- Dec 13, 2012
- 273
- 1
- 71
Hello!
I'd like to introduce myself as the BlurBusters Blog -- www.blurbusters.com
-- working on a zero-motion blur LCD project
-- hobbyist blog on my research project for a home-made monitor mondifcation
-- home-made 3D Lightboost style backlight hack for an existing computer monitor;
Short story: Done via using ultra-short, ultra-bright backlight flashes (shorter and brighter than CRT phosphor illumination), using 150 watts of LED per square feet of display: An insane amount of wattage 10 times bigger than the average current monitor at average brightness. Essentially, it a homemade turbo-charged equivalent to 3D Lightboost, but primarily designed for motion blur elimination during 2D. Allows simulating CRT flicker in an LCD, essentially.
--> zero motion blur LCD is possible as an expensive user modification hack to an existing gaming LCD monitor!!! <--
(by adding 250+ watts of LED's to a 24" monitor for a ultrashort strobe backlight)
First, a relevant Anandtech's article...
AnandTech recently had an excellent article about 3D LightBoost being a strobed backlight that also reduces motion blur. This is a good small explanation to excellent mass-market start to bringing an optional strobed mode to an LCD display backlight, even if it's not nearly short/bright enough to completely eliminate motion blur. Having worked in the home theater industry before, I have a good familiarity with displays and how LCD's work -- including how 3D LightBoost work. Also, I'm familiar with the existing commercially-developed strobed/scanning backlights found in high-end HDTV's for home theater. However, very few (almost none) computer monitors do this, and I'd like to do this myself, as a computer monitor hack -- a user modification of a computer monitor.
The dominant cause of motion blur is no longer pixel persistence
Most motion blur on today's recent LCD's is because of eye-tracking on a sample-and-hold display. LCD refreshes are continuously shining on most LCD displays. Pixel persistence is mostly in the first 2ms of a refresh on a good LCD display. (That's barely more than 10% of a refresh at 60Hz -- 2ms out of 16ms for 60Hz. And especially good 3D LCD panels that now erases most of pixel persistence after the first 2ms with so little remnants leftover nowadays). The vast majority of the remainder of motion blur is because our eyes are continuously moving even during a statically displayed LCD frame. (Proof: Science & References) Eye-tracking-based motion blur is separate of pixel-persistence-based motion blur.
The LCD panel technology is already here; some new panels is not a limiting factor.
The zero-motion-blur-*capable* LCD panel technology is here today: 3D 120/144Hz panels. These 3D 120Hz panels have been able to finally virtually erase pixel persistence effects very quickly well before the end of a refresh -- a pre-requisite for 3D shutter glasses operation, because both shutters are closed while waiting for the LCD to refresh. To make 3D possible, monitor manufacturers have worked very hard to erase pixel persistence effects as quickly and as much as possible. This has the fortunate side-effect of making the panels capable of zero motion blur, if using an appropriate super-charged strobed backlight.
But, how do we bypass pixel persistence as the motion blur barrier!?!?
You simply strobe the backlight on an already fully-refreshed LCD refresh; completely bypassing pixel persistence. Keep the backlight turned off while waiting for pixel persistence (so pixel persistence effects are 100% unseen by the human eye). THEN strobe the backlight on a fully-refreshed LCD image. Strobe 120 times per second for 120Hz. Your strobes can be shorter than the pixel persistence. With 8 milliseconds at 120Hz, and pixel persistence only 2ms, you have plenty of time to find a window to strobe the backlight for a short time period (e.g. 0.5ms or 1ms). One backlight strobe flash per refresh. Strobing like a CRT.
...But backlight technology isn't yet!
To flash the backlight for very short time periods, without a dim picture -- you need a REALLY bright backlight -- as bright as CRT phosphor. (Have you seen how insanely bright CRT phosphor is for a tiny fraction of a second? See high-speed-video of CRT scanning -- phosphor overexposes the camera).
You need....150 watts of LED per square feet to equal CRT phosphor brightness
Manufacturing such a theoretical ultrabright backlight is extremely expensive -- easily over $1000 of backlight technology for a 24" monitor -- in order to achieve a zero-motion-blur LCD.
.....Even the very few ultra-expensive monitors that use extremely bright turbocharged strobed backlights for 3D Lightboost 2 (e.g. Asus VG278H, and the upcoming Asus VG248QE), don't output enough lumens/wattage in sufficiently short enough strobes, to reduce motion blur to be less than a CRT, and most 3D Lightboost 2 features is only used in 3D mode (which is too bad -- as the optional strobing also benefits 2D motion blur reduction too, if it can be made an optional 2D feature).
...Fortunately, LED technology has advanced sufficiently
There are now several ways way to squeeze more than 250 watts of LED's in a single 24" monitor (I purchased a kilowatt worth of 6500K ultrabright white LED ribbons - see photos of my LED ribbon). As an alternate approach if it ends up easier (still in testing) -- I can do an edgelight: There are ways to cram 250 watts of light into an 24" edgelight (e.g. bulky box at top or bottom edge of an LCD panel) -- e.g. line of Cree 5-10W LED's overcurrented by 3x during strobes -- going through a long, small glass prism to focus all that LED light into a tiny slit at the bottom edge or top edge of a LCD. (This is the backup approach -- essentially, a bulky box edgelight at top or bottom edge -- to replace the puny thin edgelight -- a turbocharged edgelight 10x brighter, and simply re-use the LCD panel's existing edgelight optics like optical diffuser sheets etc).
Regardless, any of these user-modification options cost several hundreds of dollars because of the high cost of this amount of wattage of LED's.
-- LED's can be safely overdriven for short periods, so strobing can permit about 3-4x the amount of light from LED's versus steady state, to compensate for the long dark periods between strobes (e.g. for 120Hz, 0.5ms backlight flash followed by almost 8ms of darkness, every refresh). A bit of extra heatsinking and maybe a backlight fan needed (like a projector -- because my LED's are projector-bright) -- but still 10-20x more compact than a CRT, and technology will improve. (You may also notice that LED's are also used in some projectors too)
...How do you synchronize the strobes to the display refresh?
That's where a simple Arduino circuit comes in. It taps into the vertical blanking signal inside the monitor (a single solder connection into monitor's electronics), or from a DVI/HDMI dongle (VSYNC detect) with a wire going into the Arduino. There's a manual phasing adjustment (strobe timing offset) which can be adjusted on PixPerAn motion test until the motion blur completely disappears (ala CRT style). I've already created the circuit, but have not yet attached it to the LED's in a monitor.
...In fact, to eliminate any soldering into a computer monitor electronics (unmodified monitor electronics) -- I've also been able to do software-based VSYNC too from computer-to-Arduino USB signalling. It's accurate because I'm using microsecond timecoding via Direct3D programming API RasterStatus.ScanLine() (tells me how long ago the last VSYNC was) combined with programming API QueryPerformanceCounter() -- to make it 100% immune to CPU fluctuation and USB variances. The equivalent microsecond accurate timer in Arduino can then compensate for any early/late arrived VSYNC signalling. In fact, VSYNC's can be extrapolated inside the Arduino program for several seconds, so I only need to signal the VSYNC once every few seconds -- so it's also freeze-proof too). Viola, CPU-fluctuation proof, USB-ping-variability-proof, and software-freeze-proof. The hardware method is preferred (console compatible!), but the software method works & is practical for complete flexibility (if you prefer zero modifications to electronics of an LCD monitor)... It's actually open source code, will release the code sometime next year once I've got the complete desktop prototype fully built.
...What monitor am I modifying first?
Curretly testing the LED's and backlight optics on cheap old/discarded LCD's before I do destructive modifications to an expensive monitor -- first one to be modified will be either cheap Asus VG236H or Samsung S23A700D; since these are often on sale for cheap -- any good 3D 120Hz panel are an excellent candidate for a monitor hack like this. Ideally, the very best 3D panels with the most perfect refreshes (e.g. the 144Hz panels) will be the ones that perform the best with the monitor hack, but those panels are far more expensive.
...Wont it flicker??
Not worse than a CRT at the same refresh rate. If you loved CRTs, you'll love this project. If you hated CRT's, you'll not care. I'm targetting 120/144Hz strobes, so flicker should be a nonissue (for most eyes at least...) Just like CRT's at 120Hz; those don't flicker annoyingly. Besides, it's software configurable. You can turn on/off the strobing -- so it can just be a steady continuous 25-watt backlight instead of a short 250-watt flashes.
...Is it compatible with 3D shutter glasses???
Yes; it just behaves like a homemade 3D Lightboost - the extra bright strobes can be timed to occur while shutters are open (via the strobe timing phasing adjustment), and it's quite equivalent to the way nVidia seems to have apparently designed their 3D lightboost; it's just simply a strobed backlight -- very simple science.
...Does this add Input lag??
A very, very, very tiny bit. Half a frame average. A strobed backlight at 144Hz refresh, adds about 3.5 milliseconds of input lag. (You're already getting roughly this added input lag anyway, if you're enabling 3D Lightboost2 anyway) However, the elimination of 95% motion blur allows that perfect "CRT fluid motion" -- now you can identify faraway snipers in 3D FPS shooters far more than 3.5 milliseconds quicker, because you can see them sharply while turning/moving, without stopping turning/moving first (like you have to do with many LCD monitors in order to identify small details). The lack of motion blur far outweighs the approximately 3.5 milliseconds of extra input lag since you're more quickly identifying enemies that are no longer motion-blurred while you're moving about in videogames.
(Even so, playing single player games are wonderful on a CRT with beautiful 3D scenery with the immersive feel of zero motion blur on CRT ...and now the same feeling is now possible on LCD panels!)
...Is it really zero motion blur???
Not if you have robot eyes. But for human eyes; at 0.5 millisecond strobes, the motion blur is so tiny -- it is thus no longer perceptible to human eyes, and it finally looks "CRT perfect zero motion blur" -- on an LCD. Fast turns in 3D FPS becomes CRT-perfect. Nintendo/MAME becomes perfect "Nintendo smooth" pans, with no ghosting, no blurring, no trailing artifacts. And you can put the PixPerAn chase text squares 1 pixel apart, and you can see the 1-pixel-gap between the two moving square clearly; just like on a CRT. That's how "perfect" the motion can be on a zero-motion-blur-modified LCD monitor. Even when you lower the refresh rate, e.g. for 80fps@80Hz, there is still less motion blur than a stock LCD monitor at 120fps@120Hz (many CRT monitor users already know 60fps@60Hz on a CRT still has less motion blur than 120fps@120Hz on a LCD), saving a lot of GPU horsepower while still having the "CRT silky smooth motion feel" on an LCD panel.
....The only way to equal the motion blur (0.5ms) without flashing the backlight (CRT-style flicker), doing it sample-and-hold, is by doing the theoretical equivalent of 2000fps@2000Hz (1000ms/0.5ms = 2000) which is not possible to do. That's almost 20 times less motion blur than 120fps@120Hz -- meaning, display-based motion blur is no longer perceptible to human eyes, and thus you get the "CRT perfect motion" feel.
...Why aren't manufacturers doing this yet???? Where can I buy???
Cost, cost, cost!!! It's insanely expensive to put 150 watts per square feet of LED in a computer monitor, as a super-charged backlight or edgelight. Fortunately, Asus is leading the way with the newer 3D Lightboost2 system and the VG248QE monitor (coming in a few months), even though it uses strobe lengths longer than CRT phosphor decay, it's a big step in motion blur elimination (especially if Asus permits the user to adjust the monitor to strobe during 2D mode, for motion blur reduction/elimination!).
....
Keep an eye on my development progress on my blog; since being essentially a super-turbocharged home-made 3D Lightboost hack addition to an existing monitor -- it's probably of interest to some of you AnandTech readers, and I'm crosslinking to AnandTech's blog articles on these topics; thus I'm introducing myself for the first time in these forums now....
(zero motion blur LCD -- as a user modification to an existing LCD monitor)
Thanks!
Mark Rejhon
BlurBusters Blog -- My Hobby Project to Eliminate Motion Blur on LCD Displays
www.BlurBusters.com
I'd like to introduce myself as the BlurBusters Blog -- www.blurbusters.com
-- working on a zero-motion blur LCD project
-- hobbyist blog on my research project for a home-made monitor mondifcation
-- home-made 3D Lightboost style backlight hack for an existing computer monitor;
Short story: Done via using ultra-short, ultra-bright backlight flashes (shorter and brighter than CRT phosphor illumination), using 150 watts of LED per square feet of display: An insane amount of wattage 10 times bigger than the average current monitor at average brightness. Essentially, it a homemade turbo-charged equivalent to 3D Lightboost, but primarily designed for motion blur elimination during 2D. Allows simulating CRT flicker in an LCD, essentially.
--> zero motion blur LCD is possible as an expensive user modification hack to an existing gaming LCD monitor!!! <--
(by adding 250+ watts of LED's to a 24" monitor for a ultrashort strobe backlight)
First, a relevant Anandtech's article...
AnandTech recently had an excellent article about 3D LightBoost being a strobed backlight that also reduces motion blur. This is a good small explanation to excellent mass-market start to bringing an optional strobed mode to an LCD display backlight, even if it's not nearly short/bright enough to completely eliminate motion blur. Having worked in the home theater industry before, I have a good familiarity with displays and how LCD's work -- including how 3D LightBoost work. Also, I'm familiar with the existing commercially-developed strobed/scanning backlights found in high-end HDTV's for home theater. However, very few (almost none) computer monitors do this, and I'd like to do this myself, as a computer monitor hack -- a user modification of a computer monitor.
The dominant cause of motion blur is no longer pixel persistence
Most motion blur on today's recent LCD's is because of eye-tracking on a sample-and-hold display. LCD refreshes are continuously shining on most LCD displays. Pixel persistence is mostly in the first 2ms of a refresh on a good LCD display. (That's barely more than 10% of a refresh at 60Hz -- 2ms out of 16ms for 60Hz. And especially good 3D LCD panels that now erases most of pixel persistence after the first 2ms with so little remnants leftover nowadays). The vast majority of the remainder of motion blur is because our eyes are continuously moving even during a statically displayed LCD frame. (Proof: Science & References) Eye-tracking-based motion blur is separate of pixel-persistence-based motion blur.
The LCD panel technology is already here; some new panels is not a limiting factor.
The zero-motion-blur-*capable* LCD panel technology is here today: 3D 120/144Hz panels. These 3D 120Hz panels have been able to finally virtually erase pixel persistence effects very quickly well before the end of a refresh -- a pre-requisite for 3D shutter glasses operation, because both shutters are closed while waiting for the LCD to refresh. To make 3D possible, monitor manufacturers have worked very hard to erase pixel persistence effects as quickly and as much as possible. This has the fortunate side-effect of making the panels capable of zero motion blur, if using an appropriate super-charged strobed backlight.
But, how do we bypass pixel persistence as the motion blur barrier!?!?
You simply strobe the backlight on an already fully-refreshed LCD refresh; completely bypassing pixel persistence. Keep the backlight turned off while waiting for pixel persistence (so pixel persistence effects are 100% unseen by the human eye). THEN strobe the backlight on a fully-refreshed LCD image. Strobe 120 times per second for 120Hz. Your strobes can be shorter than the pixel persistence. With 8 milliseconds at 120Hz, and pixel persistence only 2ms, you have plenty of time to find a window to strobe the backlight for a short time period (e.g. 0.5ms or 1ms). One backlight strobe flash per refresh. Strobing like a CRT.
...But backlight technology isn't yet!
To flash the backlight for very short time periods, without a dim picture -- you need a REALLY bright backlight -- as bright as CRT phosphor. (Have you seen how insanely bright CRT phosphor is for a tiny fraction of a second? See high-speed-video of CRT scanning -- phosphor overexposes the camera).
You need....150 watts of LED per square feet to equal CRT phosphor brightness
Manufacturing such a theoretical ultrabright backlight is extremely expensive -- easily over $1000 of backlight technology for a 24" monitor -- in order to achieve a zero-motion-blur LCD.
.....Even the very few ultra-expensive monitors that use extremely bright turbocharged strobed backlights for 3D Lightboost 2 (e.g. Asus VG278H, and the upcoming Asus VG248QE), don't output enough lumens/wattage in sufficiently short enough strobes, to reduce motion blur to be less than a CRT, and most 3D Lightboost 2 features is only used in 3D mode (which is too bad -- as the optional strobing also benefits 2D motion blur reduction too, if it can be made an optional 2D feature).
...Fortunately, LED technology has advanced sufficiently
There are now several ways way to squeeze more than 250 watts of LED's in a single 24" monitor (I purchased a kilowatt worth of 6500K ultrabright white LED ribbons - see photos of my LED ribbon). As an alternate approach if it ends up easier (still in testing) -- I can do an edgelight: There are ways to cram 250 watts of light into an 24" edgelight (e.g. bulky box at top or bottom edge of an LCD panel) -- e.g. line of Cree 5-10W LED's overcurrented by 3x during strobes -- going through a long, small glass prism to focus all that LED light into a tiny slit at the bottom edge or top edge of a LCD. (This is the backup approach -- essentially, a bulky box edgelight at top or bottom edge -- to replace the puny thin edgelight -- a turbocharged edgelight 10x brighter, and simply re-use the LCD panel's existing edgelight optics like optical diffuser sheets etc).
Regardless, any of these user-modification options cost several hundreds of dollars because of the high cost of this amount of wattage of LED's.
-- LED's can be safely overdriven for short periods, so strobing can permit about 3-4x the amount of light from LED's versus steady state, to compensate for the long dark periods between strobes (e.g. for 120Hz, 0.5ms backlight flash followed by almost 8ms of darkness, every refresh). A bit of extra heatsinking and maybe a backlight fan needed (like a projector -- because my LED's are projector-bright) -- but still 10-20x more compact than a CRT, and technology will improve. (You may also notice that LED's are also used in some projectors too)
...How do you synchronize the strobes to the display refresh?
That's where a simple Arduino circuit comes in. It taps into the vertical blanking signal inside the monitor (a single solder connection into monitor's electronics), or from a DVI/HDMI dongle (VSYNC detect) with a wire going into the Arduino. There's a manual phasing adjustment (strobe timing offset) which can be adjusted on PixPerAn motion test until the motion blur completely disappears (ala CRT style). I've already created the circuit, but have not yet attached it to the LED's in a monitor.
...In fact, to eliminate any soldering into a computer monitor electronics (unmodified monitor electronics) -- I've also been able to do software-based VSYNC too from computer-to-Arduino USB signalling. It's accurate because I'm using microsecond timecoding via Direct3D programming API RasterStatus.ScanLine() (tells me how long ago the last VSYNC was) combined with programming API QueryPerformanceCounter() -- to make it 100% immune to CPU fluctuation and USB variances. The equivalent microsecond accurate timer in Arduino can then compensate for any early/late arrived VSYNC signalling. In fact, VSYNC's can be extrapolated inside the Arduino program for several seconds, so I only need to signal the VSYNC once every few seconds -- so it's also freeze-proof too). Viola, CPU-fluctuation proof, USB-ping-variability-proof, and software-freeze-proof. The hardware method is preferred (console compatible!), but the software method works & is practical for complete flexibility (if you prefer zero modifications to electronics of an LCD monitor)... It's actually open source code, will release the code sometime next year once I've got the complete desktop prototype fully built.
...What monitor am I modifying first?
Curretly testing the LED's and backlight optics on cheap old/discarded LCD's before I do destructive modifications to an expensive monitor -- first one to be modified will be either cheap Asus VG236H or Samsung S23A700D; since these are often on sale for cheap -- any good 3D 120Hz panel are an excellent candidate for a monitor hack like this. Ideally, the very best 3D panels with the most perfect refreshes (e.g. the 144Hz panels) will be the ones that perform the best with the monitor hack, but those panels are far more expensive.
...Wont it flicker??
Not worse than a CRT at the same refresh rate. If you loved CRTs, you'll love this project. If you hated CRT's, you'll not care. I'm targetting 120/144Hz strobes, so flicker should be a nonissue (for most eyes at least...) Just like CRT's at 120Hz; those don't flicker annoyingly. Besides, it's software configurable. You can turn on/off the strobing -- so it can just be a steady continuous 25-watt backlight instead of a short 250-watt flashes.
...Is it compatible with 3D shutter glasses???
Yes; it just behaves like a homemade 3D Lightboost - the extra bright strobes can be timed to occur while shutters are open (via the strobe timing phasing adjustment), and it's quite equivalent to the way nVidia seems to have apparently designed their 3D lightboost; it's just simply a strobed backlight -- very simple science.
...Does this add Input lag??
A very, very, very tiny bit. Half a frame average. A strobed backlight at 144Hz refresh, adds about 3.5 milliseconds of input lag. (You're already getting roughly this added input lag anyway, if you're enabling 3D Lightboost2 anyway) However, the elimination of 95% motion blur allows that perfect "CRT fluid motion" -- now you can identify faraway snipers in 3D FPS shooters far more than 3.5 milliseconds quicker, because you can see them sharply while turning/moving, without stopping turning/moving first (like you have to do with many LCD monitors in order to identify small details). The lack of motion blur far outweighs the approximately 3.5 milliseconds of extra input lag since you're more quickly identifying enemies that are no longer motion-blurred while you're moving about in videogames.
(Even so, playing single player games are wonderful on a CRT with beautiful 3D scenery with the immersive feel of zero motion blur on CRT ...and now the same feeling is now possible on LCD panels!)
...Is it really zero motion blur???
Not if you have robot eyes. But for human eyes; at 0.5 millisecond strobes, the motion blur is so tiny -- it is thus no longer perceptible to human eyes, and it finally looks "CRT perfect zero motion blur" -- on an LCD. Fast turns in 3D FPS becomes CRT-perfect. Nintendo/MAME becomes perfect "Nintendo smooth" pans, with no ghosting, no blurring, no trailing artifacts. And you can put the PixPerAn chase text squares 1 pixel apart, and you can see the 1-pixel-gap between the two moving square clearly; just like on a CRT. That's how "perfect" the motion can be on a zero-motion-blur-modified LCD monitor. Even when you lower the refresh rate, e.g. for 80fps@80Hz, there is still less motion blur than a stock LCD monitor at 120fps@120Hz (many CRT monitor users already know 60fps@60Hz on a CRT still has less motion blur than 120fps@120Hz on a LCD), saving a lot of GPU horsepower while still having the "CRT silky smooth motion feel" on an LCD panel.
....The only way to equal the motion blur (0.5ms) without flashing the backlight (CRT-style flicker), doing it sample-and-hold, is by doing the theoretical equivalent of 2000fps@2000Hz (1000ms/0.5ms = 2000) which is not possible to do. That's almost 20 times less motion blur than 120fps@120Hz -- meaning, display-based motion blur is no longer perceptible to human eyes, and thus you get the "CRT perfect motion" feel.
...Why aren't manufacturers doing this yet???? Where can I buy???
Cost, cost, cost!!! It's insanely expensive to put 150 watts per square feet of LED in a computer monitor, as a super-charged backlight or edgelight. Fortunately, Asus is leading the way with the newer 3D Lightboost2 system and the VG248QE monitor (coming in a few months), even though it uses strobe lengths longer than CRT phosphor decay, it's a big step in motion blur elimination (especially if Asus permits the user to adjust the monitor to strobe during 2D mode, for motion blur reduction/elimination!).
....
Keep an eye on my development progress on my blog; since being essentially a super-turbocharged home-made 3D Lightboost hack addition to an existing monitor -- it's probably of interest to some of you AnandTech readers, and I'm crosslinking to AnandTech's blog articles on these topics; thus I'm introducing myself for the first time in these forums now....
(zero motion blur LCD -- as a user modification to an existing LCD monitor)
Thanks!
Mark Rejhon
BlurBusters Blog -- My Hobby Project to Eliminate Motion Blur on LCD Displays
www.BlurBusters.com
Last edited:
... 
... 
... 
... 
Facebook
Twitter