Understanding Display Scan-Out Lag Via High Speed Video

Advanced display talk, display hackers, advanced game programmers, scientists, display researchers, display manufacturers, vision researchers & Advanced Display Articles on Blur Busters. The masters on Blur Busters.
Post Reply
User avatar
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 13 Jan 2019, 05:37

I've posted a new article with several 960fps high speed video of display scanout behaviours -- using the new TestUFO scan-out test at http://www.testufo.com/scanout

Article:
Understanding Display Scan-Out Lag With High Speed Video

High Speed Video Of 60Hz IPS Desktop Monitor

The screen refreshes top-to-bottom, with a fuzzy zone wiping down the screen.  This is the GtG fade zone, which appears to about one-quarter the height of the screen, which is what is expected for GtG pixel response approximately 1/4th of a refresh cycle.

phpBB [video]


High Speed Video Of 60Hz MacBook Laptop

The refreshing behaviour is very similar to a typical laptop, top-to-bottom, with a fuzzy zone representing the LCD pixel response.

phpBB [video]


High Speed Video Of 60Hz iPad Mini 4

Here's where this diverges interestingly; this iPad scans from top-to-bottom in portrait mode. When filmed in landscape mode, this becomes a sideways scan effect:

phpBB [video]


NEW High Speed Video Of 60Hz OLED Samsung Galaxy Tab 4

And for a really clean refresh cycle, here’s a high speed video of an OLED, provided by Edward on Twitter. The incredibly fast pixel response of an OLED means the fade zone is extremely tight in comparison to LCD:

phpBB [video]


For more, read this article.
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
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 20 Jan 2019, 16:07

In addition, a 960fps high speed video of an OLED has just gotten posted:
https://www.blurbusters.com/understandi ... eed-video/

As some of you already know, I've got peer reviewed work, such as pursuit camera for motion blur testing, and I've obtained/accessed/visited/seen prototype 480Hz and 1000Hz displays and experiments -- which I also write about here, the Refresh Rate Race is still worth it, we are still far away from "Retina Refresh Rates".

Manufacturers Should Also Begin Using Cheap High Speed Cameras To Test Monitors During Engineering

In the weblogs -- I've observed that some manufacturers/display engineers are visiting Blur Busters Forums! I know some of you aren't even using high speed camera yet....GO BUY ONE to help better engineer your displays. True-1000fps or True-960fps cameras only cost ~$300 nowadays in a cheap smartphone if you shop carefully for a supported camera. You don't need a $10,000 Phantom Flex if you can't afford one.

It's very surprising how a lot can be learned/analyzed about display refresh behaviour -- just merely by using inexpensive high speed videography.
  1. You can see if GtG is faster to brighter colors or faster to darker colors!
    That's very apparent by seeing a faster-looking fade to yellow and green slides (brighter colors than red and blue).
  2. Scan velocity can be calculated easily!
    You observe that it takes roughly 16 frames of high speed video for one blur edge to reach from top to bottom -- that's the 16.7ms of a 60Hz refresh cycle. Remember to time one edge of the fade zone (either the top edger or bottom edge of the blurry GtG zone)
  3. GtG consistency can be inferred!
    A messy/asymmetric fade between colors means very inconsistent GtG. You've got more input lag for certain colors than other colors!
  4. GtG speed can be estimated!
    By knowing the scan velocity (e.g. 16ms), a 4ms GtG means the band is about 1/4th screen height. (4ms/16ms). Your limiting factor is the high speed camera's resolution. 1000fps means 1ms accuracy in GtG estimates.
  5. Persistence (MPRT) as a milliseconds can be inferred!
    Persistence is pixel visibility time, either via global strobe or rolling scan time. Persistence is known as MPRT(100%). Count the number of frames that pixel of a refresh cycle stays visible for. 3 frames of a ~1000fps video means the pixel was visible for about 3ms, leading to an MPRT(100%) of approximately 3 milliseconds. Error margin is about 1 frame of high speed video (1000fps = 1ms error margin)

    IMPORTANT NOTE: Blur Busters only uses MPRT(100%) instead of the MPRT(10%->90%) formula because it is a more accurate match to human-perceived motion blur in today's fast-GtG displays, due to the display-vs-camera "refresh duration" versus "shutter" equivalency. When GtG is an insignificant percentage of a refresh cycle, the motion blur of MPRT(100%) at 120Hz provides a motion-blur equivalency to a camera-shutter photograph of 1/120sec shutter. For more explanations why Blur Busters does not use the outdated MPRT(10%->90%) formula, see Blur Busters Law.
  6. Improved ability strobe backlights (less crosstalk) can be inferred!
    A good panel is a panel that can have most of its GtG fade zone (fuzzy band) between refresh cycles. The band falls off the bottom edge of the screen sooner, and the band reappears off the top edge of the screen later. Less ghosting/overlapping of refresh cycles.
  7. Success of overdrive tuning can be estimated!
    Better overdrive tuning means a thinner, more symmetric bland of refresh cycles into each other. Film with overdrive ON/OFF and at various settings, and you will notice this.
  8. Confirm if panel scanout is in sync with cable scanout!
    In eSports, you need sub-refresh-cycle real-time sync between the cable delivery of pixels, to the panel. Most good eSports display do realtime scanout from the cable, using only rolling-line-buffered processing (rather than buffering a full refresh cycle first). If you, as a manufacturer, want your display to sell well in paid professional eSports, this is a display behaviour that you must confirm
  9. Success of large blanking intervals can be inferred!
    Manufacturers & advanced users typically need to do this to improve strobe-backlight quality (e.g. ULMB, LightBoost, ELMB, DyAc, etc). Sometimes this scan-velocity conversion is done by the panel/TCON, and sometimes it's done via a custom resolution tweak. A large blanking interval means a faster scan velocity with longer intervals between refresh cycles. The GtG fade zone should fade offscreen long before the GtG fade zone reappears on the top edge.
Most Manufacturers Are Not Yet Using High Speed Cameras For Display Research

Be the first, get ahead of your competition.

Sure, you can use photodiode oscilloscopes. You can still keep using them. But use your human eyes too to speed engineering up. In the Refresh Race Race towards future 1000Hz displays, it helps a lot. Scientifically testing with "seeing-based" technologies like cameras, can help rapidly and more cheaply improve displays, too. The maths become simpler, especially for engineers that need to "learn by seeing" in applied sciences, in addition to the abstract maths that they learn in University/College. And besides, high speed cameras are also fun work at a display engineering lab, too!

And for some manufacturers, it is possible to do 10 improvements (rapid firmware tweaks) to a display in one day, rather than 1 improvement per day (firmware tweaks), with rapid reviewing of high speed videos.

This fantastic tool -- made possible by brand new 1000fps smartphones -- now accessible by advanced users -- and display manufacturers -- makes this a revolutionary new tool for display analysis. Use the tool, manufacturers -- to improve your displays!

I know manufacturers read these forums (Even though their bosses may not allow them to reply to these posts).

You heard this public proclamation here first, in Area 51 of Blur Busters Forums!

I, Mark Rejhon, as Chief Blur Buster now recommend an additional tool in the display engineering toolbox: Use those newfound cheap 1000fps cameras as a new tool to rapidly iterate success of improvements to displays.
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
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 20 Jan 2019, 17:47

Inexpensive High-Def 1000fps Cameras

With a full 720p or 1080p resolution, instead of skimpy 64-pixel resolution.

There are now inexpensive high definition Super Slow Motion Cameras capable of 1000fps HD that works excellently for display scan behavior research with the new TestUFO Scanout Test.

Links to Amazon: Some of these only records 1000fps for only a fraction of a second, however this is plenty for refresh behavior analysis and debugging latency engineering problems in gaming monitors.

If you're a poor student writing a thesis that needs a high speed camera, the great news is you can grab an older banged-up smartphone on eBay, making sure the camera lens is pristine -- and gain a research-quality high-definition 1000fps camera for pennies on the dollar!
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
Chief Blur Buster
Site Admin
Posts: 11647
Joined: 05 Dec 2013, 15:44
Location: Toronto / Hamilton, Ontario, Canada
Contact:

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 22 Jan 2019, 01:38

Less accurate slo-mo smartphones but somewhat educational
- 240fps iPhones when filming 60Hz displays
- Huawei P20 smartphone "960fps"; it does something that reduces scientific accuracy of scanout recording. However, these are fine for filming 60Hz displays.
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!

Haldi
Posts: 9
Joined: 19 Jul 2016, 14:40

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Haldi » 26 Jan 2019, 05:49

Chief Blur Buster wrote:Less accurate slo-mo smartphones but somewhat educational
- Huawei P20 smartphone "960fps"; it does something that reduces scientific accuracy of scanout recording. However, these are fine for filming 60Hz displays.
The Samsung and Sony sensors have an on chip 256mb RAM which stores the 960fps FullHD to be encoded later on because the SD855 can't handle that fast encoding, nor can the bus from the camera to the SoC handle the bandwidth.
The Huawei and Xiaomi? Or others only use 120 or 240fps video recording and interpolate up to 960fps. So its not a true slow motion.

P.S tried it with my Sony XZ2 on a 144Hz Samsung C49HG90 but seems too fast :(

https://youtu.be/KneXApDY2UI


Sent from my GT-N8013 using Tapatalk

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

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 26 Jan 2019, 22:55

Haldi wrote:The Samsung and Sony sensors have an on chip 256mb RAM which stores the 960fps FullHD to be encoded later on because the SD855 can't handle that fast encoding, nor can the bus from the camera to the SoC handle the bandwidth.
The Huawei and Xiaomi? Or others only use 120 or 240fps video recording and interpolate up to 960fps. So its not a true slow motion.
Yeah, I have since found out that those use a Twixtor-style interpolation algorithm.
So they're not true 960fps phones (for those models).
Haldi wrote:P.S tried it with my Sony XZ2 on a 144Hz Samsung C49HG90 but seems too fast :(

https://youtu.be/KneXApDY2UI
Strobe backlight & PWM-dimming generates some camera sensor scan artifacts

I see backlight flicker interacting with camera sensor scan. Your monitor either has (A) has strobing enabled, or (B) has a PWM-dimming backlight. Disable strobing and/or set to 100% brightness, and then try again.

High speed video can be helpful for strobe, as long as some actions are taken to mitigate/minimize artifacts. This part of the famous high speed video of LightBoost that convinced many that LCDs could achieve CRT motion clarity in certain situations.

Camera Sensor Scanout Interference Artifacts

Did you know camera sensors have a scanout, just like monitors do? And did you know they can "interact" with each other? High speed video is useful for strobe backlights but is only granular (1ms increments) and can be extremely problematic when the black periods or bright period are are fractional-milliseconds; creating some camera-sensor-scan interactions. One trick is stepping far away from the monitor (to make monitor 1/4th screen height) so that camera sensor scanout is only 1/4000sec over the screen.

Also portrait camera with a landscape high-Hz (e.g. 240Hz) monitor can sometimes generate diagonal camera-sensor-scanout artifacts (due to the perpendicular scanout directions of monitor and screen).

YouTube single frame stepping

In addition, press Spacebar in YouTube and single step using period/comma . , keys.
That makes analysis easier in YouTube.
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!

vanden
Posts: 16
Joined: 08 Mar 2014, 12:42

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by vanden » 02 Feb 2019, 10:46

Here is an animated APNG that compares an IPS, CRT and OLED screen (same advancement in scanout ) :
Image
For the CRT the exposure time 1/500s (500fps) ...
With an exposure time 1/1000sec it's more like this :
Image
Attachments
IPS-CRT-OLED.png
IPS-CRT-OLED.png (1.61 MiB) Viewed 2838 times

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

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by Chief Blur Buster » 04 Feb 2019, 16:44

vanden wrote:Here is an animated APNG that compares an IPS, CRT and OLED screen (same advancement in scanout ) :
Image
Fantastic comparision, and very accurate too.
-> OLED has a very sharp GtG fadezone (0.1ms)
-> IPS has a very blurry GtG fadezone (5ms -- about 50x thicker)
-> CRT has a very brief pulse (immediate on, followed by phosphor decay)

It's a great comparision of the display technologies!

Certainly, camera exposure length (1/500sec or 1/1000sec) will produce an error margin for <1ms GtG pixel response, given camera sensor scanout. But it appears that this isn't a noticeable error margin in these stillframes. You've done an excellent job finding the specific color transition (Green->Red) at similar scanout positions.

With your permission, may I incorporate that APNG in one of the next Blur Busters posts about Scanout tests? I've been capturing a few lately.

So on that note, this produces a new interesting option to scanout analysis:

Alternative to High Speed Video Cameras!
-- Since www.testufo.com/scanout is a repetitive temporal test pattern, one could simply burst-shoot it multiple times! It is possible to visually sort 100 to 500 random burst-shoots of still photography of Scanout -- into one slow-motion video.
-- Random burst-shooting with a DSLR (or good mirrorless camera) at fast shutter speed, given sufficient samples, can be a suitable alternative to high speed video for this test. This is due to the repetitive nature of 4-refresh cycles, so you just keep shooting until you've got enough samples. Then you simply sort the photos by scanout order.
-- This will not be always perfect for large-blanking-interval research (i.e. Large Vertical Totals, or scan conversion of low-Hz to fast scan velocity), where the scanout disappears long before it reappears at top.
-- However, this is still perfectly fine for mid-scanout comparison research like this photo.
-- Simply finding similiar scanout positions (of specific color transitions) and aligning them like in this animation! Since it's a repetitive temporal test pattern, repeat-shot of still photography is a suitable alternative to a high speed camera![/b]
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!

vanden
Posts: 16
Joined: 08 Mar 2014, 12:42

Re: Understanding Display Scan-Out Lag Via High Speed Video

Post by vanden » 06 Feb 2019, 04:04

Chief Blur Buster wrote:With your permission, may I incorporate that APNG in one of the next Blur Busters posts about Scanout tests? I've been capturing a few lately.
Thank you, of course you can use the animation !

And actually it is possible to use a digital camera, with ideally a manual mode ... but even with one that does not have a complete manual mode (like mine) in the raffal mode the exposure times are quite short ( 1/500s to 1/1000s in my case).

Post Reply