Severe Crosstalk with new XL2546?

Adjusting BENQ Blur Reduction and DyAc (Dynamic Acceleration) including Blur Busters Strobe Utility. Supports most BenQ/Zowie Z-Series monitors (XL2411, XL2420, XL2720, XL2735, XL2540, XL2546)
Post Reply
robben44
Posts: 2
Joined: 06 Jun 2019, 16:49

Severe Crosstalk with new XL2546?

Post by robben44 » 06 Jun 2019, 17:09

Hi everyone,

I've read quite a bit on this site about blur reduction, crosstalk, etc.

After a lot of research, I bought a Benq Zowie XL2546 today.
I mainly play Overwatch and I wanted to get a monitor that would give me as little motion blur as possible while maintaining brightness.

I don't yet have the hardware to run OW in 240hz, so i set the refreshrate to 144hz.

To my surprise, there was a LOT of crosstalk after doing the UFO test, especially at the bottom and top. Considering I am running at 144hz, I was expecting a much clearer image - I've seen a lot of photos of this test on here and the results are usually far better, with higher refreshrates.

I've tried every setting I could think of and find - all combinations of High/Premium on Dyac/AMA, adjusting the Area (this only moved the extreme crosstalk from the top to the bottom of the image), the persistence, the OD Gain in the factory menu - nothing helped.

I've also created a custom resolution in toastys CRU program following the guide in the appendix. This did absolutely nothing to help with the crosstalk. I ended up at 2020VT total.

What could be the issue? Do I have a defective model?

Picture: https://imgur.com/a/AZFHdTf

Falkentyne
Posts: 2543
Joined: 26 Mar 2014, 07:23

Re: Severe Crosstalk with new XL2546?

Post by Falkentyne » 07 Jun 2019, 17:23

robben44 wrote:Hi everyone,

I've read quite a bit on this site about blur reduction, crosstalk, etc.

After a lot of research, I bought a Benq Zowie XL2546 today.
I mainly play Overwatch and I wanted to get a monitor that would give me as little motion blur as possible while maintaining brightness.

I don't yet have the hardware to run OW in 240hz, so i set the refreshrate to 144hz.

To my surprise, there was a LOT of crosstalk after doing the UFO test, especially at the bottom and top. Considering I am running at 144hz, I was expecting a much clearer image - I've seen a lot of photos of this test on here and the results are usually far better, with higher refreshrates.

I've tried every setting I could think of and find - all combinations of High/Premium on Dyac/AMA, adjusting the Area (this only moved the extreme crosstalk from the top to the bottom of the image), the persistence, the OD Gain in the factory menu - nothing helped.

I've also created a custom resolution in toastys CRU program following the guide in the appendix. This did absolutely nothing to help with the crosstalk. I ended up at 2020VT total.

What could be the issue? Do I have a defective model?

Picture: https://imgur.com/a/AZFHdTf
I don't have one of these monitors so my ability to help you is almost nonexistent.
But 144hz uses reduced blanking interval usually, at least on older monitors like the "Mstar" Benq Z series monitors (and XL2430T).
I have no idea how the Realtek scaler models (Realtek 2796BF) work with regards to VT tweaking, but I have heard that tweaking 144hz may not get you that far. Chief mentioned something about 144hz handled worse on 240hz monitors, than on older monitors, although on the old XL2420Z/XL2420Z, you couldn't even use a VT Tweak at 144hz (you couldn't go much higher than the default of 1098 VT), or it just black screened.

What is the default vertical total at 144hz on that XL2546?
Is it around 1098?
If you are trying to use 2020, you might not be getting anywhere if that VT is too large; it could possibly be completely discarded by the video driver or monitor.

Can you test 240hz (even though you won't use it in overwatch--just test it in windows) with a VT tweak at 240hz in TestUFO only--I don't know what VT will work at 240hz-again I don't own any of these monitors (please don't go out of range though), and then try testing 120hz with a VT Tweak (go as high as you can safely--start at VT 1500 first) and see if crosstalk is *greatly* reduced compared to stock?

Compare 120hz and 144hz and see if you get decent crosstalk.
Start at 1500 VT at 120hz and go upwards 50VT at a time until it black screens. You should also adjust the "Area" parameter in the service menu to move the main crosstalk band upwards or downwards.

if you manage to get decent VT higher than 1500 (1500 should be a slam duck) at 120hz,
try 144hz with VT 1500 and see if that helps improve things.

Post your results.

Q83Ia7ta
Posts: 719
Joined: 18 Dec 2013, 09:29

Re: Severe Crosstalk with new XL2546?

Post by Q83Ia7ta » 08 Jun 2019, 16:48

Don't use 144hz on that 240hz monitor. It's adds input lag and looks like it make overdrive worse.

robben44
Posts: 2
Joined: 06 Jun 2019, 16:49

Re: Severe Crosstalk with new XL2546?

Post by robben44 » 08 Jun 2019, 18:07

Falkentyne wrote:
robben44 wrote:Hi everyone,

I've read quite a bit on this site about blur reduction, crosstalk, etc.

After a lot of research, I bought a Benq Zowie XL2546 today.
I mainly play Overwatch and I wanted to get a monitor that would give me as little motion blur as possible while maintaining brightness.

I don't yet have the hardware to run OW in 240hz, so i set the refreshrate to 144hz.

To my surprise, there was a LOT of crosstalk after doing the UFO test, especially at the bottom and top. Considering I am running at 144hz, I was expecting a much clearer image - I've seen a lot of photos of this test on here and the results are usually far better, with higher refreshrates.

I've tried every setting I could think of and find - all combinations of High/Premium on Dyac/AMA, adjusting the Area (this only moved the extreme crosstalk from the top to the bottom of the image), the persistence, the OD Gain in the factory menu - nothing helped.

I've also created a custom resolution in toastys CRU program following the guide in the appendix. This did absolutely nothing to help with the crosstalk. I ended up at 2020VT total.

What could be the issue? Do I have a defective model?

Picture: https://imgur.com/a/AZFHdTf
I don't have one of these monitors so my ability to help you is almost nonexistent.
But 144hz uses reduced blanking interval usually, at least on older monitors like the "Mstar" Benq Z series monitors (and XL2430T).
I have no idea how the Realtek scaler models (Realtek 2796BF) work with regards to VT tweaking, but I have heard that tweaking 144hz may not get you that far. Chief mentioned something about 144hz handled worse on 240hz monitors, than on older monitors, although on the old XL2420Z/XL2420Z, you couldn't even use a VT Tweak at 144hz (you couldn't go much higher than the default of 1098 VT), or it just black screened.

What is the default vertical total at 144hz on that XL2546?
Is it around 1098?
If you are trying to use 2020, you might not be getting anywhere if that VT is too large; it could possibly be completely discarded by the video driver or monitor.

Can you test 240hz (even though you won't use it in overwatch--just test it in windows) with a VT tweak at 240hz in TestUFO only--I don't know what VT will work at 240hz-again I don't own any of these monitors (please don't go out of range though), and then try testing 120hz with a VT Tweak (go as high as you can safely--start at VT 1500 first) and see if crosstalk is *greatly* reduced compared to stock?

Compare 120hz and 144hz and see if you get decent crosstalk.
Start at 1500 VT at 120hz and go upwards 50VT at a time until it black screens. You should also adjust the "Area" parameter in the service menu to move the main crosstalk band upwards or downwards.

if you manage to get decent VT higher than 1500 (1500 should be a slam duck) at 120hz,
try 144hz with VT 1500 and see if that helps improve things.

Post your results.

Thanks for the help. I arrived at 2020 VT by following the guide and plugging in the values from 240hz, locking the dotclock, and upping the VT until I arrived at 144hz - but I think I did something wrong.
I lowered the VT like you suggested, to 1700 (first arbitrary value I tried). I could now move the worst crosstalk off the visible screen. So I now had an even image.

I was still stuck with more crosstalk than I would like, but after a LOT of playing around with monitor-settings I fixed that as well.

For some reason I cannot understand lowering the color temperatures reduces the crosstalk drastically. From what I could tell this should just makes the brightness lower, but lowering the brightness through the strobe utility persistence setting (or the intensity setting in the service menu on the monitor) does hardly anything in terms of lowering crosstalk. Lowering color-temps evenly makes the crosstalk almost non-existent.

So I found a good middle-ground between lowered color-temps and brightness (I settled at 63,63,65), upped the contrast to compensate and upped the black equalizer a bit and now I have barely any crosstalk at all, evenly throughout the display, while the image is still quite bright.

I tried 240hz as well - while the crosstalk is visible, I wouldn't say its noticeable - at least in the middle of the screen. It is quite noticable at the very top and bottom, but that doesnt really bother me at all.

I am VERY satisfied with the monitor now. It went from being quite underwhelming to much better than I expected.

Though I have no idea why changing the color temperatures could reduce the crosstalk??

New image at 144hz (crosstalk at the top/bottom is less visible irl: https://imgur.com/a/NPdVnXF

Falkentyne
Posts: 2543
Joined: 26 Mar 2014, 07:23

Re: Severe Crosstalk with new XL2546?

Post by Falkentyne » 09 Jun 2019, 01:06

robben44 wrote:
Falkentyne wrote:
robben44 wrote:Hi everyone,

I've read quite a bit on this site about blur reduction, crosstalk, etc.

After a lot of research, I bought a Benq Zowie XL2546 today.
I mainly play Overwatch and I wanted to get a monitor that would give me as little motion blur as possible while maintaining brightness.

I don't yet have the hardware to run OW in 240hz, so i set the refreshrate to 144hz.

To my surprise, there was a LOT of crosstalk after doing the UFO test, especially at the bottom and top. Considering I am running at 144hz, I was expecting a much clearer image - I've seen a lot of photos of this test on here and the results are usually far better, with higher refreshrates.

I've tried every setting I could think of and find - all combinations of High/Premium on Dyac/AMA, adjusting the Area (this only moved the extreme crosstalk from the top to the bottom of the image), the persistence, the OD Gain in the factory menu - nothing helped.

I've also created a custom resolution in toastys CRU program following the guide in the appendix. This did absolutely nothing to help with the crosstalk. I ended up at 2020VT total.

What could be the issue? Do I have a defective model?

Picture: https://imgur.com/a/AZFHdTf
I don't have one of these monitors so my ability to help you is almost nonexistent.
But 144hz uses reduced blanking interval usually, at least on older monitors like the "Mstar" Benq Z series monitors (and XL2430T).
I have no idea how the Realtek scaler models (Realtek 2796BF) work with regards to VT tweaking, but I have heard that tweaking 144hz may not get you that far. Chief mentioned something about 144hz handled worse on 240hz monitors, than on older monitors, although on the old XL2420Z/XL2420Z, you couldn't even use a VT Tweak at 144hz (you couldn't go much higher than the default of 1098 VT), or it just black screened.

What is the default vertical total at 144hz on that XL2546?
Is it around 1098?
If you are trying to use 2020, you might not be getting anywhere if that VT is too large; it could possibly be completely discarded by the video driver or monitor.

Can you test 240hz (even though you won't use it in overwatch--just test it in windows) with a VT tweak at 240hz in TestUFO only--I don't know what VT will work at 240hz-again I don't own any of these monitors (please don't go out of range though), and then try testing 120hz with a VT Tweak (go as high as you can safely--start at VT 1500 first) and see if crosstalk is *greatly* reduced compared to stock?

Compare 120hz and 144hz and see if you get decent crosstalk.
Start at 1500 VT at 120hz and go upwards 50VT at a time until it black screens. You should also adjust the "Area" parameter in the service menu to move the main crosstalk band upwards or downwards.

if you manage to get decent VT higher than 1500 (1500 should be a slam duck) at 120hz,
try 144hz with VT 1500 and see if that helps improve things.

Post your results.

Thanks for the help. I arrived at 2020 VT by following the guide and plugging in the values from 240hz, locking the dotclock, and upping the VT until I arrived at 144hz - but I think I did something wrong.
I lowered the VT like you suggested, to 1700 (first arbitrary value I tried). I could now move the worst crosstalk off the visible screen. So I now had an even image.

I was still stuck with more crosstalk than I would like, but after a LOT of playing around with monitor-settings I fixed that as well.

For some reason I cannot understand lowering the color temperatures reduces the crosstalk drastically. From what I could tell this should just makes the brightness lower, but lowering the brightness through the strobe utility persistence setting (or the intensity setting in the service menu on the monitor) does hardly anything in terms of lowering crosstalk. Lowering color-temps evenly makes the crosstalk almost non-existent.

So I found a good middle-ground between lowered color-temps and brightness (I settled at 63,63,65), upped the contrast to compensate and upped the black equalizer a bit and now I have barely any crosstalk at all, evenly throughout the display, while the image is still quite bright.

I tried 240hz as well - while the crosstalk is visible, I wouldn't say its noticeable - at least in the middle of the screen. It is quite noticable at the very top and bottom, but that doesnt really bother me at all.

I am VERY satisfied with the monitor now. It went from being quite underwhelming to much better than I expected.

Though I have no idea why changing the color temperatures could reduce the crosstalk??

New image at 144hz (crosstalk at the top/bottom is less visible irl: https://imgur.com/a/NPdVnXF
Glad you got better results.
Most likely, 2020 is simply out of range of the reduced blanking interval at 144hz. I'm not sure if it's the monitor or video driver at fault, but I've seen in the past that if the timings are too wrong, the video card drivers will simply discard your timings and use defaults. The easiest way to see if that happens is to use a default settings at whatever refresh rate, check the crosstalk and then the custom settings you did. If there is zero difference, then there you go.

Reducing color settings is probably reducing overdrive artifacts, not crosstalk. You can probably see the exact same thing by reducing contrast instead. Crosstalk and overdrive artifacts are related in a way, but occur from completely different reasons. Strobe crosstalk (Chief is going to beat me up for this!) is simply a previous frame's pixel data bleeding into the next frame's picture data. If you are looking at testufo alien invasion crosstalk test, the picture moves left to right. When you move the "Area" parameter, you should notice a very thick double image that appears on one line of pixels, and another line of pixels having the double image "fade away". What you are seeing are data from two adjacent frames, 1 frame apart, shifting positions based on the strobe phase. Since the image is moving left to right, if the entire frame data shifts to the -->right one frame, then your input lag is going to be 1 frame faster (quicker, lower) than the frame to the left. If the entire frame data shifts one frame to the left, that data on the left is going to be 1 frame behind (one frame of input lag penalty) than the frame to the right would be.

So if the frame to the right were, let's say, 8.3ms at 120hz, then if you shifted the Area parameter enough so that the frame to the "left" becomes the current frame, that frame would now have an input lag of 16.7ms. The crosstalk is where those two frames start to merge and shift one way or another (depending on which way you move the Area (Strobe Phase).

Overdrive artifacts are different as they happen regardless of the position on the screen, but when using strobing, overdrive artifacts can get amplified by the effects of the strobe phase crosstalk

User avatar
Chief Blur Buster
Site Admin
Posts: 6502
Joined: 05 Dec 2013, 15:44

Re: Severe Crosstalk with new XL2546?

Post by Chief Blur Buster » 09 Jun 2019, 16:10

Falkentyne wrote:Reducing color settings is probably reducing overdrive artifacts, not crosstalk.
[...]
Falkentyne wrote:Overdrive artifacts are different as they happen regardless of the position on the screen, but when using strobing, overdrive artifacts can get amplified by the effects of the strobe phase crosstalk
Normally, it's counterintuitive, but there is a complex interaction between strobe crosstalk and overdrive artifacts.

Some of you are long-time readers, who may find this interesting, so I'll post here.
Strobe crosstalk often includes chopped segments of overdrive artifacts.

Complex Interaction Between Overdrive and Strobe Crosstalk
(From Photograph of Whiteboard from Blur Busters Training Class at Manufacturer)

During a Blur Busters training class I gave to some display manufacturers who wanted to know how to improve their strobing (This is a little known but that's something I have started to now! See services.blurbusters.com...). Over the last two years I began teaching some vendors and display manufacturers the basics of how to strobe better to reduce strobe crosstalk. The first-tiers with inhouse TCON/scaler engineers know how to do it (NVIDIA) but many manufacturers outsource these display engineering skills (or even white-label monitors).....

So that is where I come in to help such manufacturers that do not have in-house TCON/scaler engineers.

Image
(I drew that whiteboard a few times before).

I was trying to explain to less-strobe-experienced manufacturers how strobing "slices" the GtG curves and you need to pretty much strobe the good parts of the GtG curve. This is supremely challenging because the GtG are at different stages in different parts of the screen (see high speed video of LCD scanout). So the "Line Scan" X markers will be totally differnt for different parts of the panel. So you have low crosstalk on some part of the panel surface, but lots of crosstalk in others -- the famous asymmetry between non-global LCD scanout and the global backlight flash -- is what creates non-uniform strobe crosstalk with a motion blur reduction backlight.

So overdrive artifacts, do interact with strobe crosstalk, in the "curve slicing" physics. When the backlight flashes, it's making segments of the pixel response curve visible. The world of strobing is all about trying to cram the GtG elephant into the drinking-straw-sized VBI, to keep GtG limitations (including overdrive) unseen by human eyes.

Image

The different tints of the strobe crosstalk duplicates also often contain overdrive artifacts. So you want to tune overdrive to the point where it's not too dark nor too bright, so that's why "slight amount of overdrive" has less crosstalk than "no overdrive" and "lots of overdrive".

But even "slight amount of overdrive" is often not good enough; and a manufacturer has to do advanced overdrive tricks (e.g. overdrive lookup tables instead of overdrive gain) to achieve better strobing. Thusly, LCD overdrive apparently ends up being fundamental to reducing strobe crosstalk visibility -- well-tuned well-balanced overdrive makes the strobe crosstalk faint.

As a result, impeccable strobe-optimized overdrive tuning is necessary, but many manufacturers just don't bother to even tune 50% of the way. Many don't even let their users do part of the "strobe tuning" (like it is possible on BenQ/Zowie monitors) -- at least the user has the power to strobe-tune the monitor.

So you come to a situation where you have:
(A) Manufacturers with pre-tuned strobing out-of-the-box. (e.g. NVIDIA ULMB). Easy but inflexible.
(B) Manufacturers that lets users do strobe tuning (e.g. BenQ/Zowie). Harder/fiddly but rewarding for power users.
(C) Manufacturers that don't do either of the above. Head explodingly frustrating sometimes.
And I basically focus on helping manufacturers with (B) (C).

So that covers the little-known complex interaction between overdrive and strobing.... ;)
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

User avatar
Chief Blur Buster
Site Admin
Posts: 6502
Joined: 05 Dec 2013, 15:44

Re: Severe Crosstalk with new XL2546?

Post by Chief Blur Buster » 09 Jun 2019, 16:38

robben44 wrote:For some reason I cannot understand lowering the color temperatures reduces the crosstalk drastically. From what I could tell this should just makes the brightness lower, but lowering the brightness through the strobe utility persistence setting (or the intensity setting in the service menu on the monitor) does hardly anything in terms of lowering crosstalk. Lowering color-temps evenly makes the crosstalk almost non-existent.
robben44 wrote:Though I have no idea why changing the color temperatures could reduce the crosstalk??
That's normal.

Reducing the digital contrast range (e.g. RGB 16-235 instead of 0-255) also reduces crosstalk.

Pixel response (GtG) is worse for darkest colors and brightest colors. You have slower pixel response and worst crosstalk with slowest pixels. By making adjustments, those inadvertently avoids the "slow" colors. Whether you adjust RGB gains, or digital contrast/brightness (in NVIDIA Control Panel), or via Color Temperature -- all of those do similar things in inadvertently avoiding colors that have slower GtG combos.

Good invisible overdrive is only possible if you have moderate amount of overshoot/undershoot room. But those fullbright/fulldark primary colors don't have any overshoot/undershoot room to make perfectly balanced overdrive possible. And thus, those specific bright colors have crappy strobe crosstalk. Ouch.

But adjust...slightly reduce the LCD's color gamut intentionally (even just 5% makes a giant difference to crosstalk). Now you're sticking to the faster GtG colors that is completing fast enough to be completely hidden by your human eyes in the dark periods between the strobe-backlight flashes.

I'm talking about color digital brightness....not backlight brightness. The panel (majority of GtG) and the backlight (majority of MPRT) has to work together. Fast clean GtG, fast clean MPRT. Properly engineering this and that creates CRT motion clarity.

A slight intentional color-gamut reduction is a very effective but little-known trick to reduce strobe crosstalk if you can stand some color gamut loss, in exchange for more CRT-like strobing.

That's why good old LightBoost had crappy colors; it was tuned to avoid those slow colors (saturated colors). ULMB is brighter but has more crosstalk than LightBoost, but you can make ULMB look like LightBoost simply by worsening the ULMB colors -- raise your digital black level and worsen your digital white level -- and that makes a lot of crosstalk disappear. At least it is a continuum -- you decide how much you want to worsen your crosstalk in exchange for keeping more of those brighter colors.

Image vs Image vs Image

LCDs don't have the same pixel response speed for all colors, and the worst crosstalk is occuring with those colors that have slower pixel responses. This is a big problem with strobing on VA, because it generates lots of ghosty-looking strobe crosstalk during dark games -- such colors are extremely hard to overdrive properly on a VA panel.

GtG is the strobe crosstalk.

But what if we can have cake and eat it too? Avoid those color-gamut reductions? It is also why Blur Busters is such a fan of 0.5ms GtG (not just 0.5ms MPRT) and making Min/Avg/Max GtG all equally below 1ms -- the worst GtG being no worse than 1ms. Seeing 1ms native GtG (no overdrive needed) would be wonderful. But we still need moderate amounts of overdrive to get proper good GtG numbers that fits in the VBI (blanking interval between refresh cycles) necessary to hide GtG unseen from human eyes to make the crosstalk disappear.

The more GtG curve hidden, the better strobing becomes, less crosstalk ... until crosstalk falls below human-visibility noise floor. The blanking intervals of 240Hz is often far less than 0.5ms and how can you hide both GtG and still have enough time to flash the backlight, before the new refresh cycle begins. Strobing at 240Hz is extremely tough to do in a crosstalk-free manner, the GtG enroaches strongly into the refresh cycle.

The slowest GtG needs to get faster -- the slowest GtG often dictates strobe mere existence of annoying strobe crosstalk. Ouch. It's extremely hard to do on an LCD, and succeeding that helps eliminate strobe crosstalk at higher refresh rates.

We're reaching the point where we now probably need <0.1ms GtG(90%) for very good 480Hz strobing; and about <0.25ms GtG(90%) for very good 240Hz strobing; but at that point, 480Hz+ really reduce a lot of motion blur strobelessly (blurless sample-and-hold)... and not need to worry about strobe crosstalk anymore. Strobing helped 120Hz a lot, but helps 240Hz less, and will help 480Hz even less; there comes a point where strobing will not be needed for motion blur reduction once GPUs framerates are high enough (with the help of frame rate amplification technologies)
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