Afterimages On Display and Persistence of Vision [Long!]

[Chief Blur Buster]

I received permission to post a long, informative email discussion, since I (Mark Rejhon) am one of the few people to understand strobe backlight behavior with human vision (benefits/side effects and technical explanations). This person preferred to remain anonymous, however, this information posted here contains very useful and educational content for the canon of display science, and these are very worthy topics for researchers looking for new research topics.

This is posted in this forum thread.

Discussion from other engineer-minded and science-minded people are welcome, too.

Hey Mark,
Wanted to open a line of discussion with you on talking about motion blur and and replicating retinal properties.

One thing I've been researching is afterimages/persistence of vision on the retina. One a 30 fps strobed display, this is really clear. Your 30 fps UFO test visually appears as 2 distinct UFOs on a strobed display (panasonic plasma TV). No motion blur, but definitely persistence of vision.

Given what I'm reading about the phenomenon, where the after image persists for approximately 0.04 seconds, this makes sense. A slower cadence doesn't not give the double image, but then motion is lost, and it becomes stop-go series of images.

Obviously, there is a decay rate on the afterimage, and I do not think it is linear. Most of it is probably lost early, since humans can visually distinguish past even 60 fps (0.017 sec per image). I'm trying to research this in some textbooks right now, so I'll let you know what my findings are.

I think this is an important distinction that we need to take into account, since it means there is a limit to what a strobed display can achieve, and we want to balance quality with reducing motion blur and flicker. Similarly, pursuit tracking cameras, which replicating smooth pursuit of the eye, do not replicate the effect of persistence on strobed displays.

As we increase framerates, even with strobed displays, we'll be seeing 2-4 distinct images (the faster the motion, the further apart and the more distinct they will be, and the faster the refresh, the more images we'll see). Of course, with the decay rates, the impact of these images will be variable.

One thing possibility is to digitally blend previous frames based on a decay rate. I would suggest start with a linear decay rate as an academic experiment, but then use the actual decay rate to replicate as best as possible what the eye will actually see.

I'm also looking at smooth pursuit properties, which require saccades (which is a sample-and-hold effect of the eye motion itself) to catch up to angular velocities > 30 degrees/second.

On a 1080p display occupying 40 degrees of your visual field, that's well within the 1000pixels/second standard you are using. But we should also look into upcoming 4K displays, that will be much closer in achieving pixels that are smaller than acuity limits, where 1 ms of image produces 3 pixels of blur (4K display with 40 degrees of visual field) at smooth pursuit limits (30 degrees/second, or ~3000 pixels/second).

[Chief Blur Buster]

Hey Mark,
Wanted to open a line of discussion with you on talking about motion blur and and replicating retinal properties.

One thing I've been researching is afterimages/persistence of vision on the retina. One a 30 fps strobed display, this is really clear. Your 30 fps UFO test visually appears as 2 distinct UFOs on a strobed display (panasonic plasma TV). No motion blur, but definitely persistence of vision.

That's correct, because your eyes are always moving when tracking moving objects.
Your eyes have moved in the time between the two strobes of the same frame.
So 30fps@60Hz strobed (e.g. 60Hz plasma or 60Hz CRT) = double images
So 60fps@120Hz strobed (e.g. LightBoost 120Hz) = double images
So 60fps@180Hz strobed (e.g. PWM dimming) = triple images

Given what I'm reading about the phenomenon, where the after image persists for approximately 0.04 seconds, this makes sense. A slower cadence doesn't not give the double image, but then motion is lost, and it becomes stop-go series of images.

On your plasma, it's more like approximately 5-8ms, (0.005 to 0.008 second) due to plasma phosphor decay.

Obviously, there is a decay rate on the afterimage, and I do not think it is linear. Most of it is probably lost early, since humans can visually distinguish past even 60 fps (0.017 sec per image). I'm trying to research this in some textbooks right now, so I'll let you know what my findings are.

You should read these links:
http://www.avsforum.com/t/1484182/why-w ... h-comments
And this one:
http://blogs.valvesoftware.com/abrash/d ... ing-judder
Let me know what you think.

I think this is an important distinction that we need to take into account, since it means there is a limit to what a strobed display can achieve, and we want to balance quality with reducing motion blur and flicker.

You need one strobe per refresh, in order to have the zero-blur effect. This eliminates any afterimage effects. Once you manage to hit one strobe per refresh, there's no upper limit to the motion clarity that can be achieved on a strobed display, since persistence is directly proportional to the amount of motion blur perceived.

Similarly, pursuit tracking cameras, which replicating smooth pursuit of the eye, do not replicate the effect of persistence on strobed displays.

Actually, they do, in my tests. Very accurate co-relation in terms of number of copies of images! Here is proof:
I've successfully captured the triple-image effect of PWM-dimming (60fps @ 180Hz PWM):
http://www.blurbusters.com/faq/lcd-motion-artifacts#pwm
This is true WYSIWG.

Reproduction scenario:
1. Find an LED backlit LCD monitor with PWM dimming, set to 0% brightness
2. Use http://www.testufo.com/ghosting
3. Take pursuit camera photo.

I should modify the ghosting test, to run at 30fps, so you can do pursuit camera tests on a plasma display. If you wish, I could make this modification. Do, however, make sure you use a sufficiently long camera exposure if doing tests on your plasma, to average-out the temporal color dithering behavior of plasma subfields.

As we increase framerates, even with strobed displays, we'll be seeing 2-4 distinct images (the faster the motion, the further apart and the more distinct they will be, and the faster the refresh, the more images we'll see). Of course, with the decay rates, the impact of these images will be variable.

The number of distinct images is mathematically calculated as:
copies of image = (refresh rate / frame rate)
There is no distinct images if you have frame rate matching refresh rate.

One thing possibility is to digitally blend previous frames based on a decay rate. I would suggest start with a linear decay rate as an academic experiment, but then use the actual decay rate to replicate as best as possible what the eye will actually see.

In my opinion, this would not be the best case scenario. The decay rate adds to persistence.

I'm also looking at smooth pursuit properties, which require saccades (which is a sample-and-hold effect of the eye motion itself) to catch up to angular velocities > 30 degrees/second.

On a 1080p display occupying 40 degrees of your visual field, that's well within the 1000pixels/second standard you are using. But we should also look into upcoming 4K displays, that will be much closer in achieving pixels that are smaller than acuity limits, where 1 ms of image produces 30 pixels of blur (4K display with 40 degrees of visual field) at smooth pursuit limits (30 degrees/second, or ~3000 pixels/second).

This is correct. More pixels means more opportunity for motion blur.
At 1ms of persistence = 1 pixel of motion blur during 1000 pixels/second (This is the simplified "Blur Busters Law" based on reading science papers and doing motion tests).

I highly recommend reading:
http://www.blurbusters.com/zero-motion- ... vs50vs100/

Let me know what you think.
And if it's OK to copy this reply anonymously into a new "Display Science" topic area of Blur Busters Forum.

Thanks,
Mark Rejhon

(Permission was confirmed.)

[Chief Blur Buster]

I personally would prefer a true 1000fps@1000Hz display, over a strobe backlight display.

1ms of persistence = 1ms of unique frame visiblity time = 1 pixel of motion blur during 1000 pixels/second
-- Can be done 1000fps@1000Hz flicker-free, if you don't want strobing.
-- Can be done 1/1000sec strobe flashes (at any refresh rate).

2000 pixels/sec eye tracking = 2 pixels of motion blur (e.g. approx full screenwidth per second on 2K display)
4000 pixels/sec eye tracking = 4 pixels of motion blur (e.g. approx full screenwidth per second on 4K display)
Assumption: framerate matches Hertz

Additional Assumptions:
-- Squarewave persistence (no decay, no blending between frames)
-- Strobe backlights (with >8ms cycle) on 1ms/2ms LCD's behave very accurately as squarewave, as most LCD pixels finish before next flash
-- Plasma is not very squarewave. It's more complex than CRT in that it generates multiple dithered subfields, where some colors have more persistence than others, and the colors are often temporally separated under high speed video. Thus, it will not reliably follow "Blur Busters Law" that I've come up with (1ms = 1 pixel of motion blur during 1000 pixels/sec).
-- One strobe per unique frame (notice I say "unique frame" and not "refresh rate")

Known Error Factors:
-- Decay curves. Decay adds to persistence, which increases motion blur, but can also reduce flicker.
Example: Slow radar CRT's don't flicker much but they blur/ghost a lot.
-- Leakage (ghosting) of unfinished LCD pixel transitions into next strobe flash. This creates "strobe crosstalk"; this roughly approximates the intensity of 3D crosstalk except you only see the double-image effect during motion. The more leakage, the more double-image effect occurs. And if the leakage is so bad it leaks over 3 refreshes, you get a triple-image effect.
Older LCD's had this problem, but newer fast LCD's that are able to finish most of their transitions before the next flash, they behave very accurate square-wave persistence displays that accurately follows "Blur Busters Law".
-- Weird strobe algorithms that deviate from one-strobe-per-frame, such as EIZO's Turbo240, which does a very faint double-strobe. A faint strobe followed by a large strobe. It's also a VA panel which has more leakage of unfinished transitions between refreshes.

Devices that generally behave as squarewave persistence, and thus accurately follows "1ms = 1pix tracking based motion blur during 1000 pixels/second":
-- Strobe backlight LCD's on fast LCD panels (e.g. LightBoost, BENQ Blur Reduction). There's some extremely faint strobe crosstalk.
-- Rolling-scan OLED's (e.g. Sony PVM series)
-- Black insertion DLP's, provided you calculate based on the full black frame, and ignore the high-frequency elements of the per-pixel PWM that DLP is doing.

Also remember "1ms = 1 pixel tracking based motion blur during 1000 pixels/second" is just the formula (for framerate=Hz motion on squarewave persistence displays).

It translates easily to:
5ms = 5 pixels tracking based motion blur during 1000 pixels/second motion
3ms = 6 pixels tracking based motion blur during 2000 pixels/second motion
1ms = 0.5 pixels tracking based motion blur during 500 pixels/second motion
etc.

I just simply use 1000 pixels/second as a simplification, because:
-- It's a common video game motion speed
-- This speed is faster than video
-- Blur Busters is often all about video games.
-- Video games don't always have built in motion blur (like camera motion blur), and are often crisp graphics
-- Users view a monitor at close viewing distances, and FPS game players run very fast motion, so several of us easily tell apart 1.4ms persistence and 2.4ms persistence (as seen at http://www.blurbusters.com/faq/10vs50vs100/ ) ...Consider 1.4ms = 1/700sec = equivalent to a 700fps@700Hz flickerfree display. Since that tech doesn't exist easily, we're big advocates of strobing.

____________

BTW, here's a great demonstration of an example of how duty cycle affects motion blur:
http://www.testufo.com/blackframes#count=3
The longer the dark ratio is to bright ratio, the less eye-tracking-based motion blur occurs.

Hope my email replies are interesting!

Cheers,
Mark Rejhon

[Chief Blur Buster]

Given what I'm reading about the phenomenon, where the after image persists for approximately 0.04 seconds, this makes sense. A slower cadence doesn't not give the double image, but then motion is lost, and it becomes stop-go series of images.

On your plasma, it's more like approximately 5-8ms, (0.005 to 0.008 second) due to plasma phosphor decay.

Actually, I'm not talking about phosphor decay - I'm talking about the action potential frequency decay from the neurons in your retina. Once excited, the neurons take up to about 0.04 seconds to fully come back to a resting action potential frequency.

If it was purely because my eyes have moved, theoretically, it should appear as stuttering motion of a single UFO (jumps 2 steps ahead, then 1 step back, 2 steps ahead, 1 step back), not 2 UFOs that visually appear to always be there. The limits are very close, however, as you can detect some faint flickering.

Agree?

Obviously, there is a decay rate on the afterimage, and I do not think it is linear. Most of it is probably lost early, since humans can visually distinguish past even 60 fps (0.017 sec per image). I'm trying to research this in some textbooks right now, so I'll let you know what my findings are.

You should read these links:
http://www.avsforum.com/t/1484182/why-w ... h-comments
And this one:
http://blogs.valvesoftware.com/abrash/d ... ing-judder
Let me know what you think.

OK, I've read these. I had not considered the breaking of saccadic masking, it's not well described in retinal physiology text books (the masking is, the breaking of it with impulse displays is not). I'm not sure if the effect being describe here actually is the breaking of saccadic masking, but I need to do more research before I can be sure.

Also, the 30 degrees per second (while it might be the angular velocity of a head turn), is more important because its roughly the angular velocity our eye muscles can achieve for smooth pursuit without resorting to saccadic adjustments. We don't turn our head much while watching a computer monitor.

Similarly, pursuit tracking cameras, which replicating smooth pursuit of the eye, do not replicate the effect of persistence on strobed displays.

Sorry, this [my own] is a very poorly written sentence. See below. You are correct they accurately replicate the effect of persistence on strobed display as how motion is displayed without interpolation. But that's not what I'm talking about.

As we increase framerates, even with strobed displays, we'll be seeing 2-4 distinct images (the faster the motion, the further apart and the more distinct they will be, and the faster the refresh, the more images we'll see). Of course, with the decay rates, the impact of these images will be variable.

The number of distinct images is mathematically calculated as:
copies of image = (refresh rate / frame rate)
There is no distinct images if you have frame rate matching refresh rate.

So again, for all of the above, I'm talking about images perceived by the brain and the effects of image persistence because of the retina, not images being actively displayed.

Think of it this way - even in real life, with a constant stream of photons, we can't actually see motion without any blur even in our fovea (the most visually acute part of the retina). We need to track the motion, so that part of the light entering the eye becomes fixed when it hits the same spot of the retina, to realize the object clearly. If a display has a refresh rate, we are faced with one of two effects: it blurs because of sample and hold, or we see faint after images per impulse (amusingly, a similar effect to pixel persistence).

My previous points were based on how the retina and brain interpret images, not the properties of the display itself.

I highly recommend reading:
http://www.blurbusters.com/zero-motion- ... vs50vs100/

I think I've read every article on your site a few times now lol.

[Chief Blur Buster]

Older LCD's had this problem, but newer fast LCD's that are able to finish most of their transitions before the next flash, they behave very accurate square-wave persistence displays that accurately follows "Blur Busters Law".

So I do want to talk to you about this particular statement. I agree that you have definitively proven that some LCDs have fast enough pixels, but this is one are that I am finding your conclusions possibly lacking. I've see several LCDs, even 3D 120 Hz LCDs from as recently as 2012 that show triple frames and pixel persistence. These are TN panels with advertised response times that did not accurately reflect the true response times.

Are you sure you can say that all newer LCDs are able to do this? How many can and how many can't? And we know that TN panels have many tradeoffs for their fast pixel response properties - there are other quality considerations for truly fast pixels and a strobed backlight (Anandtech did show that between Lightboost and non-lightboost, the drop in quality wasn't too bad).

But really, my main concern is that LCD manufacturer's lied about pixel response time for literally years. How can we know that we can trust them now?

[Chief Blur Buster]

Given what I'm reading about the phenomenon, where the after image persists for approximately 0.04 seconds, this makes sense. A slower cadence doesn't not give the double image, but then motion is lost, and it becomes stop-go series of images.

On your plasma, it's more like approximately 5-8ms, (0.005 to 0.008 second) due to plasma phosphor decay.

Actually, I'm not talking about phosphor decay - I'm talking about the action potential frequency decay from the neurons in your retina. Once excited, the neurons take up to about 0.04 seconds to fully come back to a resting action potential frequency.

I believe you are correct, but I can certainly confirm that doesn't explain the double-image effect of 30fps@60Hz.

If it was purely because my eyes have moved, theoretically, it should appear as stuttering motion of a single UFO (jumps 2 steps ahead, then 1 step back, 2 steps ahead, 1 step back), not 2 UFOs that visually appear to always be there. The limits are very close, however, as you can detect some faint flickering.

Actually, both you and I are correct.
The UFO is still stuttering, but the black period (of strobing) is hiding the stuttering between two positions, and only the strobes are visible, so you are seeing the two endpoints of what was a stutter cycle.

Try this experiment:
-- Pursuit video camera of 1000fps on sample-and-hold
-- Slowly introduce strobing. Slowly adjust the duty cycle of the strobing, until you got more black period than bright.

So again, for all of the above, I'm talking about images perceived by the brain and the effects of image persistence because of the retina, not images being actively displayed.

Think of it this way - even in real life, with a constant stream of photons, we can't actually see motion without any blur even in our fovea (the most visually acute part of the retina). We need to track the motion, so that part of the light entering the eye becomes fixed when it hits the same spot of the retina, to realize the object clearly. If a display has a refresh rate, we are faced with one of two effects: it blurs because of sample and hold, or we see faint after images per impulse (amusingly, a similar effect to pixel persistence).

My previous points were based on how the retina and brain interpret images, not the properties of the display itself.

The double image effect is very easy to explain in my head, and I've found it's surprisingly mathematically simple.

Human vision does play a role and it can affect the accuracy of coorelation between pursuit camera and human vision. However, for double-image effect, this definitely is not the factor from my findings, my tests, and my napkin math. (I can write the napkin math out for you -- maybe you can roll it into something more formally scientific)

[Chief Blur Buster]

Older LCD's had this problem, but newer fast LCD's that are able to finish most of their transitions before the next flash, they behave very accurate square-wave persistence displays that accurately follows "Blur Busters Law".

So I do want to talk to you about this particular statement. I agree that you have definitively proven that some LCDs have fast enough pixels, but this is one are that I am finding your conclusions possibly lacking. I've see several LCDs, even 3D 120 Hz LCDs from as recently as 2012 that show triple frames and pixel persistence.

But you're right while I'm right -- they're actually separate factors.
That's something I've lately been calling "strobe crosstalk" because it's the leakage of incomplete transitions between refreshes. This usually doesn't easily exhibit itself in very rich detailed panning images such as http://www.testufo.com/photo because the richness of details and fewer high-contrast boundaries, make it harder to see the strobe crosstalk on modern displays.

This [strobe crosstalk] mainly appears in tests such as http://www.testufo.com/ghosting but less so in http://www.testufo.com/photo .
Also, LightBoost has far fainter strobe crosstalk than televisions (e.g. 99.5% completeness of pixel transitions before the next flash).

These are TN panels with advertised response times that did not accurately reflect the true response times.

That's correct.

Are you sure you can say that all newer LCDs are able to do this? How many can and how many can't?

There's no specific "can" or "can't".
LCD's have slowly been becoming more and more complete in transitions.

Strobe crosstalk (leakage of transitions between refreshes) is a different issue, and has improved so massively in the last few years, that Marc Repnow, the display researcher, says that his ASUS VG248QE has almost perfectly finished LCD pixel transitions during the "center band" of the screen -- this would represent approximately a 1-color-off error (e.g. RGB(230,230,230) only looks like RGB(229,229,229)), which is super-faint for strobe crosstalk, and much fainter than most 3D crosstalk of most television LCD's at the moment.

In a few years from now, I expect LCD's to exist where the pixel transition incompleteness is below human perceptible threshold. It already mostly is, on LightBoost displays, when viewing cluttered material such as http://www.testufo.com/photo on some of the better displays, and only seen in solid color bright/dark boundaries (e.g. http://www.testufo.com/ghosting ...)

Now, the manufacturers have done several-steps-forward, a few-steps-back, sometimes. For example, Eizo FG2421 Turbo240 is more impressively bright and colorful, but its motion blur is not as 'clean' and there is a bit more ghosting. But it looks fine for noisy-graphics-material games rather than games of lots of high-contrast solid-color-boundaries.

And we know that TN panels have many tradeoffs for their fast pixel response properties - there are other quality considerations for truly fast pixels and a strobed backlight (Anandtech did show that between Lightboost and non-lightboost, the drop in quality wasn't too bad).

Agreed.

But really, my main concern is that LCD manufacturer's lied about pixel response time for literally years. How can we know that we can trust them now?

There's actually two different pixel response measurements:
Grey-to-Grey pixel response -- manufacturers have been quoting this
Moving Picture Response Time -- this is equal to persistence. Google "MPRT response" in Google Scholar

I prefer MPRT because it is more directly proportional to motion blur (once other factors such as strobe crosstalk continues to become fainter, and go below human perceptible tresholds)

I should clarify:
Grey-to-grey = the time pixels are spent TRANSITION.
Persistence = MPRT = the time pixels are spent STATIC.

Yeah, it can be fuzzy between both. There can be a blend between static vs transition (e.g. phosphor decay). For LCD's, pixels are never perfectly static, so one has to define a reference point such as "static pixel state means the pixel is not perceptibly transitioning".

Grey-to-grey transitions is not the dominant factor affecting tracking based motion blur today. It's more persistence. For displays that displays a perfect checkerboard at http://www.testufo.com/eyetracking#pattern=checkerboard that confirms you have a display that more resembles square-wave persistence. Only square-wave sample-and-hold displays display a perfect checkerboard at http://www.testufo.com/eyetracking#pattern=checkerboard (for human eyes without vision problems; and capable of tracking accurately that test pattern). The slower the LCD, the smudgier/messier the checkerboard becomes, as the LCD starts to behave less as a square wave persistence. Test the eyetracking pattern on your iPad, on a TN LCD, on an IPS LCD, on a VA LCD, and on a plasma. Very interesting pattern differences show up. Only on OLED, DLP or recent TN (LightBoost compatible panel but with LightBoost OFF), it shows as a perfect or near-perfect checkerboard -- and for all practical purposes, it's effectively squarewave persistence to human eye.

[Chief Blur Buster]

OK, I agree with you that tracking motion at 30 fps on a 60 Hz impulse display will result in double images. That's because at 30 fps we're into perceiving smooth motion, and as such, the 2nd non-interpolated image would be in a different location.

However, this is oddly in sync with neuronal signal decay. Assuming it is 1/25th of a second (close to 30 fps), we would see 2 images - spaced apart about the same amount as our eye would track a moving object (on a 60 Hz display). We actually might be seeing the same double image that is re-enforcing itself due to both motion and 30 fps on 60 Hz, AND retinal persistence.

Here's a potential experiment - though if its too much hassle to code for, don't bother. But if you could get the 60 fps UFO test moving twice as fast, I could describe how it looks on my plasma for you. Right now, the spatial separation is about a half of a UFO, so its very hard to see if there is a 2nd UFO on the dark background and how compared to the 30 fps version (where the distance between the 2 UFO's are greater).

Sadly, I don't have any audio-visual equipment to test out your experiment. My background is entirely in other areas (I do cancer research currently, lol), I just wanted to give you some reference points for our discussion.

Pixel response discussion:

I do agree with you that some LCDs do exist that have fast enough pixel transitions. And you are correct, it will appear like strobing crosstalk when it does happen (with lightboost) and be less than normal pixel persistence. The ASUS VG248QE is fantastic about this, I guess I'm just concerned that other panels will not be so Lightboost friendly.

I also agree that over time, LCDs will likely improve and thus this may be a moot point. We do need to consider the possibility of impulse OLED displays that could potentially bring good to excellent motion resolution along with fixed all the other LCD deficiencies. With how good OLED contrast is, the lower impulse OLED light output may not matter indoors. The Samsung OLED TV with black frame insertion (50% off) still gets bright enough, and that's really a 1st generation product (though prohibitively expensive).

[Chief Blur Buster]

Indeed -- LightBoost is an imperfect solution, but many people consider it a miracle that LCD's can do that now. The EIZO FG2421 is pretty strobe friendly, but there are improvements they can make to it (e.g. eliminating that double-strobe). Blur Busters recognized the "LCD miracle" and pushed the word very hard around, to show that displays can be better... But we're not stopping here; we are anxious to cover OLED's once good desktop OLED's hit the market and manufacturers are able to send us an OLED of some kind...

Personally, I would prefer a good, inexpensive (<$1000) adjustble-persistence rolling-scan OLED. I believe they have the promise of easily achieving CRT-quality, in a brightness-versus-clarity tradeoff. You probably saw the Japanese paper about the Sony OLED and the 7.5ms duration rolling scan, the one linked near bottom of http://www.blurbusters.com/faq/oled-motion-blur/ ... Assuming you already did, I can confirm that adjusting the amount of time the OLED pixels are on, directly influences motion blur.

Since active matrix transitors aren't infinitely fast, the OLED pixels do take time (e.g. fraction of a millisecond) to turn on and then turn off, so color distortions may occur if you attempt to turn on-then-off an OLED pixels too quickly. So there /might/ be a limit that might be harder than a strobe backlight.

However, as long as we can achieve strobes of several hundred milliseconds on any display (LCD, DLP, OLED, etc) -- I'm waiting for a display manufacturer to release a display capable of 0.5ms strobes -- as I am quite confident that I can tell apart 0.5ms and 1.0ms with my eyes during fast ultra-detail full-framerate panning motion.

Even better is going flickerfree, but that's impractical without interpolation (at the moment). So selectable persistence displays (strobing on/off) are probably here to stay for video gaming -- OLED, LCD, DLP, etc...

But for OLED we probably could easily achieve 2ms persistence without noticeable color degradation, by having an OFF scan pass chasing only 2 milliseconds behind the ON scan pass. The rolling scan would still be squarewave and does not emulate phosphor decay (but that's only a flicker concern; and whether or not there is decay, has nothing to do with double image side effects). Even a true 120Hz OLED sample-and-hold would be extremely attractive to the typical Blur Busters reader, even if it wasn't strobed, and I would still be very interested in covering that display.

We're still wondering if the OLED black frame insertion is low-lag and works in Game Mode. The big problem is a lot of TV manufacturers aren't giving motion blur reduction love to Game Mode, and televisions are more expensive, harder to review, but Blur Busters would love to test/cover more of this too...

[Chief Blur Buster]

Double image discussion:

The 60 fps on 60 Hz speed increase wouldn't help, sorry. I thought about it, and you're eyes would track it so that each image would appear in the same spot in your retina, thus no retinal persistence would appear.

However, we would begin to see double images of background while tracking an object - but that's no big deal, since we can't focus on background while tracking that well anyways.

When it comes to tracking, I believe we were actually both correct. We see a double image and not a stuttering motion due to retinal persistence, which is only visible due to the fact that your eyes have moved since the last stationary frame occurred. I believe retinal persistence and the tracking effect of smooth motion are very much tied together as a visual property. Will read more into this.

The good news is you have clarified that retinal persistence should not limit motion resolution of tracked objects (I stupidly didn't think about how they would appear on the same spot on the retina every time when being tracked). The background may become double imaged or more, but that's okay, since that isn't typically in focus.

Thanks for the discussion.

It's true that OLEDs will have a minimum amount of strobe timing, although being much faster than LCD pixels, they should be much better about strobing crosstalk, especially at faster and shorter strobes (to reduce motion blur without too much flicker).

Also, this is something I have read from several sources, that so long as you're hitting around 100 Hz with sufficiently bright flashes, you really won't get any flicker. As for 0.5ms and 1.0ms, you are most certainly correct - because that is limited to spatial acuity. As you said, 1 ms is 1 pixel, correct? But that width is not anywhere near our visual spatial acuity assuming something like a 24" LCD at normal viewing. It doesn't matter TOO much since the display itself can only show 1 pixel at its best, but the effect of 1 ms blur is 2 pixels when looked at the leading and trailing edge of the pixel, so dropping that down to 1.5 pixels (again, measuring from the initial trailing edge to the end leading edge of the pixel during the 0.5 ms) is absolutely going to be a noticeable difference. Around 0.25 ms is probably where your visually acuity will start to be challenged as you'll be trying to see the difference between 1.25 pixels and 1.5 pixels (or roughly, the width of a stationary pixel on a 4K display).