Page 1 of 1

Re: Ideas of new TestUFO tests -- SUGGESTIONS HERE

PostPosted: 21 Dec 2018, 12:29
by pneu
May I suggest a test pattern which can be used to measure motion resolution.

For example there is this test pattern from Chapter 31 of FPD Benchmark Blu-Ray


The pattern scrolls horizontally and wherever you can see interline contrast up to in the converging lines, is the amount of motion resolution you have.

Re: Ideas of new TestUFO tests -- SUGGESTIONS HERE

PostPosted: 22 Dec 2018, 20:21
by Chief Blur Buster
pneu wrote:May I suggest a test pattern which can be used to measure motion resolution.

For example there is this test pattern from Chapter 31 of FPD Benchmark Blu-Ray


The pattern scrolls horizontally and wherever you can see interline contrast up to in the converging lines, is the amount of motion resolution you have.

Great idea!
On regards to the FPD Benchmark Software Blu-Ray Disc...

We sort of already have one --
That said, it does need to be made much easier to use.

We don't like "lines of motion resolution" as that's now an archaic motion resolution index, we prefer the modern science of MPRT (google it on Google Scholar). I prefer the technique of "MPRT" (Motion Picture Response Time) or the use of "Motion Clarity Ratios" (the inverse of MPRT). I explain GtG versus MPRT in 1000Hz Journey, and MPRT is a motion resolution benchmark. Exactly twice MPRT means exactly half motion resolution. So halving MPRT means double motion clarity. I trust MPRT now, it's the new modern method of measuring motion resolution.

MPRT = Moving Picture Response Time
It's totally different from GtG response time.

GtG = pixel transition time (how long a pixel takes to transition from one color to next)
MPRT = pixel visibility time (how long a pixel is visible for)

Tests have shown that MPRT is equivalent to motion resolution, and is now the new standard for motion resolution benchmarking nowadays.

And 1/MPRT represents the equivalent refresh rate of theoretical sample-and-hold display running at full framerate, e.g. A honest 2ms MPRT = requires a 500fps@500Hz perfect sample-and-hold display to match the motion clarity of.

For many modern displays, "MPRT" means the same thing as "persistence" when quoted as a number, and the number is linearly proportional to motion blur. It is resolution independent as well, so much more useful than old "lines of motion resolution" benchmarks which I find nearly useless nowadays, and TV manufacturers should adopt MPRT benchmarking like everyone else is for motion resolution.

Complexities such as GtG curves can muddy this, but modern displays are getting more and more increasingly squarewave pixel transitions with faster and faster GtG, and as long as that remains true, MPRT is very pure mathematically and linearly proportional to amount of motion blur, aka inverse of motion resolution.

Nontheless, adding a FPD-compatible motion test may be useful with a warning "outdated motion benchmarking technique", to allow people to map the old numbers to the new numbers. And obviously, I will need to make the MPRT benchmark much easier too, as well.

MPRT is the way to go now, and it has a great progression path to future 240Hz, 480Hz and 1000Hz displays -- which makes possible strobless blur reduction (blurless sample-and-hold), as seen in my 1000Hz Journey article.

Some TV manufacturers and monitor manufacturers now include a mention of "MPRT" (e.g. google "1ms MPRT") to boast how good their motion-blur-reduction modes are, and the claim actually matches benchmark motion tests much more honestly than GtG benchmarks and, thus is now suddenly becoming the new better way to share motion benchmarking (MRPT numbers) and will become very popular in 5 years from now, since my future tests will remove ratings/stars of monitor manufacturers that don't properly advertise "GtG response" versus "MPRT response"

TestUFO Motion Resolution Test (MPRT) - FPD Test Pattern Blu

PostPosted: 23 Dec 2018, 09:34
by pneu
Apologies, I wasn't even aware of MPRT until now :oops:

; ; ... 023108.pdf wrote:MPRT is the duration of time from the point at which a blur arises in the contour of an image in motion until the time it disappears, expressed in milliseconds.

MPRT (ms) = BEW (pixel) /v (pixel/frame) x Tf (ms/frame).

BEW stands for the perceived blurred edge width, which is proportional to the object’s moving speed (v)
Tf is the frame time (unit: ms), which is the inverse of frame rate (f, unit: Hz): 1000 / Tf

On the XB271HU @144hz I get a MPRT of 6.9ms according to the Blur Busters pattern.

So it takes the entire duration of a frame for blur to disappear, which seems to imply that the image is in a constant state of blur? But then how much motion resolution do I actually have? Can I plug known values into the above formula and solve for BEW to get the number of actual pixels of motion resolution?

edit: if I do that I get:
6.9ms = BEW / 23 x 6.9ms
BEW = 23

So does this mean at a horizontal scrolling rate of 23 pixels per frame I have 1 line of motion resolution per 23 pixels? 2560 horizontal res / 23 = 111 lines horizontal motion res? 23 pixels per frame @144hz is very fast though - can we convert that to a lower rate mathematically? Do we also need to take into account the checkerboard size and thickness? (I was using 24 and 2).

Side note: for some reason on my system (Windows 7, Chrome) the up/down control on the MPRT pattern stops taking effect after more than 1 click until I move the mouse cursor away from the up/down control.

Re: Ideas of new TestUFO tests -- SUGGESTIONS HERE

PostPosted: 24 Dec 2018, 23:10
by Chief Blur Buster
pneu wrote:Apologies, I wasn't even aware of MPRT until now :oops:

; ... ved-motion ; ... 023108.pdf wrote:MPRT is the duration of time from the point at which a blur arises in the contour of an image in motion until the time it disappears, expressed in milliseconds.

MPRT (ms) = BEW (pixel) /v (pixel/frame) x Tf (ms/frame).

Mathematically simplified into what I call Blur Busters Law.

EDIT: Now, we should note we use MPRT(100%) rather than MPRT(10%->90%) measurement (See new Pixel Response FAQ), so this writings use MPRT(100%) to make it possible to massively simplify the math to a level that even bloggers can understand, an Einsteinesque formula simplification that makes MPRT more mainstream.

1ms MPRT persistence = 1 pixel of motion blur per 1000 pixels/second

It's the exact same formula, quoted in an easier way!

1ms MPRT persistence = 1 pixel of motion blur per 1000 pixels/second
1ms MPRT persistence = 2 pixel of motion blur per 2000 pixels/second
1ms MPRT persistence = 4 pixel of motion blur per 4000 pixels/second\

2ms MPRT persistence = 2 pixel of motion blur per 1000 pixels/second
2ms MPRT persistence = 4 pixel of motion blur per 2000 pixels/second
2ms MPRT persistence = 6 pixel of motion blur per 3000 pixels/second

4ms MPRT persistence = 3 pixel of motion blur per 1000 pixels/second
4ms MPRT persistence = 6 pixel of motion blur per 2000 pixels/second
4ms MPRT persistence = 9 pixel of motion blur per 3000 pixels/second

This remains constant at all framerates, since MPRT numbers can go down with black frame insertion (strobing) and black frame insertion can be asymmetric ratios (e.g. 85%:15% OFF:ON ratio, like a strobe backlight that flashes for 2ms out of a 8.3ms refresh cycle). So the MPRT formula is difficult to deal with. So the simplification into Blur Busters Law is useful. Because of this, BEW gets hard to calculate with black frame insertion especially with such asymmetric ratios.

The beauty of Blur Busters Law is that BEW remains constant at all refresh rates if the strobe flash length is unchanged, for framerate=refreshrate=stroberate motion (triple match).
2ms flash at 60Hz strobe backlight = 2ms MPRT
2ms flash at 100Hz strobe backlight = 2ms MPRT
2ms flash at 120Hz strobe backlight = 2ms MPRT
2ms flash at 144Hz strobe backlight = 2ms MPRT

i.e. Motion resolution remains constant at all refresh rates for a given strobe flash length.

(2ms MPRT assumes squarewave flash unaffected by GtG or measurable fade-on/fade-offs, such as LED phosphor behaviours -- most of these effects are well below the noisefloor of human detectability and the MPRT tests become pretty accurate between human perception and measured numbers)

As GtG approaches an insignificant fraction of a refresh cycle (whether ultrafast GtG, or hidden in the black period between strobe flashes) -- the MPRT mathematics begins to become the simpler Blur Busters Law formula which is exactly the same mathematics into a simpler sentence. Most people can understand Blur Busters Law more easily than the full complex MPRT formula. So Blur Busters Law is a much easier formula for MPRT, and the math becomes purer/simpler especially if the GtG curve is almost squarewave (e.g. <=1ms GtG on TN or OLED monitors) and for those, it's easy to mathematically calculate predicted human-perceived motion resolution for any set of variables (any frame rate & refresh rate, any motion speed, any resolution), and actual measured motion resolution actually ends up within the error margin of the math calculation amazingly so.

Re: Ideas of new TestUFO tests -- SUGGESTIONS HERE

PostPosted: 24 Dec 2018, 23:20
by Chief Blur Buster
pneu wrote:So it takes the entire duration of a frame for blur to disappear, which seems to imply that the image is in a constant state of blur?

Correct. That's the sample-and-hold effect. Doubling Hz will halve motion blur (assuming GtG does not start bottlenecking MPRT).

pneu wrote:But then how much motion resolution do I actually have?

Your motion resolution is 6.9ms or if you want a motion clarity ratio, 1000/6.9 = 144.
And yes, sample-and-hold displays have a motion resolution matching the full refresh cycle.

pneu wrote:Side note: for some reason on my system (Windows 7, Chrome) the up/down control on the MPRT pattern stops taking effect after more than 1 click until I move the mouse cursor away from the up/down control.

That's a bug in the test. The test is an experimental beta and I'll be redesigning it to be as easy to test as FPD, but I will be labelling motion resolution as MPRT instead.

This experimental test is currently not the best one to use yet for MPRT benchmarking yet (more expensive equipment is often used), but it demonstrates that it is indeed possible to measure approximate MPRT using a test pattern, and the test demonstrates that millisecond differences in MPRT are indeed human-visible. Future test patterns will need to be created for easier MPRT benchmarking, perhaps based on the old FPD test pattern but converted to MPRT numbers or MCR numbers instead.

I had to specially design the MPRT test to function with incredibly fast motion speeds far beyond FPD's abilities -- including 4000 pixels/second and faster -- including strobed at 0.5ms and 1.0ms MPRT. A test pattern as easy as FPD becomes impossible when trying to compare 0.5ms MPRT versus 1.0ms MPRT versus 2.0ms MPRT (which are human visible) and requires different special tests.

If you have an NVIDIA ULMB monitor (nearly any 144Hz+ GSYNC monitor), here's a test that makes 0.5ms MPRT versus 1.0ms MPRT human-visible. Go to TestUFO Panning Map Test at 3000 pixels/second. You'll be unable to read the tiny street name labels which are 6 point size text. ULMB is by default 1.0ms to 1.5ms MPRT (sometimes 2ms MPRT). Now go into the monitor's menus, and select "ULMB Pulse Width" and lower the number to somewhere slightly 50%. That reduces the strobe flash length down to roughly 0.5ms MPRT as seen on a photodiode oscilloscope. Now you can read the street name labels! So you're probably by now, amazed that the human eye can see the difference between 0.5ms MPRT and 1.0ms MPRT. Thanks to motion blur reduction modes on modern gaming monitors capable of having adjustable MPRT (since MPRT equals strobe flash length) and NVIDIA actually added "ULMB Pulse Width" menu option upon my request after my very old LightBoost 10% vs 50% vs 100% advocacy. Obviously, the shorter the flash length, the darker the image, so brighter monitors have been coming out to allow brighter strobing at smaller MPRT numbers. Some new 240Hz monitors can do 1ms MPRT at nearly 300 nits, and 0.5ms MPRT at well over 100 nits, thanks to voltage-boosting in their strobe backlights. Nonteheless, the point is that it's extremely difficult to create tests that are human-visible for sub-refresh-cycle MPRT numbers, unlike for easy FPD test patterns which aren't so motion-critical. eSports players can play at motionspeeds far faster than the FPD motionspeeds. And MPRT numbers are much more universal.

Currently "lines of motion resolution" is utter uselessness on Blur Busters. Why?
  • The old "lines of motion resolution" tests were calibrated for TV motionspeeds instead of other kinds of motion speeds, such as videogame motionspeeds or computer scrolling motion speeds.
  • FPD was designed for 60Hz. Motion resolution varies on the same display at different refresh rates
  • Lines of motion resolution varies depending on motion speed.
  • Pixels per frame varies depending on framerate. Maintaining the same motion speed requires bigger pixel steps at lower frame rate.
  • Available framerate varies on refresh rate. The "lines of motion resolution" was invented with the assumption of 60Hz. High Hz displays give more framerate that gives you ability to increase motion resolution via reducing the sample-and-hold effect.
  • Back in the NTSC days, people used stationary "lines of resolution" to measure sharpness of CRT tubes. Stationary test patterns.
  • A motion test (FPD) converted the static terminology to motion terminology, "lines of motion resolution" to make it familiar to other videophiles/television engineers that were used to the old stationary "lines of resolution" tests of the older test patterns. The first invention of motion resolution was a much needed innovation in a motion test that was used in the test pattern. It was important for its time, but it is now fully obsolete. It was recyclage-of-old-terminology but is intrisinically now flawed as a display motion resolution measuring method.
  • MPRT benchmarking is refresh rate independent, framerate independent, resolution independent, and motionspeed independent -- you can (fairly accurately) calculate predicted motion resolution of everything else from just one single MPRT test!

Since it was simply recycled terminology that is wholly arbitrary (since motion resolution changes depending on motion speed, and 23 pixels per frame is a very arbitrary number); the phrase "motion resolution" is an arbitrary phrase -- I've written about the outdatedness back in year 2013 in this AVSFORUM article: Standardizing Motion Resolution: "Milliseconds of motion resolution" (MPRT) better than "Lines of motion resolution. Most manufactures now use MPRT measurements instead, and in the next 2 years I will have a much easier MPRT tests on TestUFO that is roughly as easy as FPD.

MPRT or measured motion-clarity-ratios (inverse of MPRT) is more universal, because it's easy to convert to a desired motion speed. MPRT numbers are much more universal because MPRT is motionspeed independent, unlike the old obsolete "lines of motion resolution" tests.

There are some subtle variances like how strong/faint the lines are, and differences in human vision on whether they can see the lines or not. GtG pixel response will affect this greatly, but the sample-and-hold effect will be largely dominant as long as the GtG pixel response is a tiny fraction of a refresh cycle. That's why your MPRT is darn near exactly 1/144sec -- that's correct.

pneu wrote:Can I plug known values into the above formula and solve for BEW to get the number of actual pixels of motion resolution?

Yes, MPRT numbers can easily be converted to motion resolution and vice versa if you know the motionspeed that the "lines of motion resolution" was measured at. If you don't know the motionspeed, then no, the conversion cannot happen.

TestUFO currently standardizes at 960 pixels/second because that is divisible by 60Hz, 120Hz, 240Hz and 480Hz, and is very similiar to many scrolling and common videogame speeds. And the number is so close to 1000 that as a result, it's super-easy to convert TestUFO results into MPRT number with an error margin that's within a stone's throw of expensive laboratory equipment.

Now, for alternating black-and-white lines, the same number can be converted to ANY motion speed you desire. So for single-pixel-thick alternating lines you have 2 lines so divide the inverse of MPRT by 2. The conversion formula is easy as it's a pair of pixels so you simply divide the inverse of MPRT by 2.

Converting MPRT into Lines Of Motion Resolution For 1000 Pixels/Second
(1 / (MPRT in seconds)) / 2
(1000 / (MPRT in milliseconds)) / 2

Translates into your motion resolution at 1000 pixels/second. Scale accordingly for different motion speeds; motion blur always scales linearly with motionspeed. Double motion speed means half motion resolution.

1ms MPRT = 500 lines of motion resolution during 1000 pixels/sec
1ms MPRT = 250 lines of motion resolution during 2000 pixels/sec
1ms MPRT = 125 lines of motion resolution during 4000 pixels/sec

10ms MPRT = 50 lines of motion resolution during 1000 pixels/sec
10ms MPRT = 25 lines of motion resolution during 2000 pixels/sec
10ms MPRT = 12.5 lines of motion resolution during 4000 pixels/sec

For your 6.9ms MPRT, the formula at 1000 pixels/sec is (1000ms / 6.9ms) / 2 = 72 lines of motion resolution at 1000 pixels/sec

I can't remember what motionspeed and framerate the FPD test was (motion resolution will go down at faster motionspeeds and/or lower framerates on sample-and-hold displays), but if you're saying 23 pixels per frame (are you sure?) and the frame rate was 60..... Then that means if you're doing the old "lines of motion resolution" which is 23 pixels per frame at 60 frames per second = benchmark is 23ppf x 60fps = 1380 pixels/sec

So you would do the formula (1000/1380ths) x 72 = 52 lines of motion resolution at 1380 pixels/sec under the old test pattern measurement criteria.

Now, if FPD was running at 30 frames per second instead -- then 23ppf x 30fps = 690 pixels/sec and the formula is (1000/690) x 72 = 104 lines of motion resolution at 690 pixels/sec

Barf. What the **** does "lines of motion resolution" mean if the number varies depending on parameters?

Hell. I don't even remember what FPD motionspeed was. With MPRT numbers I don't even need to remember the motionspeed or pixels-per-frame or whatever. MPRT becomes a universal number. Once you know MPRT of a display, the rest of the motion resolution numbers (with different variables like different framerate, different pixels per frame / different refresh rate, etc) can be estimated pretty accurately.

As you can see, MPRT conversion is universal, motionspeed independent, resolution independent, refreshrate independent, framerate independent, and that is why I feel that old "lines of motion resolution" terminology is utter obsolete trash nowadays like measuring in 15th century "Hands" measurements (and hands varied in size). MPRT is easily convertible to predicted motion resolution on the simple Blur Busters Law formula (which is the math simplification of the more complex MPRT formula)

Heck, MPRT can even be converted to "dpi of motion resolution" if you wish to invent a new method, simply by including the display's DPI in the calculation. A 300dpi display with drops to 75dpi of motion resolution

FPD was VERY useful and VERY important back in its olden days more than a decade ago, but is obsolete with non-60Hz non-resolution-consistent monitors

I am pleased that ever since I wrote that post, more and more display manufacturers have agreed, and have begun to sometimes advertise MPRT numbers with the gaming monitor.

Regardless, there is always some subjectiveness, both with FPD tests and with MPRT tests, due to various factors like GtG curves, but the formulas begin to match human perception the closer GtG reaches 0 (resulting in more and more mathematical beautifulness). The subjectiveness means Person X sees 300 lines of motion resolution, and Person Y sees 240 lines of motion resolution. The math formulas in this post falls squarely within this arbitrary subjectiveness range, and thusly, brings a lot more mathematical beauty to subjective measurements. Neither MPRT nor FPD test manages to cover nuances such as overdrive ghosting or strobe crosstalk, but MPRT is more data-complete than old "lines of motion resolution" terminology which can vary in framerate/refreshrate/speed/resolution/etc in subsequent different motion resolution tests. So one vendor's "lines of motion resolution" test is incompatible with a different vendor's "lines of motion resolution". MPRT solves that problem as a universal constant. Although this is imperfect, this is far better and vastly superior to the obsolete "lines of motion resolution" measurement of FPD test-pattern era.

Now, in the era of GtG being a tiny fraction of a refresh cycle, MPRT matching a refresh cycle (or nearly exactly) is extremely accurate for sample-and-hold displays such as LCD, LCoS, OLED (interpolation disabled, pulsing disabled, strobing disabled, BFI disabled). MPRT only becomes sub-refresh-cycle if the refresh cycle is impulsed (strobed / phosphor / pulsed / blackframe / etc). This is also explained in my Blur Busters Law And The Amazing Journey To Future 1000Hz Displays.

During 2019, I will be releasing some very easy MPRT motion tests that makes it much easier to quote motion resolution from MPRT numbers. And perhaps an old FPD converter tool too, for the analog-era diehards. Regardless, MPRT is The Way to go (or inverse of MPRT). I will see if I can model it after FPD, while displaying MPRT / MCR numbers, to allow easy conversion of motion resolution numbers to MPRT and vice-versa. That said, the master numeric value should always be recorded as a MPRT number (or its inverse, a measured motion clarity ratio) because amount of motion blur stays the same independently of frame rate, motion speed, resolution, refresh rate.

Nontheless, despite the obsoleteness of "Lines Of Motion Resolution", this is still an excellent idea to provide a motion resolution test that provides a good easy migration path to using MPRT (or its inverse 1/MPRT) as the proper successor motion resolution standard

Re: TestUFO Motion Resolution Test (MPRT) - FPD Test Pattern

PostPosted: 27 Dec 2018, 17:12
by Chief Blur Buster

I was doing some research on FPD test pattern.
I have found out that it moves at a rate of 6.5 pixels per frame (HDTVTest)

6.5 pixels per frame is unusually slow for modern displays now -- even for LCD, especially strobe-backlight LCD. That's only 195 pixels/sec at 30fps, and only 390 pixels/sec at 60fps -- much slower than the TestUFO default motionspeed of 960pps. And with increased resolution, 4K resolution, 390 pixels/sec takes 10 seconds to cross a 3840-pixel-wide screen, and 195 pixels/sec takes 20 seconds to cross a 3840-pixel-wide screen.

Those motion speeds are way too slow to successfully measure the motion resolution of ULMB which to my math calculations, at 0.5ms and much faster motion speeds, can equivalent to far more than 4,000 lines of motion resolution even at 1080p if we don't cap it at the static lines of motion resolution. At 6.5ppf, the motion resolution caps out quickly. This is a non-sequitur because the static motion resolution is less than that, and also reveals the further ridiculousness of the terminology recylage of "lines of resolution" (good old static resolution) -> into "lines of motion resolution" (moving resolution).

For modern motion resolution testing, motionspeeds need to be vastly faster than 6.5ppf to successfully measure the differences of 2ms, 1ms, and 0.5ms -- and therefore we need videogame-style motionspeeds, not slower-than-snail-pace motionspeeds that permanently shows maximum motion resolution on FPD, even though in some of my TestUFO tests, 0.5ms MPRT is actually four times sharper (in TestUFO Panning Map Test at 3000 pixels/sec on 4K display) than 2ms MPRT. See? Another reason why the FPD style approach (which isn't even a public standard but a consortium of display manufacturers back in the era of slower-responding displays -- we need an ISO standard or open scientific standard. Like, you know, "MPRT" which is a perfect base to piggyback off)

Perhaps, maybe I can make a variant of TestUFO Panning Map Test the new MPRT test, because I have found that the speed roughly equals half 1/MPRT. For example if you can read the TestUFO Panning Map Test at 960 pixels/sec, your "motion clarity ratio" is 480 -- common for 2ms LightBoost/ULMB. So a 60Hz display can successfully keep the street name labels readable at up to roughly 120 pixels/second before it blurs into complete unreadability. There's a big error margin in both directions but it is one of the few motion tests that can very clearly tell part 0.5ms MPRT versus 1.0ms MPRT versus 2.0ms MPRT.

  • sample-hold 60Hz (16.7ms MPRT) = panning map stays readable up to 120 pixels/sec
  • sample-hold 120Hz (8.3ms MPRT) = panning map stays readable up to 240 pixels/sec
  • 2.0ms MPRT (LightBoost) = panning map stays readable up to 960 pixels/sec
  • 1.0ms MPRT (144Hz ULMB mode on 240Hz GSYNC monitor ) = panning map stays readable up to 1920 pixels/sec
  • 0.5ms MPRT (144Hz ULMB mode on 240Hz GSYNC monitor + ULMB Pulse Width 50% setting) = panning map stays readable up to 3840 pixels/sec (almost too fast to read, so I slow it down to 3000 pixels/sec)
Since 1/MPRT is:
  • 16.7ms MPRT = inverse is 1/0.0167 = motion clarity ratio of 60
  • 8.33ms MPRT = inverse is 1/0.00833 = motion clarity ratio of 120
  • 2.0ms MPRT = inverse is 1/0.002 = motion clarity ratio of 500
  • 1.0ms MPRT = inverse is 1/0.001 = motion clarity ratio of 1000
  • 0.5ms MPRT = inverse is 1/0.0005 = motion clarity ratio of 2000
So I now begin to see the pattern; the motionspeed actually matches 1/MPRT divided by two!
That creates a useful basis for a brand-new motion resolution test that is not as limiting as FPD.

Which means, inventing motion tests to measure motion resolution via MPRT (or 1/MPRT) actually becomes easy, if I use some variant of a test like this that adapts well to slow motionspeeds all the way through fast motionspeeds. The formula is easy and maps very accurately to human perception.

The legacy tests neglects to consider additional variables I've explained in Blur Busters Law And The Amazing Journey To Future 1000Hz Displays, which is confirmed by VR scientists which I've also quoted, and the NVIDIA experiments in quad-digit refresh rates.

I'd like to buy the FPD Test Pattern Disc -- please email squad [at]
[EDIT: obtained the test]

Does anybody here have the disc? Please send me a PM, I will need to invent a new TestUFO pattern that is equally as easy or easier than that one. And so that I can make sure that my calculation formulas will properly convert from legacy lines-of-motion-resolution numbers to the new MPRT or 1/MPRT numbers (measured motion clarity ratios).

Currently, I am somewhat of a fan of inverse MPRT numbers as the potential future motion resolution standard, since it's the same terminology as "Samsung Clear Motion 960" or "Motionflow 240" or "ClearMotion 480" or such. The TV manufacturers hype this, and there is now a way to invent a test pattern that actually measures the honesty of these numbers. A perfect 960 benchmark means the manufacturer was being honest in the 960-eqiuvalence. A motion clarity ratio of 960 would be 4 times the motion resolution of 240, and be mathematically easy because it's framerate/refreshrate/resolution/scaling/motionspeed independent. And it's unbounded (unlike "lines of motion resolution").

My thinking is I would simply display multiple numbers simultaneously, simply calculated off the currently selected motionspeed:

Moving Picture Response Time (MPRT): XXX ms
Measured Motion Clarity Ratio (1/MPRT): XXX
Equivalent Legacy Lines Of Motion Resolution: XXXX

That way, users have a choice of motion resolution numbers:
(A) Users can compare to legacy motion resolution numbers (as long as they tested 1080p 60Hz)
(B) Users can compare actual measured motion clarity ratios to the hyped "motion clarity ratios" on manufacturer advertisements
(C) Users can use MPRT, the modern scientific motion-resolution measurement.
(D) This method of motion resolution has no upper limit -- it can reach human limitations instead (e.g. eye tracking speed)

(A) It is now possible for "lines of motion resolution" to exceed static line of motion resolution, e.g. 4000 lines of motion resolution at 1080p. If testing using faster motion speeds that scroll one screenwidth in less than 1 second. (another reason why "lines of motion resolution" is nonsequitur to me).

For users, 1/MPRT is easier to understand because bigger measured numbers are better; double the number is always universally twice the motion resolution.

Thoughts? Opinions?

However, the legacy "motion resolution" number will be less of an apples-vs-apples comparision at different refresh rates, since re-testing FPD may actually generate a different value because FPD is not 240 frames per second. That said, it would allow users to 'ease' into new standards. Especially since I have to also invent motion patterns capable of ultrafast motionspeeds (representative of material that benefits from blur reduction modes, like videogames) which forces me to invent a test pattern that diverges greatly from FPD. However, I think I can still do that and still keep the testpattern as easy as FPD, while being properly flexible for benchmarking any resolution, any refresh, any framerate.

Re: TestUFO Motion Resolution Test (MPRT) - FPD Test Pattern

PostPosted: 03 Jan 2019, 01:29
by Chief Blur Buster
Some data:
I finally got the FPD motion resolution test pattern.

1. It's 1080p at 30fps.
2. It's 6.5 pixels per frame.
3. It's 195 pixels per second, 5 times slower than the default motion speed in TestUFO tests.
4. It clips-out motion resolution maximum really early due to its very low framerates and slow motionspeeds.
5. Its low framerate bottlenecks motion resolution.

On sample-and-hold displays, motion resolution doubles at double framerate. Likewise, there here is 8x motion resolution at 240fps@240Hz. Modern motion resolution (0.5ms MPRT displays) can go more than 20x the motion resolution than this.

With this test pattern on sample-hold displays, the motion blur (without interpolation and without impulsing) is always perpetually 6.5 pixels, no matter the refresh rate, even if the display is capable of higher motion resolution at higher frame rates via tricks like interpolation. One way to see the highest motion resolution on this FPD test is if the display interpolates 30fps to a higher frame rate, and then strobes the high frame rate.

I'll definitely be creating a new TestUFO motion resolution test this year which will be way more universal, and capable of MPRT(100%) measurements all the way down to 0.5ms.

Re: TestUFO Motion Resolution Test (MPRT) - FPD Test Pattern

PostPosted: 04 Jan 2019, 01:03
by Chief Blur Buster
Scientific Paper Note: Blur Busters Simplification of MPRT

1ms persistence (MPRT100%) = 1 pixel of motion blur per 1000 pixels/sec

This is the Blur Busters simplification of the same MPRT formula found in that mentioned scientific paper, and much easier to calculate off motion tests which standardize at a motionspeed of 960 pixels/sec (very close to 1000 pixels/sec but divisible by common refresh rates as 60Hz, 120Hz, 240Hz).

We use MPRT(100%) instead of MPRT 10%->90% (in the scientific paper) because the entire MPRT motion blur is getting close to fully equal, so it becomes unnecessary (and more complicated) to exclude the first 10% and last 90% of the MPRT curve. This is an old carryover from GtG benchmarking where GtG is often the GtG90% metric.

Unlike for GtG, the full 100% of MPRT has actually become increasingly important on the increasingly faster displays (<2ms LCDs and 0.1ms OLEDs).

MPRT(100%) is the same amount of motion blur as the equivalent camera shutter. Meaning a 120Hz OLED has exactly the same motion blur as a 1/120sec camera shutter (MPRT(100%) = 8.333ms). As a result, we use MPRT(100%) at Blur Busters for simplicity and to correctly match human-perceived motion blur on the ideal squarewave sample-and-hold display.

There are other complexities such as long GtG and strobe crosstalk that prevents the full MPRT(100%) being achieved, but for simplicity, MPRT(100%) can never be less than the duration of a refresh cycle on a sample-and-hold display. The more perfect a sample-and-hold refresh cycle becomes, the more exactly motion blur exactly matches a camera photograph of a specified shutter length.

Going forward, we'll use MPRT(100%) because the entire blur is equally strong (even the 10% and 90% edges) in fully continuous motion, and the resulting picture on 120Hz sample-hold screen is (on average) never sharper than a photograph of 1/120sec camera shutter speed. So I prefer to use the full milliseconds of true motion blur matching the equivalent said camera photograph, MPRT(100%) = 1/120sec = 8.3ms for 120Hz sample-hold.

In some situations for sample-and-hold with asymmetric pixel responses, one direction of pixel transition could have an MPRT(100%) slightly less than a refresh cycle, while the other direction of pixel transition can have an MPRT(100%) slightly more than one refresh cycle. Averaged for all GtG combinations (e.g. all 256x256 combos of 8-bit panels), MPRT (100%) can never be less than a refresh cycle.

We may have to display both MPRT numbers. MPRT(90%) and MPRT(100%), or simply use the "persistence" terminology.

Re: TestUFO Motion Resolution Test (MPRT) - FPD Test Pattern

PostPosted: 23 Jun 2019, 14:40
by Chief Blur Buster
UPDATE: Here's a new image comparing persistence(MPRT) based display motion blur to a blurry real-world photograph -- the blur mathematics is identical when we math with MPRT(100%)! Which makes it easy to explain science to readers.

This image is from the Pixel Response FAQ. There are reasons MPRT is measured from 10% to 90% but the problem is that it makes popular science explanations to readers much more difficult, and massively complicates visual blur analysis.

This makes it more self-explanatory why we prefer to explain MPRT blurring via MPRT(100%). This permits easier explanations (At the popular science level, rather than journal level) we successfully do on Blur Busters.