Is the tide turning from AE100 to 75U - Page 3  

I just had a thought.

Why the infighting? Let's all go over to the CRT forum and ask them about calibrating the convergence on their projectors :D

Robert,

I think it comes down to what is tolerable for most. The bars with 4:3 showing a 2.35:1 movie are so big that you get a large amount of light-spill and need a big screen that shows how much of the panel you aren't using. With a 16:9 one, the bars, proportionate to the content, don't light up your room as much. I see no need to mask my 16:9 PJ when I watch 2.35 but if I watched more than a couple hours a month of 4:3 material, I would probably do something to eliminate the pillarboxing on the sides.
Conversely, if I had a 4:3 PJ, I would obviously get no light-spill with normal TV but more with DVDs.

Namlemez,

Not really. Suffice to say, i disagree with most of your comments.

Define reasonable distance. I am able to see a pixel structure (aliasing) at WXGA and above easily from a 'reasonable' seating distance, which is around 1.5 times screen distance.

Cheat? Are you sure you understand why i used it?

No, its not. This is a common misconception. And you probably mean 'TV distances'?

Oh, maybe they look the same to you. If i show these images on my monitor and simulate a normal seating distance of 1.5 screen width by going the proper distance away from the monitor, i still see the difference quite easily.

Not really. As someone else correctly noted, getting rid of the aliasing is simply more important than scaling errors to odd output resolutions.

You misunderstood me there. The post had just as much to do with digital pj as with CRTs. The point of the pictures was that a DILA with ISCO can reproduce a 480p source better than a XGA unit, which in turn is better than a native resolution unit.

The 'no' you quoted simply implied that if you have a digital pj, you can't really 'choose' the highest and thus best possible resolution. You can't say, 'hey the higher the better, so i drive my XGA projector with UXGA resolution'. You should use the resolution of your panels instead. [There are seldom cases where using a higher output resolution than the panel's native res, has its advantages, though. The reason for that is a bit more tricky.]

Oh, absolutely. But if you can't see a pixel structure (with a properly focused lens) with a WVGA projector, you are either sitting way too far away to even call it a home-'theater' application, or you simply don't see that well.

If you 'see' it, but it doesn't 'bother' you, thats ok. But that doesn't change the fact that a higher resolution picture with less visible pixel structure wouldn't look quite a bit better.

Its true that there are 'sweet spots' when you scale. Probably the best known is 1440x960 (doubling X and Y), which can currently only be achieved on CRTs. Basically, problematic are resolutions that are 'too close' to the source resolution. Scaling 480 lines to 576 (XGA 1024x576) has more noticable errors than scaling to 768 (WXGA 1366x768).

But:

a) again, these scaling errors are less important than getting rid of aliasing

b) the horizontal resolution of WVGA units, 848 pixels, is SO close to DVD's 720 pixels, that it will have just as severe scaling errors in this dimension alone, as XGA units have combined in both dimensions.

c) these scaling errors mostly show up on the very highest frequencies that are contained on DVD ( in the 6.0 - 6.75Mhz band ). On most of the current DVD transfers, the frequency response is already severly down in this range. It will easier to see this on test patterns on AVIA.

Agreed.

Regards
Bjoern

Hmm a few things to respond to :) I'm probably going to go out of order.


Is this totally true? I thought I had read somewhere that there really isn't much scaling going on as it is simply switching from rectangular pixels to square pixels? I can't understand how it is "scaling" when the vertical resolution stays the same as 480. I could just be pulling this out of my a$$ though.

As for the attacks on my vision, I'm 20 years old and recently got my contact prescription checked and I'm still 20/20. Perhaps I just don't have "golden eyes" :)

Gaussian blur: Just figured out why you used. Its to show how CRTs flying spot isn't as exact as a digital grid right? I just looked fast and thought you were trying to "simulate" what 1920x1080 would look like.

Perhaps where I'm coming from wasn't clear. I don't have as much experience with this in the home theater market as in another place, 3d rendering. There seem to be some parallels and perhaps I'll state my thoughts in that frame of reference and they'll transfer over easier.

3dfx started it and Nvidia soon picked up the tactic of supersampling a 3d rendering, mostly for games. Jagged lines with stairstepping were a common visual artifact when looking at a large shape on screen. It is something akin to your picture of what the Building clip looks like "natively." As you said, by supersampling or bicubic resizing it up to a higher resolution, you are smoothing out the edges. This looks very nice on the edges, but the flipside is the insides of the shape.

The eye is good at seeing the blockiness in the lines, but everything inside your outline did not have this artificial snap to grid effect. The end result is that the lines look real nice, but everything inside gets blurred out. At first antialiasing for games and such was a big feature, but when more people noticed how it blurs out the textures on an object, it got switched away from. This doesn't show up to a degree in movies, because some of it is a side effect of the texture existing a high resolution compared to the screen. It still exists and is why I was saying scaling it up is not equal to simply adding resolution.

The advancement that ATi is working on and that I think would be good in HT as well is a smart antialiasing / scaling. They run a different algorithim on the lines versus a pattern on the inside of a shape. Not really sure at all how they would do this with raw video data, but thats for others to figure out :)

I said it was misleading because you chose a "best case scenario." Perhaps _I_ was misleading :) It is not really a fault of yours, since you were trying to show off an effect and did very well. The thing we were looking at was a high-contrast element that was basically a line. I simply didn't think it would do as much if a random sample was taken.

Does a pixel actually have a shape? I'm not talking about a displays pixel since some diplays use rectangular pixles (ie, Fuji 1024 X 1024 plasmas) but rather a data pixel as stored on a DVD, HDTV broadcast, bmp, etc. I always understood that going from 720 to 848 is scaling no matter how you look at it. Scaling can occur in the verticle and/OR horizontal axis.

I thought it was because it demanded so much processor cycles as to be useless. I alway liked the way the N64 looked as compared to its competion of the day.

Yea, normal TV uses slightly rectangular pixels (if it had strict pixels.) Computers use square pixels. Actually, I'm almost fairly positive this is the case with DVD. 720x480 = 1.5 aspect ratio. Yet anamorphic movies are supposed to be viewed at 16:9 (1.78). Guess what 856 / 480 is? 1.78. There's a reason plasmas and the AE100 are that resolution as it shows the movie how it is supposed to look. 1024 / 576 (what the 75u shows) is ~1.78 as well.

So they are essentially rectangular pixels squashed into a square box that need to be stretched out. I guess it is scaling, but the difference I think is that it is supposed to be done rather then being done to fit the native rate of the display.


New graphics hardware, algorithims (sp?) and CPUs make antialiasing doable. NVidia's Quincunx is particularily nice. I run that fine @ 1024x768 on my Athlon 1.2Ghz with Geforce 3 that isn't top of thel ine anymore. Problem is, when people noticed they could run it, they also noticed the shortcomings! :)