Why Image Retention / Burn-In Sometimes Occur On LCDs

[Chief Blur Buster]

Crossposting here.

Dang. Thats the weirdest thing ever. I've had the same static image on the screen for hours at a time, never had a issue. Wow, I'm amazed this happened in such a short time. Furthermore, omg that retention in general is loud and bold ! Most bold IR I've ever seen. How high do you have your brightness ? IMO 99% of people use their monitors way too bright, in which could be culprit to image retention quicker. It is also why they see outrageous screen bloom and backlight bleed.

Image Retention ("Temporary Burn-In")
From Temporary Static Electricity Buildup In Pixels

Flicker patterns such as:
- Emulator black-frame insertion
- 3D-glasses software
- Certain motion tests (including but not just TestUFO's Sync Track)
- Anything that does exactly a half-Hz or quarter-Hz flicker with no dropped frames for a sustained period

Can create temporary LCD image retention. It's a static-electricity behavior when a flicker goes in sync with the positive-negative voltage inversion algorithm, and creates this inversion-related image retention.

This is temporary and you just display video full-screen, or some other thing that really exercise the whole screen -- e.g.
This gets rid of the per-pixel static electricity buildups.

You also get the same problem in anything that flickers pixels in-phase with the positive-negative voltages of the LCD inversion algorithm.

https://www.google.com/search?q=techmind+LCD+inversion

Voltages inverts to try to balance the electricity buildup in the panel, but flicker patterns that go in sync with this, can cause a voltage unbalance = static electricity build up, as an LCD pixels can accidentally behave as capacitors.

The layered nature of an LCD unfortunately creates unavoidable capacitance effects that interfere with operation.



Modern LCDs try to avoid this by using spatial and temporal alternating voltage (positive voltage, then negative voltage, then positive voltage, then negative voltage, and so on)



It's often in a chessboard pattern spatially, which sometimes produces an inversion artifact. Normally this is invisible when the positive voltages are perfectly balanced with negative voltages. But the voltage balancing is not always perfect, so you see this:



However, this is also done temporally -- the voltages swap (like an inverted chessboard) at the next refresh cycle.

Now, if you flicker perfectly (at half Hz), then pixels that gets the "higher voltage" (different brightnesses = different voltages) are always getting negative voltages or always getting positive voltages = static electricity buildup = image retention.

• The chessboard artifact is the quirk from spatial component of the inversion algorithm.
• The image retention is the quirk from temporal component of the inversion algorithm.

Not all screens use the same inversion pattern / inversion algorithm, so some LCDs don't get image retention easily, while others do. In an ideal world, we would be now using error-diffusion temporal dithering or some randomized dithered inversion algorithm, to be fully immune to all material. But in the real world, the panel makers don't do that, and just settle for simple patterned inversion algorithms which are easier to do at high refresh rates (complex invisible inversion processing can be costly). Historically, this is why inversion patterns often have showed up more commonly on high-Hz monitors -- for example, the early ASUS VG278HE (one of the first-ever 27 inch 144Hz monitors) was particularly known for its chessboard artifact during 3D glasses operation.

Thusly, I am not surprised that the world's first panels of a specific refresh rate has some "inversion-related quirks". Every single 240Hz 1ms IPS panel currently has this pixel-as-capacitors quirk at the moment, that only shows up with sustained exact-Hz flicker patterns. As time passes, I'm sure this will improve, with improved inversion algorithms.

Since the pixels have inadvertently behaved like capacitors because of the layered nature of an LCD worked against proper pixel operation. Now you got to drain the charge -- the built-up static electricity stored in the pixel.

Draining the pixel static electricity charge is best done by playing highly active video material. If you want to erase image retention faster, use full screen random-color flashing (fully randomized colors).

HOW TO FIX FIX: Play highly active fullscreen video or animation.
Play pixel fixer software https://www.youtube.com/watch?v=39HUG7QrQi8 (play at 2X speed, it's too slow)
Or simulated analog TV noise https://www.shadertoy.com/view/tdXXRM -- this usually erases image retention faster.
I prefer simulated analog TV noise as a quick burn-in eraser.
Make sure to click the full screen button

While a technology quirk, it is not an RMA defect at all.

Workaround #1: Emulator/RetroArch BFI: Use an odd divisible Hz such as 180 Hz or 300 Hz

Doing this techniques is 100% immune to these problems.

If you want to keep using emulator BFI, use the new 180Hz BFI feature now found in some emulators, a 3-cycle flicker pattern never produces image retention on majority of monitors -- RetroArch is building this feature in now.

Odd cadence is harmless because both IMAGE and BLACK are getting both voltage polarities for every subpixel, as seen in this example:

33% BFI (33% blur reduction) at 180Hz would be:

IMAGE(+) : IMAGE(-) : black(+) : IMAGE(-) : IMAGE(+) : black(-)

66% BFI (66% blur reduction) at 180Hz would be:

IMAGE(+) : black(-) : black(+) : IMAGE(-) : black(+) : black(-)

As you can see, voltage polarity is balanced both IMAGE frames and black frames.

NOTE: Some old LCDs had inversion-algorithms bugs (firmware bugs or hardware bugs) that caused burn-in even with stationary images, especially at certain picture settings. Also, certain manoevers such as overclocking an LCD may cause inversion algorithms to fail, creating a more image-retention-sensitive LCD. You will immediately know if image retention occurs with stationary images instaed of moving images.

Workaround #2: Interrupt BFI Cadence With Framedrops/Framedups

For evenly divisible refresh rates (120, 240, 360) for 60Hz, just framedrop or framedup one refresh cycle every few minutes. That will swap LCD inversion polarity (positive vs negative). So you only expose the LCD to a few minutes of image-retention risk that rewinds when the voltage polarity swaps.

So, for 120Hz, you interrupt cadence by 1/120sec with either a single-refresh-cycle framedrop or framedup (pick one) -- so 60Hz black frames are now using the opposite voltage polarities instead . And now you don't have static-charge buildup that creates image retention.

In other words

IMAGE(-) : black(+) : IMAGE(-) : black(+) : IMAGE(-) : black(+)

becomes:

IMAGE(+) : black(-) : IMAGE(+) : black(-) : IMAGE(+) : black(-)

As you can see, voltage polarity is balanced both IMAGE frames and black frames too.

Swap once a minute (or few), and eliminate LCD burn in risk. You might sometimes get a very faint amount of temporary image retention that takes only a few minutes to undo after exiting your BFI mode -- even if you've been running BFI for days. But you've removed the permanent burn in risk. Because your static buildup is limited to the phase swap interval (e.g. 1 minute). Your static buildup is limited to the phase swap interval (e.g. 1 minute) instead of days, for a multi-day always-running-BFI-on-LCD 24/7 session.

[petergiokaris]

Hi there,

I was wondering if there is a list of known monitors that do not exhibit this particular type of burn-in. I am working on a project to view 3D on panels and so far I have not been lucky in finding one that doesn't cause burn-in over time with 3d content I am rendering.

Ideally, I'm looking for a newer (i.e. sourceable) 27-32 inch monitor(s) that run at 120Hz or higher. I am currently using DLP projection as my main 3D display, which does not have this problem (obviously ) and have just recently begun to explore flat-screens. I picked up an ASUS VG249Q as a starting point. With ELMB on at Turbo setting (the lowest possible persistence mode), I get some burn-in after about 15 seconds or so (this is using a fairly high contrast 3D grid scene), but when the 3D works, it's pretty good.

I am also running into some interesting cross-talk issues, where the top of the screen is out of phase with the bottom of the screen, but the center is perfectly 3D. Although this issue is a completely different topic of conversation, I thought I would add this to share a few interesting problems I am dealing with. Although I cannot go into too much detail, I have had hardware built to allow for syncing active glasses to the monitor.

Any help would be appreciated. I'm hoping to engage in some meaningful conversations about other topics related to flat-screen 3D as well, but mainly am focused on sourcing a good monitor that doesn't burn in.. Thank you!

[Chief Blur Buster]

I am also running into some interesting cross-talk issues, where the top of the screen is out of phase with the bottom of the screen, but the center is perfectly 3D. Although this issue is a completely different topic of conversation, I thought I would add this to share a few interesting problems I am dealing with. Although I cannot go into too much detail, I have had hardware built to allow for syncing active glasses to the monitor.

To understand why top/bottom edge crosstalks occur -- understand that not all pixels refresh at the same time on a panel, and the glasses are open at wrong phases -- see high speed videos of refresh cycles. The problem is that the display begin refreshing anew from the top edge long before refreshing. This can also affect strobe backlights running near its refresh rate limits (e.g. ELMB at near max Hz).

Also, the GtG of the shutter glasses (yes, LCD shutter glasses have a pixel response -- they're one-pixel shutters per eye), which is a shutter fading to black or fading to transparent. So you ideally want the screen to be doing nothing (being static or turning itself off) when your shutter glasses id doing this. So you need really good strobing, ideally.

Getting better 3D requires a display that finishes its pixel response (GtG) at bottom edge, with enough guard time before beginning anew at top. So you want a strobed display, perhaps a ViewSonic XG270 or ViewSonic XG2431 at 120Hz refresh rate (refresh rate headroom is great for reducing strobe crosstalk, so 240Hz produces better 120Hz 3D). That's why LightBoost worked so well with 3D glasses -- the display flashed only while the glasses was open. ELMB is the same.

Now that said, the image retention problem of even-divisor flicker is sometimes an issue. Odd-divisor refresh rates also sidesteps the problem too - you could also do a 360Hz monitor to solve the burn-in, using this 6-refresh-cycle cadence:

L-L-OFF-R-R-OFF

(With strobing disabled).

This is 100% burnin proof, but requires custom programming in your app to support a 360 Hz display using software-BFI at odd-divisor cadence (much like emulator 180Hz BFI mentioned to do burnin-proof blur-reduced 60Hz emulation, except you're doing 120fps at 60fps/60fps for each eye)

You could also use a lower shutter frequency, like a 144Hz display showing 48fps (24fps/24fps) per eye, at the same cadence (L-L-OFF-R-R-OFF). If you combine strobing with the 6-frame cadence, you could do L-OFF-OFF-R-OFF-OFF and get less blur, but it'll likely be very dim, albiet very high quality nearly perfectly crosstalkless 3D

Another method is simply to use an Oculus Quest 2 + Virtual Deskop / BigScreen as your stereoscopic 3D monitor using VorpX -- treating Quest 2 as a wearable computer monitor for your PC. It's actually quite amazing 3D quality (much more comfortable than stereoscopic 3D glasses).

[Supermodel_Evelynn]

So I used the Desktop BFI from github and I get these image retention with my Gigabyte M27Q P monitor

Had this weird image that appeared to be permanently stuck like a transparent rectangle with a shadow or emboss like

After 30 mins was still there so I enabled BFI desktop again and it immediately disappeared\

I am really enjoying the Desktop BFI app it's like a 100% perfect strobing function without having it built into the monitor the smoothness of Street Fighter 6 at 60 FPS was mind blowing.

I do wish that app had a setting where I could intensify the BFI to see how clear I could get the 60 FPS motion clarity.
The closest thing I found was enabling the monitor's built in BFI hardware after I have enabled Desktop BFI software

You want to know what's crazy? on it's own the Gigabyte BFI was complete trash horrible image duplication.
Yet when I enable Desktop BFI software and then monitor BFI in the OSD it looks like the BenQ Dyac at 60FPS

I have absolutely no idea how this is even possible.

the ideal thing would be avoiding the monitor's BFI hardware all together and some way to intensify the Desktop BFI app to get it even more CRT motion.

[Skyy]

Sorry to also chime in on this rather old topic. I actually registered here for it because I got a Cooler Master Tempest GP27U 4k 160 Mini-LED display and I love it to bits. I also thought I pretty much hit the jackpot for retro-games short of an actual CRT:

- Mini-LED so no uneven pixel-wear or burn risk in like on an OLED
- 160hz so enough for 60fps BFI
-4k so enough pixels for CRT shaders
- 1400 nits so enough to counter the darkness of CRT shaders and BFI

I went and fired-up Super Mario World on Retroarch with BFI and the Mega Bezel CRT Shaders and initially had lots of flicker and I quickly found out why, G-Sync wasn't able to evenly lock the display at 120hz and would overshoot to 121fps alot. I fixed this by limiting the framerate of Retroarch with RTSS at 120fps and I was in retro-heaven, I never saw SNES games looking that good on a non-CRT screen.

However after playing a few levels and exiting the emulator I noticed the image-retention on my screen. I heard of it before so I knew it wasn't permanent and it wouldn't damage the screen short term. My old MSI 144hz 1080p display never had the issue so I didn't think much of it and upon looking for solutions I arrived here.

Now I definitely don't want to damage the panel long term so I rather turn BFI off although the controlled frame-drop method to inverse the patterns interests me. Is there any way I can achieve that in Retroarch in a controlled manner, e.g. every 30 seconds? Maybe some little software that interrupts on purpose every now and then that runs alongside Retroarch?

Thanks for help in advance!

[RealNC]

- 160hz so enough for 60fps BFI

That should be 180, not 160.

[Skyy]

- 160hz so enough for 60fps BFI

That should be 180, not 160.

Oh I'm aware that 180hz bypasses the problem altogether. I just wanted to mention that the screen I bought had >=120hz so it's enough for 1-Frame BFI. I also asked over at the Retroarch forums and they aren't a fan of the Framedrop Workaround so it might be unlikely to be implemented into Retroarch itself. They have a new function where shaders can run faster than the content refresh-rate which I was told might be able to help shaders applying a similar "discharge and inverse" effect as dropping a frame would but that's on the mercy of the shader devs.

So I guess short term there's no way to purposefully drop a frame in my use-case.

[JabooMooseKeys]

Did you end up figuring out or has there been an update to how Retroarch does BFI to either drop a frame or there's a proper shader to apply now that achieves the effect we want? Otherwise I'll do the tv static workaround mentioned in the article.

[Skyy]

Did you end up figuring out or has there been an update to how Retroarch does BFI to either drop a frame or there's a proper shader to apply now that achieves the effect we want? Otherwise I'll do the tv static workaround mentioned in the article.

I sadly didn't, it seems that not many people are really interested in solving this issue.

[Chief Blur Buster]

Did you end up figuring out or has there been an update to how Retroarch does BFI to either drop a frame or there's a proper shader to apply now that achieves the effect we want? Otherwise I'll do the tv static workaround mentioned in the article.

I sadly didn't, it seems that not many people are really interested in solving this issue.

They are!!!

The problem is few people know A+B+C concurrently.
(A) The person is interested in issue
(B) The person is ALSO a software developer
(C) The person knows how to make a RetroArch github pull request

The chicken-and-egg is a software developer working on RetroArch is usually not a 240Hz+ user.

Gigantic chicken-and-egg problem if they can't test their own complex source code change that never benefits themselves.

Paid open-source bounties are available by Blur Busters for certain related RetroArch changes. Do you know A+B+C concurrently? Please feel free to help -- Blur Busters is willing to consider paid bounties on certain things such as solving RetroArch #6984, #10754, #10758, or other certain Retroarch github issue request opened by @mdrejhon or @blurbusters. Contact squad [at] blurbusters [dot] com with your github CV/resume and I'll promise/commit a negotiable bounty for certain working pull requests. Certain items will be three-figure payments, and other items will be four-figure payments available [valid thru end of 2024]. This is real money for rare open source developers with enough knowledge to match (A)+(B)+(C) above; and I even lost more than $500 of funds (held at BountySource) during the last BountySource' bankruptcy, so I'm only holding the bounties locally (at Blur Busters).