TestUFO Suggestions and Ideas

NEW for 2017: Discussion about the testufo.com Blur Busters Motion Tests. Widely used by enthusiasts, display tweakers, YouTubers reviewers, monitor manufacturers and VR headset makers!
User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

TestUFO Suggestions and Ideas

Post by Discorz » 26 Nov 2021, 16:24

Chief Blur Buster wrote:
10 Nov 2021, 12:01
Discorz, I'd love you to create a "Discorz' TestUFO Wish List" thread and aggregate all your ideas into that. I bet you have some fantastic ideas already.
Request fulfilled! :)

Suggestions are not perfect or fully thought through, but it might spark some new ideas for you.

I'll update post for more or comment for new.


Ghosting / Pursuit Camera Test
  • Add more narrower Pixel Separation options, maybe 100 pixels min. I'm aware there will be blur overlapping with slow panels, but its good for fast ones. That is why it's optional.
  • New graphic preset 'UFO Compact' with decreased row height for average smartphone cameras that can't do true high resolutions. Came across this problem many times.

    UFO Classic - 360px high
    UFO Compact - 260px high prefer this one
    UFO Compact v2 - 290px high

    - more detail fit into frame, closeup pursuits, sharper, better focus
    - more UFOs in addition of narrower separation
    - reduces chances of user capturing only middle row ufo
    - possibly reduces image cropping, reviewers might like this (ST moved to top and bottom edges)

    Here is same camera setup before and after.
  • For MBR out of fullscreen mode Pattern Positioning should be automatically adjusted to aprox. middle screen for high resolution displays (if possible). User shouldn't take MBR shots outside full screen but I've seen people doing so. At 1080p its around screen center, 1440p top-mid, 4k up too high. This is related to number and size of browser bookmark bars and windows task bar.
    1440p centered
  • Had this idea of 'Status Track', which won't work well, but here it is anyways. It is supposed to hook test setting so that user don't have to state each time when publishing shots. Needs to be in 'BFI mode' to reduce sample-and-hold blur, still looks bad at low refresh rates. Would also require hooking model name, overdrive, adaptive-sync or other settings... which is impossible at this point.

Strobe Crosstalk Test
  • Add some type of cycle ruler for better vertical screen location coping
    - option to toggle OFF/ON
  • Add Sync Track to full screen mode
    - Requires much larger ST for whole screen capturing from further distances, something like in 'Moving Photo' test. As resolution increases it gets smaller. Replace top and bottom ufo rows for ST. One problem with this is leaving areas where ST would be uncovered. One way to get around this issue is with panning Vertical Sync Track. For crosstalk screen width don't matter anyways (landscape pivot). Can add multiple of them. Not sure how effective would this be.
    - option to toggle OFF/ON
  • New Pattern 'Line 1 Pixel, 5 Screen Divisions' (or more)
    This is meant for finer crosstalk pursuits. Great for debunking duty cycles with strobing and vrr strobing. With few tricks even strobe duration can be measured from photos.
    - classic PCST could be used here.
    - default colors: line RGB 255, background RGB 63, or leave cyan and red default but exchange line color for something with longer curve for all panel types.

Video Game Motion Test (and VRR Simulation, Moving Photo)
  • Add new photos of popular fps games like csgo, rss, ow, quake
  • Option to toggle static crosshair for stroboscopic effect

Frame Skipping Test
  • Add refresh rate in full screen mode. Maybe even hook display parameters

Overall
  • Add new test somewhere (maybe in Ghosting Test) with multiple colored thick lines and backgrounds in same frame so camera could capture more of the transitions at once, the more the better. Problem with this is image can get huge. Singling out some of the more demanding ones can shrink it.
    cyan, gray - 260px high.
    gray transitions only
    - optional line thickness
  • Since TestUFO doesn't state actual moving speed it requires:
    - addition of approximate (∼) symbol to each speed preset in dropdown list
    - statement of real pps speed instead of preset
  • Add quick pause/play button next to Speed selection box that pauses/continues panning. This is to reduce nausea from watching repeating patterns for too long and reduce unnecessary screen retention from PCST. Selecting pause from selection box is slow. I found my self opening another tab as a solution to this.
    - add to all tests
  • Add quick properties in full screen mode when mouse hover screen edge. Maybe even add option to pin it. Currently changing any setting requires exiting full screen mode and entering again.
  • Exiting fullscreen mode with esc key doesn't work properly. Esc only partially exits, required another mouse click to fully exit.

I have one if not my favorite idea, but it requires VRR support... I encourage to make VRR finally work with TestUFO.
Last edited by Discorz on 28 Nov 2021, 09:53, edited 2 times in total.
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: TestUFO Suggestions and Ideas

Post by Chief Blur Buster » 26 Nov 2021, 16:39

Bravo! Fantastique!

Masterpiece of suggestions.

I'll try to get most (or all) of them added during next year's UFO software development life cycle (2022).
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Re: TestUFO Suggestions and Ideas

Post by Discorz » 26 Nov 2021, 18:50

Great!

Winter 2021 updates postponed?
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: TestUFO Suggestions and Ideas

Post by Chief Blur Buster » 27 Nov 2021, 13:43

Discorz wrote:
26 Nov 2021, 18:50
Great!

Winter 2021 updates postponed?
No. It's only the first cycle. There's at least two or three iterative TestUFO development lifecycles planned in 2022. It's just too many features to squeeze into the first cycle, but will try to do my best!
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Re: TestUFO Suggestions and Ideas

Post by Discorz » 09 Dec 2022, 04:01

I made a custom pattern that can be put into Smooth Frog and it allows to directly read out display persistence. By observing a moving single pixel wide vertical lines with a certain mutual distance we can find out what is our display's MPRT. I call it Persistence Ruler.

Its intended to be used for impulsed modes (strobing/bfi) but it also works for sample-and-hold displays. As long as GtG is not too extreme and crosstalk is minimal test will do good.

Since we are limited by physical pixels we are only able to get approximate persistence number. Margin of error for moving speed of 4000px/sec is 0.13ms, for 2000px/sec 0.25ms and for 1000px/sec 0.5ms.


How to use Persistence Ruler

To add pattern into Smooth Frog images must be in .png format, named "User1" or "User2", larger than 256x256 and smaller than 8192x8192 and placed in the "SmoothFrog0.94b\WindowsNoEditor\SmoothFrog\Content\UserPatterns\" subdirectory. Please read ReadMe for full instructions about the program. Stay in Pixels/Frame panning mode and keep vsync enabled for consistency reasons.

Embedded numbers on pattern represent persistence (MPRT) in milliseconds and for them to match reality moving speed must be closest possible to 2000 px/sec. For example at 60Hz/fps 33px/fr results in 1980px/sec and 34 px/fr in 2040px/sec, choose closer one. After every frame/refresh rate change you must manually readjust px/frame to achieve ~2000px/sec.
2000 is good speed for down to 1ms MPRT readouts, but anything bellow that requires much larger speeds and horizontal resolutions, otherwise it's very hard to eye track on normal 1080p screen. Perhaps a more accurate solution is possible but this works well for a quick check.

Now eye track the moving pattern and look for following; your display's approximate persistence will be shown as perfectly joined lines/bands with no horizontal overlapping or separation. Displays prone to pixel inversion may behave different at odd number pixels per frame speeds.

Simulated example: Pattern moving across impulsed OLED screen, eye tracked
explanation.png
explanation.png (12.93 KiB) Viewed 4860 times

Patterns

Ruler 1 - for 2000px/sec
Image

Ruler 2 - for 4000px/sec, extra pattern for detailed sub 2ms MPRT readings, hard to eye track on screens smaller than 4k
Image

Ruler 3 - universal
Image
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: TestUFO Suggestions and Ideas

Post by Chief Blur Buster » 11 Dec 2022, 17:16

Discorz wrote:
09 Dec 2022, 04:01
I made a custom pattern that can be put into Smooth Frog and it allows to directly read out display persistence. By observing a moving single pixel wide vertical lines with a certain mutual distance we can find out what is our display's MPRT. I call it Persistence Ruler.
Fantastic suggestion -- it certainly works well on OLEDs, and some fast-GtG LCDs. It's harder when GtG is so muddy that it's hard to connect the edges, but this is easy when GtG is a tiny fraction of a refresh cycle.

I have already come up with something roughly similar that I have been wanting to add to TestUFO, but your design is excellent. It works very well on OLEDs and fast-GtG displays.

I have something roughly similar (old) at www.testufo.com/chase and it can easily be modified to behave similarly to Persistence Ruler, perhaps it will be a merged test (www.testufo.com/chase and www.testufo.com/mprt). I intentionally kept the top/bottom offsetted to allow easier monitoring of how the blur exits the gap for either side, since GtG are often ugly/messy curves.

The future "merged" www.testufo.com/chase and www.testufo.com/mprt behaves very similarly to Persistence Ruler (with some key differences).

Originally it would only have a gap adjuster. But your inspiration made me think I should multiple blocks too concurrently, for quicker identification, and the gap would auto-compensate depending on the pixel motion speed, and display MPRT numbers on screen.

With math, I can make it variable speed and still calculate the correct MPRT. I need to replace the old (difficult) MPRT pattern at www.testufo.com/mprt and I like the name "Persistence Ruler", actually.

Also, having thicker left/right white lines makes it easier, as two moving edges that can be made to overlap, e.g. 2-pixel thick lines. It's why I used 2 pixel thick lines at www.testufo.com/persistence because 1-pixel-thick lines were too hard to see on many LCDs. The gap between the two is the important part of measuring MPRT, extra line thickness (in the same direction) simply makes the lines easier to see.

--> So I would also add a thickness adjuster to my MPRT-modified version of www.testufo.com/chase

On many displays that have fast-to-black GtG, but slow black-to-white GtG, this creates a new problem for many legacy LCDs. Because of asymmetrical pixel response on LCD, you also need two symmetric images -- white-to-black and black-to-white. You may read 3.5 for white-to-black and 4.5 for black-to-white, and need to average the two MPRT's to cancel out the GtG-direction distortion, resulting in the correct number 4.

--> So my tests may have two variable speeds (top half + and bottom inversed-colors half running at different speeds) with an automatically calculated average.

But for a static PNG designed at 2000pps, this is a fantastic simple slide rule for measuring MPRT with human eyes -- and very reliable on OLED since they can suddenly go bright-white and full black, meaning 1-pixel thick and symmetric pixel response, means the Persistence Ruler is reliable on OLED (both BFI and non-BFI).

But will need software-logic to make it more reliable on LCDs (line thickness adjuster, two different speeds for normal/inverse versions). You can simply give people two PNGs (one normal, one inverse color) and say that LCDs require you to run both images, and average the two numbers. This is not needed on OLED because of their symmetric pixel response universally a tiny fraction of a refresh cycle for all color combinations (and their reverse).

Your talk here made me think though, on how to make it even more user friendly, for the rewritten TestUFO version of MPRT tester -- basically a major-modified version of www.testufo.com/chase that is optimized to behave similar to Persistence Ruler. I'll have to mention you as inspirations for some idea-refinements to something I was already doing. How may I credit you for some inspirations? For example I credit shurCool at www.testufo.com/frameskipping
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Re: TestUFO Suggestions and Ideas

Post by Discorz » 12 Dec 2022, 03:27

Chief Blur Buster wrote:
11 Dec 2022, 17:16
Fantastic suggestion -- it certainly works well on OLEDs, and some fast-GtG LCDs. It's harder when GtG is so muddy that it's hard to connect the edges, but this is easy when GtG is a tiny fraction of a refresh cycle.

I have already come up with something roughly similar that I have been wanting to add to TestUFO, but your design is excellent. It works very well on OLEDs and fast-GtG displays.
It turned out good. Slow GtG will need extra attention to notice joined lines. What I found most distracting was full screen strobe crosstalk on LCDs. There are multiple lines all over the screen which can easily confuse user. So it would be nice to see adjustable vertical position so user can manually move the pattern to clearest phaze position. I'm also not sure how to treat PWM fills in today's vrr+strobing modes. It's a mixed mess with those.

2-pixel thick lines may work too but I avoided it since lines no longer looked uniformly joined with correct MPRT readouts. It looked like correct answer moved to one place beside, therefore might be misleading.
Chief Blur Buster wrote:
11 Dec 2022, 17:16
I'll have to mention you as inspirations for some idea-refinements to something I was already doing. How may I credit you for some inspirations?
Feel free to use this idea. I can't think of better place for it than testufo. Didn't expect a credit but ok. I guess Discorz is fine, or however you prefer it.
Last edited by Discorz on 13 Dec 2022, 06:42, edited 1 time in total.
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: TestUFO Suggestions and Ideas

Post by Chief Blur Buster » 12 Dec 2022, 05:30

Discorz wrote:
12 Dec 2022, 03:27
2-pixel thick lines may work too but I avoided it since lines no longer looked uniformly joined with correct MPRT readouts. It looked like correct answer moved to one place beside, therefore might be misleading.
You have to compensate mathematically, since the blur amplitude is twice, you need to subtract 1 pixel from the gap.

That's why www.testufo.com/persistence still touches each others' lines no matter what "Line Thickness" you choose! Despite spacing of left edges staying constant (e.g. every 8 pixels) and line thickness varying outwards. Adjust line thickness and observe that the lines still exactly touch each other in that specific test (when testing at full persistence = MPRT is refreshtime).

The formula is [line thickness of one line] + [gap thickness] needs to remain constant, for the same MPRT blur number.

If you thicken the lines, you need to reduce the gap proportionate to one line, for the same MPRT number.

This is intuitive to me, but may be mathematically confusing to others;

So the gap gets 1 pixel smaller for every 1 pixel thicker line, to maintain the same persistence calculation line-touch of a specific persistence. Motion blur linearly blurs in one direction, so you only need to count for one line, despite both lines being thicker. So you only need to account for the blur-inwards line (into the gap), not the blur-outwards line (beyond the gap).

Or another way to view it / calculate it mathematically (if it's easier for your brain to understand), 1 pixel of extra blur has a visual equivalent of half pixel extra blur on leading edge and half pixel extra blur on trailing edge. It's just a matter of how you're putting the mathematical zero-origin of the blur (whether you're blurring full in one direction, or blurring half in both directions -- same amount of linear motion blur blur at the end of the day).

Thicker lines, admittedly, also raises the minimum MPRT you can measure too, so a line thickness 50% of the gap, may limit your ability of measuring persistence down to half refreshtime, which is limiting. So thicker lines is easier to see, but limits your minimum MPRT because for every 1 pixel thicker, you need to move the lines 1 pixel closer together, for the same MPRT persistence number.

You can tell this is correct by experimenting with line thickness at www.testufo.com/persistence which touches perfectly whenever MPRT=refreshtime, regardless of whatever line thickness you decide to use in that specific test. You can tell that for every 1 pixel thicker, the gap is 1 pixel thinner, to maintain the same MPRT number. This is because [thickness of one line] + [thickness of gap] is always the pixels per frame whenever MPRT=refreshtime.

That's the formula to base off, and I got the correct MPRT's for 50%:50% strobing, by appropriately doubling the speed (while maintaining existing thickness & gap) or halving the total [thickness of one line] + [thickness of gap] (while maintaining the same speed).

You get the most precision out of 1-pixel-thick-lines, but at current tech, especially LCD, 2-pixel-thick and 3-pixel-thick is much easier especially when testing both directions (white on black, and black on white) to get LCD-accurate MPRTs, especially for thin black lines on white background, which often is hard to see.

That being said, it ceases to be an issue when it comes to OLEDs because of their symmetric near-zero pixel response speed in both directions, and you don't have to test inverse and average the MPRTs. (Asymmetric pixel response can increase/decrease MPRTs for certain color pairs -- because a slow GtG-in and a fast GtG-out can artificially shorten MPRTs, and a fast GtG-in and slow GtG-out can artificially lengthen MPRTs. But if you average all the color combinations, it all averages out to refreshtime MPRT100% for sample and hold display)
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

User avatar
Discorz
VIP Member
Posts: 999
Joined: 06 Sep 2019, 02:39
Location: Europe, Croatia
Contact:

Re: TestUFO Suggestions and Ideas

Post by Discorz » 12 Dec 2022, 06:57

Discorz wrote:
12 Dec 2022, 06:57
So you only need to account for the blur-inwards line (into the gap), not the blur-outwards line (beyond the gap).

So thicker lines is easier to see, but limits your minimum MPRT because for every 1 pixel thicker, you need to move the lines 1 pixel closer together, for the same MPRT persistence number.
I see what you mean. I experimented and confirmed. Thicker lines will be easier to see/eyetrack on screens with tight pixel density. And 1-pixel lines are more reliable for measuring very fast persistence (many today's LCDs go bellow 1ms).

I'm wondering if playing around with colored red-green-blue lines on black background can get us sub-pixel accuracy. Probably not?
Compare UFOs | Do you use Blur Reduction? | Smooth Frog | Latency Split Test
Alienware AW2521H, Gigabyte M32Q, Asus VG279QM, Alienware AW2518HF, AOC C24G1, AOC G2790PX, Setup

User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: TestUFO Suggestions and Ideas

Post by Chief Blur Buster » 12 Dec 2022, 16:07

Chief Blur Buster wrote:
12 Dec 2022, 05:30
I'm wondering if playing around with colored red-green-blue lines on black background can get us sub-pixel accuracy. Probably not?
If you know the exact pixel structure of the LCD, such as BGR and RGB, then yes.
You can definitely apply the ClearType-style algorithms to subpixels.
For white lines, you gain 3x horizontal resolution.

But remember those Samsung OD-OLEDS use a triangle pixel structure and this technique won't work. Even when using Windows to attempt to detect displays' current pixel structure,

Unfortunately, you cannot guarantee what pixel structure the user has, and whether or not they zoomed their OS DPI or web browser (which scales the PNG to non 1:1). So you may get very erratic results from users who don't exactly know how to confirm, before using subpixels.

Instead, one can derive off my idea of temporally varying line distance. For example, 6 pixel separation one refresh cycle, 5 pixel separation next refresh cycle, and so on. This would work fine for high refresh rates where the jitter is beyond your flicker fusion threshold, e.g. for measuring theoretical future 240Hz OLEDs that contains sub-refresh BFI. This "high frequency 1-pixel jitter blending to blur like a fast-vibrating guitar string" infills half a pixel of blur into the gap.

Another option is to can add a gamma corrected vertical line instead at half nits brightness. Remember to gamma-correct, much like www.testufo.com/gtg-vs-mprt ... as RGB(127,127,127) is not half the nits brightness of RGB(254,254,254) ... You have to go all the way down to roughly RGB(88,88,88) to get half the brightness of full white in standard linear colorspace at gamma 2.2 formula ... Doing adjusted-brightness lines has an error margin LCDs due to the unbalanced GtG heatmap as mid greys are often slower GtG than full blacks and full whites, but works great on OLEDs due to the glassfloor-flat GtG heatmap. And LCD ghosting/overdrive effects (brightenings/dimmings) will fudge the number around a bit.

And a derivative of above -- you can also try a very slanted diagonal gamma-corrected-antialiased line, as a continuous stepless MPRT ruler line that would work well on OLEDs, as an analog MPRT ruler with no discrete MPRT steps! Very hard to correctly draw a line with correctly gamma-corrected antialiasing pixels, not even GPUs always do it correctly, classic antialiasing was often un-gamma-corrected! And you need to know what gamma the display is using (usually 2.2 but can be different, and user may have tweaked monitor settings or NVIDIA Control Panel settings) [The slanted edge my idea too, so please credit me if taking my idea for your app or whatever]

BTW, it's pretty neat that human eyes can estimate MPRT relatively reliably on OLEDs with these custom temporal test patterns even all the way down to ~0.25ms MPRT and below (strobed 4K display test at 4000 pixels/sec is still easy enough to eye-track) and the bottom floor of human-eye-measurable MPRT can fall below that if the display is very bright and the gap between lines are still visible.
Head of Blur Busters - BlurBusters.com | TestUFO.com | Follow @BlurBusters on Twitter

Image
Forum Rules wrote:  1. Rule #1: Be Nice. This is published forum rule #1. Even To Newbies & People You Disagree With!
  2. Please report rule violations If you see a post that violates forum rules, then report the post.
  3. ALWAYS respect indie testers here. See how indies are bootstrapping Blur Busters research!

Post Reply