DIY 'Infinity Contrast' Display (projector-LCD-backlight mod) development thread

I wonder if the alignment issue could be solved with a simple image skew. If you initially warp a white image so that it matches all corners of the display, then wouldn't this allow for less glow filtering and tighter lines?

Is it worth it to use a 4k projector instead of 1080 counterpart? It's For 4k lcd tv.

Based on Matt's video, probably not. What you're creating is a light map and the mapped area (even of 4K is is far larger than you might imagine, so that could most likely be "fixed" in OBS.

@nonamecola This is what i was thinking. Wish i still had my old HD DLP TV. Would be a perfect fit to this idea.

I must be a little electronically challenged because i seem to be missing something. So, i get using obs to alter the video going to the projector, and i get using a hdmi splitter to get the same signal to both devices. However, what are you running obs on? Does this mean you need to put a pc in between the hdmi splitter and the projector? So, one side of splitter goes to tv, second goes to pc running obs and then to projector? Do i have this correct?

I mentioned this in the YouTube video’s comments just now, but you can reduce the throw distance if you use a convex mirror. Have the projector just below the TV pointing backwards at a mirror that reflects the image back at the screen. This reduced throw distance should also let you create an enclosure that looks a lot more polished.

Hell, you can probably use an old security mirror to see if this works before investing in something nicer.

I wonder if this setup is possible since i would prefer not to use a PC to stream stuff, the convenience of an apple TV is insane

@ejg OBS can capture outputs such as video, often used for broadcasting/streaming or recording. There is a chance he is just duplicating his display for this for the captured video. Though, the way I am understanding it, he is only capturing the display being output to the tv on OBS then putting OBS in full screen on the projector with the tv video captured and luma key filter applied. The projector is just a display output with OBS on it in full screen.

Interesting speculation, but as long-time sub, what you see with Matt is what you get with Matt. Either that or he went to whole lot trouble to fake something that would cost more than just doing it,

can i use a old led tv and a led projector as replacement for the devices used?

Looks like it. The trick is separating the panel without damaging it.

hi was thinking if can use led projector of 20euro from aliexpress

Posted by: @manos89

↑

hi was thinking if can use led projector of 20euro from aliexpress

Hard to be 100% sure with Ali prices.  The actual brightness at the projected distance is crucial. It has to be bright enough to light the "lit" pixels so the size of the donor screen is an issue. A very large screen (like Matt used in his example) means the projected light has to travel further, and it falls off (brightness is reduced by the inverse square law).

20 euros isn't going to be hugely powerful but could quite possibly be enough to light a screen up to perhaps 21"-24". I suppose at that price, you could import one and sell it on eBay for more than you paid if it's too feeble so it's virtually risk-free.

hey will someone help me with making this without obs lets say i want on the C= controler  have obs so i don,t have to use 2 outputs and  i don,t want to open obs every time and stream ( sorry for bad english) btw thanks 

Greetings everyone! I created an account on a forum for the first time in ages due to this project. What a brilliant idea! Kudos to you! I have a few concepts for my potential initial build. One involves utilizing an open-source video board to handle the filters instead of relying on OBS. This would necessitate reimplementing the filter code, but I believe it could be entirely feasible and much faster when running on an FPGA, which could significantly reduce input latency. Another idea I had was using one of the early-generation 4k monitors circa 2015. I think that a Pico or LCOS projector might provide sufficient light to achieve this idea on a monitor size scale. 

Edit: I wasn't sure how we felt about double posting here, so I decided to add to my op. Here's a link to an FPGA video dev board list with quite a few options Here. The Arty z7 and Mimas A7 both seems like good options. I plan to continue my search as I believe a board with a VGA output would make more sense for use with an office projector. 

Welcome to DIY Perks. Great to have you.

Sounds like a good option - the only reservation I have with FPGA (Field Programmable Gate Array) is the complexity of programming these little beasties.

For readers unaware of FPGAs, these are pure logic chips - and don't run programs like a CPU/MCU like x86, ARM, etc. Rather they are lower-level - closer to the metal - and therefore both incredibly efficient and fast for low-level logic applications. In effect an FPGA can be set up as a dedicated video chip with all the options that we'd use something like OBS to do using a small computer.

The general "gotcha" is that the many require a special low-level programming languages although (according to the web, I haven't verified this) some support higher level languages like C/C++ and even Python. CHISEL looks interesting although it's not yet considered "mainstream". http://www.chisel-lang.org/

Similarly this also requires a specially designed PCB or a development board.

I'm happy to help in the layout but that's about the limit of my skill-set.

I wonder if the Pi Pico could be used here? The Pico has some FPGA-like functions (state machines) that have already been used for video applications.

I'm not very familiar with the Pi pico, my first thought is whether it would have the power to process a video stream the way we want. Unfortunately, the ARM version of OBS is in quite a state and requires manually building on Raspbian, even if the Pico had the power, the state of OBS ARM would impede us, I think.

Edit: I've done some more research on HDCP, and it seems like the best option is a cheap HDMI splitter that ensures a proper handshake with the donor TV, then allows non-HDCP devices on the other ports. I found someone who used Verilog on the Mimas A7 board to create a copy of the framebuffer and send it unmodified to the output HDMI port. I also plan on consulting the OSSC codebase for example code that we know works. The OSSC has all kinds of filter features for different cores. I think I'm going to grab both boards because the Mimas has people using them for similar use cases and the A7 has a Xilinx processor that may be more performant. There's much to learn here!

Excellent. I did try Grok (of all things) and it does seem at least aware of and able to write simple code for Chisel. I agree that the Pico might be a bit weak, the state machines (4 in total) are very similar to a hardware language as the instruction set is very small and seems to run independent of the main CPU.

The main advantage of the Pico is the low cost vs. the cost of dev boards for FPGA chips which put them out of reach of the average user. That's not to say this is impossible, far from it, but we'll need someone to obtain and develop the code before we could really consider making a custom board.

Here's some more info on the Pico in case it helps.

https://medium.com/geekculture/raspberry-pico-programming-with-pio-state-machines-e4610e6b0f29

See here, This guy managed a 480p VGA interface with some difficulties and he was only doing 16 colors at half of the original resolution we are targeting, at a minimum 1024x768. Copied from the linked github:640x480 = 307,200 pixels, but the Pico only has 264 kB of RAM. Thus, each pixel can only be 4-bits (giving 16 colours) with each byte of the frame buffer containing two pixels.

Unfortunately, this is a fundamental hardware limitation as I suspected. The reason I chose the above boards is A) The Xilinx board in particular is a well-known processor specifically for video processing and should be able to handle our resolution floor with zero problems. B) both of the boards have video input and output already licensed and included with only basic passive adaptors to get to the projector. The Mimas board even has miniDP outputs which means you can output straight to VGA with the right cable

Then there's always this sort of thing https://shop.pimoroni.com/products/pimoroni-pico-vga-demo-base?variant=32369520672851 which isn't the only one and while I agree that FPGA is the "right" solution, the cost of dev. boards puts it outside the scope of most of us (myself included).

Additionally, there are other alternative like this one: https://geoffg.net/picomitevga.html

Again, I have zero experience with this hardware, I'm just looking for a (financially accessible) alternative. When we factor in the cost of the dev board, it makes better sense to leverage something with a bit more muscle, perhaps even a x86 board based on the Atom (yeuch). I think LattePanda had a crack at something like that.

However, if you want to run with FPGA, you have my full support (for what little that's worth). I can likely make a custom board given a sample layout from the manufacturer because we won't need all the bits found on a typical dev board and that reduces cost quite significantly.

I believe the Pico is not a viable option simply because of the memory constraints and maybe cpu power itself. Needs further research.  however, you are right there are a variety of SBC x86 options out there for cheap that can just run OBS, in addition a really budget constrained user could simply use a spare PC from any number of years ago. I absolutely love the ecosystem Pimoroni has on their site btw. Hey, they offer a custom Pimeroni Pico board that expands the RAM to 8MB, which might be worth considering......Obviously, it would be ideal to achieve this with such a low-cost part, but I am not sure it is attainable. it's on my radar tho

I found a guy on the Internet I'd really like to talk to; name's Paul Fleming and he is a freelance FPGA dev with tons of experience with Xilinx and video processing via FPGA. Even has a patent for quote "A method to calculate a replacement value for a defective pixel using proximate pixels targeted at an FPGA implementation.". He has a profile up on a freelance site advertising his services. Anyways I've shot an email off to this guy, hopefully he will be interested in hearing about this project! 

Yeah that's the sort of guy you'd need. I work mostly in the analogue domain now. Although I can program in many languages, I'm not classically trained and won't necessarily write efficient code.

The new Pico 2350 is faster than the original and the state machines aren't CPU bound, although they can set and read CPU registers. It's also got two completely separate architectures ARM and RISC-V on chip and it (looks) to be a good deal faster than the original.

Still, FPGA is undoubted the superior option when ultimate speed (and ease of development) is required as Cecil, Scheme, etc. are high-level compiled languages whereas the state machine is essentially an assembler language so it's long winded and programs are quite short. They are quite close to FPGA in function though, not being automatically CPU bound  and directly targetting the I/O devices.

It's not that I want to dissuade you from FPGA, quite the opposite, but the cost of developing an FPGA-based board is considerable (PCBs invariably don't work as we expect on the first run) and the Pico 2350 is just $2 a pop fitted from JLCPCB (minimum of two per order). I've designed in industry and hyperfocus on "one way of doing it" can lead us up a dark path.

The other advantage of Pico is the number of people developing for it, since it's any CPU-tasks that don't need PIO can be handled with the huge number of pre-existent libraries

While I'm bound by industrial "keep your gob shut" laws, I can tell you that I once worked at a place that employed a huge, bulky and very expensive FPGA,controlled box to do the job that little more than a handful of fairly accessible electronics could do as well, or better. The original designers had gotten lost in the forest and rather than tossing everything in the trash and starting over, they came up with something that didn't actually work. (The fact that the product as a whole worked as advertised was due to something entirely separate, but to say more would likely see me in court.)

The same (I suspect) is true of the current problems at SpaceX - I'm not a fan. It's not that Musk makes promises that engineering science simply isn't capable of (the promised Tesla Roadster needs a larger battery pack than there is room for). But that they are trying to move forward with a design that appears to be unreliable and unrealisable for a bunch of reasons that they are just not prepared for.

Sure, it looks impressive, esp. with all those camera angles and ultra-high definition drone shots. And it sounds good on paper but when you get into the weeds there appear to be inherent flaws that may simply be a problem of scale. Vibration is a serious issue in rocketry and the 30-odd Raptor engines will create a lot of vibration which has different effects as the craft enters leaves Earth gravity. I'm paraphrasing from experts in this field; if I was a rocket scientist... 

What worked in the 1960s at the size of the Saturn V, may have been the pinnacle and while NASA certainly lost a few (and some lives too) they had a lot more success. (Mars is orders of magnitude further away than the moon, and as the most recent attempt to reach the moon showed, again, that even the moon is tricky, with the most recent landing falling over on the surface.)

Again, nothing here says that your approach is necessarily wrong, but sometimes existing solutions are simpler and have had more minds to work out the kinks. The wheel existed for well over a 1000 years before someone thought about putting a tyre on it - but the basic concept didn't change. Recently some very smart designers figured they could improve on the wheel and produced a spoke-less e-bike (Reevo) which attracted a lot of money in pre-orders but never worked as promised and the few that did make it to backers are (as one person put it) an unreliable death-trap.

Perhaps the way to fund such a project would be with crowd-funding because that could (potentially) provide sufficient working capital and free everyone from the constraints of traditional open source hardware. My reservation is the limited appeal since (after all) this does require some very delicate work on the screen and a practical product would need to be fully assembled to reach mass market. I'm old enough that I creak a bit but was also around before the first practical LCDs reached the mass market. In those days we demonstrated the effect of using sticky tape on some old Polaroid sunglasses to show how light polarises at different frequencies, thus allowing specific wavelengths - colours - to appear as the polarising filter was rotated.

OLED, a more recent development, doesn't need the polariser which is why it's capable of much brighter whites and deeper blacks. Each pixel is its own light source vs. the LCD method which has to block light and some amount invariably bleeds through, resulting in muddy blacks. Matt's solution is quite brilliant in this regard.

Raspberry PI 5 and that this can be a smart TV

Edited. Please don't use all caps in future. Thanks.

Great video. I am just wondering how on earth you made the video feeds the same, but the projector feed with the filters. is it just one device that outputs both feeds, like a laptop that mirrors both screens, or hoe does it work?

Looks that way (I haven't studied the detail) but I was explaining to someone today how it worked. I looks like the split feed is processed by the OBS software in real time to provide the light sources more like an OLED, even though it's really a hybrid. Matt's ability to turn a crazy idea into a working device is just awesome. Quite a unique talent.

(@matt - just leave the unmarked bills behind in the cistern in the usual place mate.)

@yash69 Re: https://community.roku.com/t5/Channels-viewing/Software-version-14-1-messed-up-on-Model-4802x-ultra/m-p/1056263

-----------------

i would have done it different... try without light diffuse, or at least Not glue it. LOL & RGB Color Back light, for increased saturation...

adjusting geometry can be done very easy.... Linus Tech Tips has a 2000" projector video, using multiple videobeams, all require geometry adjustment.

the last idea.... is to use Fiber optics instead of a Bulb... fiber optics can be placed outside looking at the sun, or looking a light source, like a vehicle xenon lamp or a fire.. or a mega powerful 200,000 lumen portable Led flashlight "portable", outdoor camping, high efficiency low power consumption.. similar to this:

https://odysee.com/@techteamgb:7/this-monitor-is-powered-by-the-sun!-sun:2

but... image quality depends on Color Rendering Index LEDs dont have good CRI, unless very expensive, Tunsgten bulbs have 99% CRI. the quality improvement you had comes mostly from higher CRI light source. https://en.wikipedia.org/wiki/Color_rendering_index

IF using FiberOptics looking at the sun, light becomes unstable, because clouds, requires a fluorescent material in between, used as a light capacitor. and needs to be inside a mirror box.

Anyway... the reason i dont do it, its because size... 50" TV at 3 meters = 21" at 0.3meters. a 34" UltraWide at 0.3 - 0.5meters looks much bigger... 50" looks like a toy.

also most movies are designed for Super35mm, https://en.wikipedia.org/wiki/Super_35 1920x1080 ratio 1.777 is Not for movies.

2.38 ratio is much better. https://www.amazon.com/LG-34GP63A-B-Ultragear-Adjustable-FreeSync/dp/B0CC2LBTB2

a TV/Monitor used close Requires Curved screen. very large LCD screens break very easy...

i had a Samsung TV, broke when i moved out of the room, it was heavy, too much force in 1 hand, broke it. it had multiple layer LCD panels, with power hungry back light,  LOL was heavy, but had nice colors & contrast.

your videos are awesome, dont get me wrong.

but... i prefer other path. LOL Another way to improve your Projector TV, is to create a Black Box, like a Bellows...

https://en.wikipedia.org/wiki/Bellows_(photography)

using Super Black fabric.

https://www.youtube.com/watch?v=43OGgDaR2aE&t=72s

room reflections does affect image.

This projector idea got me thinking again about earlier projects such as the Artificial Sunlight, and possibly even borrowing mirror/lens adjustment ideas from the 4K Phone Projector project!

One handy online tool I have used is this ray optics simulator.
It lets you figure out optimal mirror placement to avoiding unnecessary calibration for image distortion. You can also specify exact emission angles to match your projector specifications.

There is also this handy Room Projection Calculator to help with dimensioning and projection FOV angles.

For reference the TV model Matt used is an LG 42" LB5500.
So approximately 523mm vertical screen height.

Taking into account assumptions about projector unit size, angles etc. as you can see with relative ease you could definitely condense this system to fit into existing recycled furniture!

@marcdraco I believe what he was doing in the video was plugging both the display and projector into his computer, selecting duplicate displays so that both have the same image, and then only filtering the projector's display on OBS

Yeah, looks that way, but the OBS display is for the lightmap only the actual colour rendition is purely the LCD panel. Incredibly clever.

I wonder how far this varies from making this one of those $2000 TLCD(transreflective lcd) or a RLCD (Reflective lcd). Couldn't you just use a piece of mylar or some sort of reflective material to make this work?