the deinterlacing and 3:2/2:2 pulldown detection better than a $1'000 scaler would and in the digital domain to boot. All I need now is to get rid of the linestructure.

Quote:

Still, if you dont' have a grand to invest in one, this is probably your best approach

How does on adjust the beam-size on a CRT? Can this be done on any model?

 

[Michael Grant]

Quote:

Why would there be a loss of high-frequency by simply stretching the beam vertically? There is no impact on the videosignal itself?

I'm stumped at how to explain it further. Do you believe my explanation of digital audio, even if you don't believe it for video? The points on the 720 x 480 DVD are not meant to be represented by thick beams like that. So by doing what you're proposing you will be producing an image that does not reflect the "intent" of the DVD.
Quote:

What happens to a pixel when you adjust the H-Size of your CRT? Doesn't it just get stretch in the horizontal direction?

This is a bit different, because you're making the entire image bigger. But that would make the pixel or scan line structure more obvious, too, just like sitting closer to the screen would.


Surely you have downloaded a cheap movie file from the internet that's only, say, 320x240 pixels, and used Windows Media player to blow it up to full screen size? To me it actually looks worse at full size all of the time because the pixels are blown up into big squares. Quicktime, on the other hand, does a much better job at zooming into movies---the result is less pixelated, because I believe it uses a more advanced interpolation technique.

 

[Dizzman]

with a digital camera take a picture at a resolution of 640 x 480.


Then take another one at 1280 x 960 Or whatever high res option you have.


Compare the two. THis is the difference between the DVD output and the theoretical we are trying to acheive.


Now, print the two of them as an A4 sheet. this is the difference between what you are proposing.


You are stretching things out, so the curves on a ferrari that are smooth on a small screen, now are very jagged.


Does this help?