Blurbusting Technology Suggestion: Scanning Backlight

Advanced display talk, display hackers, advanced game programmers, scientists, display researchers, display manufacturers, vision researchers. The masters on Blur Busters.
Post Reply
billburrbuster
Posts: 9
Joined: 22 Oct 2019, 04:54

Blurbusting Technology Suggestion: Scanning Backlight

Post by billburrbuster » 23 Oct 2019, 06:39

I've read a lot that CRT blinking is softer than LCD strobbing of the same frequency, while giving good motion clarity.
Why dont they imitate crt scanlines with LED backlight? Here i made a 3d visualization with proper materials (emmiters, diffuser, front filter) of what i'm talking about.

phpBB [video]


Image

if we pack LEDs denser the line would be even thinner, even more crt-like.
it looks pretty easy to make from my non-scientific point of view :?
Maybe its used somewhere already?

User avatar
Chief Blur Buster
Site Admin
Posts: 6512
Joined: 05 Dec 2013, 15:44

Blurbusting Technology Suggestion: Scanning Backlight

Post by Chief Blur Buster » 23 Oct 2019, 11:27

Fantastic idea and we've tried to work on these but these are extremely hard to do at a high quality;

Yes, this is what some manufacturers have worked on -- This is already being done in the industry. However, there are sometimes more drawbacks than benefits. It's softer but the quality of blur reduction is much worse because of multiple law-of-physics problems that is super expensive to solve.

In fact, Blur Busters was originally at the domain name www.scanningbacklight.com before it renamed to www.blurbusters.com due to my old Arduino scanning backlight experiments the idea that John Carmack blessed, that convinced me to start Blur Busters. Do a Whois, I own scanningbacklight.com since 2012 -- the very idea you describe. ;)

I've written about this in year 2013-2014:
Electronics Hacking: Creating a Strobe Backlight
Later on, this work helped a lot of manufacturers decide to add a strobe backlight, and I have helped a few manufacturers, and there's an open source 4-segment scanning edgelight in the Zisworks 4K 120Hz diplay that can do 540p 480Hz. (http://www.blurbusters.com/480hz)

For a good scanning backlight, the problem making it work well is eliminating internal backlight light leakage from the ON segments to the OFF segments. It's VERY hard to do that -- you need to scrap the entire assembly behind an LCD and create excellent light optics that focusses the backlight only to the right segments. If there's even just a 5% light leakage to the opposite end of the screen, you have a 5% strobe crosstalk elsewhere on the screen.

Image

To build a VERY good scanning backlight, you want to make sure your GtG is quite practically complete in all areas that will receive between 3-10%+ light leakage. So you have to initiate GtG far ahead of the scanning light line.

On steady state light, the LCD GtG creates a scrolling fadezone as seen in high speed videos.

phpBB [video]


Image

Light Leakage Problem In Scanning Backlights
Now adding a scanning backlight to this, you need to illuminate a row of LCDs far away from the GtG fadezone (but as close as possible, to minimize latency). More light scatter problems, the more distance between GtG zone and scan-flash zone you need (and thus more lag relative to a CRT). Too much light scatter, and no area of the panel is free of the flash and you've got unsolvable strobe crosstalk.

For years, LCDs could not cram GtG into VBI, and you got 3 refresh cycles ghosting into each other:

Image
(This screen is showing 06, 07 and 08 simultaneously from http://www.testufo.com/flicker test).

This panel is impossible to strobe (or scan-backlight) without strobe crosstalk. It was not until most of GtG pixel transitions became a tiny fraction of a refresh cycle, before strobing finally became practical.

Now, I'm obviously interested in restarting the scanning backlight experiments thanks to newer 240Hz 1ms IPS panels which are astounding quality. Best strobed quality I have ever seen in an LCD, bar none.

_____

Open invitation to hobbyists to restart scanning backlight experiments

If they are paired up with a really good voltage-boosted RGB scanning backlight, that would be sweet -- the problem is the time & expense assembling excellent optics from scratch (e.g. a full array of parabolic lens, or LEDs that have those built-in into them, combined with other optics that focuses them only on areas of interest, while NOT looking beady.

Image

Currently, it has been easier to get CRT clarity with a full-flash strobe backlight, because of the light leakage problem of scanning backlights. The best strobe backlights outperform the majority of scanning backlights. In the past, you have to spend a huge amount of money (turning a monitor from $500 to a $5000 monitor) to pull off a good scanning backlight. But there are now some clever solutions to make a scanning backlight much cheaper.

Focussing the LEDs forward means you get beads. But trying to diffuse the beady-look means you will leak light all over the panel (including the far reaches), creating annoying strobe crosstalk (faint ghost images).

Fantastic ideas and talk though and I'd like to hear solutions of ideas of how to solve the law-of-physics problem of cheaply de-beading a FALD without light leakage to far reaches of the LCD panel. (This is why good FALD HDTVs are so fracking expensive) I've come up with some inventions in display testing so that is more the Blur Busters sauce nowadays, but our legacy is indeed practical strobe backlights that has revolutionized LCD motion blur reduction nowadays.

Digital De-Beading / De-Halo Algorithms
One possible is a GPU shader that does digital de-beading of backlights. Some TV manufacturers now do this to have more focussed FALD with less halo problem. The halo problem is exactly the same problem as strobe crosstalk during scanning backlights! Light leakage. Now, if you display gray in one part of the backlight, and you predict light leakage, you can display a darker gray or even black on the LCD, and then the pixel will be the correct brightness. Now you have less leakage issue. But it's even worse for scanning backlights because you don't have LCD black helping keep things dark. Another "OFF" part of the screen that needs to stay OFF (due to need to keep GtG transitions in dark) may have LCD panel white or bright colors in an OFF part of the backlight, and that amplifies light leakage visibility compared to haloing. So light leakage is much more critical for scanning backlights than for local dimming.

Doing de-beading digitally (if you have 10-bit or 12-bit color depth) solves the problem. I think this can now be done cheaply with a GPU shader, and solve a major scanning backlight expense problem without adding light scatter. The problem is creating the de-beading mask (alpha), but this could be done by an ultrasensitive photodiode and illuminating 1 pixel at a time, and running an overnight recorder. Cheap Arduino stuff that can be done DIY. And can be done realtime to Windows/games with less than a 5% performance penalty on most GPUs nowadays (albiet not slower GPUs or laptop GPUs).

Don't Forget Importance of Overdrive for Strobing / Scanning Backlights
Also, another thing is we need better overdrive for strobe or scanning backlights. We need GtG to be as close to 100% clean as possible (no overshoot or undershoot). Overdrive lookup tables are superior to Overdrive Gain formulas when done with either strobe backlights or scanning backlights. You want the LCD pixel color at the exact correct color the moment of the flash at the specific area of the screen. Some strobe backlights use a Y-axis variable overdrive strength, to accomodate the time-differential between scanout-in-dark versus global all-at-once flash, and that helps a lot. This can be done by simply incrementing overdrive gain (e.g. Strength 50 through Strength 80 on a Overdrive Gain 0...128 scale) from top to bottom in a linear fashion. This obviously, needs to be calibrated with a test pattern such as http://www.testufo.com/crosstalk .... But this reduces strobe crosstalk even further, up to the limit of law-of-physics. There are issues near fullblacks and fullwhites (fully saturated colors) since you need overshoot room below black and above white, for accurate overdrive at the extreme edges of the color gamut. That's very hard, so some vendors reduce contrast ratio sligthly (reduce digital gamut to only 90-95% original) to improve GtG accuracy at the moment of strobe flash -- that's why LightBoost had degraded colors. But with even faster LCDs (sub-1ms LCDs) this becomes less and less necessary to reduce gamut to improve strobe. Anyway, this is less than 10% of my "Blur Busters Einstein Brain", but it underscores how complicated high quality strobing or scanning-backlight can be.

Overdrive can be simultaneously combined with de-beading / de-haloing algorithms
A more advanced implementation would combine into a hybrid overdrive that is light-leakage aware. You just want the correct number of photons emitting for a specific pixel, and so accomodating all the complex factors (light leakage, GtG inertia, time differential between GtG-start and strobe-flash, etc) to try to get as close as possible to the correct number of photons emitted per pixel. It's getting easier and easier, but the algorithms can be horrendously complex. This separate the $500 displays from $5000 displays. The best LCD HDTVs has had some really advanced Ph.D math formulas in their FPGAs / ASICs / GPUs. But we can get 90% as good with just pure hobbyist math that is understandable by a high school graduate.

----

(This talk is why I love running Blur Busters. This used to be a hobby, even as it turned into a business for me -- Blur Busters Services).

P.S. Great animated artwork! Whose animations were these? I would love help from whomever did those animations, to create some "related" animations in the future. Did you create those animated GIFs yourself? Are they of a real experiment or a photoshop? I could use such skillz to create educational animations that raise the state-of-art of the industry. There's other explanation animations I need to explain strobe crosstalk behaviours and the importance of millisecond and sub-millisecond GtG for cheap-to-engineer high-quality strobe backlights. That's why I'm so enthusaic about 240Hz 1ms IPS as excellent candidate to get closer to CRT quality (for strobe or scanning backlights)....
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

       To support Blur Busters:
       • Official List of Best Gaming Monitors
       • List of G-SYNC Monitors
       • List of FreeSync Monitors
       • List of Ultrawide Monitors

billburrbuster
Posts: 9
Joined: 22 Oct 2019, 04:54

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by billburrbuster » 23 Oct 2019, 14:23

in the first post i used horizontal, perpendicular to a screen surface bars between LED rows to prevent leakage in the second gif, with blurbusters.com on the screen .
its probably not a DIY solution due to how monitor manufacturers arrange FALD, but its something that could be done on the factory.
Here is how it works in "light simulation"
https://imgur.com/a/SPmAoeM

User avatar
Chief Blur Buster
Site Admin
Posts: 6512
Joined: 05 Dec 2013, 15:44

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by Chief Blur Buster » 23 Oct 2019, 17:01

Digital leakage prevention bars (alpha-blend gradients to dim LCD lines that are closer to the light leakage) is indeed much simpler than a full 2D gradient map. This could easily be done with a math formula and sliders combined with a calibration test pattern.

There will be some clipping issues at the extremes (full darks / full brights ) where you can't get blacker-than-black or whiter-than-white without intentional gamut reduction.

Now, you still need to be reasonably bead-free closer to the LEDs for this to work. So a slight amount of de-beading is needed with the right optics. I was thinking of full 2D de-beading algorithm to permit even simpler optics.

Also, if modifying an existing edgelight panel, the optics sheets will probably need to be swapped, since it's a edgelight-optimized optics filter in sheets, scattering side-light to forward-light. This can amplify light leakage when modifying an edgelight panel to a fully backlit panel, so a filter swap to one optimized for backlighting. Something that will spread a little but not too much. A semi-focussed but even spread. This is a lot of experimentation for hobbyists, including trying to source the right diffuser/filter sheets, but often done in a display R&D lab.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

       To support Blur Busters:
       • Official List of Best Gaming Monitors
       • List of G-SYNC Monitors
       • List of FreeSync Monitors
       • List of Ultrawide Monitors

billburrbuster
Posts: 9
Joined: 22 Oct 2019, 04:54

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by billburrbuster » 23 Oct 2019, 18:03

maybe im not fully understand how things work, but i propose not digital bars, bur rather real material (plastic or whatever) bars between LED rows :D like this
bars.gif
bars.gif (29.69 KiB) Viewed 427 times
it allows a really diffuse material (or several layers of one) to be used, smoothing out almoust all beading, and still preventing light leakage (or keeping it moderate). no special optics needed to direct the beads, just regular omnidirectional LED beads maybe of bigger than regular size. even if beading still exists, its so weak it can be eliminated with really slight 2D map permanently applied to lcd matrix. the map can be hardcoded with electronincs of the monitor, without the need of gpu.
Of course it all applies only if leds are used only as backlight\scanline, using it as an HDR support at the same time would increase complexity significantly.

billburrbuster
Posts: 9
Joined: 22 Oct 2019, 04:54

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by billburrbuster » 24 Oct 2019, 05:25

another physical light simulation for comparison:
method.jpg
method.jpg (113.33 KiB) Viewed 400 times

billburrbuster
Posts: 9
Joined: 22 Oct 2019, 04:54

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by billburrbuster » 25 Oct 2019, 02:41

Denserly packed LED maxtrix (17 rows this time) and more finetuning = narrower scan line
Untitled-1.jpg
Untitled-1.jpg (21.51 KiB) Viewed 377 times

User avatar
Chief Blur Buster
Site Admin
Posts: 6512
Joined: 05 Dec 2013, 15:44

Re: Blurbusting Technology Suggestion: Scanning Backlight

Post by Chief Blur Buster » 21 Nov 2019, 20:30

Very nice stuff!
Extreme control of scanning-backlight bleed is definitely key to a superior scanning backlight.

Are you working with any current monitor manufacturer with this stuff? I could send this thread to vendors such as NVIDIA and AMD.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

       To support Blur Busters:
       • Official List of Best Gaming Monitors
       • List of G-SYNC Monitors
       • List of FreeSync Monitors
       • List of Ultrawide Monitors

Post Reply