A small comment is also that it's /not/ a xenon-lamp, but a metal-halogen bulb. It doesn't require as much high voltage to strike and it doesn't have the same high pressure.
DLPs have always fascinated me, fitting millions of small switchable mirrors inside a chip... then again we now have chips with billions of transistors :D
the fascinating thing is that they are mechanically moving by a few degrees. I always wonder(ed) if the die itself would feel rough when they point in different directions
I would have expected fabrication defects would be more random. I wondered whether it was due to the driver chip failing.... Like when you toast LEDs in a multiplexed display when the clock stops?
@@edlakota That can quite easily happen with handling the devices. A high voltage (from static on the human body) can flip the mirrors in a row or column, the same way that they are moved when the device is in circuit. (I used to work for TI, and have a few of these devices myself.)
It's impressive how MEMS completely flips what we understand of materials on it's head. The effects on the material like fatigue and such change completely to what applies in a macroscopic manner, hence they even work to begin with, the amount of deflection, how many per second and speed of the mirror deflections would be completely unfeasible in a macroscopic scale. Let alone the energy to keep flipping the mirrors fast enough for moving pictures with different shades of light.
but all colors are only partially activated for a gray color. That means the mirrors have to switch on/off rapidly. I wonder if this generates friction, and therefore heat. According to the German Wikipedia, the PWM switching frequency can be up to 5 kHz.
@@TheRailroad99 all DMD devices are equiped with a heatsink but I think the major part of heat come from the lighting source, not from the device itself.
I believe the individual pixel light intensity is controlled by using a duty cycle on the mirrors. A mirror spends only as much time reflecting towards the screen as required for a particular brightness, then it flips away. It is amazing to me that DLP projector are even physically possible, much less common.
Ceramic chips with gold plating seriously need to make a come back. They look so classy and give off a feeling of power and capability limited only by one's imagination.
@@TomStorey96 oh come on, I fucking paid $1999 dolar in a new CPU, the least I could have is proper ceramic, what would that cost ? $30 more dollars ? I'll file a bug report for my CPU, absolutely unusable, its not pretty and made out of ceramic and gold.
Oh I know the color shade answer! So you know how the color values are made out of bits right? Full brightness is all 1s and black is all 0s, basically each bit position is assigned a "time slot". The most significant bits get the most amount of time. That also means a value of 0b0101 switches more times than a value of 0b0110, makes driving it easier but it also means those mirrors need to toggle incredibly fast!
The grayscale is achieved by rapidly turning on and off the pixels. The ratio of on time and off time determines the intensity level. This technique is widely used in electronics (e.g. controlling backlight brightness on your phone or laptop screen) and it called PWM or Pulse Width Modulation.
Agreed! For each RGB frame cycle time, the mirror can be on the entire cycle (full brightness), or any percentage down to zero. Video signals encode two values: Chrominance and Luminance. The Chrominance is used for the color information and the Luminance for brightness. The ratio between RBG (Red,Green,Blue) defines the color, then the duty cycle (on time versus off time) defines the brightness. The higher end units use 3 separate DLP's, one for each color; no color wheel needed.
The micro mirrors vibrate back and forth to create tonal values. When DMDs fail its due to the mirrors being stuck or moving extremely slow. I used to work on rear projection DLP and 3LCD rear projection TVs for over 10 years. The biggest pain in the butt with those TVs was the ballast going out after the lamps would go out. People would replace the halogen lamp themselves, but touch the surface with their oily fingers thus significantly shortening the life. Color wheels were easy to replace and easy to tell when they were going bad (the bearing would fail) and it would sound like the tv was a vacuum cleaner. Lol.
THIS is the kind of stuff that made my day👍 Thank you. I had an 800x600 Acer DLP projector back in '04. After 3 years of light usage it start getting more and more broken pixels so I replaced it to a Full HD DLP beamer. TI seems to optimize their DLP chips as this time I had no issue with broken pixels (still have the projector in use from time to time). Too bad I did not disassemble the broken DLP back then. Very beautiful!
I am an ex-TIer. I have a few of these in my collection - they are really cool pieces of technology and still amaze me 25 years after I first came across them. Although I did not work in the DMD business, I knew a couple of the DMD/DLP experts one of the regional offices Thanks for making this video.
I am also fascinated by DLP as a technology. DLP uses the * fast * mirror movement to reflect light in a direction or another (fully saturated color is 1 and no saturation color is 0). The 1 is fully reflected into the lens, the 0 is reflected away from the lens. How fast is that switching? It goes up to 9 kHz, that is 9000 times a second for each pixel (newer ones go up to 32 kHz). This is how they manage to get all the grayscale values for each color by synchronizing with the color wheel and setting the grayscale for each image 30 or 60 frames per second, that is about 90-180 frames per second to get the full spectrum of color. All the intermediate grayscale patterns are done by the rapid switching between the fully saturated state and no saturation state. This is easy because the chip is capable of up to 9000 Hz switching and being digital it can easily synchronize with the color wheel for full color rendering. DLP is capable of 1024 shades of gray for each color individually, that makes up 1,073,741,824 colors for RGB or even more for CYMK. The human brain is blending the rapidly switching images to get a fluid movement. I hope it clears your question at the end.
Yeah the video got me thinking how they make shades with this thing and as the mirrors can't do partial turns as they'd reflect light on other pixels where they shouldn't I concluded that I can't think any better than that it must be some kind of a PWM device. And from what you say, it is. Which means that those tiny mirrors keep flipping in an insane rate. First the video signal, then the color wheel, then top it all with the shade PWM value... Their control needs some crazy precise timing. And they need to be quite durable for mechanical devices too, to pull off all the switching.
Of all the tech we have, DLP is still the one that fascinates me the most. Tens of millions of microscopic mirrors packed onto a chip and being moved independently with such speed and precession is nothing less than astounding. Thanks for sharing the numbers, only makes it more impressive. Anyone not impressed by this is so sadly jaded they have my sympathy.
My daily driver (watcher) is still my Samsung series 8 LED DLP 62”. They’re LED for the light source unlike a lot that had a bulb that needed a bulb that needed replacing. Bought it new in around 2008 for ~$2000 cdn on sale. 1920*1080. About 4-5 years after lots of use it started to get white dots (stuck mirrors) and black dots (dead mirrors). When it got to be too much I tore the whole set apart and ordered a new dlp module (about $250). There was a common failure on these where they didn’t put enough heat sink compound on the module. Added some good compound under the new module, and it’s been working perfectly since then.
Interesting. At work, one specific projector model always goes bad (after many years) and it seems to always start in a corner and spread from there, so I guess they skimped on the cooling paste as well.
They are a super cool addition to a collection! I got into DMDs a bit and managed some decent microscope shots of them running, though they where the pocket projector form factor which easily fit on the microscope stage. These have always fascinated me, the engineering is absolutely mind boggling. If you have a chance check out some of the MEMs micro mirrors, they are also really cool.
The pixels are pulse width modulated, not sure of the frequency but these chips update pixels very very fast. It's amazing the speed, precision and sheer quantity of electromechanical parts on these devices. Look into the consumer "4k" projectors, they seem to be using 1080p chips but are doing some sort of pixel shifting, combined with up to 6+ updates (3 color wheel RGB x2) per 60Hz refresh you can start to see why the chip needs a big heatsink (aside from being blasted from the front with a few hundred watt lamp). Would be interesting to see a follow up investigation! Edit: Also it's electrostatic, so those pixels may not be dead, just stuck.
9:24 Xeon Lightsource, the small sister of a Ryzen Lightsource? ;-) Good video as always! And the DLP Technologie is really high tech with all these tiny mirrors which have to work flawlessly in order to get a good picture. In comparison to your 10MB Hardrives which you could repair with a hammer and a screwdriver, these devices were much more filigran.
Nice video! Intensities are done via something like PWM - effectively moving the mirror thousands of times per second to reflect to the lens or into another direction. A little thought: Your "samples" might also be named that way because they wouldn't have made it into the market with their defects ;)
Thing is, Audi uses this in their newest Headlamps. I tought about that many years ago when I had my first DLP beamer, that you could use that to modulate low and highbeams in a pixel perfect manner and create a Matrix LED system, much better than even modern ones. Tought about that around 2011, but there I was too young to make something about it, also, DLP Beamers then were expensive as hell :D They clearly won't use regular DLP's, they must be fancy high reliable ones, but nonetheless, DMD's are facinating technology that is now even used in Headlamps (Audi E-tron).
When I first started working in Germany and someone in the office mentioned the Beamer, I was confused because I hadn't learned that expression and only knew it as a slang word for BMW. Very interesting technology - thank you.
It's not analog. The mirrors have two states and are controlled with very precise digital timing. It's rather more complex than a simple PWM, but the result is the same. Your eye receives digitally timed pulses of light of different colours which your eye and brain then integrate into an analog perception of the desired colour at the desired intensity. (I have the benefit of having worked at Texas Instruments from the late 90s to mid 00s.)
Really interesting that a majority of the dead “pixels” were along the same row. Why might that be? Might one row have been over-driven for some reason?
Could be when one shorts out, the current received by the rest of them in the same row is higher, so probability of another one burning out in same row raises. Just a guess.
I didn't have that, I had such a projector fail after years of usage. First I had 1 bright pixel, a few hours later I had dozens. A day later it was like looking at the stars when I projected z black image. None of them on the same row really.
interestingly i recall seeing almost every video game manual warning section about not playing games on "projection screens" might DLP be the projection screen they warned about?
Excellent video, i had 2 dlp televisions in the past and even a lcd projection tv. The mitsubishi developed stuck pixels shortly after replacing the bulb and after doing research found most early model dlp chips where plagued with mirrors sticking due to a defective lubricant with mirrors either getting stuck on or off. As for the light intensity i understand it was achieved in similar manner as lcd pixels, a fully opened mirror/pixel is white and a fully closed one is black but just with a single set of of pixels having the color added with the light wheel
Thanks for sharing and especially for the views under the microscope. Also, I believe that the intensity of individual pixels is achieved by varying the rate of switching on each mirror per image frame.
I have a DLP chip looking exacly like yours at 4:03, from a "Liesegang ddv 2111 ultra" projector with 1024x768 resolution. It has an aluminum block attached to the 4x4 square section in the middle shown at 4:10 with some sort of ceramic glue. It was connected to the PCB by pressing it on elastomeric connectors. Although I don't have a microscope to look at the micromirrors directly, when I reflect a laser off it, I can see a rectangular grid as defraction pattern. At 15:08, the trick how these things display different intensity is as follows: The DLP pixels can only be either completely on or off (reflect to wall or black beam dump), but nothing in between. However they switch extremely fast, probably around kHz range. So to modulate intensity, they do PWM for each pixel. Also they switch between the RGBA colors of the color wheel multiple times per frame to create the illusion of color. You see this effect when displaying a white image on the wall and quickly waving your arm through the light, then you will see the rainbow colors of the color wheel.
yeah, this is still knocking my head. Small structures on a die, yes. Small structure on a die which are electro-mechanical moving? NO IDEA how to process this. 🤔
They are MEMS devices. A Bosch sensortec engineer (who is part of the team at Bosch which designs MEMS Sensors) explained it to us (roughly) a few years ago in university. As far as I remember they are build "solidly" with support structurs left intact to support the movable parts. After that I think acid was used to remove them and make the parts movable. Not sure about the acid though, as it is also possible (and mind-boggling) to produce those in an enclosed , air tight chamber (which is only a few µm in size on the die). This really is absolutely fascinating. At least a few of them get designed here in Dresden, Germany, I think the factory is also here. (For the Bosch MEMS devices that is, not the TI ones as the DLP chips)
Interesting it is very related to physical movement os microscopic mirrors, and also there is a lot of heat production to generate the final images on the screen we actually watch to.
How the DLP makes shades: Each mirror flips at a very high frequency. With that, a signal similar to PWM controls how much of the light is reflected to the lens and how much reflects to the heatsink. A fully "on" pwm duty cycle reflects all light through the lens. A fully "off" pwm duty cycle reflects all light on the heatsink. and a 40/60 duty cycle reflects it accordingly: 40% into the lens and 60% into the heatsink. The number of shades depends on the number of bits controlling the pwm dutu cycle.
Not just super-interesting, but super-FASCINATING! I wonder how focus-able those mirror arrays are. I'm in need of a very small parabolic mirror to reflect stray light back to a collimated photomultiplier tube-based detector.
Very nice microscope image. We can clearly see that dead pixels are mirror detached from their mechanical shafts. (Note that 3DLP beamers have one for each color, and no color wheels.)
Next time try to display a grayscale gradient or test pattern as the ones used in TV, so you always see the same image on the chip (otherwise you see N images, one per each color wheel color)
Or just a black and white video, which as a bonus are often out of copyright as well, so long as they predate Steamboat Billy from the mouse corporation. Plenty of them to choose from, and you will get a stable image out of them.
I never knew how DLP projectors worked. I always assumed it worked like a film projector as in the DLP chip had light shines through it. I always wonders how they could withstand so much heat. I never thought about the possibility of all those tiny mirrors physically moving. That's wild.
I wonder if it will happen tho, there's few if any disadvantage inside a cinema to use projection, the contrast ratio of individual LEDs is easily matched by DLPs (what they use now in digital cinema projection after all), and the cost would be astronomical even many years in time, think about the size of that thing, it might get down to 1% of what it would cost now but never cheap. Plus individual failure points that are far harder to deal with. Projection still has quite a lot of interesting uses, both indoors and out.
the brightness is controlled via pulse width modulation - switching the pixels on and of at a pretty high frequency. it should be possible to meassure that frequency with a photo diode, hooked up to an oscilloscope. PS: i've recently come across the avegant glyph video goggles, which use miniature DLP projectors instead of the usual LCD or OLED displays. looking foward to trying those out. compared to 720p goggles sold for FPV, they're an absolute bargain.
This video was super fascinating. I had no idea DLP projectors used chips with thousands of microscopic mirrors being actuated. Kind of insane when you think about it.
Super awesome! Now I know what to look for next at eBay….by the way, if you would have played doom on this dmd, you would have beaten your record as I think the dmd is around one inch!
I've repaired a few DLP projectors some time ago. What I've learned is that the market is smack-filled with bad knockoff DLP chips from china that develop stuck pixels just hours after powering on. (Could also be refurbished chips) Finding actual original DLP chips is really hard, and once one does find them, they're so extremely expensive, it makes the repair cost multiple times higher than a brand new projector.
The chip itself doesn't really heat up that much but only a percentage of the light gets reflected, the rest is dumped as heat in the chip case. If you stick a piece of paper into the light in front of the chip it will be set on fire immediately. The light is extremely intense.
The intensity is due to the color wheel sections. It often has multiple sections of transparant wheel usually to make it brighter. Some have multiple sections of the colors on the wheel. The mirrors can flash 10thousand times per second, it strobes extremely fast through the correct colors on the wheel. It stays reflective more for brighter image, while it blocks light more of the time for a darker one. Because it's 10000 per s, you don't see this and percieve a single image.
It's impressive, but I always preferred LCD projectors - the rainbow effect when your eyes move to a different part of the screen gave me a headache as well as ruining the movie. Some people aren't as sensitive though, and the DLP contrast was better than LCD.
Amazing, it would be cool to get in real close with the microscope while it's working, and maybe some slow motion recording to be able to see the mirrors physically move.
I still have mine from an old Samsung. The chip has what appears to be a burn in image of Xion (desktop wallpaper of Kingdom Hearts) because I used this as a desktop PC. If I hold up to the light just right, I can see it pretty clear.
So if I use a uv light source and display a static circuit trace, I can expose a uv sensitive film without going through a photo transfer process. wonder if that is patented..
@@sunnohh Wait no sunnohh is right. TFT doesn't include the silicon manufacturing process. TFT is about manufacturing transistors and other structures on an inactive substrate. Such as plastic, glass or even some kind of ferrite for manufacturing the latest nanometer sized tape heads. The DLP is made using the CMOS process on a standard silicon semiconductor substrate.
The reason why you can see different light levels from the Micromirrors is that your eye is integrating the perceived scene's light level over a given timescale, therefore if you keep flipping the mirrors at really high speeds, you can see an average light level which is proportional to the amount of time that the mirror is in the on position.
If you want to use the projector without the lamp there will be an optocoupler you can short on the lamp PSU to have the projector thinking everything is fine. I did this myself when playing around with a DLP many years ago. However, you will still need to shine a flashlight through the color wheel as the DLP chip will stop working if it isn't given a sync signal often enough, and that is done with an actual color detector.
I have a friend that still has a working Hitachi Ultravison 57" widescreen DLP projection tv they had surprisingly good picture for the time. No idea what chips Hitachi ran only downfalls were there price and only 1080i interesting technology though never seen a close up of they chips before.
There is no any problem to make a different intensity of colors, as you have notice, color wheel rotating at high RPM, so every frame gets multiple wheel rotations, as you can imagine, mirrors should turn on an off at least once per color, and for DLP device it is not a problem to turn on and off multiple times per color, so its just PWM
The change in intensity is sorta not whats going on. This is a correction taking place to the color wheel. There are two different types of color wheel. One with 2 of each (blue,green,red) and another that has transition colors (ie a blend of two colors and a solid color. When there is a variation of a color outside the color band the transition color is used to give a better color tone. This requires more light. So the gradient changes when you go from a solid primary color to a in between color. Hope that makes sense.
Man, you can hear his passion and curiosity at 13:35 hahahaha. Nice. I have 2 questions: 1-Does anyone know the name of the anime at 14:26? (Yes, asking for a friend) 2- Is it possible to hear the action from the mirrors? Like, some buzzing or vibration due the electromechanical movement? Cool video, good sir. Thanks.
Hi, The Anime is Sakura wars, here the link: ua-cam.com/video/_F8AMZhYxhU/v-deo.html and second, I don’t think that the moving mirrors are making any noise.
Usually the Lamp driver board just has one optocoupler that tells the main board if the Lamp is in Ok state. So If you want to experiment without the bright hot lamp you just need to figure out the pins of the driver board which have the transistor side connected to the main board. Then remove the driver board and short circuit the two pins leading to the main board.
No, not always. If it is an Osram Driver board, it reports the state of the lamp over serial. It also receives commands over serial. For a project of mine I had to make an emulator, just to make the main processor think that the lamp is accepting commands and it reports back that is in good operating condition.
9:25 "Xeon light source" - The side effect of being a CPU collector :D
Great video! I always wanted to try this.
A small comment is also that it's /not/ a xenon-lamp, but a metal-halogen bulb. It doesn't require as much high voltage to strike and it doesn't have the same high pressure.
DLPs have always fascinated me, fitting millions of small switchable mirrors inside a chip... then again we now have chips with billions of transistors :D
the fascinating thing is that they are mechanically moving by a few degrees. I always wonder(ed) if the die itself would feel rough when they point in different directions
@@TheRailroad99 The mirrors are insanely delicate. The second you put your finger to it, it would be the equivalent of touching sand.
I'd love to see the micro-mirrors movements captured with a high speed camera! maybe the slo-mo guys can help you with that :)
That would be a really cool collab!
That would be great. I would like to see that too
I dont think the slo mo guys are needed really. Just a high speed cam. Those guys are super cringe.
look no further ua-cam.com/video/KpatWNi0__o/v-deo.html
ironically, not even a high speed cam is needed, actually you need a slower camera
Careful if you drop one of those, it's 2 million years of bad luck ;)
LOL, this comment deserves much more thumbs ups!
😂👍
Notice the defective mirrors all in one row - perhaps a fabrication error?
I would have expected fabrication defects would be more random. I wondered whether it was due to the driver chip failing.... Like when you toast LEDs in a multiplexed display when the clock stops?
@@edlakota That can quite easily happen with handling the devices. A high voltage (from static on the human body) can flip the mirrors in a row or column, the same way that they are moved when the device is in circuit. (I used to work for TI, and have a few of these devices myself.)
It's impressive how MEMS completely flips what we understand of materials on it's head. The effects on the material like fatigue and such change completely to what applies in a macroscopic manner, hence they even work to begin with, the amount of deflection, how many per second and speed of the mirror deflections would be completely unfeasible in a macroscopic scale. Let alone the energy to keep flipping the mirrors fast enough for moving pictures with different shades of light.
Very cool video ! As you said in your video level of brightness are achieved by varying time on/off of each pixel.
Didn't know that thanks.
Does anyone know the PWM switching freqency? It needs to be in the kHz range.
Does a gray image use the most energy then?
@@TheRailroad99 a gray image is composed by equal Red, Green and Blue Light. So I don't think gray image used more power.
but all colors are only partially activated for a gray color. That means the mirrors have to switch on/off rapidly. I wonder if this generates friction, and therefore heat.
According to the German Wikipedia, the PWM switching frequency can be up to 5 kHz.
@@TheRailroad99 all DMD devices are equiped with a heatsink but I think the major part of heat come from the lighting source, not from the device itself.
Really cool .. thanks for showing this !
I believe the individual pixel light intensity is controlled by using a duty cycle on the mirrors. A mirror spends only as much time reflecting towards the screen as required for a particular brightness, then it flips away.
It is amazing to me that DLP projector are even physically possible, much less common.
yeah. I also believe that it is controlled by using duty cycles. Indeed, fascinating
It changes the brighness by changing duty cycle of mirror
Yes and the projectors have a light dump, where they are tuned to aim to dump excess light
Ceramic chips with gold plating seriously need to make a come back. They look so classy and give off a feeling of power and capability limited only by one's imagination.
Alas, too expensive 😔
@@TomStorey96 oh come on, I fucking paid $1999 dolar in a new CPU, the least I could have is proper ceramic, what would that cost ? $30 more dollars ?
I'll file a bug report for my CPU, absolutely unusable, its not pretty and made out of ceramic and gold.
For some reason it took me ages to understand what he meant by "beamer".
"Beamer" is a very common German pseudo-anglicism. Many native German speakers are unaware of that and use it when speaking English.
At first I was expecting a BMW.
@@FreihEitner Me too!
Reminds me of Snatch: give me the shooter!
It's the same with "handy" which is the german term for a cellphone :)
Oh I know the color shade answer!
So you know how the color values are made out of bits right? Full brightness is all 1s and black is all 0s, basically each bit position is assigned a "time slot". The most significant bits get the most amount of time. That also means a value of 0b0101 switches more times than a value of 0b0110, makes driving it easier but it also means those mirrors need to toggle incredibly fast!
That is called "pulse width modulation."
@@harrkev This isn't quite the same thing.
Really, really cool! Thank you for this exciting insights ;)
The grayscale is achieved by rapidly turning on and off the pixels. The ratio of on time and off time determines the intensity level. This technique is widely used in electronics (e.g. controlling backlight brightness on your phone or laptop screen) and it called PWM or Pulse Width Modulation.
Agreed! For each RGB frame cycle time, the mirror can be on the entire cycle (full brightness), or any percentage down to zero. Video signals encode two values: Chrominance and Luminance. The Chrominance is used for the color information and the Luminance for brightness. The ratio between RBG (Red,Green,Blue) defines the color, then the duty cycle (on time versus off time) defines the brightness. The higher end units use 3 separate DLP's, one for each color; no color wheel needed.
Cool! In one sense then hot in the thermal sense. Another great vid. ♥️♥️♥️♥️
I must admit, I was looking forward to seeing the action of the DLP chip through a microscope.
The micro mirrors vibrate back and forth to create tonal values. When DMDs fail its due to the mirrors being stuck or moving extremely slow. I used to work on rear projection DLP and 3LCD rear projection TVs for over 10 years. The biggest pain in the butt with those TVs was the ballast going out after the lamps would go out. People would replace the halogen lamp themselves, but touch the surface with their oily fingers thus significantly shortening the life. Color wheels were easy to replace and easy to tell when they were going bad (the bearing would fail) and it would sound like the tv was a vacuum cleaner. Lol.
Great video, so cool to see the DLP operating directly on die.
THIS is the kind of stuff that made my day👍 Thank you. I had an 800x600 Acer DLP projector back in '04. After 3 years of light usage it start getting more and more broken pixels so I replaced it to a Full HD DLP beamer. TI seems to optimize their DLP chips as this time I had no issue with broken pixels (still have the projector in use from time to time). Too bad I did not disassemble the broken DLP back then. Very beautiful!
This was freaking awesome!
Thanks for the video.
I am an ex-TIer. I have a few of these in my collection - they are really cool pieces of technology and still amaze me 25 years after I first came across them. Although I did not work in the DMD business, I knew a couple of the DMD/DLP experts one of the regional offices Thanks for making this video.
i wasnt cautious enugh and this blew my mind. nice video man good job
Superb! I had no idea how these projectors worked when it comes to projecting images! Thanks for sharing.
I am also fascinated by DLP as a technology. DLP uses the * fast * mirror movement to reflect light in a direction or another (fully saturated color is 1 and no saturation color is 0). The 1 is fully reflected into the lens, the 0 is reflected away from the lens. How fast is that switching? It goes up to 9 kHz, that is 9000 times a second for each pixel (newer ones go up to 32 kHz). This is how they manage to get all the grayscale values for each color by synchronizing with the color wheel and setting the grayscale for each image 30 or 60 frames per second, that is about 90-180 frames per second to get the full spectrum of color. All the intermediate grayscale patterns are done by the rapid switching between the fully saturated state and no saturation state. This is easy because the chip is capable of up to 9000 Hz switching and being digital it can easily synchronize with the color wheel for full color rendering. DLP is capable of 1024 shades of gray for each color individually, that makes up 1,073,741,824 colors for RGB or even more for CYMK. The human brain is blending the rapidly switching images to get a fluid movement. I hope it clears your question at the end.
Yeah the video got me thinking how they make shades with this thing and as the mirrors can't do partial turns as they'd reflect light on other pixels where they shouldn't I concluded that I can't think any better than that it must be some kind of a PWM device. And from what you say, it is. Which means that those tiny mirrors keep flipping in an insane rate. First the video signal, then the color wheel, then top it all with the shade PWM value... Their control needs some crazy precise timing. And they need to be quite durable for mechanical devices too, to pull off all the switching.
Of all the tech we have, DLP is still the one that fascinates me the most. Tens of millions of microscopic mirrors packed onto a chip and being moved independently with such speed and precession is nothing less than astounding. Thanks for sharing the numbers, only makes it more impressive. Anyone not impressed by this is so sadly jaded they have my sympathy.
My daily driver (watcher) is still my Samsung series 8 LED DLP 62”. They’re LED for the light source unlike a lot that had a bulb that needed a bulb that needed replacing. Bought it new in around 2008 for ~$2000 cdn on sale. 1920*1080.
About 4-5 years after lots of use it started to get white dots (stuck mirrors) and black dots (dead mirrors). When it got to be too much I tore the whole set apart and ordered a new dlp module (about $250). There was a common failure on these where they didn’t put enough heat sink compound on the module. Added some good compound under the new module, and it’s been working perfectly since then.
Interesting. At work, one specific projector model always goes bad (after many years) and it seems to always start in a corner and spread from there, so I guess they skimped on the cooling paste as well.
Pretty cool technology, and I love the look of those gray/gold ceramic chips.
very impressive. thanks for the video!
Excellent. Best description I've seen.
I agree,
Super interesting :-)
& very well explained
Thanks
I have a couple dead dlp... I did not realize the chip was removable. I will definitely be adding them to the collection. Thanks!
They are a super cool addition to a collection! I got into DMDs a bit and managed some decent microscope shots of them running, though they where the pocket projector form factor which easily fit on the microscope stage. These have always fascinated me, the engineering is absolutely mind boggling. If you have a chance check out some of the MEMs micro mirrors, they are also really cool.
As a projector enthusiast on his 4th dlp, great video!
I have a couple of DLP chips from some old 67inch Mitsubishi projection tvs, this is way fascinating!
Very cool video...I like it Peter!
The pixels are pulse width modulated, not sure of the frequency but these chips update pixels very very fast. It's amazing the speed, precision and sheer quantity of electromechanical parts on these devices. Look into the consumer "4k" projectors, they seem to be using 1080p chips but are doing some sort of pixel shifting, combined with up to 6+ updates (3 color wheel RGB x2) per 60Hz refresh you can start to see why the chip needs a big heatsink (aside from being blasted from the front with a few hundred watt lamp). Would be interesting to see a follow up investigation! Edit: Also it's electrostatic, so those pixels may not be dead, just stuck.
9:24 Xeon Lightsource, the small sister of a Ryzen Lightsource? ;-)
Good video as always!
And the DLP Technologie is really high tech with all these tiny mirrors which have to work flawlessly in order to get a good picture. In comparison to your 10MB Hardrives which you could repair with a hammer and a screwdriver, these devices were much more filigran.
Nice video!
Intensities are done via something like PWM - effectively moving the mirror thousands of times per second to reflect to the lens or into another direction.
A little thought: Your "samples" might also be named that way because they wouldn't have made it into the market with their defects ;)
I’m sure the samples are defective units.
Thing is, Audi uses this in their newest Headlamps. I tought about that many years ago when I had my first DLP beamer, that you could use that to modulate low and highbeams in a pixel perfect manner and create a Matrix LED system, much better than even modern ones. Tought about that around 2011, but there I was too young to make something about it, also, DLP Beamers then were expensive as hell :D
They clearly won't use regular DLP's, they must be fancy high reliable ones, but nonetheless, DMD's are facinating technology that is now even used in Headlamps (Audi E-tron).
Amazing how clearly the moving images appears on the chip!
I never seen it before. Excellent video. Amazing technology. very interesting for me. Thanks. !!
Very cool tech. Thanks for the amazing video
I didn't know about this technology. Thanks for open my eyes!
I really like the enthusiasm in your videos.
Thanks 😇
When I first started working in Germany and someone in the office mentioned the Beamer, I was confused because I hadn't learned that expression and only knew it as a slang word for BMW.
Very interesting technology - thank you.
The irony is that the DLP is in fact a very analog light processing 😀
From back in the day where anything with the word "digital" on it was cool and hip 😄
@@TomStorey96 now they are needlessly throwing “analog” into ai circuits.....full circle we are
It's not analog. The mirrors have two states and are controlled with very precise digital timing. It's rather more complex than a simple PWM, but the result is the same. Your eye receives digitally timed pulses of light of different colours which your eye and brain then integrate into an analog perception of the desired colour at the desired intensity. (I have the benefit of having worked at Texas Instruments from the late 90s to mid 00s.)
Really interesting that a majority of the dead “pixels” were along the same row. Why might that be? Might one row have been over-driven for some reason?
Could be when one shorts out, the current received by the rest of them in the same row is higher, so probability of another one burning out in same row raises. Just a guess.
@@KrzysztofC-1 seems like a sound theory!
I didn't have that, I had such a projector fail after years of usage. First I had 1 bright pixel, a few hours later I had dozens. A day later it was like looking at the stars when I projected z black image. None of them on the same row really.
interestingly i recall seeing almost every video game manual warning section about not playing games on "projection screens" might DLP be the projection screen they warned about?
Excellent video, i had 2 dlp televisions in the past and even a lcd projection tv. The mitsubishi developed stuck pixels shortly after replacing the bulb and after doing research found most early model dlp chips where plagued with mirrors sticking due to a defective lubricant with mirrors either getting stuck on or off. As for the light intensity i understand it was achieved in similar manner as lcd pixels, a fully opened mirror/pixel is white and a fully closed one is black but just with a single set of of pixels having the color added with the light wheel
Another excellent video!
Very cool =D Thanks for sharing!
Thanks for sharing and especially for the views under the microscope. Also, I believe that the intensity of individual pixels is achieved by varying the rate of switching on each mirror per image frame.
ive always always always been curious about DLPs! I remember reading about them in 1989 when TI invented them. Was a young kid
You get a thumbs up for showing the into of Sakura Taisen on the DLP. ;-)
lol. thank you 😊
I was hoping, that you show the working chip under the microscope somehow.
I have a DLP chip looking exacly like yours at 4:03, from a "Liesegang ddv 2111 ultra" projector with 1024x768 resolution. It has an aluminum block attached to the 4x4 square section in the middle shown at 4:10 with some sort of ceramic glue. It was connected to the PCB by pressing it on elastomeric connectors.
Although I don't have a microscope to look at the micromirrors directly, when I reflect a laser off it, I can see a rectangular grid as defraction pattern.
At 15:08, the trick how these things display different intensity is as follows: The DLP pixels can only be either completely on or off (reflect to wall or black beam dump), but nothing in between. However they switch extremely fast, probably around kHz range. So to modulate intensity, they do PWM for each pixel. Also they switch between the RGBA colors of the color wheel multiple times per frame to create the illusion of color. You see this effect when displaying a white image on the wall and quickly waving your arm through the light, then you will see the rainbow colors of the color wheel.
amazing... Just amazing.
Appreciate your efforts 👍
I didn't know how those work, fascinating video.
Thank you for the detailed information and sharing this:)
Also, Disney: "We have to strike this!"
my favorite video to date
That chip looks like something behind objective in my camera :) Awesome and informative review!
The one inside your camera is a CCD device. Completely different kind of technology.
@@TheUglyGnome I see! Thank you :) But visually they look similiar (which confused mě)
@@miroslavzima8856
Yes. They look very similar.
Grate video as usual! Now I want to now how those chips are made! XD
yeah, this is still knocking my head. Small structures on a die, yes. Small structure on a die which are electro-mechanical moving? NO IDEA how to process this. 🤔
They are MEMS devices.
A Bosch sensortec engineer (who is part of the team at Bosch which designs MEMS Sensors) explained it to us (roughly) a few years ago in university.
As far as I remember they are build "solidly" with support structurs left intact to support the movable parts. After that I think acid was used to remove them and make the parts movable. Not sure about the acid though, as it is also possible (and mind-boggling) to produce those in an enclosed , air tight chamber (which is only a few µm in size on the die).
This really is absolutely fascinating.
At least a few of them get designed here in Dresden, Germany, I think the factory is also here. (For the Bosch MEMS devices that is, not the TI ones as the DLP chips)
Interesting it is very related to physical movement os microscopic mirrors, and also there is a lot of heat production to generate the final images on the screen we actually watch to.
Very cool! Thanks! (DLP technology is really se yellow from se egg! ;-))
Cool, didn't know of this technology.
with this video you save the life of my old svga Benq
How the DLP makes shades:
Each mirror flips at a very high frequency. With that, a signal similar to PWM controls how much of the light is reflected to the lens and how much reflects to the heatsink.
A fully "on" pwm duty cycle reflects all light through the lens. A fully "off" pwm duty cycle reflects all light on the heatsink. and a 40/60 duty cycle reflects it accordingly: 40% into the lens and 60% into the heatsink.
The number of shades depends on the number of bits controlling the pwm dutu cycle.
Great video! I was hoping you would make a DLP video.
see, you hope and its here already 😉
I wonder what the moving mirrors look like under the microscope!
Not just super-interesting, but super-FASCINATING! I wonder how focus-able those mirror arrays are. I'm in need of a very small parabolic mirror to reflect stray light back to a collimated photomultiplier tube-based detector.
DLP TVs were nice in the early 00s and had a nice pictures. Not ideal for gaming or anything but it had a very cinematic look. I'd love a 4K DLP TV.
Very nice microscope image. We can clearly see that dead pixels are mirror detached from their mechanical shafts.
(Note that 3DLP beamers have one for each color, and no color wheels.)
Next time try to display a grayscale gradient or test pattern as the ones used in TV, so you always see the same image on the chip (otherwise you see N images, one per each color wheel color)
Or just a black and white video, which as a bonus are often out of copyright as well, so long as they predate Steamboat Billy from the mouse corporation. Plenty of them to choose from, and you will get a stable image out of them.
I never knew how DLP projectors worked. I always assumed it worked like a film projector as in the DLP chip had light shines through it. I always wonders how they could withstand so much heat. I never thought about the possibility of all those tiny mirrors physically moving. That's wild.
may have to do a video on bypassing the ballast. this would be more awesome if the color wheel was over the DMD chip.
Thanks!! I was repair movie theater projectors, it's amaze technology) But soon in theater will be a big LED screens and DLP go to the history..
I wonder if it will happen tho, there's few if any disadvantage inside a cinema to use projection, the contrast ratio of individual LEDs is easily matched by DLPs (what they use now in digital cinema projection after all), and the cost would be astronomical even many years in time, think about the size of that thing, it might get down to 1% of what it would cost now but never cheap. Plus individual failure points that are far harder to deal with.
Projection still has quite a lot of interesting uses, both indoors and out.
the brightness is controlled via pulse width modulation - switching the pixels on and of at a pretty high frequency. it should be possible to meassure that frequency with a photo diode, hooked up to an oscilloscope.
PS: i've recently come across the avegant glyph video goggles, which use miniature DLP projectors instead of the usual LCD or OLED displays. looking foward to trying those out. compared to 720p goggles sold for FPV, they're an absolute bargain.
This video was super fascinating. I had no idea DLP projectors used chips with thousands of microscopic mirrors being actuated. Kind of insane when you think about it.
The Colourwheel can be dangerous. It spins with a high RPM. I had it on a Beamer, where it is scattered.
Super awesome! Now I know what to look for next at eBay….by the way, if you would have played doom on this dmd, you would have beaten your record as I think the dmd is around one inch!
I've repaired a few DLP projectors some time ago.
What I've learned is that the market is smack-filled with bad knockoff DLP chips from china that develop stuck pixels just hours after powering on. (Could also be refurbished chips)
Finding actual original DLP chips is really hard, and once one does find them, they're so extremely expensive, it makes the repair cost multiple times higher than a brand new projector.
The chip itself doesn't really heat up that much but only a percentage of the light gets reflected, the rest is dumped as heat in the chip case. If you stick a piece of paper into the light in front of the chip it will be set on fire immediately. The light is extremely intense.
The intensity is due to the color wheel sections. It often has multiple sections of transparant wheel usually to make it brighter. Some have multiple sections of the colors on the wheel. The mirrors can flash 10thousand times per second, it strobes extremely fast through the correct colors on the wheel. It stays reflective more for brighter image, while it blocks light more of the time for a darker one. Because it's 10000 per s, you don't see this and percieve a single image.
It's impressive, but I always preferred LCD projectors - the rainbow effect when your eyes move to a different part of the screen gave me a headache as well as ruining the movie. Some people aren't as sensitive though, and the DLP contrast was better than LCD.
Amazing, it would be cool to get in real close with the microscope while it's working, and maybe some slow motion recording to be able to see the mirrors physically move.
Oh it is amazing, I have some clips of them running under a microscope in some of my pico projector teardowns.
@@Zenodilodon Oh, I will have to check some of your stuff.
just a hint: the term "beamer", americans will understand, that you are talking about a BMW, they call it projector.
I still have mine from an old Samsung. The chip has what appears to be a burn in image of Xion (desktop wallpaper of Kingdom Hearts) because I used this as a desktop PC. If I hold up to the light just right, I can see it pretty clear.
Amazing stuff people can build when they are not involved in all kinds of hatred but just focus on the job.
So if I use a uv light source and display a static circuit trace, I can expose a uv sensitive film without going through a photo transfer process. wonder if that is patented..
The MDs will have a light dump were they aim the mirror to dump the excess light to modulate the brightness
Was especting Doom!! Wold like to know how this is made, amazing tecnology!
TFT. Thin Film Technology. The same basic concept is used for ICs, DLP, LCD, OLED, even modern tape drive heads.
@@ky5666 wut? Dlp is made in foundry on a cmos process, ask Texas Instruments (the only manufacturer in the world of dlp chips)
@@sunnohh Wait no sunnohh is right. TFT doesn't include the silicon manufacturing process. TFT is about manufacturing transistors and other structures on an inactive substrate. Such as plastic, glass or even some kind of ferrite for manufacturing the latest nanometer sized tape heads. The DLP is made using the CMOS process on a standard silicon semiconductor substrate.
@@ky5666 wrong. This is not TFT. This is MEMS. en.m.wikipedia.org/wiki/Microelectromechanical_systems
The reason why you can see different light levels from the Micromirrors is that your eye is integrating the perceived scene's light level over a given timescale, therefore if you keep flipping the mirrors at really high speeds, you can see an average light level which is proportional to the amount of time that the mirror is in the on position.
per pixel/per color duty cycle
If you want to use the projector without the lamp there will be an optocoupler you can short on the lamp PSU to have the projector thinking everything is fine. I did this myself when playing around with a DLP many years ago. However, you will still need to shine a flashlight through the color wheel as the DLP chip will stop working if it isn't given a sync signal often enough, and that is done with an actual color detector.
I have a friend that still has a working Hitachi Ultravison 57" widescreen DLP projection tv they had surprisingly good picture for the time. No idea what chips Hitachi ran only downfalls were there price and only 1080i interesting technology though never seen a close up of they chips before.
The brightness is changed by modulating the duration of the mirror being on.
There is no any problem to make a different intensity of colors, as you have notice, color wheel rotating at high RPM, so every frame gets multiple wheel rotations, as you can imagine, mirrors should turn on an off at least once per color, and for DLP device it is not a problem to turn on and off multiple times per color, so its just PWM
The change in intensity is sorta not whats going on. This is a correction taking place to the color wheel. There are two different types of color wheel. One with 2 of each (blue,green,red) and another that has transition colors (ie a blend of two colors and a solid color. When there is a variation of a color outside the color band the transition color is used to give a better color tone. This requires more light. So the gradient changes when you go from a solid primary color to a in between color. Hope that makes sense.
Man, you can hear his passion and curiosity at 13:35 hahahaha. Nice.
I have 2 questions:
1-Does anyone know the name of the anime at 14:26? (Yes, asking for a friend)
2- Is it possible to hear the action from the mirrors? Like, some buzzing or vibration due the electromechanical movement?
Cool video, good sir. Thanks.
Just commenting here so I get a notification when someone answers the first question ;)
Hi, The Anime is Sakura wars, here the link: ua-cam.com/video/_F8AMZhYxhU/v-deo.html
and second, I don’t think that the moving mirrors are making any noise.
@@CPUGalaxy
Hahahaha thanks, not gonna lie, I wasn't expecting that.
I wanted to see it displaying an image while under the microscope
Usually the Lamp driver board just has one optocoupler that tells the main board if the Lamp is in Ok state. So If you want to experiment without the bright hot lamp you just need to figure out the pins of the driver board which have the transistor side connected to the main board. Then remove the driver board and short circuit the two pins leading to the main board.
No, not always. If it is an Osram Driver board, it reports the state of the lamp over serial. It also receives commands over serial. For a project of mine I had to make an emulator, just to make the main processor think that the lamp is accepting commands and it reports back that is in good operating condition.