Angles Of View
Lars A. Yoder is Product Marketing Manager for Digital Video Products, Digital Imaging, and Corporate Venture Projects for Texas Instruments in Dallas. Educated at the University of Rochester from which he received a B.S. in Optical Engineering and a subsequent MBA, Mr. Yoder is the author of numerous papers on Digital Light Processing technology. He can be reached at firstname.lastname@example.org and is interviewed here on the subject of
Digital Micromirror Devices - Reflecting on the FutureDa-Lite: Currently the DMDs your company is manufacturing contain more than 50,000 movable mirrors arrayed over the top of a silicon chip that isn't any bigger than a postage stamp. That seems like an awful lot of angels to fit onto the head of pin. However do you do it?
Yoder: Basically we start with a standard silicon memory wafer which is very common in the semiconductor industry. What that layer gives us are digital memory cells from which we can control each mirror. After we build a mirror over each of these memory cells we can plug a 1 or a 0 into it which will cause the mirror above it to tilt ą10°.
With that in mind, all we really do is accomplish a series of photo lithographic layering on top of the memory cells. To get the mirror high enough above the memory cells so that its corners have enough room to tilt down by 10°, we'll deposit a spacing layer that's appropriately thick. Then we'll etch where we want the holes, or lines, or spacings to be. Once that's done we can layer on the aluminum which constitutes the mirror and when it has formed we dissolve or wash the spacer layers away.
Da-Lite: From the illustrations in your literature and on your website (www.ti.com) , each completed mirror looks rather like a little flat topped mushroom which is connected to the structure beneath it only at the base of its stem. Is that correct?
Yoder: Yes; and it is at the base of that stem where the hinging takes place. If a charge is put into one side of the memory cell it will electrostatically cause the edge of the mirror above it to pull down toward it. If the charge is put into the opposite side of the memory cell, the mirror will tilt the other way. At these opposing sides we've placed little stops which mechanically block the mirror from tilting more than 10° in either direction. When there is no voltage applied, the mirror has no reason to tilt and it sits at 0°, which is to say parallel to the bottom of the chip.
Da-Lite: How fast can you switch each mirror through those 20°?
Yoder: 50,000 times a second.
Da-Lite: Inside a projection device one or more of these DMDs gets placed in front of a light source (a lamp) and behind a lensing system. How does it create a picture?
Yoder: When a mirror is in the +10° position light falling on its surface from the lamp will be reflected out through the lensing system and onto the screen. When a mirror is in the -10°position light falling on it will be reflected away from the lensing system.
Since a typical chip will include 600 rows of 848 mirrors, this is the resolution limit of the display. 600 x 800 of the mirrors are used as the pixels of a 3:4 aspect ratio image and 480 x 848 are used to display the 9:16 aspect ratio NTSC video signal.
Da-Lite: And of course you produce gray scale by varying the number of times, between 0 and 50,000/sec, each mirror is in its On position, correct?
Yoder: That's correct. Because we're digital and because we can switch the mirrors from On to Off in only 16 microseconds, the gray scale produced by DMDs is extremely accurate.
Da-Lite: What are the other intrinsic advantages of this technology?
Yoder: The first is that DMDs are reflective. So if you do the math for them you'll find that in video operation 60% of the incident light will get off the DMD. Now for an LCD projector, you start with 100% light but, since you have to go to polarized light, you throw out 50% right away, and by the time it gets through the rest of the system you've got only 6-10% left. But because we're not transmissive, because we're reflective and therefore don't have to use polarized light, DMDs have an inherent brightness advantage that's really powerful.
The other advantage we have (although we see that the LCD people are moving fast to narrow this gap) is what I call Fill Factor.
Da-Lite: Is that what other people call Aperture Ratio?
Yoder: Yes; it translates into the amount of useful information on the screen. I like to talk to people in terms of blocks. Imagine that if your screen was divided up into ten square blocks. With DMD technology nine of those ten blocks are full of information (which is a 90% aperture ratio). With LCD technology, today, at best six or seven of those blocks are full of information. The rest are black.
What you really want to see is 100% information but of course technology can't do that right now. So we're limited by that Fill Factor/Aperture Ratio. Although once you get that up to 90 or 95 per cent, you can't really resolve the unfilled 10% - at least from any reasonable viewing distance.
Da-Lite: What makes up that 10% on a DMD?
Yoder: Well, first there's the spacing around the mirrors. That gap equals one micron. Then, there's a very tiny hole right in the center of each mirror which we call the "via."
Da-Lite: That's the part leading down to the hollow stem of the mushroom?
Yoder: Yes. It is because of the gaps and the vias our Fill Factor ends up at "only" 90%.
Da-Lite: Is there some reason the size of the individual mirrors, at only 16 microns square, has to be so small? Doesn't that make them terribly hard to make?
Yoder: Not really, no. If you follow semiconductor technology, you see an endless push for more and more transistors per square inch and faster and faster switching speeds. A few years ago the standard was .8 microns, which meant that .8 microns was the smallest dimension you were permitted. Now they're down to .3 microns....
Now think about these mirrors. Including the gap, they're 17 microns. They're a no brainer! It's not at all difficult to create objects of this size if you're using .8 micron technology. So creating our chips is well within the bounds of the old semiconductor limits and even more so within the new.
Da-Lite: Will the dimensions of the mirrors need to change when you go to make a chip of higher resolution than 600 x 848, or will you simply enlarge the chip to accommodate, say, 1024 x 1280 mirrors.
Yoder: We could go either way. The new .3 semiconductor standards would permit us to make smaller mirrors or, as you suggest, we could keep the current size and make the chip bigger.
Da-Lite: Given what must be the great difficulty of concentrating large amounts of light onto a surface the size of a postage stamp, wouldn't a larger chip give a greater collection efficiency?
Yoder: Yes, but larger chips mean a smaller number of them can be created from a standard wafer. But that cost question aside, there may well be high brightness and professional applications where we will want bigger chips because we want to collect more light.
As you've seen, it's possible to achieve over 3,000 lumens on the current size chip, 848 by 600. So, if you do the math, you discover that chip to be about .67 inches in diagonal. Now when we go to a 1280 by 1024 array of the same sized mirrors, we jump to a chip with a 1.1 inch diagonal - which is close to four times the surface area. Obviously that will enable us to collect far more light.
Da-Lite: How bright do you think projectors using three of those chips could get?
Yoder: We're anticipating 10,000 lumens and higher.
Da-Lite: And what principal applications do you see for a projector that bright?
Yoder: Certainly the most prominent is large screen video projection in movie theaters.
Da-Lite: You mean as a replacement to film?
Yoder: Indeed I do. Here at TI we are doing a lot of work to quantify what does film give you today and to what extent can an audience tell the difference between film and electronic cinema.
Da-Lite: A DMD prototype projector was demonstrated adjacent to Infocomm this year and the quality of its video images was extremely impressive. But was it really as good as film?
Yoder: We know there's still a way to go and what was shown in Philadelphia in some critics' opinion wasn't quite there. But we're still moving forward and I believe we will get there.
Da-Lite: You believe that you can achieve the resolution, the brightness, and the color saturation of a fresh print shown in a first run movie movie house?
Yoder: Yes. Think about the phrase "fresh print."
Go to a movie after six weeks and see how fresh the print is. Whereas electronic projection is constant; constant beautiful performance. No hairline scratches, no dust, nothing. What happens when you take away all of those film artifacts while at the same time you don't introduce any digital artifacts? You have a better picture. You talk about getting to where film is?
We just might surpass where film is.