RE: Expiance’s post on RGBW, PenTile, Subpixels and Graininess of mobile displays

I wanted to take a moment to respond to Alex Taylor’s blog post on last Friday. For starters, I am very impressed at how much thought and work you put into your post, Alex. Well done!

Still, I feel I need to add some clarification and correction to a couple of things you said in your blog.
Many people like yourself who have, for so long, thought of pixels as having a fixed number of dots, typically three per pixel, so it is not surprising you look at this layout and say there are two per pixel. Certainly, on the average that is true, but it is important to think of pixels in a subpixel rendered display as logical pixels.  This is not unlike it used to be for CRTs. How many subpixels are in a CRT spot? A CRT spot is comprised of a Gaussian distribution of light about a logical pixel center. Such logical pixels can overlap, but when the modulation ratio drops below 50% one loses resolution, per the VESA specification.
PenTile works much the same way.  As many as 10 subpixels can be involved in a given logical pixel, so it is misleading to say one pixel lacks blue and the next lacks red or green. Every pixel is addressed at 8 bits/color and each luminance center is lit by the proper combination of the layout and the algorithms that analyze the image and render the display. It is nothing like compression or zipping.

You seem to agree that pictures look similar for RGB stripe and PenTile. Imaging scientists call these images bandwidth limited images. I would say that these look equivalent between RGB stripe and PenTile displays and can show this show this with MTF characterization plots.

One correction to what you said is that those of us at Nouvoyance never say that PenTile looks identical to RGB stripe. There are differences, but the differences that people point out are sometimes not correct.  For example, we render black and white text perfectly.
I know how tough it is to take a good photograph of a display, but even with the ones you show of very small, single and double stroke  black and white text prove that black and white text is not fuzzy, blurry, or otherwise defective. Look at the single pixel at the top of the “n”, where the curve joins the upright stem. You can see that black pixel every bit as well on the PenTile displays as on the RGB stripe. There is some softness at the edges of each of these that is attributable to the original anti-aliased font, which is also fully and faithfully rendered on the PenTile panels; so there should be no doubt that for black and white text it is rendering perfectly.

You say ”… edges which appear straight on a RGB stripe display will appear jagged, with odd pixels sticking out, just like on those old camera screens, and vertical lines will zig-zag across the screen.” I think that the photos you exhibited to prove this seem to prove the opposite. The same could be said for horizontal line edges on an RGB Stripe, as the red and blue subpixels appear so much more darker than the green. What makes it acceptable in either case is the fact that the resolution is chosen to be high enough that the subpixels blend together by the human eye, when viewed at the appropriate distance.

Let me add that much of the IP for PenTile is in our algorithms. There are several adaptive filters that look at many aspect of images and provide sharpening to edges, especially things like diagonal lines.
You pointed out that this image above demonstrates that the low res layout causes a grid artifact. There is, in fact, some graininess that is possible for a fully saturated green on a black background. The algorithms, for anything less than fully saturated colors, fills in the black regions with white or other color subpixels. This is the same as what is experience on an RGB stripe display that is fully saturated green on black, but in the PenTile case it appears as a checkerboard whereas on the RGB stripe it appears as green vertical stripes, albeit 30% closer together for stripe. Most people will be hard pressed to see this on a 300 dpi screen even at close range. This is especially the case since the human vision system has less resolution in the diagonals. It is also why photographic dot half tones have the same diagonal grid pattern.

As for your assessment that there is a diagonal organization to the display, I would agree, but I disagree with your conclusion. So, allow me to take it up a notch on the technical aspect of the answer. I would agree that the MTF of the display is less on the diagonal for the PenTile OLED RGBG panels than on the horizontal or vertical, but in all directions PenTile can write to the Nyquist limit. While the MTF of PenTile is slightly less than RGB stripe on the diagonal for fully saturated colors, it is still well in excess of the requirement of the VESA/IMID standard of 50% modulation, so it is not reasonable to downgrade the resolution by a factor of 1.4.  For the PenTile RGBW, the MTF, even in the diagonals, is the same as the RGB Stripe panel for black and white, that is to say, that they both will show a checkerboard pattern of equal modulation when the diagonal resolution limit is reached. The checkerboard is the result of an alias that occurs in the original data, before it reaches either display.

Let me turn to one other aspect of fitness for use, which is another name for good engineering, it is known to vision scientists that the human vision system is less capable of resolving detail on the diagonal than on the horizontal or vertical. So the slight fall-off in MTF on the diagonal nearly perfectly matches the sensitivity to detail on the diagonals. Any advantage of RGB stripe in this direction is often not seen, especially as we get into the resolution range of theis WQXGA panel.

So, I am troubled by calling the PenTile resolution claim “somewhat dishonest”. Samsung is saying that our WQXGA display is a PenTile RGBW LCD. And Nouvoyance is even directing those who are interested to our website, showing how it meets the industry standards for modulation contrast ratio when measuring Michelson contrast through a moving aperture grille. This is the test provided by the industry’s leading experts in display metrology. We have disclosed a great deal about what we do.

Do PenTile displays look the same as RGB stripe displays?

No, they look different in some special cases especially at the lower end of the resolution applications for  those with very good vision and well trained eyes, and, as stated above, with a bit more of a textured look for fully saturated green on black. At the higher end of the dpi range, e.g. products with 3.1” wVGA PenTile OLED, I have yet to see even one person who has blogged about a specific product about it looking grainy or less than sharp.

PenTile RGBW can look better than RGB stripe for things like the glint from metal and the reflections from water. It takes that white subpixel to give it that extra punch.

It is very important to apply PenTile to the resolutions where it makes sense – where pattern visibility is not visible for the bulk of the market. To us at Nouvoyance this seems to be good engineering for making a product that saves 40% of the power over the equivalent RGB stripe and meets the needs of the product. Surely, battery life and brightness are important engineering design parameters that as a package make it fit for use.

In the WQXGA product at 300 dpi in a product that is typically viewed from a greater distance in normal use than for a smartphone, so I am hard pressed to think that anyone will feel that this panel will look grainy or textured .

I hope to see you at SID to show you how good this panel can look.

One comment on “RE: Expiance’s post on RGBW, PenTile, Subpixels and Graininess of mobile displays

Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>