Copyright 2003 Society of Photo-Optical Instrumentation Engineers.This paper was published in SPIE Proceedings Vol. 4980 and is made available as an electronic reprint with permission of SPIE. One print orelectronic copy may be made for personal use only. Systematic or multiple reproduction, distribution to multiple locations via electronic or othermeans, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.Reliability, Testing, and Characterization of MEMS/MOEMS II, Rajeshuni Ramesham, Danelle M. Tanner,Editors, Proceedings of SPIE Vol. 4980 (2003)  2003 SPIE - 0277-786X/03/$15.00DMD reliability: a MEMS success storyMichael R. DouglassTexas Instruments, P. O. Box 869305, MS 8477, Plano, Texas 75086214/567-6774; 214/567-5454 (fax); e-mail: [email protected] Digital Micromirror Device (DMD) developed by Texas Instruments (TI) has made tremendous progress in bothperformance and reliability since it was first invented in 1987. From the first working concept of a bistable mirror, theDMD is now providing high-brightness, high-contrast, and high-reliability in over 1,500,000 projectors using DigitalLight Processing TM technology. In early 2000, TI introduced the first DMD chip with a smaller mirror (14-micronpitch versus 17-micron pitch). This allowed a greater number of high-resolution DMD chips per wafer, thus providingan increased output capacity as well as the flexibility to use existing package designs. By using existing packagedesigns, subsequent DMDs cost less as well as met our customers' demand for faster time to market.In recent years, the DMD achieved the status of being a commercially successful MEMS device. It reached this statusby the efforts of hundreds of individuals working toward a common goal over many years. Neither textbooks nor designguidelines existed at the time. There was little infrastructure in place to support such a large endeavor. The knowledgewe gained through our characterization and testing was all we had available to us through the first few years ofdevelopment. Reliability was only a goal in 1992 when production development activity started; a goal that manythroughout the industry and even within Texas Instruments doubted the DMD could achieve. The results presented inthis paper demonstrate that we succeeded by exceeding the reliability goals.Keywords: DLP™, DMD, MEMS, testing, characterization, reliability, picture reliability1. INTRODUCTIONThe Texas Instruments DMD has achieved a performance level that in some cases exceeded its reliability goals. Forevery new DMD as well as for each major design change, Texas Instruments performs a detailed failure modes andeffects analysis (FMEA). This process assures that all subsequent designs achieve the same high standards forreliability and performance.Because of the testing and characterization efforts expended since 1992, projectors based on DLP™ technologydemonstrate reliability and lifetime superior to competitive technologies. A lifetime estimate of over 100,000 operatinghours with no degradation in image quality is the norm. As evidence, the TI reliability department performs ongoinglife tests of both DLP™ subsystems and DMD chips. Large screen televisions continue to operate in the lab for over10,000 hours with no defects and no image artifacts. Small, portable, and lightweight conference room projectorsoperated in our reliability lab for over 26,000 hours with no added defects or image degradation. Nine (9) DMDs,placed on test in December 1995, operated for over 56,500 hours and over 3x1012 (trillion) mirror cycles (the equivalentof over 100 years of typical office projector applications) with no added defects. These demonstrated results, pairedwith modeling predictions, support the conclusion that the DMD is exceptionally robust and reliable. For example:• DMD MTBF > 650,000 hours• DMD lifetime > 100,000 hours• Hinge lifetime > 3x1012 mirror cycles (equivalent to >120,000 operating hours)• Environmentally robustThis paper will highlight some of the DMD-specific metrology, including, characterization tests (normal, accelerated,and environmental), unique DMD life tests, test equipment development, packaging, modeling and failure analysis. Thepaper will also discuss how characterization tests are essential to achieving our reliability goals.Proc. of SPIE Vol. 4980 - 2 -2. SETTING AGGRESSIVE, ATTAINABLE GOALSTexas Instruments invented the DMD in 1987.The concept was refined through the next fewyears and entered full-scale product developmentin early 19921. At the time, Texas Instrumentsanticipated that the DMD provided superior image quality due to its digital operation and reflective approach tomodulating light. Unknown was how long the DMD maintained its image quality and how long it operated beforefailing. The first commercially produced DMD consisted of 840 micromirrors in a linear array. Its application was in alow-resolution printer. Numerous other potential applications existed for the DMD, ranging from printers to high-definition TVs to telecommunications to movie projectors. Understanding each market's unique needs and aligning ourgoals to satisfy these needs was step one on the road to developing reliability. Every market being considered by TexasInstruments had reliability as a priority. In order to enter each market, this concern for reliability had to be addressed.Some of the earliest applications required only 5000 hours but at high temperatures. Although at the time DMDs couldonly function for about 100 hours at 65oC before failing, we established what appeared to be the very aggressive goal of5000 hours at the maximum operating temperature of 65oC. As an organization, we agreed that we would not startshipping products until we achieved this minimum goal. All teams associated with DMD development focussed onachieving this goal.In addition to the minimum goal, the product development team understood that future markets, such as home theater,consumer television, business projectors, and telecommunications, had much higher expectations for reliability and lifetime. Therefore, a secondary goal was to assure the DMD was capable of supporting these applications where lifetimesof 50,000 to 100,000 hours would be considered more typical. If the DMD were to achieve these two goals, not onlywould it meet the market needs but also it would provide