Digital Light Processing, So Many Tiny Mirrors

Did you know there are a million little mirrors flickering back and forth, reflecting light within some modern projectors; like a flip-dot display but at the micro level? In his video, [Ben Krasnow] explains the tiny magic at work in DLP, or digital light processing technology with a scaled up model he constructed of the moving parts.

LCD projectors work much like old slide projectors. Light is shined through a transparent screen containing the image, which is then focused and enlarged through a lens. DLP projectors however achieve the moving image in a slightly different way. A beam of focused light is shined onto a chip equipped with an array of astonishingly small mirrors. When the mirror is flipped in one direction, it reflects the light out through the lens and creates a visible pixel. When the mirror is tilted the opposite direction, no light is reflected and the pixel is dark. All of these tiny moving parts are actuated by means of static electricity, and since a pixel can effectively only either be in an on or off state without any range of value in-between, the pixel must flutter at a rate fast enough to achieve the illusion of intensity, much like pulsing an LED to create a dimming effect.

In addition to slicing open the protective casing of one of these tiny micro-mirrored chips to give us a look at their physical surface under a microscope, [Ben] also built his own functioning matrix from tiles of mirrors and metal washers sandwiched around pieces of string. A wound electromagnet positioned behind each tile tilts the pixel into position when a current is run through the wire — although he didn’t sink the time needed to build out the full array in this manner (and we don’t blame him). If you do have the time and add in a high powered flash-light, this makes for an awesome way to shine messages on your roommate’s wall.

29 thoughts on “Digital Light Processing, So Many Tiny Mirrors

  1. Excellent overview on DLP technology. The SEM photos were first rate – don’t see that every day. You nailed it. The competing technologies are: LCD, LCOS and Laser-MEMS, all of which have their pros and cons. DLP is great for front projection, but can only be made so small, so won’t be embedded inside phones anytime soon. Whereas, Laser-MEMS projection can be shrunk down to very small sizes. just make sure to wear a patch over one eye when staring at the projector beam. :-)
    As movie theathers migrate to digital projection technology, expensive film reels will no longer need to be sent to the thousands of theaters in the US and worldwide – just download the movie to a large hard disk and project. TI had a long-term idea of where the market would go and made it happen. That’s visionary.

    1. I work volunteer at a new theater that just opened in my town. It indeed has a DLP projector, and the movies we show arrive on hard drives. We then copy the movie and associated trailers from the hard drive to the projector and send the hard drive back. Then we can show the movie for as long as we’ve paid the fee.

        1. The theater stuff is heavily DRMed. The content is encrypted. The “projector” is a “trusted system” that will only decode the movies when provided with the proper keys. Tamper proof and all that. I’m sure they have though about you resetting the time on the projector to some time in the past when you DID have a licence to play that movie.

  2. Good demo. I dig it. Would be cool to see the model actually functioning.
    Also, quite cool to see the mirror-mounts in the SEM images. Simple torsional springs. Incredible!
    It’s interesting how light-wheels are considered de-facto for DLPs, whereas they most-certainly could also be used for LCDs, instead of having three separate light-paths. Though, likely, dimmer… But it’s kinda like they have two different lines-of-thought with the *entire* design… one largely-mechanical, the other largely-solid-state… yet “never the twain shall meet”

    1. LCD isn’t nearly fast enough for color wheels. You might think so because there’s 1 -5 ms panels being marketed, but those are achieved by cheating with the measurement standards that wouldn’t cut it for a projector. Not enough contrast, not color-accurate, slow on gtg transitions etc. You only need to reach 90% of the target transition to qualify – in reality reaching the exact value will take many times longer.

      You need less than 5 ms actual genuine response time to get an LCD switch fast enough to do sequential color at 60 Hz. The worst kind of TN matrices just about achieve that in reality.

      1. Sorry, that’s actually wrong. It takes 5 ms for the LCD to reach the target value, but then you’re already switching to another color and the average value during the 5 millisecond timing window will never be correct.

        The response time has to be way faster still, because each color in the wheel is only present for 5 milliseconds at a time.

      2. While this seems logical based on the de-facto state-of-the-art goin’ round… It seems to me there’s some inherent limits in this logic… E.G. LCD-TVs used for 3D by switching pixels between every frame… 120Hz… Unless my understanding is wrong (which it very likely is), that’s only 1/3rd slower than necessary for three colors (assuming you’re not doing 3D)… Then there would seem (to me) to be a scaling-factor… e.g. LCD pixels on a matrix only 2cm-square would have quite a bit less inherent capacitance(?), seems they’d be more likely to be able to switch faster than larger pixels… Again, I could be wrong.
        And, again, the consideration, in my mind anyhow, is that it’s funny that these two technologies seem *completely* mutually-exclusive… e.g. in reverse, why use a color-wheel with a MEMs mirror-array? Why not use three different arrays and merge ’em as they often do with separate LCDs? Or how ’bout using one array with mirrors that’re alternately-coated in RGB filters…? These aren’t really questions of *why* or *how* regarding the technologies, exactly, more a question of *why* regarding apparent-mutual-exclusion.

        1. 3-chip DLP projectors do exist, and they work just how you would expect, with 3x color filters and beam combiners. They cost a LOT.. but they are the benchmark in projection. This is the type used in digital theaters.
          A home 3DLP projector costs >$20k, for example the Runco LS-10d

    2. It’s been a few years since I’ve been in it, but the high end DLP projectors used to not use a color wheel, and had 3 dedicated DLP chips and light paths. I believe the color wheel was mainly a ‘cost effective’ option for lower to mid-level units because of per-chip DLP cost. Personally DLP tv’s with color wheel’s always looked ‘shimmery’ to me and were hard to watch, but I guess this doesn’t effect everyone. It’d be interesting to see some research (probably exists somewhere) on ‘frame rate’ of sorts of our eyes, and variances between people…

  3. Wow, what an excellent video. Thanks to HaD and Ben for providing this to the masses – I bookmarked it. When [if] I have kids this would be a part of their TV schedule – not Teletubbies or whatever brain melting nonsense is out there these days.

    1. Manufacture of DLP devices is derived from standard semiconductor processes. Keep in mind that the dimensions of DLP mirrors and support structures are relatively large compared to modern processors and memory devices, so minor defects are perhaps less significant. However, a a dead mirror = a dead pixel, and it can’t just be replaced by substitution. Some info at, although TI don’t give too much on the manufacturing and low-level details.

  4. I’ve been a projectionist at one of the largest cinemas in the netherlands. 12 theaters with the biggest screen about 24m wide. We’ve switched to Barco DLP’s about 5 years ago. So I’ve seen both worlds. It’s amazing the stuff that can come out of one of those things. When properly calibrated it can be just breath taking. The DRM on the DCP’s (Digital Cinema Package) is just insane. At one point the distributors wanted to give out licenses on a per film, per machine, per show(!) basis. That meant that for a single day you had 4 shows per projector per day = 48 different licenses. And every day had different programming so different movies were shown. It was a management hell. Which drove me into learning to program in python just to manage those XML keys (which got me my current awesome job at a VFX house). Luckily we convinced them to just give out licenses for a couple of weeks per projector. But aside from that the projectors themselves are great! Especially when you can play in 2 theaters 4 against 4 call of duty with everyone having a portable radio and the sound set astronomically high ;) great fun those days were.

    Ever since I’ve been wanting to recreate one of those DMD’s that I used to switch out from those machines. I’ve moved and I’ve got a great back yard that would fit 50 to 60 people easily and a shed that is positioned in such a way that it could itself function as a projector! It would be really awesome to pull it off. Just have to find the time for it.
    And have a way to DIY build those dmd’s to actually produce 24-25 fps.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.