240 Hz strobe BenQ?

Everything about displays and monitors. 120Hz, 144Hz, 4K, 1440p, input lag, motion tests & TestUFO, monitor decisions, and more. Questions? Just ask!

240 Hz strobe BenQ?

Postby Vega » 21 Apr 2017, 11:11

http://www.benq.us/product/gaminggears/xl2546/

So I've had the Xl2735, which only strobes the back light with their marketing term "DyAc". According to the above reference, the monitor strobes at 240 Hz. I thought even the fastest LCD transitions wouldn't make this possible without serious motion degradation?
Vega
 
Posts: 107
Joined: 18 Dec 2013, 21:33

Re: 240 Hz strobe BenQ?

Postby Chief Blur Buster » 21 Apr 2017, 11:30

The short answer: 182Hz strobing looks better on XL2540 (and probably XL2546 too). 240Hz has too much strobe crosstalk.

The long answer: Good 240Hz strobing is mathematically theoretically possible on 1ms TN LCDs.
It's not yet being done, but it will be hugely expensive (and still probably not good colors).

(if a good low-diffusion scanning backlight is used, and using ultra-well-optimized strobe-friendly nonlinear overdrive tables like NVIDIA uses, but optimized for a scanning backlight -- but such a backlight hasn't been built yet for a 1ms TN panel)

Nothing stops a panel from being strobed at a frequency far beyond LCD response limitations. The problem is if you do so, you get nasty strobe crosstalk (worse double image effect) -- see Motion Blur Reduction FAQ

Understanding this subject matter better requires a better understanding of how an LCD monitor refreshes.

It's not just an LCD transition issue, but also scanout, Vertical Totals or other internal logic (the ratio of LCD scanout time versus waiting between refreshes).

For example, most LightBoost monitors internally scans out their LCD in approximately 1/175th of a second and pauses between scanouts (to let GtG finish before flashing the backlight/edgelight). The scanout appears as a top-to-bottom "vertical wipe" effect seen in high speed video. The thickness of the blurry fade is the 1ms-wide blurry zone (the GtG zone) seen in the video. Scan fast enough, and you create enough time to let GtG finish between refresh cycles, and spends the rest of the time pausing between refresh cycles.

The slower the LCD GtG, the thicker the blurry "vertical wipe" zone becomes in seen in high speed video. Eventually, the vertical wipe zone becomes taller than the screen -- like they were on old 33ms LCDs, and they overlapped refresh cycles, which meant multiple refresh cycles overlapped in an ugly way.

Photo of http://www.testufo.com/flicker on old 33ms LCD
3-refresh-cycle overlap on old 33ms LCD -- "06", "07" and "08" overlap.
Image

It only became possible to do good Blur Reduction once most of LCD GtG became a tiny fraction of a refresh cycle. Even with 240Hz LCD (4.1ms refresh cycle), there is still ways to transition a TN LCD pixel in 0.5ms-1.0ms. The problem is it's very hard to do so accurately. You need devilishly well-optimized overdrive tables that gets LCD pixels as close as possible to its exact final value (no undershoot, no overshoot).

Modern gaming monitors now easily get a TN LCD pixel near its final value within ~1ms, but doing it accurately is nearly impossible. I have seen improvements of ~10% GtG error to approximately ~1% GtG error (middle of screen, before strobe flash before next refresh cycle) with a lot of complexity (points of diminishing returns + expensive engineering). Some vendors such as NVIDIA has accomplished wonderful things with strobe-optimized overdrive tables (in conjunction with a fast scanout) in order to have LCD pixels darn nearly at its exact value. Sometimes contrast is sacrified to improve GtG accuracy (which is why LightBoost often has poor contrast, in exchange for low strobe crosstalk). On the ASUS VG248QE LightBoost, you can achieve >99% GtG accuracy on TN LCDs with a well-optimized strobe-optimized overdrive table (ghosting-free/corona-free) that's precisely tuned to correct GtG value, at the time of the strobe-backlight flash. Often, sacrifices are made (worse-than-non-strobed-TN colors) to get darn near zero strobe crosstalk, but the situation is gradually improving. However, not all manufacturers use strobe-optimized overdrive tables (BENQ does not) so you need to use huge Vertical Totals to compensate (faster scanout + longer pause between scanouts, to let GtG finish before strobe-flash)

The pause between refresh cycles is what is needed.

Mathematically, a 1ms TN LCD is rated GtG from the 10% to 90% point (takes much longer to do 0->10% GtG and 90%-100% GtG) and we're already pushing limits to eliminate ghosting during strobing. But as we already know, the ASUS VG248QE and a few other LightBoost displays manage to successfully eliminate most strobe crosstalk (double image effect) with just a mere ~2-3ms pause between refresh cycles.

Now, a 240Hz LCD is 1/240sec which is 4.1ms per refresh cycle.

Most of them can't scan out faster than 1/240sec, so there's no pause between refresh cycles to wait for the 1ms GtG (really 2-3ms for well-strobe-optimized overdrive tables) to finish in total darkness before flashing the backlight.

Now, mathematically, if you can finish a 240Hz refresh cycle in approximately 1/500sec (2ms), you create approximately ~2ms pause between refresh cycles to successfully achieve reasonable-crosstalk 240Hz strobing.

The current problem for 240Hz LCD is none of them can do a 1/480sec scanout at full resolution, necessary for nearly-crosstalk-free strobing.

Now, if you use a high-quality scanning backlight (good optics & low light leakage to other parts of screen) instead of a global flash-all-at-once, you can tighten the margin a lot more. You can flash a fully-refreshed part of the display (the part that doesn't have the blurry-wipe zone seen in high speed video) while waiting for a different part of display to finish refreshing (aka the blurry wipe-zone you see in high speed video). So in theory you don't need a long pause between refresh cycles. The problem is a lot of scanning backlights are not very well made, with lots of light leakage (see the very old 2012 Scanning Backlight FAQ) between the ON-part of the display versus the OFF-part of display. High-end scanning backlights use good optics to maximize contrast between the ON area and the OFF area, to minimize strobe crosstalk. The better the light control, the lower the strobe crosstalk (double image effect) of a scanning backlight.

So in theory, if a monitor manufacturer doubles the price of a monitor by adding a hugely expensive local-dimming scanning-backlight, you can in theory strobe a 240Hz 1ms TN LCD with roughly similar strobe crosstalk as a globally-strobed-all-at-once 120Hz 1ms TN LCD.

Confused? I could write a book, but I've also created Electronics Hacking: Creating a Strobe Backlight for the advanced DIY geeks wishing to understand strobe backlight technology better.

So, in other words, with a good scanning backlight or ultrafast scanout (one or the other), you can have good 240Hz strobing on a 1ms TN LCD. Unfortunately, neither has yet been achieved (either a expensive backlight or an ultrafast 1/480sec scanout of a 240Hz refresh cycle), so we aren't yet at a point on relatively crosstalk-free 240Hz strobing.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter!
User avatar
Chief Blur Buster
Site Admin
 
Posts: 3010
Joined: 05 Dec 2013, 15:44

Re: 240 Hz strobe BenQ?

Postby Falkentyne » 21 Apr 2017, 11:37

So how is this "XL2546" monitor different than the XL2540 that's already out?
Simply because the XL2540 has blur reduction disabled in the service menu and "hidden" from normal users, while the XL2546 has it ENABLED in the service menu and renamed as "DyAC?"

Can someone please tell me just what BS Benq is trying to pull here?

P.S. nice post, Chief, as always!
Falkentyne
 
Posts: 2279
Joined: 26 Mar 2014, 07:23

Re: 240 Hz strobe BenQ?

Postby spacediver » 21 Apr 2017, 13:57

Falkentyne wrote:P.S. nice post, Chief, as always!


indeed :)
spacediver
 
Posts: 472
Joined: 18 Dec 2013, 23:51

Re: 240 Hz strobe BenQ?

Postby Chief Blur Buster » 21 Apr 2017, 14:09

Here's a simpler version that's easier to understand!
I've now edited-up a diagram of proper quality strobing.
A picture is worth a thousand words.

Image

First, observe the high speed video. It is a high speed video of http://www.testufo.com/flicker where the screen flashes rapidly between two colors, while displaying incrementing numbers. This is very educational to learn how an LCD refreshes.

Video that educates you how an LCD panel refreshes!
It's a high speed video of an LCD panel refreshing. In both LightBoost and non-LightBoost modes.
phpBB [video]


Now that you roughly understand how an LCD panel refreshes, you understand that only the fully-refreshed part of LCD (no double images) should be flashed.. Which means you can do one of two things:

(1) Use a global flash, all at once. This is possible on a cheaper LCD panel with edgelight/backlight.
....You need an LCD panel with a GtG much faster than the time period of a refresh cycle. Scan the LCD top-to-bottom faster so you have a longer idle period between refreshes (at the same refresh rate). The longer the idle period between refresh cycles, the less likely/visibly that the LCD GtG fade zone doesn't overlap into the next refresh cycle (weaker strobe crosstalk for top/bottom edges). Flash the LCD when the whole screen is simultaneously clearest.
Requirement: You can use existing cheaper LCD panels. Just do a fast scanout and do a long pause between refresh cycles. And use good overshoot-free overdrive. You're essentially trying to squeeze the whole LCD GtG response into the blanking interval between refresh cycles -- "that pause between scanouts".

OR

(2) Use a scanning backlight with good optics.
....This can help slower LCDs, or push closer to GtG limits. You can even use reduced blanking intervals which makes crosstalk-free global-strobe impossible. Scanning backlights often require full-array backlights (which can have other benefits such as local-dimming and HDR capability, with higher contrast ratios). You flash a row of LEDs behind the fully-refreshed part of LCD, while keeping the "GtG wipe zone" in total darkness. You also need good full-array backlight optics to prevent leakage of light from the illuminated section into the dark section. By doing this, it is possible to reduce strobe crosstalk on slower LCD panels (or a fast LCD at even higher refresh rates, e.g. 240Hz TN with a theoretical FALD scanning backlight).
Requirement: An expensive full-array backlight, although it can do double-duty for HDR and local-dimming (Potentially 1,000,000:1 contrast ratio in some cases). Good diffusion-free backlight optics to keep rest of panel as totally-black as possible, while flashing a different part of LCD. Very difficult. Dramatically raises cost of monitor. Already successfully used in some high-end HDTV sets. Only recently starting to arrive in gaming monitors, including the new "4K 144Hz" HDR-capable panels.

NOTE: Little known, but scanning backlights can also potentially reduce input lag over global-strobe backlights -- and potentially reduce some of the strobe-amplified VSYNC OFF microstutters (due to the time-delta inconsistency between LCD scanout and global-strobe, for top/center/bottom edges during VSYNC OFF).
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter!
User avatar
Chief Blur Buster
Site Admin
 
Posts: 3010
Joined: 05 Dec 2013, 15:44

Re: 240 Hz strobe BenQ?

Postby Vega » 21 Apr 2017, 18:15

The question would be is BenQ really going to release a native 240 Hz strobing monitor with massive cross-talk? Or how are they running it above the magic 182 Hz number?
Vega
 
Posts: 107
Joined: 18 Dec 2013, 21:33

Re: 240 Hz strobe BenQ?

Postby Chief Blur Buster » 21 Apr 2017, 18:52

Vega wrote:The question would be is BenQ really going to release a native 240 Hz strobing monitor with massive cross-talk? Or how are they running it above the magic 182 Hz number?

First, let me simplify things --

The good news is that XL2540 and XL2546 has an adjustable strobe timing/phase support, and supports large Vertical Totals via Custom Resolution Utility. The XL2540 already has strobing support (like XL2546) but it's hidden in a Service Menu. The original XL2540 also strobes at all refresh rates, including 240Hz.

1. There's almost no strobe crosstalk at lower refresh rates (120Hz and less) when using large Vertical Totals (via ToastyX)
2. Single strobing is supported at all refresh rates in sub-1Hz-increments above ~75Hz.
3. It's an analog continum of gradually worsening strobe crosstalk, the higher the refresh rate you go.
4. Therefore, it's a user choice. You use the highest refresh rate that you can tolerate the strobe crosstalk on.

De-facto, the Blur Busters Community has declared 182Hz as a compromise for good strobing on the XL2540 -- after proper calibration of Strobe Phase ("Area"), the strobe crosstalk is roughly similar to current ULMB monitors running at their maximum refresh rates. For some refresh rates, the bottommost 1/10th sliver of screen and topmost 1/10th sliver of screen may be worse than ULMB -- while center of screen has almost zero strobe crosstalk (i.e. sometimes better than ULMB). It really depends on refresh rate and Vertical Total tweaking. Some people will prefer lower (even less strobe crosstalk), and some people will prefer higher (more strobe crosstalk).

Example: You tested a "X" Hertz and it looks good. You think you can tolerate maybe 2% more strobe crosstalk? Try bumping Hz upwards by say "X + 1Hz" or "X + 5Hz". Then you keep trying until the strobe crosstalk begins to look slightly too visible. Then you back off a bit. And stick to your preferred Hz. Above ~180-200Hz(ish) the crosstalk will rapidly get worse and fiddlier as your ability to use Vertical Totals slowly diminish (VT becomes smaller and smaller till it's a reduced blanking interval) -- you're forced to use reduced totals at the maximum refresh rate (i.e. display cannot scan-out faster than 1/240sec at 240Hz). You're pretty much increasing scanout speed as much as possible via the huge Vertical Total trick -- basically configuring your scanout to scan 1/240sec, no matter what refresh rate you're targetting. The higher your refresh rate, the less time between the 1/240sec scanouts, which is incrementally less-and-less room for the ~1ms LCD GtG (to mostly complete) -- and we know "1ms GtG" hype isn't really "1ms-perfect-complete" for all color combos. You're see increasing intensity of strobe crosstalk a percent (or few) for every 1Hz step upwards that you test....as you test 190Hz,191Hz,192Hz,193Hz....in the ever-marching testing towards 240Hz. There's a sweet spot compormise, and the Community here decided (give or take) that best compromise is occuring at (approximately) 180Hz(ish) for what looks roughly ULMB-quality strobing.

The larger the Vertical Total, the more the strobe crosstalk disappears. (For example, every 0.1ms added to Vertical Total can fade the strobe crosstalk by another 10%). Until you bang hard against the dotclock limit of your monitor. >VT2000+ is supported at 120Hz (which means it scans out a 120Hz refresh cycle in approximately 1/240sec -- creating a healthy 4ms pause between refresh cycles to let 1ms GtG finish), and >VT1500+ is supported at 182Hz, which are excellent. If you don't mind say, 10% more strobe crosstalk, you can go to 190Hz or 195Hz, for example -- and so fourth. It's a continuum of strobe crosstalk intensity, as XL2540 supports refresh rate in roughly sub-1Hz increments. You simply choose a refresh rate and then maximize the Vertical Total, and adjust the strobe timing (via "Area" setting in Service Menu) so that the flash is earlier/later -- minimizing the double-image effect at either top or bottom edge. If you adjust strobe timing (Area setting) while watching a full-screen TestUFO.com/photo -- you'll notice the horizontal "strobe crosstalk zone" (LCD GtG zone) move upwards and downwards. And the smaller the Vertical Total, the more the horizontal "strobe crosstalk zone" (LCD GtG zone) disappears off the bottom edge of the screen before wrapping-around to the top edge of the screen.

The higher the refresh rate, the taller the height of the strobe crosstalk zone becomes. So it's great user choice that BENQ let the user be able to choose how much strobe crosstalk they can tolerate -- the more the user tolerates it, the higher the refresh rate. There's almost near-zero strobe crosstalk at 1080p@120Hz (better than LightBoost) when strobe timing is properly manually adjusted in Service Menu (remember: BENQ is one of the few manufacturers with a user-adjustable strobe flash timing adjustment in Service Menu), and the strobe crosstalk gradually worsens the more you increase refresh (in 1Hz steps) towards the maximum. At 240Hz, you're stuck with Reduced Blanking (That means Vertical Totals that are barely bigger than vertical resolution -- this is bad for strobe crosstalk), but the more you lower the refresh rate, the bigger your Vertical Total becomes, the less strobe crosstalk occurs, in an analog continuum of user choice. That's the great thing...

Yes, it's kinda complex to adjust, but at least the user choice is there.

BENQ XL2720Z users have been doing this for years now, with Strobe Phase ("Area") / Strobe Length support. Fortunately, adjustments still remain in the XL2540 for optimizations. Obviously, you will get lots of strobe crosstalk out-of-box, and you have to use ToastyX and Large Vertical Totals to make strobe crosstalk magically disappear.

The XL2546 makes it much easier to enable and access strobing. It might even support even larger Vertical Totals than XL2540 (untested), which may reduce strobe crosstalk on XL2546 over XL2540.

If you don't have an hour to adjust strobe phase, don't bother getting the "DIY Vertical Total" method of reducing strobe crosstalk -- it has a learning curve involved in order to manually adjust to make things look as good as ULMB at the highest refresh rate possible (~160-200Hz(ish) depending on user threshold of tolerance for strobe crosstalk). NVIDIA ULMB makes it easy by automatically doing everything for you, but you're limited to certain refresh rates.

The most problematic monitors were the Version 1 of the XL2720Z which had an unadjustable strobe timing that created an ugly fixed (non-adjustable) strobe crosstalk zone (LCD GtG zone). A stationary double-image zone as a horizontal band that stayed stationary about 2/3rds down from the top edge of the screen.

Image

That's what happens if you flash a strobe backlight (Blur Reduction) while the LCD GtG Zone is still on the screen. Not good for an image-quality aim, although it might slightly reduce lag at the crosshairs (This might have been what BENQ was originally tying to do, as input lag is lower just above the crosstalk zone, than below the crosstalk zone -- I'm imagining BENQ probably put a lag sensor in the middle of the screen, and adjusted strobe timing until minimum lag -- creating this visually-flawed XL2720Z setting).

Many users complained, and BENQ released a firmware update (Version 2) letting the user adjust strobe flash timing so that this artifact can be fixed. Adjusting while watching a TestUFO test, causes the crosstalk-zone (LCD GtG Zone) to move upwards/downwards, as the flash timing is changed. You adjust until you push this zone off the bottom edge of the screen. Unfortunately, when Vertical Totals get tight, the crosstalk will reappear at top edge long before most of the crosstalk is pushed below bottom edge of screen. Now, the bigger you can make Vertical Total, the more you can push the crosstalk zone off the bottom edge of the screen without affecting the top edge of the screen.

(ULMB, while it does things far more automatically out-of-box -- has the same top-and-bottom-edge crosstalk problem too, and NVIDIA makes compromises by choosing specific fixed settings, usually in combination with a Y-axis-compensated LCD overdrive table that uses more aggressive overdrive at the bottom edge of the screen... That's a topic for another day)

At this time, BENQ has kept strobe timing adjustable on all XL Z-Series (Version 2 and later) and their Zowie variants, on XL2420G, on XL2735, on XL2540 (complex adjustments via Service Menu), and XL2546. It's possible to manually tweak timing/phase until it looks as good as ULMB (and better, if using a lower refresh rate and massive vertical totals!). It's a lot more work, but very rewarding when you're rewarded with a better-than-ULMB after 1 hour of manual tweaking.

Since then, BENQ now lets users adjust most of their newer 144Hz-240Hz monitors to preference -- priority to lag versus priority to minimizing strobe crosstalk. Not all users realize that they can fix strobe crosstalk on their BENQ monitors via the service menu, and via Large Vertical Totals in CRU. Some users have bought a BENQ, saw terrible strobe crosstalk, and returned the monitor -- without realizing that spending an hour adjusting this can make it look practically as good as ULMB (As long as you "back off" the refresh rate by approximately 25%-ish).

The only BENQ strobed monitors (post-LightBoost) with unadjustable strobe crosstalk XL2720Z (Version 1) and XL2730Z (all versions). All the other BENQ/Zowie can all successfully strobe in 1Hz increments. Including all 240Hz BENQ monitors, have strobing that becomes very good when combining BOTH (A) proper Vertical Total adjustments and (B) Adjusting "Area" setting, and (C) Choosing a refresh rate where strobe crosstalk doesn't bother you.

Anyway... XL2540/XL2546 has totally adjustable strobe crosstalk, which is a Great Thing. They actually let you CHOOSE which refresh rate you begin to hate strobe crosstalk at. (Usually that's >180Hz for most people, at least on the earlier XL2540).

Understand better?

...Yep, Blur Busters will probably need to create an easy "How To Reduce Strobe Crosstalk on BENQ Monitors" instruction page. There's already sort of instructions at http://www.blurbusters.com/strobe-utility -- but it doesn't work on XL2540s, but the same TestUFO "Alien Invasion" test can be used while adjusting the BENQ Service Menu "Area" setting -- while experimenting with new large Vertical Total modes in ToastyX CRU. (To speed things up, there's already a 182Hz ToastyX CRU screenshot in one of these threads)
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter!
User avatar
Chief Blur Buster
Site Admin
 
Posts: 3010
Joined: 05 Dec 2013, 15:44


Return to General - Displays, Graphics & More

Who is online

Users browsing this forum: Bing [Bot] and 6 guests