`
`INNOLUX CORPORATION v. PATENT OF SEMICONDUCTOR ENERGY
`
`LABORATORY CO., LTD.
`
`IPR2013—OOO66
`
`
`
`
`
`SOCIETY FOR INFORMATION . DISPLAY
`
`INTERNATIONAL SYMPOSIUM
`
`DIGEST OF TECHNICAL PAPERS
`
`.
`
`’ SAN 505E, CALIFORNIA
`
`VOLUME XXV
`SAN IC)SE MCENERY CONVENTION CENTER
`
`JUNE 14~16,1994
`
`
`
`31.2: FPD Test, Inspection, and Repair Technologies: State of the Art
`and Future Directions
`
`F. J. Henley, 7'. Knuth, H. Mui
`Photon Dynamics, lnc., Milpitas, CA
`
`ABSTRACT
`
`Flat—panel display (FPD) manufacturing requires
`sophisticated Test, Inspection, and Repair (TIR) equipment
`to become practical and cost effective. Today‘s state-of—the-
`art TlR equipment has advanced measurement technologies ‘
`to measure pixel performance in a non-contact manner with
`automated repair vectoring . This paper will introduce
`current equipment capabilities and outline the
`improvements in resolution, throughput, and cost of
`ownership required to fully integrate modern TIR
`equipment in FPD mass-production lines,
`
`I.
`
`Introduction
`
`The flat-panel industry is moving towards mass-production
`with products rivaling CRTs in picture quality, sharpness,
`and brightness. The leading technical approach for
`producing these flat—panels, Active-Matrix LCD
`(AMLCD), is currently experiencing yield problems,
`resulting in higher prices and lower than expected market
`penetration for high-definition flat-panel products. Yield
`continues to be a major problem preventing AMLCD
`manufacturers from serving their markets with products at
`a reasonable cost. Building redundant rows or columns
`similar to the RAM industry is not a viable option since the
`‘diSplay is visual and no pixels can be substituted. Other
`methods of redundancy often complicate the active plate
`and lower the yield gains.
`
`Industry analysts agree that should the cost of these flat—
`panel displays fall to the ¥50,000 level, an explosion in
`demand will be realized. The potential applications include
`laptop, notebook, PC, and workstation computers,
`monitors, data displays, consumer appliance displays, and
`televisions.
`
`1]. Current Mass-Production TIR Needs
`The need for better TIR equipment was realized in the late
`19805 when first generation AMLCD manufacturing lines
`were not yielding with reproducible results. Running such
`a complex process "open~loop" invited long and frequent
`line shut-downs, erratic display quality, and low overall
`yield. The early open/short probe systems borrowed from
`simple-matrix array manufacturing which manual repair
`had proven inadequate for modern FPD manufacturing.
`
`FPD mass—production lines need fast, reliable information
`to make go/nogo decisions in actual production and to
`analyze and correct any yield and cost issue which surfaced
`during manufacturing.
`
`MM
`
`
`
`
`i
`’ re:
`
` :i 9%.
`
`Figure 1: Simplified AMLCD manufacturing line including
`TIR equipment.
`
`The steps where TIR equipment can be utilized is shown in
`Figure 1. Although the AMLCD process is shown in the
`example, the fundamental use of TIR equipment as a
`manufacturing feedback and go/nogo tool holds for any
`FPD technology. in the figure, tesUinspection and
`sometimes repair is made available at each important
`manufacturing step. The intent is to "divide and conquer"
`the complex process through intermediate process data
`which correlate and estimate final display quality without
`waiting to complete the FPD. Significant cost savings are
`possible once information with sufficient reliability can
`decide production quality early in the process.
`
`A complete FPD TIR infrastructure as shown in Figure l is
`not available. The missing capabilities in Table l. The
`Table shows that ARPA and USDC has launched certain
`
`programs and request for proposals (RFPs) to support core
`technology development and its integration in practical
`TIR equipment. To date, most funded work has yielded
`new approaches which promise mass-production
`capabilities in areas such as array test and repair. Most
`actual equipment from these efforts are pilot production
`capable but need more work to yield true mass-production
`capabilities.
`
`ISSN0097-0966X/94/2501-0687-$1.00 + .00 © 1994 SID
`
`SID 94 DIGEST . 687
`
`
`
`
`
`
`
`Through
`
`
`
`mation
`
`
`
`_-
`
`
`
`Test
`Coslof
`
`Pro--ram CoveXra;e Ownexrshi
`1. Glass
`‘
`
`Ins ection
`
`
`
`2a. Pattern
`Test
`
`2h.
`
`Particulate
`Inspection
`None
`2c Plate
`
`
`
`
`
`
`
`
`
`
`
`
`
`Indirect
`Measure-
`menl
`
`None
`
`ARPA
`
`X
`
`O
`
`X
`
`O
`
`0
`
`O
`
`0
`
`
`
`X
`
`3. Array
`Open/
`Shon
`4. Array &
`Color
`Filter
`Re . air
`
`5. Color
`
`Filter
`
`Ins ection
`
`6. Cell
`
`Ins - cation
`
`
`7. Module
`ins - ection
`
`
`
`
`Other than 0/8 testers there are three classes of array
`testers: Optical, Electrical and Electro—Optic. The optical
`test systems survey the plate and flags defects (optically
`different areas). The main advantage ofthis technology is
`its general use as a process improvement tool. It has little
`test coverage; however, its high throughput is useful in
`mass-production test. The electrical array testers use more
`traditional high pin count probe cards with added
`electronics to infer indirectly the pixel performance by its
`electrical interaction with the data line. [1,2] Detection
`schemes can be time domain (IBM technique) or frequency
`domain (Genrad); but all must use large pin count probe
`frames Test coverage is higher than 0/8 but limited to
`specific FPD plate designs. Although requiring less pins
`for test, E-Beam test technology is also of the indirect
`measurement type. The electro--optic array tester measures
`pixel voltages directly through the use of a process known
`as Voltage ImagingT The technology measures tme
`voltage and therefore has the highest test coverage. The
`technique does not require high pincount probe frames,
`thereby significantly reducing changeover time and other
`related costs.[3]
`
`Cell and Module Inspection Systems
`Currently, most manufacturers employ human inspectors
`using equipment originally developed to inspect solar cells,
`the early predecessor to the modem or-Si TFT. Using
`computer generated test patterns and complicated probe
`card systems, operators typically spend 3~5 minutes
`visually inspecting displays for pixel defects, line defects,
`and other gross anomalies. Unfortunately the results are
`often inconsistent, subjective, and incomplete.
`Compounding this problem are the inherent limitations in
`the human visual system, especially when inspection time
`is limited. Figure 2 shows the correlation between
`inspection time and defect detection sensitivity.
`mom
`
`
`Gross
`
`Non—Uniformity
`
`
` Line Defect Region
`
`0.10m
`
`.
`Waite
`
`0,00»
`
`
`
`
`Bcsl
`_ Human
`Virion
`_ Machine
`Vision
`
`won
`
`0.1
`
`I
`.
`.
`Spllialquuancy(cycluIOag)
`- I I I '- - ' Trends in decreased human inspection time
`
`VGAXGA
`
`
`0: Approach has advantages in this area
`-: Approach is neutral in this area
`X: Approach has disadvantage or currently inadequate in
`this area
`
`It is expected that this core work can be extended to other
`TlR areas not currently being sewed, such as cell test
`technology being fundamentally applicable to module test
`or glass inspection technology being applicable to color
`filter and particulate inspection. In this respect, the current
`ARPA/USDC programs should yield a large portion of the
`TIR infrastructure if the chosen approach has broad
`applicability.
`
`Ill. Types ofTIR Equipment
`In-Process Array Test Systems
`
`Array test systems are to be distinguished from the simpler
`line test systems, usually called open/short testers (O/S
`testers). An array test system has the added capability of
`detecting not only line defects but also the functionality of
`the TFT pixel. O/S testers are usually not desired for
`serious AMLCD mass-production because of its inability
`to predict final display quality through pixel functional test
`and as a result cannot be used as an SPC tool.
`
`688 . SID 94 DIGEST
`
`Figure 2: Machine Vision vs. Human Visual Response in
`defect detection
`‘
`
`
`
`
`
`Further, display end-users are demanding that the LCD
`industry move from VGA to XGA, EWS, and beyond.
`Confronted with these challenges, it is clear that human
`inspection can no longer provide effective product
`inspection and manufacturing process control. Quietly,
`human vision is being replaced by machine vision. The
`cell and module inspection system currently can utilize one
`of the following four technologies: scanning, step/repeat
`camera, multiple cameras, or N—AliasingTM - single camera.
`
`Scanning technology scans the display much like a copying
`machine to detect cell/panel defects, The main
`.‘
`disadvantage is its inability to measure mura with precision
`because time variations on panels cause measurement
`artifacts. Another limitation is its perpendicular
`measurement geometry which cannot provide any viewing
`angle data.
`
`Step/Repeat camera systems use a high-magnification
`camera which is mechanically stepped across the display.
`High cost and complexity prohibits this system from being
`commercialized. Viewing angle changes and any panel
`drift makes mura detection and accurate measurement
`
`difficult. Due to the complexity and low throughput only
`R&D systems are available.
`
`The multiple camera approach can give the required
`resolution and accuracy with a cost and complexity
`penalty. Although taking all data simultaneously, viewing
`angle changes make mura detection and accurate
`measurement difficult.
`
`Single camera technology inherently provides higher
`system reliability, lower cost, and easier system calibration,
`alignment, and maintenance, as well as reducing system
`hardware and software complexity. Additionally, software
`techniques such as N-AliasingTM have been developed to
`enhance the ability of a single camera to substantially
`improve its spatial resolution detection. N-AliasingTM is a
`technique which significantly reduces the systematic error
`associated with the moire patterns generated by the overlap
`of the VGA pixels on the camera‘s CCD array, allowing a
`test system employing this technique to provide much
`higher spatial frequency resolution than a native camera.
`N-AliasingTM provides the system with a higher capability
`to accurately place the location of the defect on the panel.
`These technology strides have made machine vision
`effective, usable, and reliable.
`
`Repair Systems
`
`Low yields of LCD panel manufactures have created a high
`demand for an effective laser process to salvage the
`defective displays during LCD panel production. The LCD
`repair system is designed to be installed in a manufacturing
`
`process line and will do the various laser functions of
`welding, ablating, cutting and writing, depending on the
`repair strategy used by the manufacturer. There are
`fundamentally two kinds of repair strategies, and thus two
`types of repair systems: Deposition/cutting systems and
`Cut/Weld systems. The deposition-capable systems can
`effectuate open repairs at the plate level by depositing a
`suitable conductive material. This repair process is
`compatible at the plate repair stage only. Depending on the
`deposition system—gas chamber, vacuum chamber, or open
`environment-deposition is necessary. The cut/weld type
`systems can both out material (usually for short repair) and,
`in some systems, can weld two conductive traces for open
`repair [4].
`In this case the array structure needs to be
`modified to incorporate repair rings outside of the active
`area. For the cut/weld system, plate and cell repair is
`possible since no material deposition is utilized.
`IV. Future TIR Directions
`
`The future direction of TIR equipment is in automation and
`networking. Utilizing cluster tool automation concepts with
`seamless factory automation and information management
`will realize the advantages of using TIR equipment.
`Throughout this section the cluster concept with network
`links for Statistical Process Control (SPC) and Computer
`Integrated Manufacturing (CIM) utilizing Automated
`Guided Vehicles (AGV), will be discussed,
`
`The next step for the array testers previously mentioned
`will be to include SPC so that cost control can be
`
`implemented at the critical stages in manufacturing. System
`selection criteria include operating costs (with all running
`costs such as probe cards), sensitivity, test coverage,
`footprint, automation, and second/third generation LCD
`plant compatibility.
`
`In the cell and module test arena there is also been a trend
`to move towards automated, quantitative cell and module
`inspection to allow for SPC and factory automation for
`mass-production. [5] This is seen in addition to large
`growth in both the pixel density (making visual inspection
`increasingly difficult) and general market growth. Table 2
`estimates this growth requirement and an inherent need to
`move to automation.
`
`The above inspection load increase has a parallel in the
`semiconductor industry. During the late l970s,
`semiconductor mask inspection was done using a line of
`human visual inspectors using microscopes, As the device
`count increased, the time to inspect lengthened
`unreasonably and the inspection error rates also increased.
`With the growth of the industry, human inspection was
`eliminated and replaced by automated optical inspection.
`
`
`
`SID 94 DIGEST - 689
`
`
`
`
`Panel Density
`
`(Mpixels)
`
`AMLCD Panel
`
`Table 2
`
`3.6
`
`4.5
`
`
`
`5.5
`
`
`
`
`
`
`
`
`
`
`
`Shipments
`
`Inspection
`
` Load Increase
`from 1992
`
`The similar market dynamics in cell and module inspection
`is causing the manufacturers to shift inspection to
`automated solutions.
`
`7
`
`
`
`
`
`ROBOT
`HANDLING
`
`_SYSTEM
`
`
`
`
`
` IN-PROCESS
`
`TEST SYSTEM
`
`Figure 3: "Cluster" concept for AGV in—line operation
`V. Conclusions
`
`Achieving consistent, reliable yield in mass production
`requires modern TIR equipment. TIR equipment has been
`successfully employed for array test, cell and module
`inspection, and pre—assembly repair and has proven itself
`cost effective. Future systems will offer automation and
`system networking capability which are necessary
`requirements for the next generation of display
`manufacturing and true mass application of Ah/TLCD
`technology.
`
`VI. References
`
`[I]
`
`[2]
`
`[3]
`
`[4]
`
`[5]
`
`H. P. Hall and P. R. Pilotte, "Testing TFT-LCD
`Substrates with a Transfer Admittance Method",
`SID ‘91 Digest pp. 682-685 (1991).
`
`R. L. Wisnieff, L. Jenkins, R. J. Polastre, and R.
`R. Troutman, "In-Process Testing of Thin-Film
`Transistor Arrays", SID '90 Digest, pp. 190-193
`(1990).
`
`F. J. Henley, "Voltage Imaging: A New Method
`to Locate Point and Line Defects”, Technical
`Proceedings, SEMICON/Kansai-Kyoto 91
`Technology Seminar, pp. 84-88, June 1991.
`
`‘
`
`King C. Sheng, et. al.; "A Fast In-Line Repair
`Process for LCD Panel Mass Production", SID '92
`Digest (1992).
`
`H. Asakura, "Image Quality Inspection
`Equipment Rivaling l-Iuman Eyes to be
`Introduced in TFT Color LCDs Production
`
`Fabs", Nikkei Microdevices, pp. 58-63, February
`1994.
`
`For an effective next generation repair system for mass-
`production, manufactures need specific capabilities. These
`include rapid plate load/unload with no gas or vacuum
`chamber for fast, computer driven writing and cutting
`repair operations. Additionally the system would be
`vectored by the array test system and would allow many
`levels of automated repair. Repair could be hands free
`where the system would choose from a library of repair
`solutions depending on the test system classification of the
`defect and the repair system's own optical pattern
`recognition system. In this arena the mass-production
`repair capability must be physically and electronically
`integrated into the production line. Figure 3 shows a
`"cluster" concept AGV in-line robotic integration of a
`robot handling system with a repair and array test system.
`The material flow is made between the systems using the
`robotic handler. The systems are networked together and to
`a CIM system for tracking and process control. Similar
`cluster concepts can be utilized for cell inspection and
`repair.
`
`In addition to the future concerns of automation and
`network links for SPC and CIM, there also needs to be
`consideration and focus on the cost incurred in the
`production of FPD's. Drivers, backlight, assembly,
`overhead, and administration account for over 60% of the
`final cost, making in-process and cell/module testing a very
`important element in cost reduction and reliable display
`manufacturing Additionally there are operating cost that
`need to be considered on many of the technologies such as
`probe card replacement cost on electrical array test
`systems.
`
`690 - SID 94 DIGEST
`
`
`
`