Originally Posted by HoustonHoyaFan
I am not sure what you mean by "dimming". The Sony uses a standard sub pixel mapping function. Each pixel is transformed into a 10x10 virtual
pixel matrix. The shifting is done in this virtual pixel grid. The resulting shifted pixels are then downsampled into the real pixel grid.
As Greg Roger's explains in his review of the Sony VPL-VW60 (excerpt pasted below), suppose your red pixels are .3 pixels to the right of the green pixels (due to misalignment). With the subpixel adjustment, if you wanted to display a single-pixel width white line, the Sony, instead of illuminating the native red pixel at a value of, say, 1.0, would illuminate two red pixels, one with a value of .3 and the other with a value of .7, to give a total illumination of 1.0, whose luminosity-weighted center would be at the center of the green pixel. This will reduce color fringing, but at the obvious expense of sharpness (as Mr. Rogers points out), since a single-pixel width red feature is now illuminated using two red pixels! The effect of this subpixel adjustment is qualitatively different from that of the address remapping that, as I understand it, the RS1 (and, it would appear, the VW60 also*) does to compensate for integer-pixel misalignment. For instance, if you have exactly one pixel of misalignment, with integer pixel remapping your resolution is increased
to the level it would have been at had there been zero misalignment. By contrast, if you have, say, 0.3 pixels of misalignment, using the subpixel adjustment to correct this actually decreases
your resolution to below what it would be without the use of that feature.
Originally Posted by AKuan
Does it make any difference between "dimming 100% current pixel and brigthening 100% next pixel" and "turning off current pixel and turning on next pixel"?
There should be no difference between them assuming its DSPs have enough computation power to handle it without slowing down any other tasks.
I do however believe that it does both remapping and dimming. You can actuall cleanly shift pixels up to two pixels distance.
*Roger's description of how the VW60's alignment correction works supports HoyaFan's contention that, when one pixel of adjustment is needed, the VW60 does the same simple remapping as the RS1: i.e., if you need one pixel of adjustment, the two pixels are illuminated to values of 0 and 1.0, which is equivalent to simple address remapping. So yes, Akuan and HoyFan, you are correct, in this case there would be then no difference between the Sony and JVC for integer-pixel remapping. But my concern about the deleterious effects of the VW60's subpixel remapping upon sharpness remain. In line with what reincarnate points out, Sony knows well the concerns people have with panel misregistration are not only for color fringing, but also loss of sharpness. Yet they offer this subpixel adjustment feature without explaining that it helps one at the expense of the other.
If there is an error in Roger's description of Sony's subpixel mapping or (more likely) my understanding of it, please let me know.
Excerpt from Greg Roger's review of the Sony VPL-VW60, Widescreen Review, Issue 125, November, 2007:
"It is easy to electronically shift an image by single pixel
increments on a projector’s red or blue display panels to
improve convergence; Sony and other manufacturers have provided
that feature on other products. But since panel misalignment is a
mechanical tolerance, the red and blue images really need to be
shifted in sub-pixel increments to match the green image and fully
correct convergence. Since the panel’s native pixels are fixed in
place and can’t be moved electronically, Sony uses two native pixels
to represent each adjusted pixel, and varies the brightness of the two
pixels to make it appear as though the adjusted pixel is located
between the native pixels. For instance, if the adjusted pixel needs to
be halfway between two native pixels, both of the native pixels are
illuminated to the same brightness. By varying the brightness of two
native pixels, the adjusted pixels appear to move between native pixels
in 0.1-pixel increments. Sony apparently expects to hold the
mechanical panel-mounting tolerance to 2 pixels because that is the
adjustment range they have provided in 0.1-pixel increments.
Because single pixel blue and/or red lines are spread out across
two pixels when the Panel Alignment feature is used, some color
fringing remains unless the image position is moved by exactly one
or two pixels. However, the color fringing is much less than it would
have been without correction, since one of the two pixels will have
much lower brightness than the unadjusted single pixel. The worse
case occurs if you adjust the positioning by 0.5 pixel or 1.5 pixels,
since the two “new” pixels have equal brightness. I adjusted the red
image by 0.2 pixels vertically to optimize convergence, and the color
fringing went from just noticeable at my normal viewing distance to
Most of our perception of sharpness and detail comes from the
green image, which isn’t altered by the Panel Alignment function. But
perceived sharpness and color detail may be reduced, and that will
vary based on whether you apply correction to both the red and blue
images, whether you use both horizontal and vertical correction, and
the precise amount of correction as discussed above. Since the panels
were exceptionally well aligned in the review projector, I only had
to apply 0.2 pixels of vertical correction to the red panel. That was
perceived as a slight reduction in sharpness standing close to the
screen, but it was nearly insignificant from my normal viewing position.
But it is impossible to estimate how much the sharpness and
detail would have been affected if I had needed other levels of correction
in both directions and on both panels."