A collection of interviews of Dr. Larry Hornbeck, innovator of TI Digital Light Processor technology.

EE Times

R. Colin Johnson, DLP Pioneer Tells how TI Did It with Mirrors  Direct link to the article in EE Times

Solid-state physicist and Texas Instruments Inc. fellow Larry Hornbeck won an Emmy for his invention of the digital micromirror device (DMD), the microelectromechanical system at the heart of TI's digital light processing (DLP) technology for projection displays. In addition to advancing the state of the art for digital cinema, front projectors and HDTV, Hornbeck's MEMS micromirror is enabling 3-D metrology systems that measure with finer detail, confocal microscopes that eliminate the out-of-focus "haze" normally seen around fluorescent samples, and holographic storage systems that write data in three dimensions instead of just two. Hornbeck told EE Times' R. Colin Johnson that TI has still more applications up its sleeve for digital micromirrors.

EE Times: Texas Instruments has perhaps the industry's longest history in MEMS development.
Larry Hornbeck: TI's history with MEMS goes back to 1977.

EE Times: Back then, you couldn't have been aiming at digital light processing. Were you even focusing on an application?
Hornbeck: Even back then, we were interested in using MEMS to modulate light. I had only been at TI for a few years after receiving my PhD from Case Western University as a solid-state physicist. We got started in MEMS back then because of a Defense Department contract to make a spatial light modulator using deformable mirrors. This was an analog technology that the DOD wanted to use for optical computing.

EE Times: I guess you had to invent the fabrication techniques you needed to even get started with MEMS.
Hornbeck: In 1981, MEMS meant bulk micromachining of single-crystal silicon, which made the devices expensive to manufacture. But the universities were already experimenting with surface micromachining of polysilicon, which was much more economical and has become the traditional way of doing MEMS today.

EE Times: As your micromirrors became more successful, did you dedicate a fab to MEMS, as Analog Devices did in Cambridge [Mass.]?
Hornbeck: We've never needed a special fab but have always made our MEMS fabrication compatible with our conventional CMOS manufacturing areas. We finish out all the transistors and metallization layers for interconnecting the transistors, and then we use a low-temperature process to put MEMS on top of the completed CMOS chip.

EE Times: I see; you finish the whole chip but leave an area open to add the MEMS last.
Hornbeck: That's a radical departure from the way others do MEMS—I believe we are still the only company that does it this way. The way I implemented the process was by choosing aluminum alloys for the mechanical elements and conventional photoresist to act as a sacrificial spacer. All of this is done at temperatures under 200°C so that metallization is not affected, the transistors are not affected, none of the finished CMOS circuitry is affected when we add MEMS to a chip. To this day, this forms our standard basis of manufacturing MEMS micromirrors, but it was a radical departure at the time.

EE Times: Only Bosch Sensortec and its spin-off SiTime seem to put their MEMS structures down first—using high-temperature processing, which they swear by. But most of the other MEMS startups, such as Discera and Silicon Clocks, have gone to putting the MEMS down last—using either polysilicon or, in Silicon Clocks' case, silicon germanium. And whether they favor MEMS-first or MEMS-last, all the MEMS vendors seem to agree that the holy grail is seamless integration of MEMS structures onto the same CMOS chips as the circuitry to which they interface. Akustica claims to be there already with its MEMS microphone, though they do add a step at the end to remove sacrificial oxide layers. All told, I think you were way ahead by integrating MEMS with CMOS from the beginning.
Hornbeck: We think so. This is one of the pillars we are resting on as a basis for our success in DLP.

'Second radical departure'
EE Times: But in the beginning, you were trying to make analog micromirrors.
Hornbeck: Yes, we struggled for several years trying to get enough uniformity and optical efficiency to do simple xerographic printing with a linear array of 2,400 analog micromirrors. But by 1986, it became apparent we were not going to be successful. The uniformity just wasn't there, our analog voltages were too high—as big as 30 volts—and still there wasn't enough mirror deflection angle. This was all because we were trying to make analog micromirrors. So the second radical departure to anything that anyone had done before was to go digital.

EE Times: What year was that?
Hornbeck: I invented the digital micromirror device in 1987 and applied for a patent that, when issued, formed the basis for all subsequent DMD architectures. Instead of continuing to develop analog MEMS micromirrors that depended upon a delicate balance between electrostatic attractive forces and the restoring forces of a flexure, I developed a micromirror that would flip between two digital states, where contact was made to stop the micromirror in the positive and negative directions. This technique made it easier to control the angles compared with our analog micromirrors, which had no stops.

EE Times: Are there other benefits to going digital?
Hornbeck: Yes. By operating the micromirrors in a bistable mode, we could go to much lower operating voltages since the micromirrors could be triggered into either stable state. So this new digital architecture enabled larger rotation angles with better uniformity and lower operating voltages compared with analog micromirrors. That's been the basis ever since for our success with MEMS.

EE Times: Did you apply the digital design to the page-printing application you mentioned earlier?
Hornbeck: Actually, the first commercialization of DMD was in an airline ticket printer. We had a very successful impact printer for airline tickets in those days, but the industry was converting from the old-style red carbon copies to individual coupons. Printing individual coupons upped the speed requirement, and our impact printers couldn't keep up. So to maintain market share, we decided to go to higher-speed xerographic printing. Instead of using a conventional polygon scanner, TI made the decision to use a linear DMD—an 840 x 1 array of micromirrors. The first product, the DMD2000 airline ticket printer, went to market in 1990.

Road to high def
EE Times: The HDTV application for DMD seems quite a stretch from the ticket printer application. What influenced TI to move in the HDTV direction?
Hornbeck: In 1989, the Defense Advanced Research Projects Agency started an initiative to spur the development of HDTV technology in the U.S., and TI was awarded a multimillion-dollar contract to develop a prototype high-definition DMD chip.

EE Times: And that helped you make the transition from printing to light projection?
Hornbeck: Well, that was only the beginning. Rank-Brimar, a subsidiary of the Rank Corp. in the U.K., was looking for a way to project high-definition TV in very large formats for theaters and auditoriums. In 1989, they invested money to help us develop prototype 3-chip DMD projectors.

By 1991, TI itself decided to start a corporate venture project where we brought together the critical mass of people and resources to fund our own initiative, which we called the Digital Imaging Venture Project. Its goal was to develop high-definition television, which sounds strange for 1991, because back then TV was analog, nobody was doing anything with MEMS at that level, there wasn't a product or a standard or anything. So what we decided to do was to go for digital high-definition TV, but to begin with projectors because we could already build them and had a customer. By 1996, we had our first DLP products.

EE Times: What kinds of projectors were available back then?
Hornbeck: In 1994, a traditional projector weighed 35 to 40 pounds, was relatively dim, and cost between $15,000 and $18,000. So we thought we could dramatically impact the weight, brightness and cost of projectors. Front projection became the means to our early success. We started off with just three customers in 1996—InFocus, nView and Proxima—and now we have 75. Epson had an LCD projector at that time that was our main competitor.

In those days, we provided a complete digital light engine to OEMs, because no engineers were familiar with designing digital light projectors. Now we provide chip sets and software to our OEMs. The chip sets consist of one or three DMDs, an ASIC for image processing and formatting, and a waveform chip to drive the DMD. The software, DLP Composer, allows OEMs to design customized projectors.

Today, DLP has about 50 percent of the worldwide front projection market, with over 350 products available. DLP is the market leader in 1,080p HDTV technology for displays 40 inches and above.

EE Times: How did TI get into the digital-cinema business?
Hornbeck: In 1997, the year after our introduction of business projectors to the market, we introduced the first high-brightness three-chip DLP systems for large-venue applications. DMD is naturally suited to these applications because of its ability to take the heat loads from very bright projection lamps. We started talking to moviemakers about what kind of technology would work for them, and eventually we won them over.

In the meantime, our three-chip systems were being used by the TV broadcast studios as monitors behind news anchors and for game shows because of their color stability.

And that's how it came about that both TI and myself were awarded Emmys in 1998 from the Academy of Televisions Arts & Sciences. TI got its Emmy for DLP TVs, and I got mine for the invention of digital micromirrors. I have my Emmy at home on a shelf. It makes for quite a conversation piece.

EE Times: What's next?
Hornbeck: TI has begun equipping digital cinemas with 3-D versions of its DLP Cinema projection technology, and at the other end of the spectrum it's showing a prototype of a tiny DLP chip that is small and cheap enough to add a projector to a handheld device, enabling relatively large displays to be projected from, say, a cell phone.

Regarding the future of digital micromirrors, just use your imagination. We are developing reference designs for almost anything a DLP system could possibly be used for, and we have several important announcements in new application areas that we plan to make very soon.

Der News-Service von
"Eigentlich wollte ich einen Fotokopierer erfinden"
Direct link to the article

1: Did you think about revolutionizing the cinema when starting the development of the micromirror display?
Larry Hornbeck: No, I didn’t. When I invented the DMD(tm) microchip technology in 1987, my intent was to produce a chip-based digital light modulator for electrophotographic (or xerographic) printing applications. The first experimental chip contained 512 pixels (digital micromirrors) and the first production version contained 840 pixels. The number of pixels required for DLP Cinema(tm) projectors is more than 1000 times greater, i.e. millions of pixels are required.
Although I had thought about the possibility of using the DMD(tm) microchip for display applications, it was beyond my imagination to conceive that we would ever be able to manufacture the number of pixels or achieve the image quality needed for digital cinema applications. Of course thirteen years of intense development by Texas Instruments transformed the inconceivable into a practical reality.

2: Didn’t people say you were crazy making a projection system out of millions of moving mirrors?
Larry Hornbeck: Yes they did. The most interesting example of this occurred in August 1993 by an editorialist writing in a well known PC magazine. He had discovered that Texas Instruments was developing a new display technology based on tiny, movable mirrors. He described the ”dinky mirror” technology as ”the weirdest technology every invented.” He went on to say ”I found the whole thing so bizarre and weird that, to me, it proves that what the nutballs have been saying is true: The U.S. government has captured some bug-eyed aliens. We are using them to design this stuff.”
Needless to say, I found his article very stimulating and posted it outside my office in hopes of creating even more determination by our engineers to bring to market the first ever, all-digital, projection display products.

3: Is a DLP(tm) projector superior to 35 mm film in any respect or are there some points where film is better?
Larry Hornbeck: In terms of image quality, DLP Cinema(tm) projectors have been described as having an image quality as good as film. We are delighted with that appraisal, but our goal is to continue to develop the technology such that the DLP Cinema(tm) projector image is superior to film.
There are two areas where DLP Cinema(tm) projectors are superior to film projectors, mechanical registration artifacts and film degradation. Let me elaborate.
a) The DLP Cinema(tm) projector image is free of the mechanical frame-to-frame registration artifacts inherent in a film projector. These artifacts include jump and weave, resulting in an image that moves erratically on the screen. (Remember, the basic projection mechanism for film was developed in the 1890s and has changed little over the course of the last 100 years!). This film-projector mechanism, operating at 24 frames-per-second, uses a mechanical shutter & pull-down assembly to spatially locate each image of the perforated film stock to a fixed position on the screen.) Translated into audience experience the DLP Cinema(tm) projector image has no registration artifacts so the image has a natural look to it with no extraneous motion.
DLP Cinema(tm) field trials are currently being conducted in thirty-one theaters around the world, including one in Dallas. One of the things I like to do is to watch the digital version of a movie first and then for a short time, immediately afterwards, watch the film-based version. As soon as I enter the theater with the film version, I am always startled by the erratic motion of the image on the screen. I guess in the past I had no basis of comparison, so I got used to this fundamental film limitation.
b) The DLP Cinema(tm) projector image is stable with time. The first showing looks as good as the last. Because the projector is all-digital from the digital master all the way to your eye, whatever the colorist sees in the digital mastering studio is what the audience sees, again and again. How many times have you been to see a movie two or three weeks after opening night only to be disappointed by faded colors, scratches and other marks on the film? I remember seeing a block buster movie last year where the film was badly degraded. One of the emotional highs of the movie occurred at night and you could see in the dark sky, thousands of tiny white specks moving through the scene. My wife, who usually ignores film artifacts, commented to me later that the condition of the film really ruined the emotional impact for her.

4: What are the next big steps in DLP technology - daylight projection, Imax, laser, 3D, near-to-eye-displays?
Larry Hornbeck: DLP(tm) projectors are uniquely adaptable from ultralight applications to ultrabright applications. The near term ”next big steps” are direct results of this adaptability. In the near future we will continue to see DLP (tm) projectors lead the way in mobile applications. DLP(tm) projector technology has driven the definition of two new classes of mobile products within the industry, first the ultraportable and recently the microportable class, projectors weighing under three pounds but having the brightness of heavier projectors. Only four years ago, portable projectors weighed more than 20 pounds! The trend to ever lighter DLP(tm) mobile projectors will continue, bringing more and more business travelers into the ranks of the so-called ”road warriors.”
On the high-brightness end of the spectrum, DLP Cinema(tm) field trials are under way to aid the movie industry in establishing standards, technology and business models for digital mastering, distribution and exhibition of movies. The next big step for Texas Instruments and its three partners who will be manufacturing the cinema projectors will be the gradual process of replacing in theaters around the world, a more than one-hundred year old technology. This to me is the most exciting, next big step that we could ever take, dramatically improving the entertainment experience for the movie-going audience.
In between ultralight and ultrabright, the next big step is high-definition TV. DLP(tm) high-definition rear projection displays will be introduced by three manufacturers later this year. The advantages of superior image quality compared to CRT-based rear projection displays and a thinner cabinet profile, will bring the all-digital advantage to home viewers.

5: Do you have a DLP(tm) projector at home?
Larry Hornbeck: Well ... no!
But I do have a 19-inch CRT television made in 1980 by one of the companies that will introduce DLP(tm) high-definition TV projectors to the market later this year. My 1980 TV has an analog tuner, the brightness and contrast controls hardly work, and the colors are never quite right. If this sounds strange that the ”father of DLP(tm) projection technology” doesn’t even have a ”modern” TV, let me just say that my personality is as digital as the micromirrors I invented. I stick with the old for a long time and then suddenly switch to the new. Maybe it’s time to switch?

EE TIMES Asia
Colin Johnson
Direct link to the article in EE Times

Solid-state physicist and Texas Instruments (TI) Fellow Larry Hornbeck won an Emmy Award for his invention of the digital micromirror device (DMD), the MEMS at the heart of TI's digital light processing (DLP) technology for projection displays. In addition to advancing the state of the art for digital cinema, front projectors and HDTV, Hornbeck's MEMS micromirror is enabling 3D metrology systems. It is also enabling confocal microscopes that eliminate the out-of-focus "haze" normally seen around fluorescent samples and holographic storage systems that write data in three dimensions instead of just two. Hornbeck told EE Times that TI has still more applications up its sleeve for the digital micromirror.

EE Times: When TI began its MEMS development in 1997, were you focusing on a particular application, such as light processing?
Larry Hornbeck: Even back then, we were interested in using MEMS to modulate light. We got started in MEMS back then because of a U.S. Department of Defense (DOD) contract to make a spatial-light modulator using deformable mirrors. This was an analog technology that the DOD wanted to use for optical computing.

I guess you had to invent the fabrication techniques you needed to even get started with MEMS.
In 1981, MEMS meant bulk micromachining of single-crystal silicon, which made the devices expensive to manufacture. But the universities were already experimenting with surface micromachining of polysilicon, which was much more economical and has become the traditional way of doing MEMS today.

As your micromirrors became more successful, did you dedicate a fab to MEMS?
We've never needed a special fab, but have always made our MEMS fabrication compatible with our conventional CMOS manufacturing areas. We finish out all the transistors and metallization layers for interconnecting the transistors, and then we use a low-temperature process to put MEMS on top of the completed CMOS chip.

So you finish the whole chip, but leave an area open to add the MEMS last?
That's a radical departure from the way others do MEMS. I believe that we are still the only company that does it this way. The way I implemented the process was by choosing aluminum alloys for the mechanical elements and conventional photoresist to act as a sacrificial spacer. All of this is done at temperatures under 200°C so that the metallization, transistors or any of the finished CMOS circuitry are not affected when we add MEMS to a chip. To this day, this forms our standard of manufacturing MEMS micromirrors, but it was a radical departure at that time.

For MEMS vendors, the holy grail is seamless integration of MEMS structures onto the same CMOS chips as the circuitry to which they interface. I think TI was ahead by integrating MEMS with CMOS from the start.
This is one of the pillars we are resting on as a basis for our success in DLP.

But at first, you were trying to make analog micromirrors.
Yes, we struggled for several years trying to get enough uniformity and optical efficiency to do simple xerographic printing with a linear array of 2,400 analog micromirrors. But by 1986, it became apparent that we were not going to be successful. The uniformity just wasn't there, our analog voltages were too high (as much as 30V), and there still wasn't enough of a mirror deflection angle. This was all because we were trying to make analog micromirrors. So the second radical departure from anything that anyone had done before was to go digital.

What year was that?
I invented the DMD in 1987 and applied for a patent that when issued, formed the basis for all subsequent DMD architectures. Instead of continuing to develop analog MEMS micromirrors that depended upon a delicate balance between electrostatic attractive forces and the restoring forces of a flexure, I developed a micromirror that would flip between two digital states where contact was made to stop the micromirror in the positive and negative directions. This technique made it easier to control the angles compared with our analog micromirrors, which had no stops.

Are there other benefits to going digital?
Yes. By operating the micromirrors in a bistable mode, we could go to much lower operating voltages, since the micromirrors could be triggered into either stable state. So this new digital architecture enabled larger rotation angles with better uniformity and lower operating voltages compared with analog micromirrors. That's been the basis ever since for our success with MEMS.

Did you apply the digital design to the page-printing application you mentioned earlier?
Actually, the first commercialization of DMD was in an airline ticket printer. We had a very successful impact printer for airline tickets in those days, but the industry was converting from the old-style red carbon copies to individual coupons. Printing individual coupons upped the speed requirement, and our impact printers couldn't keep up. So to maintain market share, we decided to go to higher-speed xerographic printing. Instead of using a conventional polygon scanner, TI decided to use a linear DMD—an 840 x 1 array of micromirrors. The first product, the DMD2000 airline ticket printer, went to market in 1990.

What influenced TI to move in the HDTV direction?
In 1989, the Defense Advanced Research Projects Agency started an initiative to spur the development of HDTV technology in the United States, and TI was awarded a multimillion-dollar contract to develop a prototype high-definition DMD chip.

That helped you in the transition from printing to light projection?
Well, that was only the beginning. Rank-Brimar, a Rank Corp. subsidiary, was looking for a way to project HDTV in very large formats for theaters and auditoriums. In 1989, they invested money to help us develop prototype three-chip DMD projectors. By 1991, TI itself decided to start a corporate venture project, where we brought together the critical mass of people and resources to fund our own initiative, which we called the Digital Imaging Venture Project. Its goal was to develop HDTV, which sounded strange in 1991 because back then, TV was analog, nobody was doing anything with MEMS at that level, and there wasn't a product, standard or anything. So what we decided to do was to go for digital HDTV, but to begin with projectors because we could already build them and had a customer. By 1996, we had our first DLP products.

How did TI get into the digital-cinema business?
In 1997, the year after we introduced business projectors to the market, we launched the first high-brightness three-chip DLP systems for large-venue applications. DMD is naturally suited to these applications because it can take the heat loads from very bright projection lamps. We started talking to movie makers about what kind of technology would work for them, and we eventually won them over.

In the meantime, our three-chip systems were being used by the TV broadcast studios as monitors behind news anchors and for game shows because of their color stability. And that's how it came about that both TI and I were awarded Emmys in 1998 by the Academy of Television Arts & Sciences. TI got its Emmy for DLP TVs, and I got mine for the invention of digital micromirrors. I have my Emmy at home on a shelf. It makes for quite a conversation piece.

What's next?
TI has begun equipping digital cinemas with 3D versions of its DLP Cinema projection technology. At the other end of the spectrum, it's showing a prototype of a tiny DLP chip that's small and cheap enough to add a projector to a handheld. This enables relatively large displays to be projected from, say, a cellphone.

Just use your imagination for the future of digital micromirrors. We are developing reference designs for almost anything a DLP system could possibly be used for. We have important announcements in new application areas that we plan to make very soon.