DIY Backlight Strobing [must read for experimenters!]

Advanced display talk, display hackers, advanced game programmers, scientists, display researchers, display manufacturers, vision researchers & Advanced Display Articles on Blur Busters. The masters on Blur Busters.
HeLLoWorld
Posts: 33
Joined: 07 Jan 2014, 18:44

Re: DIY Backlight Strobing [must read for experimenters!]

Post by HeLLoWorld » 15 Jan 2014, 10:37

Thank you very much for explaining this lcd inerty, to my knowledge it is not well known and it is very important information.
More ideas come to mind then, regarding the algorithms and strategies one can develop to do better reaction times, and what do overdrive and such do.
Adding the strobing/shutter (if shutter opening is intentionally small) to the equation, one could imagine other pixel transition schemes where the desired value is deliberately overshooted (ie does not settle to correct value but higher), but is correct (but still rising) precisely at the moment of the flash, and with different target values for top and bottom of screen (even if already using accelerated scanout) (ie, bottom pixels overshoot more, because they must rise faster, because they are commanded later than top pixels). So the whole screen changes at different rates, is simultaneously correct only for one precise ms, and we flash during that very ms. Copyright HeLLoWorld :) (I just realize that at that point even the 1ms flash duration will maybe be too long if the values are rising fast) .I guess the overshooting algos (like overdrive probably is) must be complicated enough to additionnaly take in acccout the value of the desired color of the individual pixel and maybe the previous value can be a valuable information too.
This kind of things must have been quite thought of already of course.

HeLLoWorld
Posts: 33
Joined: 07 Jan 2014, 18:44

Re: DIY Backlight Strobing [must read for experimenters!]

Post by HeLLoWorld » 15 Jan 2014, 10:50

I had already seen all this by the way, I just had not the part about slow pixels and longer VBL.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: DIY Backlight Strobing [must read for experimenters!]

Post by Chief Blur Buster » 15 Jan 2014, 10:51

HeLLoWorld wrote:Schemes where the desired value is deliberately overshooted (ie does not settle to correct value but higher), but is correct (but still rising) precisely at the moment of the flash, and with different target values for top and bottom of screen (even if already using accelerated scanout)
Unfortunately the exact crossing occurs at different times for different GtG combinations, and at different temperatures (hotter = faster). Don't forget the frozen/cold LCD knowledge too. So you've got a big engineering challenge to make the zero crossing occur simultaneously for the whole screen. You'll need 3D or 4D buffers (X times Y times multiple refresh history), and probably thermosistors, to keep track of how much ghosting is occuring where, and compensate for that, as well as using temperature information to create such precise voltage/timing information, to aim the voltage crossing at correct times for all pixels simultaneously. Ugh. Precise control of a wide voltage range.

Why not do something simpler such as enlarging the blanking interval instead? :D
It's easier to just wait-out the overdrive rippling. Create a 4ms pause between refreshes, for a 1ms LCD, for example.

Or why not do a double-pass refresh? (heavily overdriven refresh, immediately followed by a clean-up refresh)
Some strobed LCDs use this approach.
HeLLoWorld wrote:(ie, bottom pixels overshoot more, because they must rise faster, because they are commanded later than top pixels).
Only if you use Y-axis-compensated overdrive algorithms, like LightBoost does.

But not all strobe backlights does this, and it is also often beyond the scope of hobbyist hacking.
Easier is to simply use a high bandwidth LCD (capable of, e.g. 240Hz scanout -- like some HDTV LCDs are) -- and just enlarge the blanking interval as large as possible. Many new 240Hz LCDs now exist in the television industry, but refreshed from interpolation engines rather than direct-driven from a computer. But this is new DIY hacking territory that cirthix has finally accomplished (240Hz LCD), and hopefully manufacturers will follow suit. But 240Hz LCDs present immense opportunities for accelerated-scanout (e.g. refreshing a slower refresh cycle in a 1/240sec scanout). Direct-driving a 240Hz-capable LCD is geek territory. Ordering bare panels from places like Alibaba (china factory bazaar), designing your own driver circuit to drive the panel, etc. DIY territory finally broken by the likes cirthix.
HeLLoWorld wrote:So the whole screen changes at different rates
Only for Y-axis-compensated overdrive. Not for most LCDs, but only on some LCDs.
It is not used for the purposes of timing the strobe for overdrive overshoot, but to attempt to speed up the LCD pixels along the bottom edge of the screen. During Y-axis compensated overdrive, it appears to sometimes have quantitization bands that occur due to rounding errors: e.g. viewing http://www.testufo.com/flicker in LightBoost (Height -> "Full Screen") reveals the bands of Y-axis compensated overdrive.
HeLLoWorld wrote:I guess the overshooting algos (like overdrive probably is) must be complicated enough to additionnaly take in acccout the value of the desired color of the individual pixel and maybe the previous value can be a valuable information too.
They already do, overdrive by definition, always uses previous color (at the minimum). That's what the memory chips in an LCD screen is often used for -- overdrive color processing. Overdrive always use previous values, and sometimes the previous 2-3 values, if the LCD is slow enough that multiple-level history helps. Sometimes they also buffer so they know the NEXT value too. That's why some LCDs add input lag when you enable overdrive because lookforward requires adding input lag. Fortunately, this does not happen with Asus Trace Free or BENQ AMA, they use only history buffers, not lookforward buffers.

The VG248QE is pretty good at ghost-free strobing. The overdrive ghosting is almost below human perceptible levels (<1% delta between visible value and correct value) along the horizontal center band of the screen.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

HeLLoWorld
Posts: 33
Joined: 07 Jan 2014, 18:44

Re: DIY Backlight Strobing [must read for experimenters!]

Post by HeLLoWorld » 15 Jan 2014, 11:21

Thank you very much. By the way I was not planning to modify the refresh process myself.
However, while discussing this the other day, the idea came out that the processing could be done by the gpu instead, and the panel stripped off of a bunch of things, if the panel can declare its properties. Can't decide if it's elegant or ugly :)

And by the way, I still think the curtain effect is nondesirable (so the strobing is good) for visual quality, so fastest possible accelerated scanout is desirable, even for continuous backlight, even if that means 1 link period lag (not necessary 1 frame if link is faster) for the top pixels.
Solution is always the same: more megahertzs on the link, less ms in the scanout, more hertz for the total refresh cycle, more mips on the gpu :)
Last edited by HeLLoWorld on 15 Jan 2014, 11:30, edited 1 time in total.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: DIY Backlight Strobing [must read for experimenters!]

Post by Chief Blur Buster » 15 Jan 2014, 11:28

HeLLoWorld wrote:Thank you very much. By the way I was not planning to modify the refresh process myself.
However, while discussing this the other day, the idea came out that the processing could be done by the gpu instead, and the panel stripped off of a bunch of things, if the panel can declare its properties. Can't decide if it's elegant or ugly :)
Yep! The GPU can process overdrive, it's already been done by AMD years ago:
http://www.techpowerup.com/forums/threa ... 7-5.32134/

It may add a slight amount of input lag unless you do it in realtime as the pixels get output. And you can't have below-zero voltage or above-white voltage like you sometimes can with real-world LCD overdrive. Sometimes having that is advantageous, to eliminate ghosting in dark greys and very bright near-white greys. You can compensate by narrowing your greyscale to leave unused room below your black and above your white, to have better-looking software-based overdrive.

It's better to try to do it in the hardware, tho. It is also frequent in strobe backlight LCDs to reduce the dynamic range, in order to have better overdrive overshoot margin near the top/bottom ends. That's why LightBoost often only has 500:1 contrast ratio, versus non-LightBoost having 1000:1 -- extra overdrive margin. However, you want to be able to adjust the amount of margin you want, for a ghosting-versus-contrast-ratio tradeoff. I tend to prefer more contrast ratio, even if it means a very slight amount of ghosting. Very few LightBoost monitors let you do this (not the VG248QE, but the VG278H does -- I can gain >950:1 contrast in LightBoost mode via monitor OSD Contrast=90 to 94% on the VG278H original, not the HE-suffix or QE-suffix)
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: DIY Backlight Strobing [must read for experimenters!]

Post by Chief Blur Buster » 15 Jan 2014, 11:29

HeLLoWorld wrote:so that fastest possible accelerated scanout is desirable, even for continuous backlight
Accelerated scanout isn't needed for continuous backlight.
It does nothing to speed up individual LCD pixel transitions, and actually degrades color quality by spending less time juicing voltage into pixels (this is why 144Hz has lower color quality than 60Hz).

The main purpose of the accelerated scanout is simply to create more idle time between refreshes, which as explained, is quite useful for high quality strobing.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

HeLLoWorld
Posts: 33
Joined: 07 Jan 2014, 18:44

Re: DIY Backlight Strobing [must read for experimenters!]

Post by HeLLoWorld » 15 Jan 2014, 11:32

No, I mean it creates the curtain effect, I think it is better to have entire images simultaneously for movement quality. No proof though.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: DIY Backlight Strobing [must read for experimenters!]

Post by Chief Blur Buster » 15 Jan 2014, 11:35

HeLLoWorld wrote:No, I mean it creates the curtain effect
Aha, you mean the tilting effect seen at http://www.testufo.com/blurtrail (Height->Full Screen).
Scanning on CRTs and LCDs definitely create a tilt effect, especially at 60Hz.
I've never heard of anyone call it the "curtain effect", but you're right.

Yes, accelerated scanout helps fix that. And strobing eliminates that.
But you get the tradeoffs such as lower color quality during accelerated scanout.
So it's really a pick-your-poison.

Fun Academic Expercise: The tilt effect occurs on iPads too in landscape mode. :D
View http://www.testufo.com/blurtrail and pinch-zoom until fills the full screen.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

HeLLoWorld
Posts: 33
Joined: 07 Jan 2014, 18:44

Re: DIY Backlight Strobing [must read for experimenters!]

Post by HeLLoWorld » 15 Jan 2014, 12:01

Not exactly, more simply, the fact the image is not refreshed everywhere simultaneously, although the frame represents an infinitely brief moment.
I called it like that because I'm French.

Effect 1. Let's say you've got a crt. It refreshes top-down. If you track a fast moving object, your eyes move while it draws, so it might appear compressed on your eyes if it's moving vertically.

Effect 2. Your eyes track a background moving left to right. your eyes have moved between top refreshed and bottom refreshed, but the picture is from the same moment in time. This must cause distortion in your perceived image.

This doesnt appear with strobing.

Something similar must happen on lcds refreshing top-down as slow as a frame, although this effect is maybe washed out by the constant backlight already, maybe it's something subtle that becomes significant only when constant backlight problem is solved (example: crts)

These are thought experiments btw.
I deduced them from the eye tracking problem and the idea from Abrash to adapt the rendering for each pixel as they are physically scanned.

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: DIY Backlight Strobing [must read for experimenters!]

Post by Chief Blur Buster » 15 Jan 2014, 13:02

We're talking about the same thing, actually. :D
HeLLoWorld wrote:Effect 1. Let's say you've got a crt. It refreshes top-down. If you track a fast moving object, your eyes move while it draws, so it might appear compressed on your eyes if it's moving vertically.
Yes, that's the vertical compression/expansion effect. This also shows up in camera photographs if you slowly tilt them upwards/downwards while taking a picture of a CRT.
HeLLoWorld wrote:Effect 2. Your eyes track a background moving left to right. your eyes have moved between top refreshed and bottom refreshed, but the picture is from the same moment in time. This must cause distortion in your perceived image.
Yes, that's the tilting effect, I was talking about. It also occurs in photographs of CRT if you tilt your camera sideways slowly while taking a picture of a CRT, too. The tilting effect is caused by your eye movement perpendicular to the scan direction.
HeLLoWorld wrote:This doesnt appear with strobing.
Correct.
HeLLoWorld wrote:Something similar must happen on lcds refreshing top-down as slow as a frame, although this effect is maybe washed out by the constant backlight already, maybe it's something subtle that becomes significant only when constant backlight problem is solved (example: crts)
Yes, on LCD, for left/right motion. You can see it when dragging windows left/right on a 60Hz LCD. You will see the window become a parallelogram, that's the curtain effect / tilting effect.
There is no noticeable compression for up/down motion on LCDs due to the continuous nature.
But the tilting effect is the same intensity on CRT/LCD for left/right motion.

It's funny when we call strobing as a solution to constant backlight, when real life isn't strobed -- the problem is we need strobing to fix the persistence problem of finite-refresh-rate displays. The theroetical perfect display is framerateless & strobeless (infinite framerate, continuous motion) but we can't easily engineer that.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

Post Reply