Standardizing Motion Resolution: "Milliseconds of motion resolution" (MPRT) better than "Lines of motion resolution" - Page 2 - AVS Forum | Home Theater Discussions And Reviews
Forum Jump: 
 3Likes
Reply
 
Thread Tools
post #31 of 76 Old 10-04-2013, 08:44 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

I wasn't questioning the accuracy of pursuit camera - just saying that it would be nice to have a test that didn't rely on one.
I never said that.
Pursuit camera tests are an additional optional tool for MPRT measurements.
They just simply give photographed proof of the motion blur, and makes it possible to measure MPRT (or "lines of motion resolution") from the photographs.
Read my above new post.

You can skip the pursuit camera, and even use MPRT with existing "lines of motion resolution" patterns (as limited as they may be).

MPRT just makes it easy to analyze the photos, knowing that mathematically, 1ms MPRT = 1 pixel of motion blur during 1000 pixels/sec. For my pursuit camera Sync Track invention, the tick marks in the Sync Track is a convenient pixel ruler too to calibrate scientific motion blur measurements (as they're a known pixel separation value). It's not the only way to skin a cat -- there are numerous ways to analyze MPRT not even written about in this thread (as this carefully-worded google search reveals).

Note: Sure, ghosting/decay effects do complicates things a bit in giving a number to motion blur. So we choose a standardized rise/fall cutoff, science observations is suggesting 50%. Otherwise we come to situations where a certain CRT with a certain phosphor may be creating a motion blur that shows up as 1.2ms MPRT under a certain measurement cutoff standard, but 1.6ms MPRT under a different measurement cutoff. (the non-linear phosphor decay curve -- do we cutoff at 50%? cutoff at 90%?) In an attempt to get the objective measurements to more closely agree with subjective observations. (now that display motion blur is now cheaply accurately photographically capturable using the method I invented) Even these numbers, 1.2ms MPRT and 1.6ms MPRT is like splitting hairs (nitpicking, essentially), when we're comparing a 1ms strobe backlight to a 5ms plasma phosphor, to a 16ms MPRT sample-and-hold 60Hz LCD. I point out contrast ratios are also complicated too. Display non-uniformity. Backlight bleed. Ambient light. Dynamic contrast ratios. Static contrast ratios. Different size of contrast ratio checkeboard (ANSI vs non-ANSI). Halo effects in local dimming. Plasma voltage sag (dimming of excessive bright image) affecting contrast. Etc. Etc. These variables changes the contrast ratio numbers.

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
Sponsored Links
Advertisement
 
post #32 of 76 Old 10-04-2013, 08:59 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Can you explain what "lines of motion resolution" is? Before this thread, I'd never heard of it, and I can't seem to find an answer on google. I imagine it's something about how many moving lines you have have before they blend into each other?
spacediver is offline  
post #33 of 76 Old 10-04-2013, 09:07 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

Can you explain what "lines of motion resolution" is? Before this thread, I'd never heard of it, and I can't seem to find an answer on google. I imagine it's something about how many moving lines you have have before they blend into each other?
Ok, I thought you knew before you read this thread.
Let Me Google That For You.

Search: "1080 lines of motion resolution" -avsforum (excludes AVSFORUM posts)
Search: "1200 lines of motion resolution" -avsforum (excludes AVSFORUM posts)

Examples:
- C|Net Review of HDTV measuring motion blur by "lines of motion resolution"
- HDGuru talks about motion blur by "lines of motion resolution"
- HDTV Magazine: Panasonic Viera touting "lines of motion resolution"

Hundreds of home theater magazines sitting here, blabbing about "lines of motion resolution" in their HDTV reviews. Ugh.
Mainstream reviewers are still using "lines of motion resolution" which is not future-proof, is arbitrary, and is not apples-vs-apples.

Shall I continue?

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
Sponsored Links
Advertisement
 
post #34 of 76 Old 10-04-2013, 09:09 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
edit: nm, thanks.


key part:
Quote:
The main pattern used to test motion resolution from the FPD Benchmark disc is called the monoscope, which scrolls across the screen at a set, medium-speed rate. A series of vertical wedges consisting of four parallel lines, each with numbers corresponding to a vertical resolution between 100 and 1,200, is set in the middle of the pattern. While the pattern scrolls across the screen the lines blend together and blur, to a lesser or greater extent depending on the display in question. To arrive at a number that corresponds to the display's vertical resolution, the viewer (er, reviewer) must judge when the lines become blurry and blend into one another. The number closest to the place on the wedge just before the blending occurs is the motion resolution.
spacediver is offline  
post #35 of 76 Old 10-04-2013, 09:16 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

edit: nm, thanks
Don't forget that FPD is not the only method of measuring motion blur in "lines of motion resolution". Various different Blu-Ray motion tests have muddied this. Several of them use different methods of measuring "lines of motion resolution", and create different measurement numbers that are not comparable to each other.

See, the incompatible standards are part of the problem.

Blur Busters will at some point (e.g. 2014 or 2015) declare a war against "lines of motion resolution". (I have a several display reviewers lined up on my side now already. For example, TFTCentral and pcmonitors already agree with this too, for example.) I'm lining more people behind the scenes. Right now, my priority is focussing on beta-testing the motion tests, as well as on the upcoming Blur Busters Input Lag Tester amongst other things. Above and beyond my primary job (which isn't Blur Busters).

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #36 of 76 Old 10-04-2013, 09:17 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
right, it's just hard to find an actual explanation of what lines of motion resolution is. Hundreds of references to it, but very little description of what it is.
spacediver is offline  
post #37 of 76 Old 10-04-2013, 09:20 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
anyway, when I'm feeling motivated to do so, I'll probably program in that rectangle gap test, and see how it looks on the fw900, the vpixx display, and a regular crap lcd. I'll report back when I do so.
spacediver is offline  
post #38 of 76 Old 10-04-2013, 09:23 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

right, it's just hard to find an actual explanation of what lines of motion resolution is. Hundreds of references to it, but very little description of what it is.
Yep. You found it; FPD's explanation is the best one.

Don't forget to test www.testufo.com/mprt-fast on the Viewpixx LCD. My prediction for the Viewpixx Scientific LCD, is that the lines becomes a solid block during ppf=2 in non-strobed mode (MPRT=8.3ms), and approximately ppf=16 in strobed mode (MPRT=1.0ms). It may be a bit higher or lower. I'm basing my predictions based on the data I already know from the monitor's specifications.

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #39 of 76 Old 10-04-2013, 09:36 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

anyway, when I'm feeling motivated to do so, I'll probably program in that rectangle gap test, and see how it looks on the fw900, the vpixx display, and a regular crap lcd. I'll report back when I do so.
I'd love to hear your findings. Or give me the specifications of the test, and I will attempt to create a hidden/experimental TestUFO test based on it.

I do not think gaussian curves is an appropriate fit for the sample-and-hold effect, because motion blur isn't gaussian on many displays. (e.g. plasma, DLP, strobe-lit LCD, CRT). Strobe backlight LCD's give you linear motion blur where the beginning of the blur and the end of the blur is rather sharp (due to the on/off of strobe backlight). The same thing happens with DLP's too, and fast-response OLED's. Only traditional crap LCD's have a motion blur that more resembles gaussian because of the pixel transitions, but it's not a perfect gaussian. The motion blur curves vary a lot between different displays. (I've already explained that, but this is me saying this again in a different way).

- Linear motion blur (square-wave light output displays) -- DLP, strobe-backlight LCD, OLED (Note: DLP can add a bit of PWM 'noise' into the blur. As does LCD dithering / inversion artifacts.)
- Asymmetric motion blur curve (ghosting/blurrier on only one edge of motion) -- slow LCD, high-persistence CRT, plasma yellow trailing
- Gaussian motion blur -- a medium-speed LCD with symmetrical pixel transitions
- Near-linear motion blur -- Fast TN LCD running at 60Hz with well-tuned overdrive. (1ms-2ms transition is almost square wave when compared to 16ms refresh, so effectively 16ms of linear motion blur)
- Noisy motion blur -- Plasma during dim colors, especially in Game Mode (less processing)

Good example test: www.testufo.com/blurtrail
On a 1ms LCD without strobe backlight, the motion blur nearly perfectly linear (solid grey bar with fairly sharp leading/trailing edges)
On a older 8ms IPS, the motion blur looks gaussian (soft leading/trailing edges). Often asymmetric, since the the pixel transition speeds aren't symmetric in both directions.

All of the above, can, of course, overlap, to a composite motion blur, that looks neither linear nor gaussian. Anything that blurs (bright ghosting) is part of the MPRT equation since it is interfering with motion clarity. Plasma yellow trails are a form of bright ghosting that actually affects motion clarity of bright material on plasmas, and thus affects MPRT. Strong ghosting is perceived (by eye) or captured (by pursuit camera) as a form of motion blur that interferes with motion clarity (e.g. ghosting more than 50% as bright as original moving object). Faint ghosting will typically not affect MPRT, as it's just a peripheral artifact much like 3D crosstalk (ghosting less than 1% as bright), it's well below the motion blur curve cutoff. So here, ghosting is a close family relative to motion blur, it's just asymmetric because of different rise/fall speeds of the leading/trailing edges of moving objects (CRT faster illuminate and slower decay, LCD faster to one color than back to original color, etc).

MPRT is mathematically perfect on linear motion blur (square-wave-light-output displays).

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #40 of 76 Old 10-04-2013, 10:09 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Quote:
Originally Posted by Mark Rejhon View Post

Yep. You found it; FPD's explanation is the best one.

Don't forget to test www.testufo.com/mprt-fast on the Viewpixx LCD. My prediction for the Viewpixx Scientific LCD, is that the lines becomes a solid block during ppf=2 in non-strobed mode (MPRT=8.3ms), and approximately ppf=16 in strobed mode (MPRT=1.0ms). It may be a bit higher or lower. I'm basing my predictions based on the data I already know from the monitor's specifications.

If you want to visit the lab some day you can bring a laptop and hook it up to the display and test it out yourself. A wknd would probably work better as there won't be anyone around. Lemme look into it.
spacediver is offline  
post #41 of 76 Old 10-04-2013, 10:26 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Quote:
Originally Posted by Mark Rejhon View Post


I do not think gaussian curves is an appropriate fit for the sample-and-hold effect, because motion blur isn't gaussian on many displays. (e.g. plasma, DLP, strobe-lit LCD, CRT). Strobe backlight LCD's give you linear motion blur where the beginning of the blur and the end of the blur is rather sharp (due to the on/off of strobe backlight). The same thing happens with DLP's too, and fast-response OLED's. Only traditional crap LCD's have a motion blur that more resembles gaussian because of the pixel transitions, but it's not a perfect gaussian. The motion blur curves vary a lot between different displays. (I've already explained that, but this is me saying this again in a different way).

I now understand what you mean about the decay function affecting the nature of the blur trail. As such, different decay functions will result in different parts of the gap being filled at different levels of motion blur.

Hm... perhaps one approach is to try and quantify motion blur as the area under the decay function curve. This would avoid the cutoff issue you discussed earlier. Think of it as "luminance energy per unit time". One one extreme, you'd have a pixel that instantaneously reached full brightness and remained at that level on for the entire duration of the frame, and then took its sweet time decaying. On the other extreme you'd have a pixel that instantaneously reached full brightness and stayed on for an infinitesimal duration.

Now how could you measure this psychophysically without a photodiode...
spacediver is offline  
post #42 of 76 Old 10-04-2013, 10:45 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Now how could you measure this psychophysically without a photodiode...
Pursuit cameras can do it.
Example:
Millimeter #1 in photographed motion blur trail is the first 1ms of pixel transition
Millimeter #2 in photographed motion blur trail is the first 2ms of pixel transition
Millimeter #3 in photographed motion blur trail is the first 3ms of pixel transition
Millimeter #4 in photographed motion blur trail is the first 4ms of pixel transition
Millimeter #5 in photographed motion blur trail is the first 5ms of pixel transition
etc.

So along a linear axis along the pursuit camera photograph, you can create a pixel response curve! biggrin.gif
Just like a photodiode. The pursuit camera photograph is a linear record of the pixel response curve. It's beautifully simple.
Tickmarks (e.g. Pursuit Camera Sync Track) can create the necessary calibration to synchronize the physical millimeters of the captured photo, to the actual pixels, and then create a pixel response graph from the pursuit camera photograph.
Beautifully simple, assuming you make sure you properly adjust your camera's dynamic range to fit the whole gamut of your monitor (a matter of simple photographer's / scientist's skill)

Commercial MPRT cameras already do this. The homebrew pursuit camera method, that I invented, won't be as accurate as certain photodiode or commercial methods, but it will still graph pixel response curves accurate to ~0.25ms or thereabouts (in my tests -- e.g. when adding an undocumented experimental Sync Track to TestUFO Blur Trail), while simultaneously providing a WYSIWYG photo that can be studied -- this creates several other superior advantages -- something a photodiode cannot do.

I believe that once a human finally understands when display motion blur CAN be interpreted as the pixel transition curve, and understands how framerate/sample-and-hold/strobing interacts with it, one can finally be considered a hobbyist vision researcher who "who gets it".
Quote:
Originally Posted by spacediver View Post

Hm... perhaps one approach is to try and quantify motion blur as the area under the decay function curve. This would avoid the cutoff issue you discussed earlier. Think of it as "luminance energy per unit time". One one extreme, you'd have a pixel that instantaneously reached full brightness and remained at that level on for the entire duration of the frame, and then took its sweet time decaying. On the other extreme you'd have a pixel that instantaneously reached full brightness and stayed on for an infinitesimal duration.
The area under the entire off-on-off curve, is an excellent suggestion. We're talking about the entire curve, not just the rise-fall. That means sample-and-hold would be accounted for, whether square-curve or gaussian-curve. Thinking about this more, the 50% cutoff method will often resemble this quite well, but can fall apart for strange curves (e.g. plasma rampup-and-then-decay). The area method is more complex and requires you to use the pursuit camera photograph method more.

Photodiode method -- is very simple, yes. Much simpler than pursuit camera. It could technically be done as a photodiode on a one-frame flash (off-on-off). The curve of the flash will often accurately predict the perceived motion blur of the display, though not always (due to various complications like complex plasma processing, and different pixel transitions being different speeds, etc). And you don't have blogger-friendly photographic proof.

Nontheless, motion test patterns for determining MPRT is still useful as is a much needed improvement to motion testing.

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #43 of 76 Old 10-04-2013, 11:06 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Quote:
Originally Posted by Mark Rejhon View Post

Pursuit cameras can do it.
Example:
Millimeter #1 in photographed motion blur trail is the first 1ms of pixel transition
Millimeter #2 in photographed motion blur trail is the first 2ms of pixel transition
Millimeter #3 in photographed motion blur trail is the first 3ms of pixel transition
Millimeter #4 in photographed motion blur trail is the first 4ms of pixel transition
Millimeter #5 in photographed motion blur trail is the first 5ms of pixel transition
etc.

Yep I understand this. But wouldn't it be great if you could do it with pure psychophysics? Unless of course it was convenient for every reviewer/consumer to set up a proper pursuit camera!
Quote:
I believe that once a human finally understands when display motion blur CAN be interpreted as the pixel transition curve, and understands how framerate/sample-and-hold/strobing interacts with it, one can finally be considered a hobbyist vision researcher who "who gets it".

hey, what about enlightened cephalopods?
spacediver is offline  
post #44 of 76 Old 10-04-2013, 11:29 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

Yep I understand this. But wouldn't it be great if you could do it with pure psychophysics?
I don't think I could pull off a test pattern that allowed a human to estimate the surface area below a pixel response curve, but throwing things out in the wild:
-- Two simultaneous motion tests running simultaneously (e.g. moving edge tests)
-- One optimized for the leading edge, one optimized for trailing edge
-- Some interaction between the two create an effect
-- Comparision with solid bars above/below test pattern (e.g. dim grey, mid-gray, bright gray) -- though that may require a very specific gamma calibration (so that 50% gray is really 50% the luminance of 100% gray) to determine approximate area

It's almost beyond my gray matter's ability to figure out how to create a human-eye motion test that can even approximately calculate the area under a curve, but my brain is telling me it's not quite impossible (People said the human eye couldn't tell apart 1ms vs 2ms, but I've successfully created www.testufo.com/mprt-fast that easily allows me to tell apart a 1ms strobe backlight, a 1.5ms strobe, and a 2ms strobe).

Realistically, we're nitpicking over how accurate the human-eye version of the MPRT test needs to be. I think it only needs to be accurate in its approximate logarithmic territory. e.g. we don't care about 0.4ms-0.5ms-0.6ms as much as we care about 0.5ms-1.0ms-2.0ms-4.0ms. Loosely speaking, a good practical a human-eye MPRT pattern is accuracy to within 25% at high MPRT values, and accuracy to within 50% at low MPRT values. It's like caring a lot about contrast ratio 5000:1 vs 1000:1, but not caring much about 1000:1 vs 1200:1. Specifically for AMOLED and for LCD (non-strobed or long-length strobes), the www.testufo.com/mprt have already shown to show accuracies within +/- 5% when doing the six-pass test on an LCD. (black->white, white->black, black->50%gray, white->50%gray, 50%gray->black, 50%gray->white, and then averaging all six MPRT / MER values). Even the 2-pass test (white->black, and black->white) show pretty practical accuracies too.

I've come up with a single, unified test pattern (not yet published) that merges the best of www.testufo.com/mprt with the best of www.testufo.com/mprt-fast so that the same test pattern can now actually measure MPRT's from slightly below 1ms (short-strobe LCD's), all the way to ~40ms (24Hz sample-and-hold), at which point moving edges strobes too slowly to be perceived as motion blur. It just becomes somewhat more inaccurate with complex curves like plasmas. The "shape" of the plasma response curves is a rather bizarre "shape" compared to CRT response curves and LCD response curves. Estimation becomes a lot more difficult for plasma due to plasma's response curve behavior, certainly definitely a much bigger error margin but probably like 25% error ("subjective" measurements diverging away from "objective" measurements). More tests are needed. The 25% error in motion blur measurements (human eye versus equpment) is like caring about whether a display is 5000:1 contrast, versus 6250:1 contrast; not a huge difference at all. The relative difference appears to matter more during FPS game use cases; e.g. we care about 1ms-versus-2ms, but we don't really care about 1.5ms-vs-1.6ms, or 16ms-vs-17ms. Certainly at faster MPRT's, we are hitting diminishing points of returns, but the same thing is happening with ever higher-and-higher contrast ratios. (Yet we still continue to see further improvements in certain situations, such as space movies for contrast ratios, or highdef FPS games / highdef virtual reality for motion blur)

[To future readers: If you're reading this thread in Year 2014, all MPRT tests may be merged into www.testufo.com/mprt .... with the www.testufo.com/mprt-fast deleted]

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #45 of 76 Old 10-04-2013, 11:48 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Good thoughts Mark.

Sounds like the best thing to do is to run these tests on displays for which you already know a lot about the pixel transition curve. I wonder whether a cheap photodiode connected to an oscilloscope could provide a reference measurement.
spacediver is offline  
post #46 of 76 Old 10-05-2013, 12:02 AM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

Sounds like the best thing to do is to run these tests on displays for which you already know a lot about the pixel transition curve. I wonder whether a cheap photodiode connected to an oscilloscope could provide a reference measurement.
It does, for pure-strobe displays (Strobe-backlight LCD, OLED)

I already did some brief tests: My prototype 4K-compatible / 120Hz-compatible Blur Busters Input Lag Tester (arduino+photodiode) already has a mode where it forecasts MPRT numbers from a sensor is put on a flickering rectangle. It's very accurate on certain displays, and less so on others. (the Arduino lag tester is running as a 10KHz oscillscope, which conveniently allows me to do other things other than testing input lag). For simplicity at this time, it's currently using a 50% rise-fall cutoff point, but could easily be reprogrammed to calculate the area under a curve. It's uses the 3-flashing-square method, similar to Leo Bodnar, using the computer as the video generator (rather than self-contained, as in the Leo Bodnar unit). A few sites have already signed up to be beta testers of the lag tester (familiar name sites). Meanwhile, I'm looking for a manufacturer to package a small run of nice handheld sensors rather than barebones Arduino's (Though I prefer Teensy, due to lower accurate USB latency). Something similar to a Leo Bodnar accessory, but one that plugs into a computer (e.g. laptop) which you connect to the display.

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #47 of 76 Old 10-05-2013, 12:14 AM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
wow, awesome!

Also, what are your thoughts on the scrolling text test that pixperan uses? Can it be made to be resolution independent? I see the strength of that sort of test as a real world test of the functional consequence of motion blur, and it would be a valid test across all sorts of pixel response functions (including the complex/noisy plasma ones). The disadvantage is that it doesn't lend itself well to quantitative generalization like MPRT. But I do think that this sort of test is an important complement to the tests that spit out an MPRT.
spacediver is offline  
post #48 of 76 Old 10-06-2013, 09:04 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

Also, what are your thoughts on the scrolling text test that pixperan uses? Can it be made to be resolution independent?
Yep. Just keep the text-versus-speed constant. If it's scrolling at 10 average-letter-widths/second, just keep it scrolling 10 average-letter-widths/second. If you double the text point size, just double the speed of text, the motion blur obscurring will stay relative, and the readability factor stays constant on the same display.

The display's own motion blur always scales proportionally with motion speed (it's scientifically provable with math formulas) (we're excluding source-based motion blur, including sourced from interpolators or compressors, and we're excluding human eye saccade limitations -- so we're strictly solely measuring display motion blur, as the guaranteed minimum motion blur that a display will be forcing upon your eyes)

Good see-for-yourself proof: View www.testufo.com/eyetracking#pattern=checkerboard on an iPad 4th generation (2nd and 3rd generation may work, but may be a bit slow). Zoom in and out with the pinch-zoom. The motion blurring checkerboard illusion stays proper at all zoomed sizes, it doesn't even have to be pixel-perfect.

The TestUFO Panning Map Test (my analog to PixPerAn's readability test) is another great example, too. Go to the fastest panning speed that you can still read the map labels at. I haven't "gameified" this yet, like Pixperan did, though. Doesn't matter what resolution this test runs at, or if you zoom the browser or not (ideally you want 1:1, but that specific test is not even 1:1-dependant for the purposes of testing human readability). If you view the map labels at higher/lower zoom settings, the motion blur is proportional to the text size, as zooming in means the panning becomes faster on bigger text, and the motion blur trail enlarges to catch up.

So the answer to your question, is, "YES". (And it's mathematically provable, too!)

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #49 of 76 Old 10-06-2013, 09:23 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
nice. I should also mention that, at least on my computer, the testufo is still quite choppy. Even when it syncs, etc. I'm on an intel core2duo, with a geforce gtx 660. I'll do some more experimenting soon - gonna upgrade the mobo/processor in very near future. Pixperan, on the other hand, is silky smooth.
spacediver is offline  
post #50 of 76 Old 10-06-2013, 09:29 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

nice. I should also mention that, at least on my computer, the testufo is still quite choppy. Even when it syncs, etc. I'm on an intel core2duo, with a geforce gtx 660. I'll do some more experimenting soon - gonna upgrade the mobo/processor in very near future. Pixperan, on the other hand, is silky smooth.
Agree that TestUFO pushes the limits of web browsers, as web browsers were not originally designed for motion perfection. But today, ~80% of mainstream visitors now are visiting with browsers that run TestUFO perfectly with no framedrops if only running one browser tab with no apps in background (and not moving around the mouse during the test). This improves to about 90-95% of HardForum users, since they're already using fast systems.

-- Also, make sure you meet the pre-requisites at www.testufo.com/browser.html ....
-- If using multimonitor, temporarily disable multimonitor.
-- If using chrome, make sure everything is enabled in chrome://gpu
-- The core2duo will be hobbling TestUFO, but you can fix this by temporarily switching browsers too.
-- FireFox also has stutter bugs. Try Chrome or Opera 15+, not FireFox .
-- Chrome and Opera 16+ stutters least, followed by Internet Explorer (good), then followed by FireFox (worst, though FF24 is light years better than previous FF), followed by Safari (buggy on several Mac's, perfect on other Mac's).

If you still get stutters even in Chrome or Opera, can you please do a screenshot of www.testufo.com/animation-time-graph and www.testufo.com/#test=animation-time-graph&measure=rendering and email them to me at mark[at]blurbusters.com ? I need to make it automatically display "Browser Stutter Detected" whenver a visible stutter occurs. (As JavaScript doesn't have a direct framedrop detector, I detect framedrops via heuristics on this graphed data internally. It's calibrated to be very reliable in most versions of Chrome on Windows, but I may have to calibrate the synchronization detector for other browsers).

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #51 of 76 Old 10-06-2013, 09:59 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Quote:
Originally Posted by Mark Rejhon View Post

-- If using chrome, make sure everything is enabled in chrome://gpu

Problems Detected
Disable 3D (but not Stage3D) in Flash on XP: 134885
Hardware video decode is only supported in win7+.: 159458
Accelerated 2D Canvas is not supported on WinXP.: 175149
Accelerated 2D canvas is unavailable: either disabled at the command line or not supported by the current system.


guess I need to upgrade from xp
spacediver is offline  
post #52 of 76 Old 10-06-2013, 10:06 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Bingo -- there's your problem.
Actually the more I think of it, core2duo isn't the problem. An iPad has a slower CPU than a core2duo, yet works perfect on iPad.

TestUFO requires accelerated Canvas2D. And, yes, and might as well buy a whole new computer at this stage. (maybe recycling a few parts such as your 660 and your SSD). I probably need to add code to stop displaying "VALID" on XP systems, or at least a subheading warning about XP.

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #53 of 76 Old 10-28-2013, 01:01 AM
AVS Forum Special Member
 
Elix's Avatar
 
Join Date: Jun 2011
Location: Dungeon, Pillar of Eyes
Posts: 1,574
Mentioned: 6 Post(s)
Tagged: 0 Thread(s)
Quoted: 253 Post(s)
Liked: 106
Kudos to you, Mark, for moving the display science forward smile.gif I vote with both hands for the MPRT standard.
You got me very interested in Optoma DLP. Do you know what MPRT it has in 120 Hz + BFI mode? Also, any other models/brands of 3D DLP projectors compatible with that tweak?
Elix is online now  
post #54 of 76 Old 10-28-2013, 01:33 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
just had an idea about measuring MPRT psychophysically. involves a method of adjustment, where observer follows a moving square, observes the blur trail, and while doing this, adjusts two independent parameters of a static representation of a blur trail. The two parameters are the length of the blur trail, and the sharpness of the drop off.

The static representation of the blur trail is simply a simulated visual depiction of what the blur trail looks like. The observer would have to switch from fixation to smooth pursuit repeatedly, using visual short term memory, and adjust the static trail until the actual blur trail (perceived during smooth pursuit) matches the static trail (perceived during fixation).

Will try to program it up tonight.

I believe this will only work for impulse driven displays. Sample and hold displays have a "pixel decay" of effectively 1/refresh inbetween frames, but may have a much smaller decay time at the offset of a final frame. So if you have a white circle that is displayed for 10 frames, at 100 hz, and then turns off, the pixel "decay" during those 10 frames will be 0.01 seconds (10 milliseconds). But the time it takes for the circle to decay from the end of the last frame may be smaller than 10 ms. Blur trail will only measure the inbetween pixel decay.

For an impulse display, like a CRT or strobed backlight, the between-frame decay value will be the same as the offset-decay value. (I'm probably using horrible terminology here, I'm sure there are better terms).

Mark, am I understanding this correctly?
spacediver is offline  
post #55 of 76 Old 10-28-2013, 01:35 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by Elix View Post

Kudos to you, Mark, for moving the display science forward smile.gif I vote with both hands for the MPRT standard.
You got me very interested in Optoma DLP. Do you know what MPRT it has in 120 Hz + BFI mode? Also, any other models/brands of 3D DLP projectors compatible with that tweak?

DLP's, without black frame insertion, create square-wave persistence (on average), and generate MPRT's almost exactly equalling the refresh length.

With black frame insertion, the MPRT at least halves (for 50%:50% black frame duty cycle), so you have at maximum, one-quarter the motion blur of 60Hz, if you're playing native [email protected] sources (e.g. video games). Keep in mind that the "true 120Hz" capability of the Optoma's only work well for video games, since television and movie material do not natively output 120 frames per second during 120 Hertz refresh, without the use of interpolation. Black frame lengths varies from technology to technology, and display to display. Longer duty cycles (bigger black frame relative to visible frame) creates lower persistence, and less motion blur. People hoping to have less motion blur for movies with a 120Hz DLP, would need to use software such as SmoothVideo to convert 24fps to 120fps, for displaying on the DLP, to get the soap-opera effect. However, the purpose of Blur Busters is very video-game oriented.

Note that Optoma does not call it "black frame insertion", since it's the 3D mode. The black frame insertion in the Optoma was originally intended for 3D glasses use (a waiting time period for shutters to switch eyes) but eliminating 3D crosstalk on LCD shutter glasses, simultaneously also greatly reduces/eliminate motion blur, as a positive side effect. (Just as LightBoost 3D has now become more popular for motion blur elimination during LightBoost 2D, just by googling "lightboost").

Alas, it will take a long, long time to convince test pattern makers to include a "milliseconds of motion blur" scale than to use "lines of motion resolution", but I've successfully convinced a few people, one person at a time. During 2014, I know one other HDTV review website is planning to adopt this method of measuring motion blur (milliseconds), abandoning the "lines of motion resolution" method. I don't mind if it takes 10 years, as long as "lines of motion resolution" no longer increases in popularity, going forward, from now on.

Thanks for the kudos!

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #56 of 76 Old 10-28-2013, 02:11 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
lol mark, our posts were two minutes apart - this is just in case you missed it smile.gif
spacediver is offline  
post #57 of 76 Old 10-28-2013, 02:24 PM - Thread Starter
AVS Forum Special Member
 
Mark Rejhon's Avatar
 
Join Date: Feb 1999
Location: North America
Posts: 4,168
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 98 Post(s)
Liked: 275
Quote:
Originally Posted by spacediver View Post

I believe this will only work for impulse driven displays.
Not necessarily, it depends on the test you create.
Quote:
Sample and hold displays have a "pixel decay" of effectively 1/refresh inbetween frames, but may have a much smaller decay time at the offset of a final frame.
That's true, the hold and the transition can blend into each other, but the MPRT should average out to extremely close to a full refresh cycle, as long as transition cycle is less than half a refresh. MPRT's will only really noticeably begin creeping upwards as transitions start taking longer, especially longer than a refresh cycle.

But going forward, gradually, very few displays will have noticeable persistence effects created by transitions. From now on, most displays transitions will be a tiny fraction of a refresh. As we now know, the vast majority of motion blur on modern displays is caused by overall persistence (including hold) rather than the transition.

Accurate blur trail measurements include *both* the hold and the transition, even if the transition takes more than 1 refresh long. You just have to design the test to accomodate the additional contribution of motion blurring caused by transitions, since longer pixel transitions increases persistence of a display. That means if measuring using a photodiode or motion tests, you need to accomodate for this; (e.g. sample the resulting motion blur for more than one refresh cycle), to capture the full ghosting/motion blur trail length.
Quote:
For an impulse display, like a CRT or strobed backlight, the between-frame decay value will be the same as the offset-decay value. (I'm probably using horrible terminology here, I'm sure there are better terms).
Blur Busters has had the challenge of using the correct terminology sometimes, when converting the known scientific research to friendly explanations that are suitable for the major population. (Not too vastly different from converting "Journal of Physics" explanations into "Popular Science" explanations)
Quote:
Mark, am I understanding this correctly?
Mostly, I think -

Thanks,
Mark Rejhon
www.BlurBusters.com - Everything Better Than 60Hz(tm)
Mark Rejhon is offline  
post #58 of 76 Old 10-28-2013, 07:22 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
Ok almost finished coding - so far this works very well. It's astoundingly easy to match the observed blur trail with a static reference. You don't even need visual short term memory - you can clearly see the static trail simultaneously with the actual motion blur trail, since I've put them close together.

So far I'm testing on FW900 @ 120hz, 1280x800. I find shifting the mean of the gaussian function first, and then adjusting sigma to fine tune works best.

About to write some code to transform the result into a decay function.

So basically, once the observer has adjusted the static trail to match the observed motion trail, the software then looks at the luminance of the static trail and, based on the speed of the moving object (in pixels/sec), graphs the inferred pixel decay function.

A good test will be to see whether I get the same final result independent of resolution, frame rate, and object speed.

Will provide screenshots to explain once I'm done.

edit: of course, this only approximates the actual decay function, as the actual function may not be gaussian.
spacediver is offline  
post #59 of 76 Old 10-28-2013, 09:13 PM
AVS Forum Special Member
 
spacediver's Avatar
 
Join Date: May 2013
Location: Toronto
Posts: 1,033
Mentioned: 4 Post(s)
Tagged: 0 Thread(s)
Quoted: 172 Post(s)
Liked: 86
all done - I noticed that the blur trail looked quite green. I ran tests using different colored objects, and found green objects left the longest trail.

I looked back at my last set of colorimetric measurements, and indeed, the green primary has about 2.5 x the luminance of red, and 7.5 times the luminance of the blue (this is due to the D65 white balance I'd calibrated to).

I'll report results hopefully by tomorrow, with graphs, screenshots, and an explanation.

I can say that it seems that a gaussian might not be optimal, at least for this CRT's decay function, but what I can say is that it definitely seems possible to address this psychophysically. The challenge is to develop the appropriate control interface for the observer, so that they can finely match the stationary trail with the motion blur trail.
spacediver is offline  
post #60 of 76 Old 10-28-2013, 10:43 PM
AVS Forum Special Member
 
Elix's Avatar
 
Join Date: Jun 2011
Location: Dungeon, Pillar of Eyes
Posts: 1,574
Mentioned: 6 Post(s)
Tagged: 0 Thread(s)
Quoted: 253 Post(s)
Liked: 106
Quote:
Originally Posted by Mark Rejhon View Post

one-quarter the motion blur of 60Hz, if you're playing native [email protected] sources (e.g. video games).
Err... I am in trouble converting it into ms of motion blur. Does it mean those optomas outperform all other projectors in terms of motion resolution (except CRTs)? What number of "lines of motion resolution" equivalent would that be?
Elix is online now  
Sponsored Links
Advertisement
 
Reply Display Calibration

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off