LCD monitors & analog connections (Get rid of LCD flicker.*)

[Ed Oscuro]

If you have viewed this Forum on a LCD monitor connected via a VGA adapter (the traditional blue one), you might have noticed some odd patterns marching up and down the screen - looking like little bunches of lines.

So let's quickly refresh our memories on a key difference between analog CRT monitors and digital LCD monitors: The analog CRT monitor's pixels are wherever the beam scans, but on a LCD monitor, there is already a grid of pixels. Unsurprisingly, the frequencies of various LCD monitors and the computer graphics adapter feeding them a signal can be mismatched. The result is generally a pattern of alternating bands of brightness and darkness marching up the screen. (I am unsure so far about whether individual dark bands shifting up and down the screen on the Shmups background, the one I have open right now while writing this post, are a phase or voltage problem, but they might just be the very beginning of a phase issue.) You can see this on the Shmups background, or when looking at a gray image in a video game. (Killing Floor seems to display this pattern a lot.) You might not notice this in many normal situations, like viewing a solid white or black screen, but wouldn't it be nice to make it go away?

Most LCD monitors have clock and phase settings. Clock is a somewhat related phenomenon, but so far I have not seen any monitor that needed a clock adjustment, so don't mess with it unless you see it is needed.

At the same time, every LCD monitor I have studied so far has needed a phase adjustment.

Fixing this issue is very simple if your monitor allows you to adjust the phase (in increments of 1, from 0 to 100, I've found; most appear to be set at 0 by default).

First, go here and follow the instructions. Shift the monitor's phase setting until any apparent movement is eliminated. Next, go to this page and click through the patterns, again watching for any movement up the screen - but beware that some of the patterns flicker in a way that might cause epilepsy (if you are prone to this condition, I fear for your safety at this website). Again, look for the same patterns you looked for on the clock and phase test, adjusting until you get a good result on each pattern (aside from the whole image flickering at once) and back at the clock and phase test image.

More questions and research to be conducted:
-Do all monitors of the same model, running identical resolutions off identical hardware (i.e. clone computers in office or computer lab settings), require the same phase setting, or can it vary?
-Do any other patterns other than regular bands of light and dark moving up or down the screen get fixed by a phase adjustment?
-Do any LCD monitors for 15KHz applications benefit from phase adjustments? Do LCD TVs benefit from phase adjustment, and if so, is it needed for different source inputs?

Lesson learned: Match digital standards to digital hardware! VGA is dead; long live VGA.

*There are at least a few phenomena that may be called "LCD flicker;" some LCD backlighting types may cause "flicker." This only seeks to remove one cause. Thankfully, there appear to be no harmful effects to fixing a monitor's phase.

[Fudoh]

Most monitors auto-adjust the analogue VGA signals and I've seen hardly any which don't do this perfectly. It's been years since I've seen a LCD monitor which needed manual clock and phase adjustments.

-Do all monitors of the same model, running identical resolutions off identical hardware (i.e. clone computers in office or computer lab settings), require the same phase setting, or can it vary?

it can vary because the initial state is based on some kind of auto adjustment and that can turn out one way or the other.

-Do any other patterns other than regular bands of light and dark moving up or down the screen get fixed by a phase adjustment?

those are extreme examples. The most common effects are moire patterns in checkerboard patterns.

-Do any LCD monitors for 15KHz applications benefit from phase adjustments? Do LCD TVs benefit from phase adjustment, and if so, is it needed for different source inputs?

sure thing. Every time you try to match to a analogue resolution (which is technically only defined by it's number of lines) to a digital pixel-based display, those adjustments have to be done - either manually or automatically.

Here a Full HD pattern you can use to do 100% perfect phase/clock adjustments on any monitor. Just open it up in any picture preview and make it's displayed in 1:1 size. http://pms.hazard-city.de/50grey_1920x1080.gif

It's not a checkerboard pattern, but a 50% grey pattern, but that's ok since clock/phase only adjust the horizontal alignment anyway. With perfectly adjusted phase/clock settings, the pattern above should be 100% stable. If you see quirks or tremors anywhere, you need to work on the settings.

[Fudoh]

PS:

but so far I have not seen any monitor that needed a clock adjustment, so don't mess with it unless you see it is needed.

clock and phase settings are related. One is basically a horizontal stretch and the other's a shift. One's coarse, the other's very subtle. You'll see it when you use them with the needle stripe pattern from above.

[Ed Oscuro]

Auto Adjust has the unfortunate consequence of throwing your centering and other settings off (which can indeed be an issue with LCD panels and VGA inputs), and you end up with a bunch of things to re-set or to unentangle with a new round of testing. It isn't a terrible idea to try it when you first get a monitor, but it's no problem for me to spend the couple minutes it takes to fix it to your liking.

Phase is also relatively easy to set, so it should be one of the last things you mess with after calibrating other settings (sharpness, centering, etc.). I know quite well that it didn't work reliably on this monitor - being calibrated after a sufficient warmup period - so I essentially had to re-calibrate all the relevant settings by hand, and it didn't save me from fixing the phase.

There's also the visual feedback - "the eyes have it." No auto control is going to convince me my eyes are wrong when I can see it's selected a phase setting that flickers.

PS:

but so far I have not seen any monitor that needed a clock adjustment, so don't mess with it unless you see it is needed.

clock and phase settings are related.

Sure they are, but the takeaway here is that so far I have never seen a monitor that needs a clock adjustment, which is different from the many monitors I've seen which require a phase adjustment. Clock settings are often at "50" out of 100 by default, and I have never had to move them from that. Phase may be auto-selected, but it almost never is right (using an analog connection) at its default setting. It's not really surprising that monitors would have the correct clock settings for their supported resolutions and refresh rates (especially if you're using them at their native resolutions).

There is some explanation of this here for those inclined. As they note:

The "automatic" adjustment of many LCD monitors is not always successful, and manual tweaking may be needed.

Also:

It is probably wise to allow the monitor (and PC) to warm up for 5- or 10minutes before running this test.

Some people would say 30 minutes.

Also:

Here a Full HD pattern you can use to do 100% perfect phase/clock adjustments on any monitor. Just open it up in any picture preview and make it's displayed in 1:1 size. http://pms.hazard-city.de/50grey_1920x1080.gif

It'll be correct for displaying that image. One thing that has been somewhat surprising to discover is that you can still get notable banding in different patterns (the inversion patterns I linked to above) even after fixing it (as far as possible) for other images, like the standard clock/phase test patterns. Patterns like the one you link, and Lagom's test, have appeared acceptable within a wide range of phase settings on the monitors I've tried this with, yet Lagom's inversion patterns and solid dark grays can still be notably off.

Given that, the clock and phase test patterns aren't doing as good a job for me as the other patterns, and having a solid gray section is a lot less "extreme" than having an alternating pattern for moire. I also find the test pattern has never looked worse after fixing for those other patterns first. It is good for a rough first go, and a final check, but if I want to get the phase as close to perfect as possible, the other patterns have proven to be more easier to judge, more realistic, and more reliable.

Edit-

-Do all monitors of the same model, running identical resolutions off identical hardware (i.e. clone computers in office or computer lab settings), require the same phase setting, or can it vary?

it can vary because the initial state is based on some kind of auto adjustment and that can turn out one way or the other.

This is not what I mean - forget about that auto adjustment, I'm talking about the actual number dialed into the phase adjustment. There is a 0-100 scale (or so) visible to the user on monitors that have this feature, so where it ends up "by accident" isn't of obvious interest. Whether the auto adjustment or the user dials it in is immaterial. There may be some difference if you run the test on the cold monitor versus after it's warmed up, but if I make the test even more similar, you'd expect swapping one monitor for another (that has been on as long) to result in the same phase setting to display that signal accurately.

There may be some way in which sample variation affects this, but I would be kind of surprised if it was a notable amount. I'd almost expect electronic interference to have more of an effect, just about...

[Fudoh]

Oh, and one more note on those settings: The LCD display will have to sample the incoming signal and that only works properly if the resolution is actually supported. That's interesting, if you - for example - feed a 720x480* signal to a LCD which only knows 640x480 (and that's true for 99% of today's TVs via VGA). Then you scrub around the phase and clock settings as much you like, you will never get a 1:1 resolution-pixel match.

* any component signal transcoded to VGA.

[Ed Oscuro]

A little hasty on my part: The lowest-end of the three monitors (ha, "all the monitors") I've tried this on (although I will be testing many more, including many of a couple types at least) had perfect clock and phase settings, the old Samsung SyncMaster 204b, 1600x1200 and 4:3. I also noticed that it doesn't label them as clock and phase, but coarse and fine, and the current reading is 2000-something and 20. I didn't mess with it, but if there are fine increments on the clock, at least, it might allow finer control than the others I've tried so far.

[Fudoh]

For high resolutions (above XGA) VGA isn't a good choice anyway, because of the decreasing signal quality with higher bandwith. Or is there any particular reason why you would want to use a UXGA display using an analogue connection ?