Blur Busters Making Of: Strobe Engineering Behind The Scenes

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:

Blur Busters Making Of: Strobe Engineering Behind The Scenes

Post by Chief Blur Buster » 18 Jun 2021, 18:00

Crossposted one of my status updates from the EVE Spectrum Community Forum:

_____________

Behind The Scenes Technicals About Strobe Engineering

<Technical>
The common situation for many monitor brand names you love, is that they often work with a supplier that helps with the firmware/scaler/TCON. (such as MSTAR, Skyworth, TPV, Lehui, Samsung, etc). Eve fits this standard Development Life Cycle chain of a monitor vendor, as they are using Lehui

In the panel programing chain, there may be multiple suppliers involved (e.g. MSTAR working with a AUO panel) or the same supplier (e.g. Samsung working with a Samsung panel). In this case, Suzhou Lehui (Eve's supplier) works very closely to LG, the manufacturer of the Eve panel.

Now, the fun begins for strobing.

For many scaler vendors, they don't even (yet) know what they are doing until I've given them exact instructions, and then they are amazed crosstalk dropped 90%. A lot doing simple DELL 60Hz panels, blindly told to add special algorithms to high-Hz backlights that are more precise than 10 microseconds (10 microseconds actually creates human-visible flicker, since 1ms MPRT varying versus 1.01ms MPRT is a 1% brightness difference!).

Overdrive Gain is simply huge numbers of overdrive settings -- sometimes as many as 256 levels -- but may be 32 level, 64 level, 100 level, or 128 level, depending on panel vendor. The LG scaler/TCON standardizes on 64 levels of overdrive internally, even though only 2 or 3 levels are usually exposed in the menus. Mind you, Blur Busters always admonishes making all levels accessible to user, because of temperature reasons and panel lottery reasons, sometimes a 2/64th change makes a big difference in a cold winter room versus hot summer room.

For those who usually don't care about milliseconds, but interested in some easy Popular Science explanations, read the Milliseconds Matters Thread in the Blur Busters Forums. It's shocking how many specific kinds of sub-milliseconds can become human visible.

The complexity is higher for VRR strobing than non-VRR strobing. There's lots of exceedingly exact scientific instructions involved in VRR strobing. This is what Blur Busters is increasingly cable of doing; we are able to successfully communicate instructions for multiple generic high quality strobe algorithms at better-than-NVIDIA-ULMB-quality, for extremely low cost. That's why EVE hired Blur Busters to help them with this work -- because we can do it at only a very tiny fraction of the monitor's development budget.

Now if a hardware limitation is discovered, it has to be discovered in early prototypes, in the EVT, DVT, PVT prototype sequence.
  • EVT aka Engineering Validation Test (prototype)
  • DVT aka Design Validation Test (prototype)
  • PVT aka Production Validation Test (mass production unit)
So if a firmware bug preventing a specific strobe feature (e.g. lack of 100% strobe phase range capability), it ideally has to be discovered in EVT or early DVT, then the electronics circuit board fixed to prevent a strobe backlight tuning limitation. Thankfully, EVE successfully fixed all hardware limitations before PVT. The only thing I couldn't fix was the KSF phosphor built into LG NanoIPS panels, but it's at least tunable to be better than plasma displays (less phosphor trails than plasma).

Occasionally, esoteric strobe tuning stuff such as 256x256 Overdrive Lookup Tables replaces the old 17x17 Overdrive Lookup Table, or there's a Y-Axis Overdrive Gain Formula (overdrive increases along the vertical axis of the display, to speed up LCD GtG near bottom edge).

Mandatory Strobe Engineering by Blur Busters At Blur Busters, the standard "*minimum-needed-to-beat-NVIDIA-ULMB*" formula Blur Busters commonly works for strobe tuning mainly focuses on these adjustments:
  • Strobe Pulse Phase (full 100% adjustability of PWM relative to VBI)
  • Strobe Pulse Width (1% to 25% minimum)
  • Overdrive Gain (as many clicks as the scaler/TCON lets me have)
  • Strobe-Any-Hz Support (no fixed strobe presets are allowed at Blur Busters, we usually refuse to work with vendors that only provide fixed strobe refresh rates)
Eve successfully met all the Blur Busters Mandatory Minimum strobe-programmability specifications, and the shipping firmware includes all of this Blur Busters mandated minimum capability.

Optional Additional Strobe Engineering Enhancements Now, further possible optional additional strobe-improving algorithms that can widen the crosstalk-free strobe zone taller:

* 256x256 overdrive lookup tables instead of common 17x17 overdrive lookup tables.
Right now, we usually work with existing 17x17 OD LUTs. This can lead to certain colors that are suboptimally overdriven for panels with very odd GtG heatmaps at finer granularity than 17x17 OD LUT can solve. Fortunately Fast IPS usually is forgiving of 17x17 OD LUTs, while VA panels more badly need 256x256 OD LUTs for strobe-mode.
* Y-Axis overdrive gain formula (faster GtG for bottom edge of panel)
This is because not all pixels refresh at the same time, and the bottom edge refreshes late, meaning less time to finish LCD GtG before the global strobe backlight flash
* Refresh Rate Headroom / Large Vertical Totals / Internal scan conversion to create large multi-millisecond blanking intervals between refresh cycles to hide LCD GtG
The world's first zero-crosstalk IPS LCDs panels successfully achieve it with an approximately ~10ms VBI. A 1/240sec scanout on a 72Hz panel creates a blanking interval of 9.7 milliseconds between refresh cycles -- sufficient enough to hide most of real-world GtG of many LCDs in the total darkness cycle between strobe backlight flashes.

Most monitors use outdated 17x17 OD LUTs because papers found they were "good enough" for a typical 60Hz panel, coming from this Outdated Year 2012 Paper. But we're in the 360Hz era with strobe backlights. Blur Busters intuitively know things such as how 17x17 OD LUTs (fine for non-strobe) sometimes to worsen backlight strobing due to a human-visibility-amplifying interaction researchers did not test for, that only happens in strobe-backlight mode, on panels with sharp-cornered GtG heatmaps (especially VA panels at specific temperatures: a real-world 15C bedroom in middle of winter).

Most factories just QA test panels at 20C, but real world rooms are colder/hotter -- like artic or tropics. It's part of why we're huge advocates of exposing the entire OD GAIN range to the end-user onscreen menu (at least as an optional "User Defined" Overdrive option). Very major quality problems of 17x17 begins to appear as the refresh rates and strobing a pushed concurrently, simultaneously with real-world temperatures. Just to save a few bytes of scaler RAM, versus a 256x256 OD LUT (64 kilobytes). This ain't year 2012 in Kansas, toto.

These generic cheap strobe-improving techniques are now fairly widespread for many years. It made possible the world's first zero-crosstalk IPS LCD panels (at least for non-KSF panels) when using sufficient refresh rate headroom to hide LCD GtG pixel response completely away from human eyes in humongous VBIs between refresh cycles (Such as 1/240sec scanouts strobed at 60Hz-to-120Hz).

Much of our generic strobe tuning work now out-tunes many patented strobe algorithms. A lot of vendors reverse-engineered techniques that many 3D monitors and 3DTVs used (e.g. Y-Axis Overdrive Formula for increased OD Gain near bottom edge of panel), and Blur Busters have followed suit with new inexpensive tuning methods that formerly cost a lot to do.

Especially with higher-Hz, clearer motion, less motion blur, lifting the fog that formerly hid tiny millisecond-dependant image artifacts. The scientists that say "Humans can't tell apart events 10ms apart" are measuring only 1 item out of 100+ possible effects. Blur Busters, instead casts that massive mile-wide scientific net (which you've seen of our research writings at www.blurbusters.com/area51 already) and know which of those specific milliseconds actually is noticeable to Grandma that can't tell apart DVD-vs-HDTV. We filter the FUD and target the important milliseconds.

The confidence level, currently is pretty high, but I wanted to publish the usual Blur Busters caveats, as Blur Busters is always cautious about predicting capabilities. As most readers of Blur Busters already know, Blur Busters is always carefully more conservative about talking about features until they hare successfully developed. We are not exaggerating when exceedingly exact strobe algorithms can get lost in communication difficulties (e.g. English-Mandarian), and often I am hired to fly out over the Pacific Ocean to teach a class room at a display vendor [PHOTO].

A lot of overseas engineers are very new to motion blur reduction physics, or the more complex variable refresh rate strobing physics.

From end, I can confirm that the strobe-tunable firmware (both the easy preprogrammed tuning & the advanced Strobe Utility tuning capability) made it to the mass production (MP) model.
</Technical>

<Shameless Plug>
Blur Busters often is ahead of outdated scientific papers (Google Scholar on 17x17 Overdrive LUTs), which is why mere Blur Busters Area 51 Articles has now become cited by more than 20 peer reveiwed conference papers -- a lot is far ahead of industry, thanks to our ability to simplify general-purpose display motion blur knowledge. The flagship Blur Busters Area 51 articles are literally elevated to textbook reading status by the display industry.

Blur Busters one of the few companies in the world who can assist firmware vendors in successfully adding major strobe features enhancements post-shipment. We are one of the world's best debuggers of strobe bugs, and most firmware vendors save a lot of money hiring Blur Busters services, even as a third-party strobe quality validation laboratory. In many cases, we have successfully prevented a few product recalls with post-shipment firmware fixes, and even fixing fatal frameskipping bugs (240Hz pluague) for multiple vendors before shipment, etc.
</Shameless Plug>

Hope this was useful technical reading about behind-the-scenes of strobe quality improvement![/quote]

_____________

Prospective display vendors wanting to contact Blur Busters for services, contact Blur Busters Services via your corporate email address to obtain glossy PDFs of the services we offer. We normally do not send PDFs to gmail addresses. We work with multiple AR, VR, TV, laptop, mobile, monitor vendors.
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