Plasma display, native resolution and internal scaler: what happens to HD video?

resolution for the display will always be the native res; for the pioneer, eg., that is 1280 x768. this is independent of input.


there are basically two situations regarding vertical resolution (number of horizontal lines of pixels)

if you send the plas an image that contains less than the native res, 480i or 480p or 540p for example, first the signal is de-interlaced if required and then the scaler in the plas executes its function which is to interpolate and create information to fill in the gaps between input points and display pixels. what is finally displayed is 768 lines of pixels

if you send the plas an image that contains more than the native res, 1080i for example, the plas first de-interlaces and then scales that to match the native res. If you had 720p for the HD signal, that would get scaled to the plas rate of 768 (for the pioneer) or 600? (for the 42" plasmas).


i think the 1024 line displays (Sony) don't really have 1024 lines but use some sort of trickery...they require a different measure for true resolution.


the goofy part is if you take a low res NTSC image, for example and blow it up to fill a 50" plas 1280x768, you have a lot of artificially created pixels in the image, so you get more of what is already poor enough.

Which raises the question of what happens to 720p or 1080i HD when it is sent to a 42" plasma with a native resolution of 848x480 (like the Panny 42"). Obviously, the Panny's processor downconverts the image to the native resolution -- but how does this explain why the picture is so amazingly good?


One explanation I've heard is that the HD signal has much more information, particularly in terms of color depth, and that this extra information allows the panel to display a much better picture. In other words, there is much more information per pixel than with a standard-definition NTSC signal. It's analagous to adjusting your computer display from 16 bit color to 32 bit color.


Is this a gross simplification? Totally incorrect? Or close to the truth?


One thing is certain -- the HD image on my 42" Panny is simply amazing, nothwithstanding the low resolution.


Please step in here, video experts.


Thanks,

Ken

nobody ever said the situation was optimized.


the point of scaling is to convert from an input resolution to a different, one assumes more appropriate, display resolution.


As far as image quality for up-scaled material, think of it this way:


if there are only 480 lines coming in, the nearest vertical spot of informtion is one of those lines away. if you scale that to 768 lines, the distance between those original spots is now 768/480 lines away, about a line and a half. You have more lines, but no more information to put on them, so the scaler invents more data and fills out the display. if the ratio was more like 1:2, the explantion would be easier...the scaler would be inventing 100% more data.


the pq problem is that the scaler's invented data is an approximation; that new data was never in the image and so might look fishy. Add moving images to the mix, then film and video sources with editing and cadence changes, etc. you can see why some scalers might be better than others and some material might scale better than other material.

that is the theory, mark. however, in practice there appear to be substantial challenges to getting a PDP to not "process" an input.


either (a) the "dot-by-dot" input is in PC mode with different (usually sub-optimal) settings for video (ie: it expects a PC screen, not a movie); or (b) the unit still keeps its electronics in-line if it's not in PC mode so you can only hope it's "not doing too much".


i too am interested in learnign more about this, so if i'm wrong, please won't somebody more educated beat me over the head!!!!


doody.


ADDENDUM: btw, this is why the Runco units are theoretically interesting. one assumes that they rip out all the internal NEC video processing electronics and what you're left with is a raw matrix of pixels that the companion outboard box can feed with data. so if you're gonna buy a PDP and an NR maybe you're better off buying a Runco PDP system. i dunno.

I'm by no means and expert doody, but I am sure that trying to match up input resolutions (well lines, anyway) and pixels (vertical) is a waste of time.


I don't know of any plasmas that bypass the internal scaler in any circumstances. The video output from your satellite box (or whatever) is overscanned so that you don't get black lines on any edge of the screen. And so you need the h-pos, v-pos, h-size and v-size controls to make sure the picture is centred etc. These control are always in circuit and the plasma is always scaling.


So to try to get it all to "match up" pixel for pixel would be virtually impossible and probably yield no benefit.


Only when you have a 848x480 *digital* output, feeding a 848x480 panel *digitally*, via VDI or SDI, will this make any sense.


Chip

Good question. What does happen to a high-resolution 1080i signal on a 1024X768 plasma panel? Say you're looking at a live U.S. Open tennis broadcast and CBS's HDTV court camera has zoomed in on Anna Kournikova's ear for a second as she serves.


According to CBS engineer Bob Ross (his 8/10 post here ), the camera isn't filtered and should deliver full-resolution 1/30-sec frames with 1920X1080 active pixels. In bright sunlight the camera might capture an ultra-fine blond hair, say, 1 or 2 pixels wide near Anna's ear.


Say that Kournikova blond hair signal (KBHS) makes it all the way to your HDTV receiver, despite MPEG-2 compression. Since there's little movement during that one-second shot, MPEG-2 isn't filtering high-resolution details such as the KBHS.


But the KBHS may not make it past the next step. According to the mid-1990s FCC's experts committee , whose report I simplified here (8/22 post; forum conversion scrambled chronology), your receiver's built-in filtering will whack some 20 percent off the horizontal resolution to minimize interlace artifacts.


So, the FCC's experts measured 800 X 1638 pixels for a stationary black-and-white test pattern. No doubt they used CRT displays with a 1080 horizontal-line raster. But their 1080 X 1920 test pattern, after filtering, only displayed 800 lines by 1638 pixels (see the 8/22 post above). A 1080-line CRT raster remains constant, but an interlaced video signal on such a raster usually delivers about 70% of its static vertical resolution to viewers because of the so-called Kell factor.


Say the KBHS makes it through, taking up two of the 1536 horizontal pixels per line involved (1920 - 20%). Next, circuits in the plasma panel combine all the pixels in the two 1/60-sec fields making up each 1/30-sec frame into a single 1/60-sec progressive frame.


Then the scaler has to squeeze this image into the plasma panel's 1024X768 pixels. Numerous algorithms can be used to manipulate the HDTV pixels, but the KBHS undoubtedly won't survive in its original form. Perhaps adjacent pixels would be blended that create the appearance of a single hair. Perhaps the KBHS would be filtered out.


Higher-grade CRTs can display the resolutions provided with 1080i signals after receiver filtering. High-end monitors can display an unfiltered full-resolution 1080i signal. Also, D-ILA systems use a hybrid progressive/interlace technique outlined here (my 9/17 post). Notice not everyone favors this hybrid technique. -- John

many of these issues were hot a month or so ago; try a search for: "not-so native rate" in the processor forum. I too was anticipating the grail of "1:1 pixel mapping", but like chippy have decided that was a fruitless search (not necessarily a waste of time...I learned a bit).

The issue was provoked by the Faroudja, who used the 1280x768 Pioneer to demo its so-called "Native Rate" scaler, but could only get the system to work when they sent the pioneer the near-miss rate of 1280x720.


ericbee and I tested this business out with our pioneers and native rates and finally gave up. We couldn't get Faroudja to make a 1280x768 scaler, we found the pioneer had pre-set problems when it was fed its native rate of 1280x768 (from the vigatec) or 1366x768 from a Faroudja. And finally, we felt the display simply looked great with the near-miss of 1280x720.

so much for the grail.


there was one interesting response (i forget from whom) which blew off the importance of a digital image anyway, claiming that once the number of vertical lines was set, it didn't matter if there were a discrete number of image points in the signal or a continuous analog signal.