DLP: an ideal display technology for HMDs?

[MSat]

You guys have really piqued my interest in VR, and now my OCD wont let me sleep!

I admit, I'm pretty much a noob when it comes to HMDs, and optics in particular, but I know my way around a circuit board and engineering concepts enough to be dangerous. With that said, if I just sound stupid, then let me know - maybe I'll actually be able to fall asleep.

So I decided to do a bit of reading and brain racking on the subject, but repeatedly found that LCDs seem to leave a bit to be desired for various reasons. Then there's OLEDs, which for most intents and purposes are not really any different. That leaves us with a few other exotic technologies, and DLP. I had a vague understanding of how DLP worked, so I decided to read up on it and peek at some of TI's introductions and datasheets, and found that this might actually be a very promising solution. Here are a few reasons why:

1. No individual color elements per pixel. Primary colors are reflected off the mirrors in sequential order at a very high rate (in the KHz). Current tech uses LEDs.
2. Slight pixel blurring due to the physical movement of the mirrors, which eliminates the "screen door effect"
3. Very high refresh rates possible, at the expense of luminance detail.
4. All mirror elements have a corresponding 1-bit frame buffer (my own term for it), and all mirrors refresh simultaneously.
5. Up to 1920x1080 resolutions available

I think it might even be practical to get away with using a single DMD for stereoscopy. When energized, each mirror has two possible states (ignoring the fast transitions): -12/+12 degrees. In TV/projector designs, the on state typically reflects the light into the optics, while the off state reflects it onto a heatsink. But what if the DMD was placed towards the front of the HMD with the light source perpendicular to it? That means depending on the state of the mirrors, the light would be reflected either to the left or the right at a 12 degree angle. Like I said before, I'm a dummy when it comes to optics, but I'd imagine it's quite workable (maybe not much more than a well positioned plano concave lens since even a 1080 DMD is less than 1" diagonal?). Of course, you wouldn't want the light that's being reflected away from one eye to be seen by the other, but that could be resolved by simply having an LC shutter on each side. It should even be possible to use a standard DMD controller IC, the only necessary processing would be that one of the stereo frames would need to be a negative of itself.

I got a few more ideas for it, but I'll just shut up for now.. Let me know what you think!

[cybereality]

I also had an idea to use DLP for an HMD, but what I want to do is take 2 pico-projectors (one for each eye) and use back-projection screen. Should work, and looks like Palmer is working on something similar. DLP is great though. I have a DLP PJ in my living room and the 3D is awesome. No ghosting at all, very nice image. They are also pretty affordable.

[PalmerTech]

Yep, DLP is pretty sweet!

The main problem is that you end up needing pretty heavy support electronics and optics. Check out this page and navigate to the Nvis SX111: http://www.nvisinc.com/technology.php It uses LCOS, pretty similar idea.

Unless you are willing to bulk things up, it is hard to get high FOV out of microdisplays of any kind.

[MSat]

@cybereality

I did read a few threads about the pico projector setup. Doing a google search for DLP HMD took me right to this forum... Imagine that! I would believe that they would be workable, but it seems that they do need quite a bit of optics - maybe partly due to the tiny DMD they no doubt use?

My friend has a monstrous Mitsu DLP RPTV, and I always liked how much "softer" the video looks, especially up close compared to my LCD RPTV.


@PalmerTech

Very fancy devices, and indeed quite a bit of parts stuffed in there. Following some links mentioned in other posts led me to 'leep on the cheap', probably made possible by utilizing a display that's not too big nor small? Still, another member was able to use a single optic per eye with an impressive FOV, but again, this probably has more to do with using the right size display?


I'm thinking that most or all of optics typically used in DLP projection design could be eliminated, except possibly what's needed to collimate the light source (then again, maybe two offset and angled conical light sources might even be more ideal), and of course, the eye piece optics.

Granted, the DMD and the primary driver IC aren't exactly cheap (~$500 and $100 respectively for 1024x768), but seeing that a 1080p DLP TV comes in at under $1k, I would think something similar can be achieved for an HMD that does away with many of the drawbacks found in LCD units.

Now to find a broken DLP TV/PJ with salvageable parts on the cheap....

[PalmerTech]

Before you go to all that effort: Have you seen the SiliconMicroDisplay ST1080? It uses LCOS, very similar to DLP in a lot of ways.

[MSat]

Besides reading up on it, I have no personal experience with the ST1080. LCOS does seem interesting, but of course it would need either two displays for a practical HMD. I suppose that's not really an issue if they're substantially cheaper than DLP - given the price of projectors using them, I'm assuming they are. However, if I understand the technology correctly, it would seem that for a large FOV HMD, they would require more optics than a DLP-based device. I'm really interested in what the effects would be from an offset and angled conical light on the resulting pixels - would they be distorted? The geometry shouldn't be too difficult to solve. I can't always help being stubborn and doing things the hard way!

[MSat]

Ok.. I busted out the protractor and a pencil just to visualize what I was thinking. Using a diverging light source reflecting off a DLP DMD, it should project an expanding image that's always in focus without any optics except those necessary to combine the point sources of the LEDs. With the projected image size being customizable, is it safe to say that the remaining optics could probably be a simple 1-piece eyepiece?

So while the DLP components might be pricier than other display devices, the reduction in parts along with the other benefits of DLP mentioned previously could lead to a net win. One of the biggest drawback that I can think of right off the top of my head is that depending on the configuration, the unit might look a bit beaky, but it should still be fairly light especially if an enclosure of light plastic is mounted to a frame.

[Chriky]

Can you draw a picture I'm really can't visualise this!

[MSat]

Ok, Ignore the typos or that parts of words that got cut off..lol

This was the simple 2D drawing I did on paper, but added some text to clarify it for you. If you don't know what a DMD is, then I suggest you google "digital micromirror device".

In the picture, I only calculated the angled and diverging light source against a flat reflective surface, where in an actual DMD, each mirror would be oriented +/-12 degrees. Regardless, I believe the concept should still hold true.

Consider any point between two letters (such as A to B, B to C, and so on) to be a mirror which represents one pixel. As you can see, the further the pixel reflections propagate from the mirror surface, the larger they become. Hopefully this clarifies it a bit.