Why aren't DIY Full Array Locall Dimming mods a thing? + Loose instructions to do so.

[Anonmon]

I made this account and making this post today (if this should go elsewhere, please let me know / move it.) specifically because of both the previous LTT video on the ASUS PA32UCX, and the one posted today about the MSI Creator 17. I've followed along with what has been released with FALD displays and continued to think it's absolutely ridiculous that the marketing for such displays tout "~1152 dimming zones!" and the like, when, in the case of the ASUS PA32UCX when I tried running the numbers anyway, turns out to be a resolution of somewhere close to 45x26. For a 4K display. What?! With even the MSI Creator 17, the LTT review had a line with "2cm large dimming zone", which somehow is impressive? 2cm in this context is absolutely MASSIVE, which even later on in that video there was a brief line of "and there's bloom" near the end. And all of that's somehow supposed to be acceptable, when LEDs come in absolutely tiny form factors that can be packed relatively easily with full control, as it is with those arduino kits and similar.

So, after looking around for a solution to this, here's the parts list and a general rundown of how someone could make a display have FALD that blows the socks off of anything commercially available with relative ease. (Not listing things like power adapters or HDMI cables, as that'd inflate this list like crazy.)

• The monitor itself (duh)
  • It can be virtually any kind of size you'd like, even ultrawide or 4:3 if you really want, though anything not 16:9 will create even more work later, and a 4K monitor could make things easier for reasons that'll be explained later on.
  • The backlight is gonna be removed and entirely redundant for this, so if you can find one with a broken one, that would be even better.
• Flexible light pipe (a ton) (https://www.aliexpress.com/item/32859443535.html)
  • The exact size and lengths required will be dependent on what resolution you want your backlight, and how physically big it is.
  • There may be a need for a new diffuser, don't have it listed as I'd try to reuse the one the display already if it turned out to be necessary.
• HUB75 LED Panels (+ cables to connect them together) (https://www.aliexpress.com/item/32998038844.html)
  • How many you'll need will depend largely on the aspect ratio you're wanting, 4 of them for 16:9 and 5 for ultrawides looks to be about right for a normal monitor. Could go bigger with x16 / x20 for even higher resolution, but then the chunk of panels behind the display would be massive and incredibly unwieldy.
  • Also, I could not find monochromatic HUB75 panels for the life of me, so hypothetically, this also means a completely RGB backlight. Meaning even better colour performance of things were configured right, as long as funky colour science between LEDs and LCD crystals doesn't nip that in the bud.
• Raspberry Pi 3
• PiCapture HD1 (https://lintestsystems.com/products/picapture-hd1)
• Adafruit Matrix HAT for RPi (https://www.adafruit.com/product/3211)
• Some sort of hardware to mount everything together, maybe 3D prints to mount the display and the HUB75 in place with black foam board as the means to mount all the light pipe into place both infront of the HUB75 panels and behind the LCD display?

And as an optional extra for convenience with a 4K display:

• HDMI 1 to 2 splitter with scaler (Literally any from Amazon)

The general idea is:

1. Video comes out of the GPU over two video cables, one of which is HDMI, and clones of each other as though you're running two monitors duplicating the same output. One of which goes to the display as normal, and the other to the PiCaptureHD1. (Or one cable to the HDMI splitter that splits the video from there, though with convenience means less control from the PC it's connected to.)
2. The PiCaptureHD1 dumps what it captures to the Pi's HDMI out, which the Matrix HAT uses to drive the HUB75 array. The PiCapture uses the camera interface to send video to the Pi, and the Matrix HAT uses the GPIO pins to send video, so there shouldn't be any conflicts there, the HAT just needs to be stacked on top of the PiCapture.
3. The HUB75 array sits behind the display, after the latter of which is torn down to the bare PCBs with no backlight with the flexible light pipe affixed in front of every single pixel of the HUB75 LED array, every single one being piped to the equivalent spot behind the display. This is necessary because there's practically no way to have an off-the-shelf HUB75 array that is perfectly sized for whatever display you'd want for it.

And with everything configured such that it gets to one end to the other properly, hey presto, you'd have a display with FALD at a resolution of 128x256, or 128x320 in the case of a ultrawide. 32,768/40,960 dimming zones if you want to count them like commercial FALD displays do. Some notes about this hypothetical approach:

• Latency, I have no idea how bad this may be. The PiCapture+Matrix Hat+Pi combo together doesn't seem particularly bottlenecked, so I'd be confident that at worst it'd be decent, even if there may be the ever most brief flash between pure white and pure black.
• Framerate of the PiCapture is 60FPS, so if this is being paired with a display that runs faster than that, technically it won't be 1:1, but given the distances of something on screen between two frames at 120fps, and the increase in screen area between a single FALD pixel and the collection of display pixels, I don't think it'll be too much of a worry.
• It will be chunky. No way around it. I've never cared for thin devices and still regularly have CRTs as active displays so that's a complete non issue personally, but I know some people care about that sort of thing with their active use displays, so do keep that in mind.
• There will be some funky scaling needed in the pipeline between the GPU outputting video, the Pi, and the LED array for any configuration, as 128x256 isn't perfectly 16:9. (OBS may be a good way to customize squish/stretch on the PC side, but figuring that out in Pi settings would make life way easier in the long run) Though especially the farther away from 16:9 you get given the fact that the PiCapture only captures at 16:9 1080p, which is the hard limit for resolution the backlight can be, if panels suddenly became available. So there will be some non-square pixel action going on to make use of all the backlight pixels.

All in, I'd estimate the bill of parts to be somewhere around $350, which if you happened to have a panel on hand or got a busted one with no working backlight, you can practically guarantee it'll be cheaper than what a off the shelf FALD display would be.

The reason I'm posting all of the above is because I'd like to see this happen, even if it's not by my own hands. Because I want to see a FALD that actually has a decent resolution be a thing, before MicroLED makes all of this entirely redundant.

Thank you for reading my short novel, Anonmon.

[Senzelian]

5 minutes ago, Anonmon said:

which somehow is impressive

Considering we can print organic LEDs, none of the current LCD technologies are impressive.

 

In the case of the MSI laptop and the ASUS display, the excitement is coming from being actually able to buy those products.

Before no one even bothered with it.

[Anonmon]

From the beginning it seemed a little disingenuous to mask double digit resolutions for backlights with "Over 1000 dimming zones!". Sure it's not a thing that was in commercial products before now, but driving a ton of LED's is relative child's play in the grand scheme of electronic things, given video walls, of which the above hypothetical project makes use of, have been a thing for years, and driving them with simple microcontrollers is hardly the most mind bending thing either.

It's not even the lack of commercial products that made me scratch my head more, it was the fact that no one seems to have tried DIYing FALD onto a display, at pitiful resolutions or otherwise. Beam splitter stereoscopic displays were / are a thing people do all the time, and making all sorts of hardware mods is the bread and butter for many, but no DIY backlights?

[mariushm]

The leds in those 64x128 / whatever panels can be shitty cheap leds, with bad color reproduction (low CRI), variations in quality (some more bright, some less, different color temperatures, different binning).

They're also relatively low current, low power, and some of those panels will dynamically adjust the brightness depending on how many leds are on, in order to keep current below some threshold.

 

You can't just use any led for backlight, if you want to have good color reproduction, good gamma etc.

 

Some of those panels also work by multplexing rows and lines, basically at any point in time only a limited number of leds are actually turned on. They just get turned on and off so fast you don't see it.

That can cause issues with the way image is updated on the screen

 

You'll have  a hard time refreshing the light information 60 times a second ( every frame sent to monitor) ... ex user moves mouse and your backlight will have latency tracking the pointer turning pixels on and off. There's latency when "talking" to each of those panels, the connection is not super fast speed...

 

Everything sounds easy when you look at the surface of things.

[MadPistol]

With OLED 4k 120hz TVs already on the market (and mature) and new technologies on the horizon that will offer even better performance, it's fair to say that FALD LCD is already outdated.

[Anonmon]

Latency was the biggest thing I was somewhat worried about for such an endeavor, with how ambiguous such a thing would be between all the hardware involved. A way to get a HDMI signal into a DIY LED array would be ideal, so the exacts of what LEDs you're using and such would be tighter controlled, but I don't know how you'd go about doing that, and sitting there soldering over 10's of 1000's of LEDs doesn't sound like a fun time.

Brightness, the panels claim ≥1500-2000 nits, so dimmer by the time it goes through the LCD sheet probably, but by how much would require testing.

Quote

Some of those panels also work by multplexing rows and lines, basically at any point in time only a limited number of leds are actually turned on.

That's, in a certain sense exactly how CRT's work, they only have a single point of a beam on the surface of the screen at a given time, with a little bit of a residual glow at most for anything solid on screen in a given moment. It's fine, because it scans across fast enough that you don't see it. And the panel claims  ≥1920Hz refresh rate, so at least on that end of things (assuming refresh rate actually means refresh rate as it's normally defined) that should be perfectly fine.

 

It's a DIY solution with all the jank and inelegance that typically entails, I don't think there was ever any expectation it wasn't going to have any caveats.

 

Quote

it's fair to say that FALD LCD is already outdated.

If you're only looking at the absolute bleeding edge probably. But people wouldn't be making and selling high end FALD displays right now if there was zero merit to producing and selling them.