No because each pixel is only on 1/4 of the time. The LED is lit the whole time but it sweeps round 4 positions, so only 1/4 of the projected pixels are lit at any time - 3/4 are dark
DLPs are already being switched on and off to reproduce colors (which are reproduced sequentially by switching the leds) and again to reproduce shades of intensity. What pixel shifting does is to add yet another level of modulation.
Indeed, but interlacing was done at half the frame rate. This is done at 4x the frame rate.
I don’t think we’re fundamentally disagreeing. Just postulating if there are circumstances the modulation might have any visual impact.
Unless I have fundamentally misunderstood, there is a block of 4 pixels served by a single LED (probably actually a group of LEDs to give colour range but conceptually a single LED).
240 times a second the physical LED is updated: 60 times it will illuminate one projected pixel, 60 times it will illuminate the second projected pixel in the group etc.
The human eye persists the vision - continuing to see the illuminated projected pixel even after the actual physical LED light source has physically shifted on to the next.
Sort of, except these are not LEDs. These are very tiny mirrors.
Mirrors don’t emit any light, they can just reflect the light out of the projector or inside it, against an absorber.
(a) They can’t do shades of gray. These are obtained by rapidly moving the mirrors to modulate the light that goes out. The longer they spend turned in the “on” position, the brighter the pixel is.
(b) Mirrors also can’t do colors. Each mirror projects a monochrome dot. To get it colored, the light source LEDs are turned on in sequence. And each mirror is modulated as in point (a), for each of the primary colors, in order to represent the intensity of that color for that dot.
Both (a) and (b) happen for each of the four states representing each pixel shifting subpixel. And that is 240 times each second.
As you can see there’s a lot of persistence of vision going on. 
(this is theory, in practice I bet they do something even more complicated)
3 Likes
Thanks. Really helpful explanation. Thanks for indulging me!
When viewed very up close in 1/8 slowmo, the stable pixels do seem to wobble:
When viewing a rapid pan at 1/8 from viewing distance, there is no wobble visible, in fact the image seems quite stable and responsive:
Meanwhile, my MacBook Pro display shows a bit of ghosting, making for a less smooth pan:
5 Likes
Awesome!
VERY nice! I’ve been looking for slo-mo footage. I can’t wait to do 960fps tests when I receive my PPM.
The reason for the MBP being blurry is the fact that it’s an LCD screen, they can’t turn off the pixels nearly as fast as a DMD.
2 Likes
I managed to see it with my camera (that unfortunately only does 120fps).
I can clearly see the edges of the content wobble, but there’s still a visible fixed screendoor which I have no explanation for. (I believe it can be very faintly seen in the video above)
For anyone curious about the pixel shift magic, here’s a 240 fps video showing how the technology works:
And a regular-speed video (30fps), which is what you end up seeing:
These videos were taken in broad daylight btw, in case you’re wondering why the blacks are grey.
8 Likes
What is confusing about this is when you look at the image up close, you can see distinct squares with slim black borders. They form a grid of 960x540 on the surface of where you are projecting. This gives the impression that the display is not actually full 1920x1080. If I understand correctly, each square is actually made up of 4 pixels from the slight mirror shift 240 times a second.
I did actually try a 960fps video capture with my S10+ phone, but the result was not what I was expecting. Slowed down it is just color flashes and a complete image. You can’t actually detect the image itself being drawn since the light that is being reflected off the DLP chip is nearly continuous. I think one would need a true high speed camera that can detect in the 10s of thousands of FPS or higher to see anything interesting.
You can see above 240 fps footage showing the exact mechanism of action, and then 30fps showing what your eyes perceive.
1 Like
If it’s true 960fps capture it should show results similar to the 240fps capture mentioned above. That one clearly shows 240 frames where the pixels shift. If the pixels stay in one place all the time but you’re able to see individual colors (which will be at least 720 times per second) I wonder if your actuator actually works as it should?
I think because my 960fps camera is slightly faster than the 720fps of each color being drawn plus pixel shifting, the colors I am seeing when slowed down are just out of sync so it looks like a flashing color image. I should try doing it at 240fps to get a cleaner video. This is the result of a 960fps capture… YouTube Don’t watch if you are prone to strobe induced seizures. Use the slow play option on youtube if this is still too fast.
1 Like
If you shoot a smaller area of the screen it’ll be easier to see the half-pixel shifting action, there’s definitely four distinct sub-frames presented but it’s difficult to see if the pixels shift correctly. Your 960fps capture is definitely good, it’s supposed to cycle the colors like that.
Look like companies are starting to implement this Pixel shift technology into Cameras. Just think of the improvements if they could shrink it down and put it in Mobile Phone camera’s.
Why does the browser not showing 1080p I go to whats my resolution website and it just says 960x540
It should be 1080p right ? So why cant the browser see this. How can we know that we have 1080p …
1 Like
It’s 1080p with 60hz (60fps), most of the affordable 4k Projectors on the market are using the same technology with a 1080p dlp chip. The PPM team made a big change to the interface screen to make the icons bigger. I would guess the web browser screen size is being effected by this recent change to the OS interface.
2 Likes
I know its pixelshifted but why not use 1080p in the menu and browser and just enlarge everything insted ?
And why is the browser using same res as the ui
Tried to go to screen calibration site but it says my resolution is to low
2 Likes
Android browsers show the “logical” resolution. The UI is running at 2x high-dpi mode that’s why it shows 960x540.
The same will happen e.g. if you check Display settings - Lagom LCD test on a MacBook Pro 13", it shows 1280x800 instead of 2560x1600 which is its native resolution.
If you want to verify the true resolution, download or make a 1920x1080 grid pattern (192 x 108 lines, 10 pixels apart). Then you can actually count those pixels on your wall.
As for why not use 1080p, it did use it initially but the user experience was not good. If you check the #picopix-max:feature-suggestions forum you’ll see it was heavily requested to scale the UI down to HighDPI. We chose 2x, but there is a lot of demand for 1.5x scaling (1280x720 simulated resolution). We are looking for ways to add that as a user option.
10 Likes