Visual Analysis of a

Texas Instruments

Digital Micromirror Device



John D. Jackson





The Institute of Optics

University of Rochester









Site best viewed with IE, 1280 or 1024 horizontal resolution.


Overview of the technology



The Digital Micromirror Device (DMD) is the micro-optical-electrical-mechanical system (MOEMS) which drives Texas Instruments’ Digital Light Processing technology. It is a reflective spatial light modulator which can digitally control the direction of reflected light with a rectangular array of individually addressable square micromirrors. Each mirror can tip toward or away from a light source in a matter of microseconds. Thus, not only can high resolution moving pictures be realized, but the temporal duty cycle can produce over 1000 shades of gray. The main application for DMDs is front and rear projection for business and entertainment. However, research in many fields of science such as biomedical optics utilizes the chip as well.



A simplified setup for projection with a DMD is shown in Figures 1 and 2. The major components are an illumination system, a video front-end (which accepts signals), the DMD chip mounted on an electronic driver which has a digital signal processor and a digital formatter, and collection/projection optics. The illumination system can be very complicated in itself, using custom reflectors, integrating rods, prisms, and lenses. One chip systems also usually employ a color wheel to generate color frames. They actually only project one color at a time, but do so fast enough so that our eyes can integrate the colors for each frame. Larger systems, such as those used in theater and other large venue projectors, have three chips, one designated for each primary color, and a large, complicated prism to combine the beams reflecting from the three chips. DMDs are made in most common resolutions: SVGA, XGA, SXGA, and different HD models. The DLP website has very nice multimedia presentation which talks more about the technology. Click on “Launch our demo” on the right side of the page.




Figure 1 – One chip DLP projection


Figure 2 – Digital reflection by DMD pixels




The microstructure of the mirrors on the DMD, as diagrammed in Figures 3 and 4, is built up by several sequences of deposition and etching of materials on typical silicon wafers. Aluminum is used for the actual reflective surface because it is cheap and has good reflectivity performance across the visible range. The post which holds the mirror is attached to what is called the “yoke.” It is the yoke which rotates with the torsion hinge rather than the mirror. The mirror address electrode and the yoke address electrode pull one side of the structure down simultaneously. On the feet of the yoke are spring tips to dampen the impact and facilitate transition to the opposite tilt.





Figure 3 – DMD pixel: exploded view






Figure 4 – 3D rendering of 2 tilted pixels,

showing microsctructure





Analysis in the SEM



            Preparation of a DMD for imaging in the scanning electron microscope is relatively simple. Most surfaces of interest are sufficiently conductive, so the largest obstacle is exposing the mirror array. There is a relatively thick piece of glass welded to the chip, usually called the “window.” (Fig 5)


Figure 5 – DMD side view; window encased in metal top


            Removal of the window was accomplished by a Dremel equipped with a metal grinding bit. This is certainly not the cleanest way to remove the window, because as the wheel cuts into the air space within the chip, it flings metal particles into the array. As a result, many areas of mass pixel destruction were observed in the microscope. One drawback of having a DMD array exposed air is that it collects dust very easily, especially when the chip is hooked up to a voltage, which will be discussed more in detail later.



Initial Images


Mirror array – Mirrors were measured with overlay markers to be 16 microns square.



Mirror array – By the end of the project, the entire array was littered with particles like this.



Mirror via (square in middle) showing texture of aluminum mirror surface.



Mirror at corner of array. Chamber stage tilted 30 degrees.



Colorized image of the microstructure under a mirror – Mirror and post have been ripped off the yoke, as evidenced by diamond-shaped residue in center. The yoke has fallen to the substrate, leaving the hinge connections severed.  



Tilted (30 degrees) view of (broken) pixel microsctructure. Contrast this pixel type with the colorized image above. This pixel is located in what is called the “sea of mirrors.” The mirrors do not tilt to communicate information, but to create a dark border around the portion of the DMD which is addressed to create images. They stay landed in one direction a majority of the time; thus, they don’t need spring tips or a bias reset bus.



Tilted (30 degrees) view showing the continuous address electrode sheet in the ‘sea of mirrors’ and the lack of a bias reset bus.



Closeup of a hinge post, missing a hinge, and again showing the texture of surrounding mirror surfaces.



Colorized foot of an electrode which is used to address the array



Mirror stuck in either ‘on’ or ‘off’ position (+- 10 degrees)

Mirror stuck in opposite position (-+ 10 degrees)



Impaled mirror

Mirror fracture by particle impact



Abrasion by paper clip – charging is evidence that some conductive layers were scraped off

Same area, magnified – notice the smearing of the surface coating on the bias-reset bus




In-chamber activation of the micromirror array


            Before removing the window, the DMD was tested with a 5V voltage supply in an attempt to find a combination of pins which would activate some portion of the micromirror array. Under normal room lighting and inspection by eye, it is evident when a portion of the array tilts. The goal was to catch this action within the SEM. The machine shop on campus (U of R) graciously made us a custom flange with a mount for a 7 pin cinch connector. Wires were soldered to the back of the DMD on appropriate leads and then connected to the cinch connector in the chamber. The pins on the outside were matched up by testing the resistance, and the appropriate wires were hooked up to a voltage supply. The following images show the results of trying to flip the mirrors in the SEM chamber during operation.



No bias

5V bias



5V bias addressing several adjacent rows of pixels

‘Sea of mirrors’ between the end of addressed region and electrodes.



Unaddressed pixel electrodes

Addressed pixel electrodes (Top left shows evidence of an electron sink)



This clip is a video recording which shows the effects of voltage triggering at many magnifications. Watch for contrast reversals of the electrodes as the triggering frequency changes. Also, see if you can detect any mirror movement. Jumping of the entire image is an artifact of current flux.


          While some images (like the zoomed out ones) appear to have tilted pixels, upon increasing the magnification, it becomes extremely difficult to detect mechanical movement of the mirrors. It is not clear whether they are just biasing or deflecting much less than +-10 degrees. It is quite perplexing, as tilt can be observed with the naked eye when addressing the device outside of the chamber.




Post-Processing Techniques


A DMD is the type of sample that is ideal for vivid colorization because it has very distinct parts. In MEMS technology, colorization is a particularly useful aid for identifying and labeling features .


1 intact pixel substructure, 2 broken. Area in ‘sea of mirrors,’ showing missing bias reset bus. Green section actually part of the address system.



Normal pixel with mirror, yoke, and most of hinges removed. Shows typical electrode layout on the CMOS. Green=bias reset; yellow=address system.



The images below show a considerably poor image (low contrast, high brightness), and a couple of many steps done to improve it. Image processing was performed on a PC with standard filters and effects in Jasc Paint Shop Pro 9. The end product is much better, but a taking a better picture to begin with would help, such as one without streaking from charging.

Original – pixel substructure (stage tilted 30 degrees), in ‘sea of mirrors’ area

After brightness decrease, contrast increase, “clarify”=5, gamma decrease

Colorized previous image






            Digital micromirror arrays are good candidates for analysis in a scanning electron microscope due to their conductivity and structural uniformity. If a marking or coordinate system is known, characterization of parts by electron microscopy is ideal because of the SEM’s superior resolution over light microscopy. One difficulty in working with unpackaged MEMS, however, is keeping them clean. Especially when passing current through the devices, they tend to attract dust very quickly. With an array of delicate micromirrors, good sample areas for imaging are lost every second of exposure.

            In order to successfully tilt the mirrors in a recognizable fashion, it would be beneficial to obtain the pin-out of the DMD chip to understand exactly where and how much voltage should be applied. The image is supposed to look like the image below, although it is unknown whether the mirrors are permanently stuck or they are being addressed in the chamber.


Landed mirrors (not my micrograph)


          I’d like to thank Brian for his continuous help (setting up the flange and video) and prodding, and infinite patience! Also, the numbered figures and the last image are from Texas Instruments’ DLP website and are free for public use.















Please enter any comments, criticisms, questions, etc. below.

Your name:

Email address: