Pixel patterns that distort color on whole pixel line

[Falkentyne]

Yes

[dclon]

Very strange. I've found other reports of this problem, but not in both axis. If you get your hands on a camera at some point and could take a good photo of the whole display, it would be interesting to compare how the distortion looks.

[Chief Blur Buster]

This is what I see:

This is very common on many high-refresh-rate displays, and I've even seen this on some HDTVs too as well (when connecting computer & displaying the same patterns).

Another good test to create weird artifacts is http://www.testufo.com/inversion -- and the links (the Techmind link) explains what LCD inversion is, and this is the culprit of these artifacts.

LCD inversion is the rapidly alternating polarity to keep the LCD voltage balanced. The inversion pattern is often a checkerboard pattern, and when you display an image containing this, the LCD is unable to do inversion efficiently anymore. Then you've thrown the voltages temporarily out of whack, and the LCD responds very strangely, creating these strange artifacts.

Sometimes certain LCDs switch to a different inversion pattern when it detects this happens (e.g. "Inversion Uniformity Test" when resizing browser window on an ASUS VG278H original, non-HE model). So you see the distortions suddenly disappear, or colors change (e.g. green/purple color distortions or orange/blue color distortions).

Alas, this is quite normal for a lot of LCDs, since you're essentially displaying a magic pattern that perfectly aligns with the LCD inversion, throwing it off whack.

The inversion could be done using random temporal dithering to solve this, but that causes "snow/noise" artifacts. The inversion could be done using a different pattern, but that will just mean a different "magic pattern" to screw things up. Pick your poison.

Faster LCDs will often have amplified inversion patterns. It happens to all 120Hz monitors to varying extents -- some much better than others. 95% of the monitors use a checkerboard pixel pattern as an inversion pattern, so you would see color distortions at http://www.testufo.com/inversion and sometimes outside-of-browser-window distortions when the inversion pattern is big enough.

If you hate this, switch to a slower LCD such as IPS 60Hz -- they have only very faint inversion issues -- but you don't get the fast responsiveness anymore. Switching between panel technologies (e.g. 24" panel versus 27", or 1080p versus 1440p) can reduce the appearance of the artifacts, but can also amplify them.

For a scientific explanation, see http://www.techmind.org/lcd/index.html#inversion
It works great when it works... but it messes up when you intentionally display images that interfere with inversion.

TL;DR: What you're seeing is, unfortunately, par for the course.

[dclon]

The above was shot on a TV. It did not happen on a closely related model with the a very similar panel type. Both are 120Hz panels, but I'm not sure they are driven as such.

Are you sure this is related to inversion? The distortion goes along the whole line, not just on the grid pattern. It happens while the image is static, and doesn't appear as flicker but as color distortion. Techmind doesn't seem to describe this problem. The description of crosstalk there is a bit closer, but if I understand it correctly it's not something that happens over whole lines, and it's not relevant to TFTs.

Does it happen on your monitor(s)? Whether it does or not, what monitors are they?

[Chief Blur Buster]

Are you sure this is related to inversion? The distortion goes along the whole line, not just on the grid pattern. It happens while the image is static, and doesn't appear as flicker but as color distortion. Techmind doesn't seem to describe this problem.

Yes, it is related to inversion.

No, they do not -- but LCD panels are addressed by row/column addressing, and when a voltage goes so extremely out of balance (not properly balanced in inversion: rapidly alternating positive/negative) that it reaches a high level of static charge (either positive or negative) that is so strong it leaks along the rest of row/column -- so the electrical charge can bleed to the rest of the pixel row/pixel column. It would be nice to have a technical explanation of this phenomenon somewhere, but this happens when the voltage is so unbalanced from a pattern (typically these fine checkerboards) that defeats the voltage balancing effect of inversion. For these quirky displays, the wider you make the browser window, the stronger the interference outside browser window occurs, because of a stronger voltage charge along the pixel row accumulated from the test pattern that prevents the voltage-balancing acts of inversion...

I have six 120Hz monitors, and all of them produce various inversion artifacts. About three of them (I think, especially the PG278Q and FG2421) overloads to an extent that bleeds to the rest of the pixel row/column. Manufacturers need to try to avoid this condition, but fortunately this is something that doesn't happen in typical games as they don't use this precision pattern that defeats successful balanced inversion.

[dclon]

This raises a few questions. How come only some panels have the problem? Why some colors cause it more severely?

There are a few others, at least for my specific monitor. Why is it not as severe in some monitor settings? Why is the color that's mixed into the affected areas green? Why does it cause vetical color gradients in some cases? (I see that in one case so far. It's not similar at all to the horizontal problem, but I'm assuming it's related.)

Even if it is an inversion-related artifact I wouldn't group it into the same class as static-grid-shimmer or motion-grid-shimmer. I find it worse, both because it bleeds outside, and because in the uniform areas it distorts color in a way that's more difficult to identify as an artifact. If the presence of this issue is related to the refresh rate or the inversion mechanism, I'd appreciate manufacturers providing the option to set which method to use. Some could be better for some uses or scenarios.

If you don't mind, could you tell me the models/panels of the affected and unaffected monitors you have? I want to start a list of affected and unaffected panels. If all LCDs suffered from this I'd chalk it up as another item in the ever-increasing list of annoying LCD problems, but not all of them are affected, so the ones that are can be avoided proactively. Probably too late for me with this one. :-/

If on some monitors the Notepad on magenta test image isn't as severe as on mine:


Maybe the magenta grid still clearly distorts the surrounding areas: