How Optical MEMS Won an Oscar
If you want to watch the video first, here it is:
Optical MEMS are a special subset of MEMS - Microelectromechanical Systems - that steer, modulate or direct light.
Today, people are trying to use optical MEMS for products like LiDAR. There is a lot of potential there.
But the first major commercially successful optical MEMS product was the DMD. A system made up of millions of extremely tiny mirrors moving in concert to reflect high powered light.
Today, optical MEMS-based systems dominate the high-end theater projection industry.
For this video, let us take a look at the micro-mirror array that took over the image projection industry. And won an Oscar to boot.
The bigger the room you need to project in, the more light your system has to push through. These needs have to be balanced by cost, size, heat, and weight concerns.
For some of the largest rooms - theaters, basically - we need a system capable of producing a picture that has to be very bright - as in thousands of lumens - while at the same time retaining very sharp details.
Various projection systems have been in use throughout the century. But For 50 years, the primary system for an extremely sharp and bright image in a very big room was an oil film system like the Eidophor projector.
This projector has a fascinating history. I want to credit Mike Harrison from the YouTube channel Mikeselectricstuff for doing some amazing work on it. I recommend his talk at Hackaday for further information.
The original concept was dreamt up by the Swiss engineer Fritz Fischer, then at the Swiss Federal Institute of Technology in Zurich.
Fischer was studying a projector proposal from the optician Bernhard Schmidt. Schmidt's system proposes using a spherical mirror to collect the light from an extremely bright Cathode Ray Tube to project an image onto a movie theater screen.
Fischer did not believe that cathode ray tubes can ever supply enough light to illuminate an entire movie theater. If they tried, the resulting projected image would not be crisp.
Legend has it that Fischer sketched out the new concept on the back of an empty cigarette box while commuting to Zurich via train. Convinced that it would work, he eventually set about to build an industrial prototype.
He named it Eidophor, which means Image Bearer in Greek.
How Eidophor Worked
Inside the projector, you have a light source and a light valve to vary and modify the light.
The light valve in this case was a mirrored disk - later a concave mirror - covered in a thin film of oil about 0.02 millimeters thick.
When the system is in use, an electron gun fires a beam of electrons at the mirrored disk, affecting the oil and deforming its surface.
The Eidophor system used a very powerful light source - the Super Ventarc lamp. It is basically, like, the Sun. The Super Ventarc was a carbon arc lamp capable of producing a stunning 375,000 lumens.
When the light passes through a deformed section of the oil film, it diffracts off its normal path and enters a system of mirrors before entering the projection optics and finally out to the screen.
Fischer and his team eventually built two prototypes, but on Christmas 1947 Fischer died unexpectedly from a heart attack at the age of 47.
Without Fischer, work on the Eidophor languished until it caught the eye of his friend Dr. Edgar Gretener.
Gretener owned a company - later to be called Gretag - that seemed to specialize in this weird engineering stuff. Other things Gretag owned was a carbon arc lamp and a cryptography engine.
He saw that the Eidophor had a lot of potential and in 1948 he purchased the rights to the invention. But it needed some finishing touches like projecting in color. So Gretener licensed a disk-based RGB color system from CBS and integrated it into the Eidophor.
Gretener took the product to the American movie industry. Movie executives looked at the super-bright 15 foot movie screen and immediately saw a weapon they can use against the growing television industry.
20th Century Fox purchased the first production Eidophor and installed it in the Pilgrim Movie Theater in New York City. In 1953, LIFE magazine wrote it up, calling it the "15-foot weapon against TV".
Over time and for decades the Eidophor became your best option for any projection screen where you needed to deliver 4,000 to 10,000 lumens to the customer.
They were used in the NASA space program, universities for lectures, religious events, and TV studios.
The closest competitor was probably the GE Talaria projector, which worked on similar principles to the Eidophor. GE built the Talaria as their own version after working with Gretag for 3 years.
Overall, the later Eidophor systems were reliable and easy to maintain. Gretag produced Eidophors with vacuum tubes, germanium transistors, and finally silicon integrated chips. As late as 1991, there were 600 units in use around the world.
That being said, they were massive, expensive ($500K!), and over-engineered. The electron gun itself was a formidable piece of work and at one point required a vacuum to operate.
There were also technical issues with the oil. Maintenance issues associated with bombarding an oil film with electrons for so long.
People felt that that might be some opportunity to replace this awkward oil film system.
Mirror Matrix Tube
In 1975, RN Thomas, Jens Guldberg and Harvey Nathanson - inventor of the first MEMS - published their work announcing the first micro-mirror array.
At the time referred to as a Mirror-Matrix Tube. Fabricated on top of a sapphire substrate, the array featured aluminum-coated mirrors about 50 micrometers wide and 450 nanometers thick. There were about 500 mirrors per square inch.
The team described this device as a new "light valve" system that can potentially be used to project images onto screens of varying sizes.
This first identified the projection display market as a very large potential market for an optical MEMS system to enter. Interestingly enough, the company that eventually entered this market never intended to do at the start.
In 1977, the Defense Department hired Texas Instruments to research a deformable-mirror MEMS for modifying or controlling light. The MEMS was to be used for optical computing.
TI chose Dr. Larry Hornbeck to lead this research team. And after a few years, the team eventually produced what they called then the Deformable Mirror Device, or DMD.
The original DMD was an analog thing. An early patent in 1980 describes how a transistor can trigger an analog voltage that can deform a suspended metal-ish membrane.
This concept would greatly evolve over the years. Manufacturing concerns led the team to turn the membrane into a discrete micro-mirror with a sort of cantilever design.
The TI team envisioned that it could be used to build a better copier or printer. But how to control it? The team found it difficult to consistently control all these micro-mirrors using analog signals.
Despite working with very small arrays of just 2,400 mirrors and pushing a lot of voltage through them, the mirrors did not reflect light together in a coordinated fashion.
Hornbeck and his team struggled with this problem for a decade, but the system never met expectations. So they pivoted from using analog signals to digital ones.
Digital, meaning 1 or 0. The mirror can either be ON or OFF.
When it is ON, it turns and reflects light onto a specific spot. When it is OFF, it turns off and no more light.
Simple as that. It kind of reminds me of those card stunts they pull at football games. And there is a piece of folklore floating around saying that was what inspired the whole thing. Larry Hornbeck has denied it, though.
The DMD abbreviation was later retconned from "Deformable Mirror Device" to "Digital Micro-mirror Device".
How it Works
Fabbed using conventional CMOS processes, the micro-mirror device is made up of many thousands of mirrors. The original had about 840 micro-mirrors, but modern DMDs can have millions.
Each aluminum alloy mirror spans about 16 micrometers on each side.
The mirror is mounted on top of a yoke, which is then connected to two torsion hinges.
At the very bottom is a SRAM memory cell which controls the individual mirror's behavior.
When activated, electrostatic forces cause the mirror to tilt about 10-12 degrees.
One of the most exciting things about the modern DMD is just how reliable it is. Those are mechanical hinges and yokes under each individual mirror. Yet the mirror has to flip billions or even trillions of times.
Furthermore, conditions can be quite harsh. Theater-quality light sources can get very hot. The massive light source for the old Eidophor projector produced so much infrared heat that if you were to hold a piece of wood into the light beam, it would burst into flames.
The original DMD lasted just 100 hours at 65 degrees. TI had to develop entirely new testing equipment to inspect and test for failures. Today, modern hinges can go through over 3 trillion mirror cycles without failing.
Texas Instruments first used the DMD for printing airline tickets. These used to be done with red-colored carbon copy paper, but were switching over to digital paper printouts.
TI used a micro-mirror array of 840 mirrors lined up in a single row to do fast printing: the DMD2000 airline ticket printer.
Then in 1989, DARPA set up a multi-million dollar research initiative to explore future HDTV technology. TI signed on and created a high-definition DMD chip. Ultimately though, going the television route at the time didn't make sense - everything back then was still analog.
However, a subsidiary of Rank Corporation, a British film conglomerate, liked what they saw and commissioned TI to help them build a projection system using DMD technology.
Hornbeck and his team immediately abandoned the printer idea and started retooling their work to better suit the projection industry. TI built a dedicated wafer fab to produce all those DMD optical MEMS.
A New Generation
TI started producing this new technology at the right time. A new generation of technologies were sweeping through the projection industry - starting with LCD.
LCD projection systems - which work by sending bright light through poly silicon panels - had been in development ever since the 1960s. For instance in 1976, Westinghouse had a 6 inch by 6 inch LCD display.
But it was not until the late 1980s when the technology finally began to become competitive. In 1986, Seiko-Epson released an LCD projector with a resolution of 320x220 pixels and a 70 lumen output.
Certainly nothing for the Eidophor to be afraid of, right? However, these technologies continued evolving as the industry obviously wanted a successor technology to these old and extremely expensive light-valve projection systems.
In 1992, the BarcoData 5000 - named the Light Cannon - came onto the market. This was the first LCD projector bright enough to match up with the oil film systems, with a pixel count of 756 by 556 and putting out 1,250 lumens.
Light Cannon warmed up far faster than an oil film system and was disruptively cheaper too. A Talaria capable of 1,250 lumens cost about $80,000-$90,000. Light Cannon on the other hand cost just $47,000.
GE was so concerned about this new product that they bussed a third of their division employees to go see it in action. A flood of new LCD projectors soon entered the market, forcing GE to exit the business in 1993.
The Eidophor quickly followed. The larger company Gretag went out of business in 2002. The last Eidophor was removed from service in 2000 - a rapid decline from 600 in less than a decade.
Similar transitions occurred in the consumer, business, and home theater markets - with little LCD projectors finally replacing older incumbent CRT technology.
So by 1995, the $4.6 billion projection market was going through a massive technology transition. So it made a lot of sense for what TI would call the DLP chipset sub-system to make a splash.
TI eventually produced two versions of the DLP sub-system - a one-chip and three-chip projector.
The one-chip version came out first in 1996 for the business conference room projector market.
Here is how it worked. First, the input signal is broken down into three RGB versions - red, green, and blue.
White light is then shone through a spinning color wheel to create a colored light, which then falls onto the DMD. Let's say red for this example.
At that exact moment, the DMD flips over its micro-mirrors to reflect back the red version of the split input signal.
That primary color image then gets flashed onto the wall or whatever for 1/60th of a second.
Your slow and incompetent human visual system is too slow to perceive the individual primary color.
So instead it mashes the string of single color images into a full colored image.
TI worked with projector OEMs like InFocus Systems, nView, and Proxima, providing component kits for the OEMs to incorporate into their products. Kind of like how Nvidia provides the chips for its board partners to make into graphics cards.
DLP For the Theater
Then in 1997, TI released its line of high-brightness DLP sub-systems, which use three DMD chipsets.
The subsystem uses a prism to split white light into three primary colors which are then routed to its color-specific DMD chip. This allows for much greater perceived brightness, since there is no color wheel so each color gets its own continuous light stream.
Because the DMD chipset is a light valve that relies on reflecting light rather than filtering it like with an LCD projector, more light from the light source eventually makes it to the wall.
In some cases, the overall light efficiency can get over 60% - which means a brighter image.
These digital projectors arrived at the perfect time, with the movie industry making the transition to digital. They worked closely with the major Hollywood studios to round out the specs, tuning the product for blacker blacks and higher resolutions.
In June 1999, the DLP Cinema system displayed its first full-length feature film - "Star Wars - the Phantom Menace".
George Lucas ran a head-to-head bake off between a DLP system and an LCD projector from the leading competitor at the time - a perhaps bizarre joint venture between Japan's JVC and the American Hughes Aircraft Company called HJT.
DLP won the test with HJT's projectors showing unstable color reproduction and streaking artifacts. The industry rapidly shifted to DLP and HJT has since receded from the theater market.
Today DLP optical MEMS power a significant portion of our cinema theaters, with laser projection as probably its most significant competing technology. It has become one of TI's more profitable businesses.
In 1998, DLP and Texas Instruments received an Emmy for their work.
In 2015, Larry Hornbeck won an Oscar for his work in developing the DMD chip.
Optical MEMS still have much to do for humanity down the line. But we should appreciate the massive impact the space has already had on our everyday life experiences - at least in this tiny aspect of it.