[doody]

the issue that the pioneer units do not have square pixels has come up before for discussion previously in this forum on a number of occassions.


i have done some tests on my PRO1000HD, and it appears that when in dot-by-dot mode, there is indeed a stretching of the resultant image. this is unfortunate.


a perfect 20" diameter circle measures just under 21.25" horizontally. an increase of about 6.2%.


the individual pixels measure 0.858mm wide by 0.808mm high. a horizontal increase of, drum roll please, 6.2%.


i tried to drive the PDP at 1280x720 to see if i could then force it to scale the image vertically to 768 using the internal hardware and see if it maintained a perfect circle. unfortunatley, the DVI input won't support that resolution. i tried it again with analog 15pin VGA input. i could put it into 1280x720, but it was over-filling the PDP - the screen stretched off the plasma in all directions. however, the circle was only off by about 3%, although the loss of detail when it's not run at NR is not subtle.


so, i'm curious to hear any suggestions folks might have on how to combat this when driving at native rate. when push comes to shove, there are worse problems to have, but i'd prefer to figure out a fix - preferably on the digital side of things. this is a bit of a bummer.


jlm suggested i look for any tweaks in the video card settings. apparently the vigatec and rock allow stretching (and they're just pc's), though we weren't sure whether they did that in the analog or digital realms. jlm also suggested i poke around the DILA areas - apparently they run into similar problems.


thoughts?

 

[jlm]

doody: i did a searech on "non-isometric" and found this chestnut:

http://www.avsforum.com/avs-vb/show...ht=nonisometric

 

[Alexander]

So we're sucking up the 6% difference? The threads matching "non-isometric" didn't seem to have much relating to DVI. If I got their gist right, it was that for VGA, we should suck it up and use 1280x720, but apparently that doesn't work on the DVI port.


What would be really nice for everyone but us Mac users would be correction in software. Is anyone concerned about this issue for anything other than DVD playing?


Alex

 

[doody]

concerned? yes.


solutions? none yet.


switching to 1280x720 won't help. the problem is that the pixels are physically 6% wider than high. any dot-by-dot addressing of the panel will sufer as a result.


as you opined, the only real fix is to compensate in software for the 6% skew and output a "correct" image dot-by-dot.


doody.

 

[Gaber]

jlm commented in a thread on the Leeza over in the video processors forum that he experienced no distortion feeding his plasma native rate from the Leeza. It would appear that this distortion is really a function of the HTPC and how it is scaling the output signal.


Doody - Did you ever play the widescreen res pattern from Avia on your HTPC?


Gabe

 

[doody]

gaber - i did my math simply by drawing very large circles in windows using visio and some other drawing app and then physically measuring.


i have buzzed jlm about this. that would certainly be interesting data!


doody.

 

[jlm]

using the leeza in-putting YUV from the proceed dvd playing avia, widescreen resolution pattern and outputting rgbhv to the pioneer at 1280x768, i got perfect circles, no 6% distortion. The pio display indicated "1280 x768, dot to dot"


Please note this was not through the DVI interface, which i did not have at the time. Now i have DVI hooked up and tonight will repeat the measurement.

 

[jlm]

when i put up the widescreen enhanced resolution pattern from avia, i get exactly round circles using the dvi path. same results that i saw with the rgb path. the screen is also perfectly filled.

 

[void]

So, do you feed the Leeza pixel size or does the Pioneer "do the right thing" when being fed YUV? I suspect the later is happening.

 

[erehm]

If the Leeza (or any scaler) can be told to output a particular aspect ratio (e.g., 16:9), and can also told to output a particular resolution (e.g., 1280x768) which is not 16:9, my guess is that the scaler is smart enough to scale differently in horizontal and vertical, yielding the correct native rate, but also the correct geometric scaling (i.e., perfect circles).


The folks at Inline told me they could do this....so I'm assuming the other scaler folks are smart enough to figure this out.


/eric

 

[doody]

jlm - this is very interesting.


how does the leeza know that your pixels aren't square?


how big is the avia circle? can you get a 20" or so diameter circle up there and physically measure it (a seamstress tape works well for this)?


doody.

 

[jlm]

in the avia pattern, the center circle measures 7.5", h and v, +/- 1/8". the leeza is set for 1280x768 and the display indicates it is receiving the same. my guess is that in the final stage of its electronics addressing the pixels, the plasma compensates for the array, no matter what you feed it, independently of its own scaling engine.


doody: try out the avia test pattern.

 

[doody]

jlm - i will try the avia pattern, but 7.5" is very small. i used other tools because i wanted a 20" circle to test on.


my guess (with all due respect) is that you have the 6% distortion, but aren't seeing it 'cuz your test is too small. at 20" diameter, it's almost an inch and a quarter of horizontal stretch - visible even to the naked eye at a distance.


if there's no way to tell the leeza "my pixels are not square - they are 6% wider than high" i don't see how it can be compensating.


unless my test was in error somehow. i had a pc and tested a GeForce3 card and an ATI 8500 card and got a disparity of the circle exactly equal to the pixel ratio disparity to two decimal places (in inches).


my operating theory remains that the scaler (leeza, htpc, etc.) needs to fix this (preferably in the digital domain). in dot-by-dot mode the monitor is going to show the dots and not futz much, if at all.


i hope i'm wrong!!!!!!! i will test avia when i get my new htpc into the rack. if you can test a big circle that'd be great.


thanks!

doody

 

[MAB]

If you are telling the leeza that your display is 16x9 and your NR is 1280x768, it may correct the geometry accordingly. That's how it figures your pixels are not square. Correct?

 

[doody]

MAB - oh, i get it now. yeah, that makes sense. interesting.


doody.

 

[MAB]

With an HTPC, it is sending a true 1280x768 picture but keeping whatever the aspect ratio is (14:9 or 15:9) so the PRO stretches out to fill its own 16:9 area. Is there a way to tell the HTPC you are using a 16x9 area?


Leeza knows you have to fill a 16:9 area with, say 15:9 (1280x768) image so it acts accordingly.


To make sure this is all correct, connect the Leeza rgb or DVI output to a computer monitor. You should see a distorted image since the computer monitor can rescan to 14 or 15:9 area, but the leeza will output 16:9 at 1280x768.

 

[doody]

Quote:

With an HTPC, it is sending a true 1280x768 picture but keeping whatever the aspect ratio is (14:9 or 15:9) so the PRO stretches out to fill its own 16:9 area

just to clarify, the pro isn't stretching it out per se - it is simply displaying it accurate on non-square pixels, which has the effect of stretching it out 6% horizontally. the pio isn't "doing" anything actively - in fact it's just doing the "dumbest" thing it knows how to do


doody.

 

[jlm]

i stand by my measurement. 6% of 7.5" is about 1/2", i might have been off by 1/8.


MAB: sounds like a reasonable explantation. the Leeza can set input and output aspect ratios as 1:1.78(HD=16x9) and 1:1.85 while still outputting 1280x768 (1:1.66), but only the 1:1.78 fills the screen side to side; it does sound like the Leeza is doing the compensation

 

[void]

I haven't understood why the Pioneer has rectangular pixels until now. Thanks guys.


The unit is 16x9 (1.77) with a NR of 1280x768 (1.66). To get the NR of 1.77 they need to stretch the pixels (to quote doody "drum roll please") by 6.2%. 1.666 * 1.062 = 1.769.


Now the question is, why didn't they pick a NR of 1.77 to begin with? Is there something about their plasma technology that is requiring the rectangular pixels? If so, why not also make the pixels wider to compensate on top of picking of 1.77 NR?

 

[jlm]

i asked a high muckity-muck at pioneer about this and he said that to get the brightness they want, they need to deep-mount the pixels, which requires more space. the layout is rectangular because there are actually three color cells at each pixel location and they got better space usage by not using a circular pattern. apparantly it is a better idea to reduce the horizontal count than the vertical, hence 1280, not 1366 x 768, and 1280 x 768, not 720.