Extractor Pattern Design for 3D Textured Displays and Light Pipes

1、Extractor Pattern Design for 3D Textured Displays and Light PipesDr. William J. CassarlyOptical Research AAbstract:The design of backlit displays and light guides used in instrument lighting often includes the use of 3D textures on the flat surfaces of a complicated plastic part. This talk will desc

2、ribe an iterative technique that can automatically define the texture pattern for a desired spatial luminance distribution by using illumination simulation results. Numerous examples using LEDs and CCFLs will be demonstrated. LT Seminar, Novi, June 20052OutlineDisplay devices Backlights Instrument P

3、anels AppliquOptimization Elements of an optimizer Merit Function and NoiseMesh Feedback Optimization ExamplesVariable size and placement 1D (CCFL) and 2D (LED) SummaryLT Seminar, Novi, June 20053What Is A Backlight? Backlights LCD backlight Instrument Panels Appliqus etc Nominally Consists of A lig

4、ht guideUsually made of plastic A light sourceCCFL and LED are typicalOptic to increase collection efficiency Other elements often usedDiffuser Brightness enhancing filmsBackplane reflector Some elements may not be used, depending on size, cost, and other requirementsTypical Backlight Light sourceLi

5、ght guideDiffuser + BEFReflectorReflectorLT Seminar, Novi, June 20054Typical Design GoalExtract light in a direction perpendicular to the direction of propagationSpatial light modulator (e.g., LCD or Appliqu) is placed over the backlightTypically want uniform light extractionLight SourcePreferred di

6、rection of light extractionDirection of light propagationLight GuideLight extraction from a light guideLT Seminar, Novi, June 20055Design Process: UniformityAvailable power varies within the light guide Extraction efficiency varies along length to compensate Ways to vary extraction efficiency includ

7、e Paint density, Texture size, Texture spacingNeed a means to design the extraction patternDistance from the sourceAvailable powerExtraction efficiencyUniform outputLT Seminar, Novi, June 20056OptimizationOptimization Ability to automatically refine the performance of a system based upon a user spec

8、ified performance criteriaParameterization Definition of system variablesChoose variables that influence performanceMerit Function Quantifies system performanceSmaller value means better performanceAlgorithm Various possible methods Local versus globalOptimization AlgorithmMerit FunctionModelParamet

9、erizationUser Interface ties these elements togetherLT Seminar, Novi, June 20057Optimization FlowchartEvaluate Merit FunctionCompare MF With Termination CriteriaSetupAlgorithm Computes New Variable ValuesUpdate ModelDoneLT Seminar, Novi, June 20058Merit Function Optimizer needs quantity to determine

10、 goodness Called the Merit Function, MF Weighted Sum of the Squares is typical Smallest MF when Vi = Ti for all i Weights, Wi, control relative importance of the Vi - TiMF = Wi2(Vi - Ti)2Wi = Weight of ith MF item Vi = Current Value of ith MF itemTi = Target of ith MF itemLT Seminar, Novi, June 2005

11、9Termination Vi can be obtained using Monte Carlo simulations Values have noise, i, and therefore the Merit Function has noise LightTools computes the noise in the Merit Function Typically, reduce noise by tracing more rays Optimizer terminates when Noise becomes large relative to MFLuminance with10

12、,000 RaysLuminancewith1,000,000 RaysMF = Wi2(Vi + i - Ti)2MF = Wi2(Vi - Ti)2 + NoiseLT Seminar, Novi, June 200510Feedback OptimizationMost backlights represent a very special type of illumination system There is a 1:1 correlation between the extractor position and the spatial luminance distribution

13、Properties of paint or 3D texture extractor elements (size, depth, position, etc.) affect the luminance distribution in a relatively localized manner An efficient optimization algorithm is possible because of this correlation This approach is very powerful and will be described over the next few sli

14、desLT Seminar, Novi, June 200511Mesh Feedback OptimizationUniformity obtained by iteratively changing the paint or texture pattern Reduce extraction where (il)luminance is too high Increase extraction where (il)luminance is too low Amazingly robustSTARTING PATTERN CAN BE FAR FROM FINAL PATTERNEvalua

15、te (il)LuminanceAdjust Extraction PatternUniformity OK?NOYESLT Seminar, Novi, June 200512ExamplesChanging Size 3D texture placement does not change, but extraction efficiency depends upon feature sizePaint density (size of paint dot)Prism Width/DepthSphere or Cone Tip DepthChanging Placement Feature

16、 size stays constant, but extraction efficiency depends upon relative spacingTypically chosen to avoid step of painting the backlight by integrating placement features into the moldLT Seminar, Novi, June 200513Example 1, CCFL with Painted LP6 merit function evaluations to obtain reasonable solution

17、for 2D texture pattern with uniform grid at startSmall Source(e.g., LED)2D DisplayTexture PatternOutputLT Seminar, Novi, June 200514Example 2, LED with Painted LPSingle 120 LED, air space between LED and light pipeMax density near the input (left) edgeStartFinalLT Seminar, Novi, June 200515Example 3

18、, LED with Painted LP and CutoutSingle LED with circular cutout near the inputMax density near the far (right) edgeCircular cutoutStartFinalLT Seminar, Novi, June 200516Example 4: Prism Width ExampleLambertian LED located along left edge of backlightSurface composed of +/-45 prisms with the width va

19、riable. Prisms are rotated 45 with respect to the backlight edges.Resulting prism width distribution is very non-uniform, but output is uniformPrism Width Distribution StartFinalTexture PatternOutputLT Seminar, Novi, June 200517Example 5: Sphere DepthLambertian LED located along left edge of back li

20、ghtSurface composed of equally spaced spheres with constant radius No paint applied to the textureSurface to vertex of sphere is variedSurface to VertexSurface to Vertex DistributionFinal LuminanceLT Seminar, Novi, June 200518Example 6, Paint Density with NonUniform TargetThis example illustrates a

21、target distribution that is not constant. The letters L and T have target values that are twice the background.Texture PatternOutputStartFinalLT Seminar, Novi, June 2005193D Textures, Placement Changing 3D texture spacing is more complicated than changing feature size Typically need to avoid overlap

22、 of 3D textures Spacing changes in one region can alter spacing in another Total number of 3D textures can change 2D extraction pattern is more complicated than 1DLT Seminar, Novi, June 200520Prism Spacing with CCFLPolynomial prism spacing coefficients controlled00.070.140.210.280.350.420.490.560.63

23、0.7-200-1000100200P(i)00.0350.070.1050.140.1750.210.2450.280.3150.35-200-1000100200P(i)Prism Spacing, AfterPrism Spacing, BeforeOptimize with 5000 ray, Results shown below with 50,000 raysLT Seminar, Novi, June 200521Specify 2D DensityZone is divided into Nx by Ny bins Area = BinAreaBins contain 3D

24、textures Number = Nij3D textures fit inside rectangles Rectangle Area = ADensityij = Nij(A/BinArea)LT Seminar, Novi, June 2005222D Placement2D spacing is defined using 2D mesh. Within each bin of the mesh, spacing is either uniform in X and Y, or neighboring bins are used to compute a slope term and

25、 the spacing is varied linearly across a bin.Number of bins equals number of mesh bins, by default, but user can over-ride2X24X44X4 with linear changeLT Seminar, Novi, June 200523Dither Within a BinDither allows textures to have random shifts with respect to their nominal centers, but randomization

26、is controlled so that textures do not overlapDither X,Y=1,0Dither X,Y=0,1Dither X,Y=1,1Dither X,Y=0,0Dither X,Y=.2,.2Dither X,Y=.5,.5LT Seminar, Novi, June 200524Placement with Hemisphere TexturesNonuniform placement of identical textures is used to vary the extraction efficiency This example uses s

27、mooth hemisphere textures (i.e., no diffuse paint on the textures)Texture DensityOutputLEDStartFinalLT Seminar, Novi, June 200525Two LEDs, No SymmetryPlacement exampleTwo LEDs No symmetryLEDsStartFinalTexture DensityOutputTexture DensityOutputLT Seminar, Novi, June 200526Two LEDs, Some SymmetryPlace

28、ment exampleTwo LEDsSome symmetryTexture DensityOutputLEDsStartFinalTexture DensityOutputLT Seminar, Novi, June 200527Zero Target Edge BinsIt is not always desired to have the edge region of the light pipe lit By using zero targets around the edge, uniform output within the center region is obtained

29、 Notice that optimization increases density near edges of the targetTexture DensityOutputLEDZero target edge binsTexture DensityOutputStartFinalLT Seminar, Novi, June 200528Nonrectangular Shape ExampleA new feature in LT computes area of partial bins areas on a surfaceThis example uses the partial areas as the optimization targetsTexture DensityOutputLEDCutoutsMatching Targetpartial edge binsTexture DensityOutputStartFinalLT Seminar, Novi, June 200529Prismatic FilmsPrismatic film