Strobe Crosstalk: Double Image Effect (razor-sharp ghosts)

[Chief Blur Buster]

What is Strobe Crosstalk?

Strobe backlights such as LightBoost and ULMB, allow LCD displays to have low persistence, much like a CRT. They successfully bypass most of a GtG pixel transition, by flashing the backlight/edgelight only on fully refreshed frames. It was only in the last couple years, that LCD panels have successfully been able to refresh fast enough for good strobe backlight operation. Many articles have been published including LightBoost media coverage that explain the principles, along with Photos: 60Hz vs 120Hz vs LightBoost.

Incomplete LCD Pixel Transitions

However, even to this date, LCD can have a lot of difficulties in finishing pixel transitions before the next refresh cycle. This is a common cause of 3D crosstalk, but it can also affect motion quality by causing a double-image effect during fast motion. They look like ghost images, but the ghost is razor-sharp (crystal clear), albiet hopefully faint. Blur Busters now calls this strobe crosstalk, which is the portion of LCD pixel transitions that are still unfinished by the time a subsequent strobe occurs. It is similar in intensity to 3D crosstalk, but is a faint double-image that chases after the object. It's not the same double-image effect as 30fps@60Hz or 60fps@120Hz on impulse-driven displays (CRT/strobed) where both images are equally strong. In this case, you are getting one strong image with one (or more) sharp but very faint afterimages chasing behind the motion. Even a single millisecond difference in real-world GtG can be enough to greatly amplify strobe crosstalk. Strobe backlight operation is one of the few cases where GtG transition speed becomes critical, to fit pixel transitions within the dark periods between strobe backlight flashes.

Good panels can complete well over >99% of GtG pixel transitions during the dark period between strobe flashes. On these, strobe crosstalk is typically very faint, and occasionally, well below human detectable levels.

Different models of displays have more strobe crosstalk than others, and it is normal for many LightBoost displays to have more strobe crosstalk at the very bottom edge of the screen.

Testing Strobe Crosstalk

I have created new TestUFO tests to test strobe crosstalk. You may see a very faint copy of the image behind the UFO.



On your strobe-backlight monitor, display this test full-screen.
Go to http://www.testufo.com/photo#photo=alien-invasion.png and set Height->Full Screen.

Typical Strobe Crosstalk Behaviors

-- You will typically observe more ghosting along the very top/bottom edge of the screen, than at the screen center.
-- Some displays have it slightly unbalanced (only ghosting at bottom edge) while others have it very nasty (e.g. unfixed BENQ XL2720Z monitors).
-- Some models of monitors (e.g. XL2411T) are almost nearly perfect along the center band of the screen.
-- Also, lowering the OSD Contrast setting on some models of LightBoost monitors, can also reduce strobe crosstalk massively. This effect is observed on the ASUS VG278H (original).
-- Warming up a monitor can help reduce strobe crosstalk, especially on strobed VA monitors (it especially helps the EIZO FG2421 massively). Cold LCD panels respond more slowly, which creates more strobe crosstalk.
-- Strobe crosstalk is usually not noticeable in most games, but can manifest itself as artifacts similar to this one during high contrast material, or if the strobe crosstalk is really bad.

[Black Octagon]

Razor sharp ghosts...I'd buy concert tickets to a band with that name.

Curious to try this on my OC'd IPS. Thanks Mark.

Sent from dumbphone (pls excuse typos and dumbness)

[Zenairis]

My PG278Q suffers pretty badly on this test. Lots of ghosting almost half of the image in motion.

[PoWn3d_0704]

The bottom of my ULMB screen is SO bad.

[flood]

which strobed lcd monitor does the best with regards to this artifact?

[Falkentyne]

Any Lightboost monitor running in Lightboost mode, by far.
Since lightboost has perline overdrive and accelerated scanout. It seems like ULMB isn't as robust as lightboost, even though ULMB at least doesn't trash the colors or lock you out of the controls.

Lightboost on the 27" monitors get a much higher contrast ratio but at the cost of noticeably more ghosting. Lowering the contrast down to 0 can bring the ghosting to a similar level as the 24" lightboost monitors.
(No idea how this applies on the Swift, or even if Lightboost even exists on the Swift..)

[jaranis]

Thanks for posting this Chief, been looking for an answer for this. My BenQ XL2420T and Asus VG248QE exhibit this with Lightboost enabled.

Would this effect be even worse on a strobed IPS panel? Because the pixel response time are much longer? By "much longer" I mean 5ms haha. I am really looking into getting one of these GSYNC IPS panels when they hit the shelves, but am wondering if the strobe crosstalk will be worse (with ULMB). If it's worse than what I see on my current monitors, I don't think I will be buying one.

[Falkentyne]

As far as I know, the Gsync IPS panels do NOT support ULMB/Strobing because of their response times. They only support gsync afaik. Someone correct me if I'm wrong.

[jaranis]

TFTCentral says Acer confirmed ULMB on the upcoming XB270HU (IPS & G-Sync), but yeah an extra 4 or 5ms response would change things quite a bit if we are experiencing this crosstalk at 1ms GtG. Maybe ULMB will be terrible on an IPS panel.

http://www.tftcentral.co.uk/news_archiv ... b270hu_ips

[Sptz]

My BenQ XL2411Z does show this an awful lot at the bottom. I did tweak the crosstalk utility to get it further away from the center. The center is indeed pratically perfect, it's just a bit annoying when playing CSGO if someone is running right next to you within your FOV and their legs are located at the bottom of screen you do see that. Although much more subtle it's kind of like that multiple leg Sonic running effect

http://www.dotemu.com/sites/default/fil ... 3920_1.jpg