DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

[0022] One form of the present invention involves the use of a stack of light directing layers disposed above a plane of separated emitters arranged either as an array of parallel stripes or as a two-dimensional array of bounded emitting regions such that a directed output beam of even uniformity is created as if from a continuous emitter of area equal to that to the output aperture. One (or two) of the light directing layers are prism sheets whose geometry and elevation above the plane of emitters is chosen uniquely so as to create the required overlap and diffusion of emitter images.

[0023] 1.0 One-Dimensional Emitting Array

[0024] An optical system constructed in accordance with one principal form of the invention is indicated generally in FIG. 1 and represents a side elevation. The optical system 10 embodies a structure and method, which uses various optical elements, disposed in a compact geometry relative to an intrinsically discontinuous light source 1 formed by an array of emitters, and manipulates this light to generate uniform output illumination over a wide range of illumination directions on an output screen 28 placed the minimum possible distance from the plane of light source 1 . Light from this output screen then provides the required even field of featureless back illumination, either continuously white in color, or pulsed rapidly and sequentially in periods of red, green and blue emission for a directly viewed image display device 3 placed against it, device 3 which may be a spatial light modulator (SLM) image display such as a conventional liquid crystal display (“LCD”) imaging device or other active image display devices that do not generate light of their own but rather modulate brightness of the tiny component parts of the image known as pixels. The image display device 3 may also be any one of a variety of passive or static image sources such as photographic transparencies, wherein for example the system 10 can be used as an improved illuminator for medical x-ray films.

[0025] The behavior of the system of FIG. 1 and that of each of its elements is described in greater detail below. In summary, the height of prism sheet 7 is used to form overlapping virtual images 26 , 27 of the output plane 34 of light source 1 . Light from the overlapping emitter images is then used to fill in the non-emitting spaces 25 between emitters as evenly as can be arranged, and thereby reduce the maximum and minimum brightness that would otherwise be observed. Subsequent conventional light scattering layers 28 and 30 are elevated above the vertex points 12 of the image-displacing prism array 7 by distances G2 and G2+G3 respectively to add further spatial mixing to the result and also to widen the range of output angles that exit the effective output screen 28 . As will be shown below, the exact height of the prism sheet 7 above the output plane 34 of light source 1 , whether this is the discrete emitting channels themselves, or a diffusively scattering layer above the emitters, depends on the geometry of the prism units and on the degree of image displacement that is desired. The prism's apex angle and size can each be varied so that the distance 18 , G1, is the smallest possible, which in some cases might be zero.

[0026] Back-reflector 46 in FIG. 1 is composed of a metal or plastic support substrate 48 , a reflecting layer 50 which may be diffusely reflecting, and a gap 52 , whose medium (air or dielectric) and thickness 52 are adjusted to provide a balanced transfer of back-reflected light back through the channels and through the non-emitting gaps 25 . When the support substrate 50 is an electrically conductive one, it becomes a capacitive part of the electrical equivalent circuit of light source 1 . When the support substrate 50 is a thermally conductive one, it equalizes the distribution of heat throughout the lamp in a manner that has become traditional in many light source systems when spatial light modulator 3 is an LCD screen. The conductive plane provides a means for preventing LCD image contrast changes caused by its exposure to local heating. When the support substrate 50 is an electrical ground plane, the purpose of separation distance 52 is also to prevent (or minimize) electrical power losses to ground from leakage of the current flowing in light source 1 through the plane's distributed capacitance. This plane can also be used to isolate electronics used to control the spatial light modulator 3 from the electrical drive fields of light source 1 . Generally, the performance of this conducting plane 50 is maximized when made of the most electrically conductive metals such as stainless steel and aluminum. The diffusely reflecting layer can be any material loaded with highly reflective white light scattering particulates such as those plastic sheets called REFWHITE manufactured by Kimoto Co., LTD. of Japan.

[0027] The prism sheet 7 may be one of the 90-degree offerings manufactured by the Minnesota Mining & Manufacturing Company (3M) under the trade name BEF, as in brightness enhancement film. The prism sheet 7 may also be a more specifically suited custom designed prism array molded, embossed or cast in an optically transparent material such as acrylic, polyester or polycarbonate, wherein the apex angle of the prism is not 90 degrees, but rather an angle matched to the exact performance required, and perhaps the prism shape modified, as described below, to fine tune the particular image displacement characteristics needed. Since direct view of the prisms themselves is obscured from view by the diffusive layers 20 the width of the prisms, any cosmetic defects on their surfaces and any dead space 16 between the individual elements, is cannot be seen. Widespread commercial applications of 3M's BEF products in backlit LCD screen systems place the BEF sheets just behind the LCD screen, where any discontinuities or defects in its optical performance are seen directly, even through weak diffuser sheets. Consequently, those principally brightness enhancing features of the 3M prism sheet materials require extreme levels of cosmetic perfection in both manufacturing and handling.

[0028] For practical applications, the total system 10 thickness 56 , T, in FIG. 1, G 3+G2+G1+L, is made the smallest possible commensurate with suppressing visibility of the discontinuous nature of light source 1 . A practical example will be given further below for a new flat, parallel-emitting-channel fluorescent lamp developed by Corning, Inc.

[0029] Another embodiment of the invention of FIG. 1 is given in FIG. 2 , which also represents a side elevation. In this case, the single light source 1 , which as above emits light from its entire internal surface, in both forward and rearward directions, does so surrounded by a completely symmetric image display system 10 featuring both forward and rearward spatial light modulators 3 and 4 positioned on opposing sides of light source 1 , each with its own interior and exterior intervening diffusing layers 20 . The result is a particularly thin two-sided display device whose bright and uniformly illuminated images can be seen from either side. In this case, half the total lumens produced by the light source are routed through each side's set of intervening multi-layers 7 and 11 .

[0030] The configuration of FIG. 2 is exactly that of FIG. 1 with its structure disposed symmetrically on each side of the system's mirror plane 6 . Virtually identical emitting patterns 24 are produced on the outermost light scattering surfaces 34 of light source 1 , and it is these light patterns that are displaced as virtual images 26 and 17 by the prism sheets 7 , governed by their apex angles 8 and their relative heights 18 above the object planes 34 . In this two-sided structure, any light reflected back towards light source 1 by the upper prism sheet 7 is either re-scattered by the upper side of light source 1 or transmits through light source 1 and becomes a part of the light emitted by the lower side of light source 1 .

[0031] Practical applications of this double-sided invention format include two-sided televisions, two-sided desktop computer monitors, two-sided commercial signs such as “EXIT” signs, and two-sided passive signs displaying a different message depending on the side viewed.

[0032] 2.0 Two-Dimensional Emitting Arrays

[0033] A two-dimensional emitting array is formed by arranging rows and columns of discrete square (or rectangular) emitting apertures, as opposed to the rows of one-dimensional emitting stripes involved above. In this case, the discrete emitting regions are separated from each other by non-emitting regions, and the need is for a means to provide light evenly distributed over the entire array aperture. Such means is provided in the present invention by a bi-layered extension of the elevated single prism sheet method of FIGS. 1 and 2 , as well as by arrays of discretely tapered micro reflectors. Both two-dimensional approaches couple light collected from the discrete emitting elements in the array to a collective and spatially unified output beam that appears to have come from a single output aperture encompassing the entire array.

[0034] 2.1 Elevated Prism Sheets

[0035] The precise elevation of two orthogonal prism sheet layers 58 and 60 is applied to create a two-dimensional array of virtual images, four virtual emitter images associated with every emitting object in the underlying emitting array. By taking this bi-layered, rather than mono-layered prism sheet approach, a completely contiguous output array of emitter images can be achieved without any appreciable non-emitting regions between them. As one example, square emitters, W millimeters on a side, separated from each other by non-emitting gaps W millimeters wide are so converted into a contiguous array of W millimeter square emitter images, each conveying about a quarter of the original light flux emitted (less the transfer efficiency of the light through the two prism layers 58 and 60 ). Moreover, this organized output light is constrained to a concentrated range of output angles characteristic of the prism geometries used within each prism sheet, and relatively independent of the considerably wider angular range of input light emitted. In this manner, an emitting array whose original emitting area is only 25% of the overall array aperture, converts to an output array whose emitting area becomes 100% of the overall array aperture, and whose emission is contained within a reduced range of emission angles. The practical advantages of beams with such uniformity and directionality will be illustrated in a set of examples to follow below.

[0036] This bi-layer prism sheet approach is implemented in one of two illustrative two-dimensional configurations related to the one-dimensional method of FIG. 1 . A first approach, shown schematically in FIG. 3 relays the contiguous virtual images created by the two elevated prism sheets to the output plane by means of an array of micro lenses. A second approach, generally preferable to the first with regard to compactness, is illustrated schematically in FIG. 7 , and uses the two prism sheets alone, with light projecting outwards to the output plane from the underlying virtual images themselves. The spatial relationship between the virtual images created by any given emitting aperture is illustrated graphically in FIG. 4 . One example of a compartmentalized spacer layer between the emitting array and the prism sheets, showing one isolated compartment per emitter, is conveyed in FIG. 5 . Then, the form of a tool enabling fabrication of this spacer layer is given in FIG. 6 .

[0037] 2.1.1 Tri-Layered Prism Sheet Illuminators

[0038] A cross-sectional view is given in FIG. 3 of one two-dimensional multi-layered emitting array consisting of two prism sheet layers and one micro lens layer. In this example, output screen 28 is arranged to display a contiguous or slightly overlapping array of well-defined and controllable image (or illumination) elements (sometimes referred to equivalently as pixels) that result from the system's manipulation of light from, in this case its two-dimensional array of individually controlled emitting entities 72 whose square (or rectangular) emitting apertures 24 form the base layer of light source 1 . Two prism sheets, 58 and 60 , are used in tandem, both sets of prism vertices pointing towards the viewer, with the planar axis of each sheet (x axis 116 for sheet 58 and y axis 116 for sheet 60 ) made perpendicular to each other and the relative spacings G1′ and S1, 19 and 34, adjusted so that, as shown in the cross-sectional perspective of FIG. 3 , two orthogonal sets of shifted virtual images 106 are created, thereby converting every emitting area 110 into a cluster 106 of four essentially contiguous virtual images 26 , 27 , 108 , and 109 of the original emitting region 110 (shown more clearly in the perspective drawing of FIG. 4 ). In this case, the lower prism sheet 58 creates a pair of shifted virtual images 26 and 27 in the plane of the cross-section in FIG. 3 and similar to those described in FIG. 1 , while the upper prism sheet 60 splits this pair into a pair of virtual image doublets shifted into and out from the plane of the cross-section, one doublet composed of images 26 and 27 , the other composed of images 108 and 109 , as in FIG. 4 . FIG. 4 provides a three-dimensional view of these spatial relations, with emitting region 110 shown as a white square in relation to its four shifted virtual images 26 , 27 , 108 and 109 , and the surrounding eight emitting regions 112 , each of these shown shaded. The spatial boundary of the resulting virtual pixel is highlighted with a black frame 114 for additional clarity. Each of the four virtual images of emitting region 110 are shifted a distance W′, 120 , from center 122 in each of the two orthogonal directions x, 116 and y, 118 . The plane of the cluster of virtual images 106 resides at a height, 124 , G1′-V above the emitting plane 122 , where V is the depth of this plane beneath the plane of the lower prisms 58 , which will be established more exactly later on.

[0039] A viewer looking at the output side of the prism sheets 58 and 60 in FIG. 3 sees the virtual image plane as a contiguous array of discrete regions, each consisting of the 4-image cluster 106 ( FIG. 4 ) of each underlying emitting region 110 . While this alone might be suitable for some direct viewing applications, there are some limitations to take into consideration. One limitation is that output light from the virtual image plane is confined to a narrow cone of viewing angles (+/−22.5 degrees to half peak power) that is an intrinsic feature of transmission through two orthogonal prism sheets. The second limitation is that demarcation lines within each 4-image cluster and the demarcation lines between the individual 4-element pixels themselves, might give rise to a visible pixel structure distracting to a direct viewer.

[0040] Practical applications of the two-dimensional illumination systems of FIGS. 3 and 7 , however, relate both to those involving direct view of the illuminator and those wherein the system's output light beam 100 is used to provide illumination to external elements that are themselves viewed. In some applications, it is preferable that light from the illuminator's aperture be smooth in its spatial uniformity, but the emission confined to a narrow range of angles. In other applications, not only must the spatial uniformity be smooth, but the illumination must also be made visible over a wide range of viewing directions.

[0041] In direct view applications, one solution to the demarcation lines and the viewing angle restrictions is provided within the multi-layer structure of FIG. 3 by the array of micro lenses 62 that are used to relay a real image of the virtual image plane at unity magnification to an output scattering layer 94 placed within (or on) output viewing screen 24 . FIG. 3 symbolizes only one lens unit per pixel region 102 , but sub-arrays of lenses may also be used as dimensions require and allow. The exact height 32 of the viewing screen, G3, can be adjusted to defocus the system just enough blur to soften the appearance of the demarcation lines. And, the more Lambertian the behavior of the scattering mechanism involved, the wider becomes the associated output viewing angles 100 . The generalized behavior of lens array 62 is illustrated in FIG. 3 with rays 88 from virtual image point B (corresponding to one set of rays from point A on emitting region 24 ) collected and imaged as rays 96 forming real image point C on the lenses image plane 94 , whereupon they are scattered by viewing screen 28 into a fan of output rays 100 .

[0042] In direct beam illumination applications, a solution to the demarcation lines between virtual images is to defocus the system, so that the relayed output images are not sharply focused on output screen 28 (used with little or no scattering layer 94 ).

[0043] Lens array 62 may be a two-dimensional array of plano-convex aspheric surfaces; each having square (or rectangular) boundaries as needed to correspond to pixel dimensions, which in the examples given are 2W by 2W. The lens array 62 may also be composed of aspheric Fresnel lens elements, or two parallel sheets of aspheric plano-convex or Fresnel cylinder lens arrays, each cylinder lens having a width corresponding to that of the corresponding pixel width (2W). In this latter case, the two sets of cylinder lens arrays are oriented so that their cylinder axes are orthogonal to each other. For shortest possible focal length, however, a stack of two aspheric plano-convex lenses (bulk or Fresnel), vertices facing each other, might be used for each pixel. This would mean using registering two parallel sheets of lens arrays in place of the single sheet depicted in FIG. 3 . For larger sized pixels, more than one shorter focal length lens can be used within each pixel region. Regardless of the lens format used, the effective lens focal length is set so that it is approximately half the lens's elevation above the virtual image plane 66 , which can be no closer than G2+S1+V. At the same time, the viewing screen 28 is elevated above the lens plane an equal distance G3. In this situation, the total thickness, 26 , T of system 10 , becomes 4F+G1′−V, where F is the minimum focal length of lens array 62 , G1′ approximately the width of the emitting region 110 , and V the depth of the virtual image plane 66 below the plane of prism sheet 58 (which as will be proven later is about 0.75W).

[0044] When the emitting regions are taken as 8 mm squares, the thickness of two prism sheets, 0.3 mm, and the spacing between them, S1, near zero, the minimum focal length of lens array 62 , when composed of only one lens element per pixel, becomes about (0.75W+0.3)/2 or 3.15 mm, which is shorter than practical for a single pixel-sized lens element. The shortest practical single lens focal length covering the entire 16 mm×16 mm aperture (22.6 mm diagonal) would be about twice the semi-diameter or 22.6 mm, making total thickness 26 , more than 90 mm. One practical way of achieving the more preferable focal length of 3.15 mm is to use a 7×7 or 49 lenslet sub-array covering each 16 mm×16 mm pixel area, with each lenslet in the sub-arrays truncated for this example to 2.28 mm squares each having an effective spherical radius of curvature of 1.57 mm. If this were done, total thickness 26 becomes about 12 mm plus the thickness of light source 1 , which is more suited to the applications of interest.

[0045] In this manner, the arrangement shown in FIG. 3 converts each emitting area 24 on light source 1 to a corresponding emitting area or pixel ( 102 in FIG. 3, 108 in FIG. 4 ) on output screen 28 with the spaces between the original emitters ( 24 in FIGS. 3 and 110 in FIG. 4 ) effectively eliminated by the optical manipulations. System performance is improved considerably, however, adding a physical structure 84 in between and otherwise bounding all emitting areas ( 24 in FIGS. 3 and 110 in FIG. 4 ), both to minimize pixel-to-pixel cross talk and to serve as a natural spacer of thickness G1′ for the lower prism sheet 58 . The processing of virtual images 26 , 27 , 108 and 109 in FIG. 3 and FIG. 4 is independent of the presence of structure 84 and its sidewalls 85 . The sidewalls serve to restrict any high angle light rays from the emitting region 24 itself, and any initially reflected light from prism sheet layers 58 and 60 from reaching or returning to prism sheets 58 or 60 outside the boundary lines 102 of the corresponding output pixel. When these sidewalls 85 are made reflecting, such rays scatter within the cavity until they are randomly converted to one of the correct angles for transmission as output light from prism layers 58 and 60 within the pixel boundary.

[0046] 2.1.2 Compartmentalized Prism Spacing Layers

[0047] A generalized three-dimensional view of one such structure 84 is given in FIG. 5 showing a hollow faceted version wherein the sidewalls 85 surrounding any discrete emitting region 24 are tilted at an angle 87 , φ, relative to the system axis 5 , Tan ⁻¹ (W/2G1′). If this hollow isolating structure is made of (or coated with) a white diffusely reflecting material, which is preferable, the sidewalls of the structure 58 and the base layer of prism sheet 58 form boundaries of an effective integrating cavity whose multiplicity of reflections improves the percentage of light transmission through the prism sheets and the light distribution uniformity within any pixel's output area ( 102 in FIG. 3 or 116 in FIG. 5 ). The structure shown in FIG. 5 can be compression or injection molded in a plastic such as acrylic or polycarbonate that has been filled with particles of a finely divided white light scattering material such as titanium dioxide. The structure can also be formed, in plastic sheets up to about 10 mm in thickness, by an embossing (or casting) process wherein the pyramidal tooling needed to generate each of the four common sidewalls, one version shown in FIG. 6 , is made thicker than the film (or resin) itself so that it punches through the molten sheet (or resin) material to a non-molten carrier layer (or air above the resin), and so generates the array of clear holes 126 needed in FIG. 5 to permit efficient light transfer from the emitting regions 24 of light source 1 . The molded, embossed or cast material can also be a composite of polymer and any second phase including glass, ceramic or metal, so as to achieve specific mechanical, thermal and/or optical properties.

[0048] The compatibility of this structure with the image-shifting function of the prism sheets themselves, as well as some other beneficial forms for this layer, will be covered in more detail below. Qualitatively, however, the most important concept is that any light scattered from the sidewalls that then appears as if emitted by sidewall surfaces themselves, contributes to virtual images of those same sidewalls that shift only inwards towards the center of the cavity, otherwise overlapping the shifted virtual images 26 , 27 , 108 , and 109 of the emitting region itself. As will be explained more thoroughly below, the distance light from any point is shifted by the prism layers 58 and 60 , relates to the specific depth of from the base of the prisms of any point of emission, and as mentioned earlier, the apex angle 6 of the prisms themselves. The closer the particular emission point is to the base of the prisms, the smaller is the shift, the further the point from the base of the prisms, the larger the shift, for any given apex angle 6 . Because of this, sidewall light is unable to shift into neighboring cavities, and appears as the tilted images 104 in FIG. 3 . This is beneficial to image-forming performance of the pixels as it virtually eliminates pixel-to-pixel cross talk.

[0049] 2.1.3 Bi-Layered Prism Sheet Illuminators

[0050] A thinner bi- rather than tri-layered alternative to the arrangement of FIG. 3 is indicated generally in FIG. 7 , where micro lens array 62 of FIG. 3 has been eliminated, and output screen 28 arranged to display or convey light from a contiguous or slightly overlapping array of well-defined virtual images 102 that result from the system's prismatic manipulation of input light. A chief advantage of the bi-layered approach is that by eliminating relay lens layer 62 of FIG. 3 , the system of FIG. 7 can be made significantly thinner. Total thickness 22 of the bi-layered system, T, reduces to G4+G1′ plus the thickness of light source 1 . With 16 mm square output pixels and the 3.15 mm focal length relay lens array used in FIG. 3 , the total thickness in FIG. 7 depends primarily on the prism offset G1′ needed to make the 8 mm emitting regions appear contiguous on the output plane 94 . When using a single 90-degree prism sheet, the condition of contiguous displacement occurs when the offset G1 is substantially equal to the emitter width W. When using two orthogonal 90-degree prism sheets, however, the offset G1′ is somewhat less than W. By both ray trace modeling (using the optical system modeling software ASAP™ produced by Breault Manufacturing Organization) and direct laboratory experiment, it is determined that G1′ is approximately 0.625W. This means that with prism sheets having prism elements with standard 90-degree apex angles can be less than about 5 mm plus the thickness of light source 1 , about a 2.5× thickness reduction over the 12 mm thick system of FIG. 3 . Then since the prism sheet offset distance, G1′, for the perfect image displacement of FIG. 4 can be reduced by means of adjustments to the prism element's apex angle 6 , even thinner systems 10 can be created, when so desired.

[0051] Being able to truncate the illuminator system thickness at the height of the upper prism sheet contributes considerable thickness reduction. The compartmentalized spacer layer 84 , whose sidewalls 85 can be made diffusively reflective, reduces visibility of virtual image demarcation lines, as does any scattering layer 94 used within output screen 28 .

[0052] The multi-layer arrangements of FIGS. 3 and 7 are generally preferable for illumination applications involving tightly spaced emitters arrays, where the spaces between emitting elements is about equal to or less than the size of the emitting apertures themselves. When applications call for considerably larger area output pixels, the prism sheet layers are replaced by an array of micro reflectors whose input apertures match the emitting apertures, and whose output apertures are contiguous by design.

[0053] 2.1.4 Multi-Layered Micro-Reflector Illuminators

[0054] An alternative output array structure is illustrated in FIG. 8 in which the virtual image forming prism sheets 58 and 60 of FIGS. 3 and 7 have been replaced by a two-dimensional layer of micro-reflectors similar to the compartmentalized diffusely-reflecting spacer layer 84 shown previously in FIGS. 3, 5 and 7 . In this instance, however, the reflecting sidewall 136 is a specularly reflecting one and its mathematically derived shape, critical to function. Wide angle input light from each emitting aperture 24 enters the associated specular reflecting cavities of FIG. 8 , wherein it is transformed by the series of specular reflections that result, into output light whose angular spread is reduced from the input spread in a determinable manner. Since the reflecting elements are themselves made two-dimensionally contiguous, as in FIG. 5 , the output light emitted from the array is itself similarly contiguous. The micro reflector boundaries do form visible demarcation lines in the output light, but these generally fine boundaries can be blurred in numerous ways, including allowance of some purposeful light leakage or cross-over to occur at the reflector boundaries.

[0055] FIG. 8 shows the cross-section 123 of several emitting pixels as well as the three-dimensional perspectives of single pixel units 121 and 127 . The pixel design of perspective 121 crosses appropriate two-dimensional sidewall shapes in the orthogonal (x and y) meridians, whereas perspective 127 is for a reflector having spherical symmetry. On the other hand, when called for, physical boundary walls 133 can be added to isolate the light and its reflections within one pixel from another, thereby substantially eliminating the number of light rays crossing over from one pixel's reflector into the space of a neighboring pixel.

[0056] By means of micro reflectors, it is possible to magnify emitting region areas beyond the fourfold expansion achieved using bi-layered prism sheets 58 and 60 . Because the micro reflector's sidewalls are made sloping outwards towards an enlarged output aperture 102 , in principle every input light ray is transmitted as output. No input rays can be trapped internally within specular reflectors 130 , as they can by total internal reflections within prism sheets 58 and 60 . The reason for this is that there is no combination of specular reflections from outward sloping sidewalls 136 that prevents any input rays from becoming output rays. Then if the outward sloping reflecting sidewalls are shaped purposefully, substantially all output rays can be made to behave in an organized manner.

[0057] There are at least two advantageous ways of shaping the outward sloping reflector sidewalls for an efficient conversion of input light to output light. One advantageous sidewall shape is that of a concave conicoidal reflector used in conjunction with polarization-selective mirror plane, such as has been described for other illumination applications in U.S. Pat. No. 6,213,606. In this case input light is injected through a small aperture made in the reflector's vertex, and multiply reflected output light exits through the reflector's outermost aperture. Another advantageous sidewall shape is provided by tapered non-imaging optical concentrators similar to integrating bars. In this case, input light enters the smaller end of the reflecting element and exits the larger end.

[0058] 2.1.4.1 Hyperboloidal Reflecting Elements

[0059] A hyperboloidal reflective sidewall shape is shown schematically in FIG. 9 for the side view of any single pixel unit of what is actually a contiguous two-dimensional array. In this case, the output aperture of the pixel, 150 , is made considerably larger than the size of the input aperture, 152 , to prevent or minimize any losses associated with light return and re-radiation by this aperture 152 . As above, the input emitting aperture may be either the output emitting-surface of a light emitting device such as an LED (or OLED) 70 , or the diffusive aperture 24 of a cavity 72 (such as in FIG. 8 ) containing one or more light emitting devices 70 . If the input aperture 152 were 2 mm in diameter, the output aperture would be preferably 10-20 mm in diameter or larger. The total lumens flowing out of input aperture 152 becomes substantially the total lumens flowing out of the output aperture 150 less any losses due to absorption and total internal reflections along the way.

[0060] The optical path of a given extreme ray 154 leaving from point O on input aperture 152 to point D on the output aperture 150 is in its simplest configuration a 3-step process, as illustrated by rays 156 , 158 , and 160 in FIG. 9 . Input light ray 154 may be polarized or unpolarized. When it strikes the output screen 131 at point D it is either polarized linearly by the reflective polarizing layer 162 itself, with one linearly polarized ray reflecting as 156 , the other linearly polarized ray transmitting as ray 160 , or just reflecting as linearly polarized ray 156 . In either case, the reflected ray 156 proceeds back towards the shaped concave reflecting sidewall 136 , as if from a front focal point 166 of the hyperbolically shaped concave reflecting sidewall 136 . This ray 156 , on reaching the concave reflecting surface at point B reflects back towards the output screen as if emanating from the reflector's rear focal point 168 , and passes through all layers comprising output screen 131 , including a wide band quarter wave retardation film 170 , the reflective polarizer 164 and the screen 172 . The screen 172 may contain a diffusely scattering layer to spread out the light, a collimating lens to narrow the angular output or both. The reflecting sidewall 136 may be smooth and purely specular, may have a stippled surface to create a range of reflected directions or a smooth reflecting surface and an external scattering layer to create a range of reflected directions. The purpose of adding some light scattering on or near the reflective sidewall, whether by stippling of its surface or by an external scattering layer near its surface is to soften any non-uniformities associated with the input aperture, thereby improving the spatial uniformity of the output light.

[0061] A special case is presented when input ray 154 is un-polarized. Selective reflecting layer 164 linearly polarizes the directly transmitted output ray 162 , and the multi-step (O-D-B-C) reflection process converts output ray 160 to the same polarization as ray 162 . As a result, there is a composite output distribution with half the lumens spread over +/−θ as if emanating from point O, and the other half spread over +/−Ψ as if emanating from point G, 168 . Uncorrected, such an angular (and spatial) mix may not be appropriate for every lighting and display application, but may have special benefits for others, particularly in general lighting when providing directed and flood illumination simultaneously.

[0062] The principal purpose of reflective sidewall structure 131 is to spread out the lumens delivered by the input aperture 152 over a geometrically expanded output aperture (the contiguous pixel) 150 with the smallest loss of light, the least cross-talk with neighboring pixels, and in the thinnest overall multi-layer system thickness, T, possible. When this sidewall shape is made hyperbolic (or approximately so), as in the cross-section FIG. 9 , the input light rays follow the deterministic paths described when the reflective polarizer plane 164 is placed a distance 174 , H, above the plane of the hyperbola's vertex point O, H being equal to 0.5 times the distance between the front focal plane and the vertex plane, F2. This positioning extends the total optical path length significantly without increasing the system's thickness. Even though the light originates at point O, its exit through the output aperture 150 at point C is as if the light actually originated at the hyperbola's back focal point 168 , a distance F+A further below, F and A being the parameters of the hyperbolic function. When A becomes very large, the hyperbolic function behaves more and more as a parabola, the output rays 160 appear to come from infinity, and are nearly parallel rather than diverging.

[0063] The mathematics of a hyperbolic reflector is summarized by equations 1-3, which describe the hyperbolic reflector's concave sag 178 , Y, as a function of the effective semi-diameter 180 , X, which can be thought of as the reflector's radial coordinate, and the salient reflected ray. 1 $\begin{array}{cc}X=B\sqrt{\frac{{\left(Y+A\right)}^2}{A^2}}-1& \left(1\right)\end{array}$ X=Tan θ( F 2 −Y ) (2) 2 $\begin{array}{cc}\mathrm{Tan}\Psi =\frac{\mathrm{Xo}}{\mathrm{F1}+\mathrm{Yo}}& \left(3\right)\end{array}$

[0064] The parameters A ( 190 in FIG. 9 ), B, C, F1 ( 186 in FIG. 9 ) and F2 ( 176 in FIG. 9 ) are the hyperbolic constants, θ, the angle 182 that an extreme ray 152 makes with the system axis. The concave sag at any point B, Y _o is determined by equating equations 1 and 2. Once solved for Y _o , the corresponding X _o is determined by substitution in equation 2. Then, the resulting maximum output angle 192 , Ψ, is set by equation 3.

[0065] The salient hyperbolic parameters, A, B and C are given in equations 4-7 for the system of FIG. 9 . When F1 approaches—infinity, the reflector shape of equation 1 is parabolic, and the output rays 160 proceed generally along the system axis, the angle Ψ approaching zero.

A=K 1 −F 2+2 H 1 (4)

C=F 2 −H 1 +A (5)

B =( C ² −A ² ) ^0.5 (6)

K 1 =[|F 1 |+|F 2|]/2 (7)

[0066] H1 in these equations is the location of the hyperbolic vertex point O on the Y-axis (nominally 0), F1, is the location on the Y-axis of the back focal point 168 , and F2 is the location on the Y-axis of the front focal point, 166 . F2 is a positive number and F1, a negative number. The true focal length, F, is F2+A ( 188 in FIG. 9 ). The reflector's eccentricity, E, is as always, F/A.

[0067] The wider the angular spread of emitted light, θ, the larger the output aperture satisfying the above conditions. Choosing the extreme ray of the emitted input light determines the reflector size and thereby, the size of the illuminating pixel.

[0068] Examples of reflector sizes can be calculated directly from these expressions, with reflecting plane 164 at the prescribed height 174 above the vertex point O, F2/2. As one illustration of this, suppose we wanted to place a 0.5 mm×0.5 mm LED having a +/−60-degree angular spread in input aperture 152 of a hyperboloidal surface of revolution with rectangular (or square) truncation. With hyperboloidal parameters F=38 and A=32, the reflector reaches a rim height of almost 2 mm at the final reflection point B in FIG. 9 . The semi-diameter of this point is about 7 mm, meaning that the output aperture is approximately a 10 mm by 10 mm square. In this case, the reflective polarizing plane 164 is at a height of 4 mm above the reflector's vertex. This configuration would be accomplished when the phase retardation substrate thickness is about 1 mm. In this case, and without an output lens, the maximum output angle, Y, is just less than +/−6 degrees, and the reflector, almost parabolic. The conic constant is −1.09, which if an ideal parabola would be −1.

[0069] The pixel's output brightness to a viewer depends, as always, on the angular spread over which the lumens are distributed. The wider the angular spread, the wider the range of possible viewing directions, but the lower the brightness. The narrower the angular spread, the higher the brightness, but the more limited the range of possible viewing directions. Layer 184 in FIG. 9 , a light spreading light scattering layer, is used to set both the pixel brightness and the angular extent.

[0070] 2.1.4.2 Non-Imaging Optic Reflector Elements

[0071] Specularly reflecting sidewalls 136 mathematically shaped so that the number of sidewall reflections a ray experiences between input and output apertures are minimized, and so that an even distribution of output power is created throughout the aperture, are generally known in the prior art as non-imaging concentrators (and sometimes as compound parabolic concentrators). A two-dimensional array of such reflectors arranged similarly to the array conveyed in FIG. 5 can be used to collect light from an array of input light emitters while generating a cohesive output beam 100 whose angular range has been restricted by the details of the sidewall design. More generally, such ideal power transfer can be arranged to behave as an array of θ _i /θ _o concentrators, in that the collective array transforms input light of maximum angle θ _i to output light of maximum angle θ _o by virtue of the well established Sine Law: A _i Sin ² θ _i =A _o Sin ² θ _o , where A _i is the area of each individual emitting region, A _o is the area of each individual output aperture, and θ _i and θ _o the respective input and output half angles. Such ideal etendue preserving designs, even for an array, transfer the brightness (and uniformity) of the source apertures, which in this case are the set of well-separated emitter regions 24 , to the collective output aperture made up of the sum of individual output apertures 102 . Less ideal sidewall designs, such as for example, the linearly tapered walls of FIGS. 3 and 7 when used without prism sheets 58 and 60 , may transfer nearly the same output power as the ideal designs, but spread that power over a larger than ideal angular range, and show greater levels of spatial non-uniformity than the ideally shaped sidewalls would.

[0072] Such non-imaging micro reflector array configurations are most beneficial when each micro reflector's emitting aperture is relatively small (less than 5 mm on diagonal) and when asymmetric output angles are desired in the two output meridians. When the emitting aperture is larger than this (i.e. more than 5 mm on diagonal), the non-imaging concentrator approach leads to reflector depths that may be considered too large for practical multi-layer systems in many situations.

[0073] One example of a potentially beneficial use of a non-imaging reflector shape is provided by a two-dimensional array of 0.5 mm square emitting apertures 24 , such would result from the wide-angle outputs of light emitting diode (LED) chips. When output angles of +/−22.5 degrees and +/−17.26 degrees are required in the two output meridians, the reflector's output aperture in accordance with the Sine Law becomes 0.5/Sin (22.5) or 1.31 mm and 0.5/Sin (17.26) or 1.69 mm. This aperture size imposes a limit on the emitting array's density, which becomes in general, A _in /A _out , and in this example, only 11%. By comparison, emitter densities possible by means of the method of FIGS. 3, 4 and 7 are greater than 25%, with A _in /A _out becoming (W ² )/(2W) ² . Yet as will be discussed later, the throughput efficiency of a non-imaging reflector is potentially much greater than that of the two prism sheets 58 and 60 of FIG. 7 , which is approximately 50%. When the non-imaging reflector is a transparent dielectric whose reflecting walls are created by its air-dielectric boundaries, throughput efficiency as high as 90-95% is possible. When the non-imaging reflector is formed by metal-coated sidewalls 136 , throughput efficiency is lower, but often as high as 80%-90%. Ideal rays leaving the reflector's input aperture 24 strike only one sidewall a single time, leading to the high efficiency. Non-ideal rays may strike more than one sidewall, reducing theoretical efficiency. When each LED in the array contributes 20 lumens to the input apertures, the air-filled illustrative non-imaging array with 85% efficiency achieves 7.68 lumens/mm ² . Lumen density increases to 9.9 lumens/mm ² when the output aperture is 1.31 mm square. The same array covered by prism sheets 58 and 60 spaced for contiguous virtual images achieves about 10 lumens/mm ² .

[0074] Hence, despite the enlarged output aperture of a non-imaging reflector, the net lumen density possible for the same output conditions is about the same as that achieved with prism sheets 58 and 60 . The main tradeoff is layer thickness. The depth (or thickness) of a non-imaging reflector is governed by its input and output aperture sizes and the output angle along the same meridian. For the case where the output angles are +/−22.5 degrees in each output meridian, the reflector length becomes 2.86 mm. While this result is almost 5 times thicker than the comparable two-prism sheet alternative (which can be as thin as 0.3 mm for the two sheets plus the preferable (0.625)(0.5) mm gap spacing), it is still relatively thin for many application examples that follow.

[0075] Concentrator length can be truncated, but the larger the truncation, the greater the negative effect on power transfer efficiency and uniformity. The best way to reduce concentrator length, without compromising power transfer efficiency or uniformity, is to decrease the ratio of W to W′, W being the emitter width, and W′, the width of the non-emitting spaces between emitters. In the above example, W′/W is 2. If this ratio were reduced 33% to 1.5, and the 8 mm emitter width was maintained, the ideal pixel size would fall from 16 mm to 12 mm, making the space between emitters, 4 mm instead of 8 mm. The associated concentrator length drops 86% to 15.8 mm, which is still thicker than preferable in many applications.

[0076] The concentrator length can also be reduced by a folded hollow lens mirror combination much like that drawn in FIG. 9 , but with polarization conversion layers 164 and 170 replaced by a lens element. In this approach, some non-imaging ray paths are folded back towards the mirror by total internal reflection from the lens.

[0077] 2.1.4.3 Micro Reflector Fabrication

[0078] Whether using linearly tapered sidewalls or the mathematically shaped sidewalls nearer to those of either the hyperboloid or the ideal concentrator, light leaving the emitting region 24 enters the pixel medium 144 (air or dielectric) that fills the volume defined by specularly reflecting sidewalls 136 . When this medium 144 is a transparent dielectric such as acrylic or silicone, specular reflection occurs by total internal reflection at the sidewalls provided the gray shaded volumes 130 in FIG. 7 or FIG. 8 are filled with air or another lower refractive index material. When the gray shaded volume 130 in FIG. 7 is made of a transparent material, an opaque material, or one that has low specular reflectivity, it must be coated with a thin specularly reflective layer or film such as for example, aluminum, enhanced aluminum or silver, to provide the basis for efficient specular reflection. Once the smoothly shaped sidewalls 136 are coated, all light rays 140 that strike them will be reflected with an efficiency determined by the reflectivity of the coating, and these rays 142 will generally exit without further reflection through the structures output aperture 138 within a prescribed range of output angles as bounded by the Sine relation above.

[0079] As described earlier, the reflective spacer structure 84 (in FIG. 3 and FIG. 7 ) or 130 in FIG. 9 can be fabricated as a plastic sheet using a forming tool 146 such as the one represented schematically in FIG. 6 . Whether casting and curing, embossing injection molding, or compression molding, the cured or cooled plastic or plastic-composite sheet can be pulled away from the linearly or functionally tapering sidewalls 148 of the tool 146 without interference. Each element in the tool has a base width 154 , W+W′, and a top surface 150 of width 156 , W. The salient molding tool dimensions 154 , 156 , and 158 in FIG. 6 are traditionally made slightly greater (or less than) the target dimensions reflected in FIG. 3 , FIG. 7 , FIG. 8 , and FIG. 9 to allow for any process expansions and shrinkages. When casting or embossing, top surface 150 is made to extend slightly beyond the specified spacer heights G1′ and G6 as given in FIG. 7 and FIG. 8 (i.e. L-G1′ or L-G6). The reason for this is to assure that the process yields a clean clear hole in the molded sheet that matches the size of the emitting region. When casting, the casting material is filled to stop line 159 in FIG. 6 . When embossing, tool 146 actually protrudes through the (L-G1′) or (L-G6) mm thick sheet to be embossed into a compliant carrier film material attached to it.

[0080] 2.2 Types of Emitting Arrays Covered

[0081] In general, the present invention applies to one-dimensional arrays of emitting stripes (FIGS. 1 - 2 ) and to two-dimensional arrays of emitting regions (FIGS. 3 - 9 ).

[0082] Preferable one-dimensional emitting arrays are sets of parallel fluorescent tubes or channels, parallel fluorescent tube lamps, a lengthy single fluorescent tube lamp bent (or molded) into a serpentine pattern whose major sections run parallel to each other, or a planar device within which a gaseous plasma has been forced into concentrated stripe-like or zone-like regions. This emitter type is best suited to specialized LCD backlighting applications, as will be illustrated by subsequent example.

[0083] Preferable two-dimensional emitting arrays are spatial arrangements of discrete emitting regions, including planar arrays of pre-packaged LEDs or bare LED chips. These discrete arrays may be a single line of equally spaced elements or a series of equally spaced lines of equally spaced elements.

[0084] Emitter elements within the array whether fluorescent stripes or discrete LEDs, are powered (separately or in groups) by external controlling electronics. The controlling electronics for fluorescent stripes is a ballast supply that provides the high frequency form of high voltage needed to create and maintain the emitter's gaseous discharge. The controlling electronics for LED arrays is a switchable source of low voltage dc interconnected to sets of LEDs having the same color, leading to widespread uses in lighting and illumination—applications that will be described by specific examples below. The controlling electronics may also be applied to individual LEDs via an image processor circuit (or circuits) that determines proper timing, emission duration, and power-level (color balance) for each LED (or LED sub-group) in the array. Individually powered LED arrays lead to applications in the display of two-dimensional images.

[0085] The range of lighting applications enabled by LED arrays within the present invention are extensive and will be considered in detail, including preferable packaging arrangements and application examples. After this, an example will be given for the use of fluorescent stripes and tubes in the special cases of LCD and transparent image backlighting.

[0086] 2.2.1 Pre-Packaged LED Arrays

[0087] Commercial LEDs can be arranged in arrays, but the output of the array is ordinarily very non-uniform and spread over a wide range of angles. Lenses and diffusers can be used to improve uniformity and directivity, but such improvements reduce efficiency.

[0088] With the present invention, in the form of FIG. 7 ( 8 or 9 ), commercial LED arrays can produce uniform beams of light in thinner structures and with higher efficiency than conventional alternatives.

[0089] A wide variety of LEDs are presently manufactured in two types of packages: clear plastic elements of a wide variety of sizes and shapes, or 1.5-3.0 mm square ceramic surface mounts suitable for conventional printed circuit boards. The physical package size determines how many lumens can be supplied in a given application area. The package's optical design determines the characteristic with which the lumens are emitted. Some packages have a lens that causes a more directional emission of light. Other packages emit light in all directions. All present packages are larger than their emitting apertures, making light strongest from the center of the package and giving conventionally packaged LEDs a point-like appearance.

[0090] FIG. 10 provides one example of the way commercially packaged LEDs can be used within the present invention. Discretely packaged LEDs (or groups of LEDs) 157 can be used as the array elements themselves (i.e. 36 in FIG. 7 ) by soldering their discrete packages 161 in equally spaced rows and columns on a printed circuit board 163 or a ceramic circuit substrate, and then arranging appropriate spacer layer 165 and diffuser layers 167 , so as to best implement the cross-section of FIG. 7 ( 8 or 9 ). Bus bar circuitry 169 (anode) and 171 (cathode) is provided for each type of LED used. For simplicity, the circuit illustrated in FIG. 10 is for a single type of LED, such that all LEDs in the array are powered simultaneously. More complex circuitry is provided when each package 161 contains separate red, green and blue LEDs.

[0091] The specific example of FIG. 10 presumes the use of commercially available 3 mm square ceramic surface mount packages 161 such as those manufactured by Nichia whose 2.3 mm diameter circular cavity 173 contains an interconnected LED chip and is encapsulated with optically transparent plastic epoxy 175 . Exploded detail 141 shows the structure of an idealized 6-package by 6-package array where the spacing between packages 161 is equal to (or less than) their physical width, as described above in conjunction with FIG. 4 . Preferably, cavity 173 is better shaped as a square. When this is not possible, diffusive reflecting layer 167 is combined with a matching array of diffusing screens 177 disposed just above each package 161 such that diffusion screens 177 become the actual illumination source from each underlying package 161 .

[0092] Exploded detail 141 in FIG. 10 also shows the sequence of multi-layer optics arranged according to the approach of FIG. 7 that is used to create the uniform output beam being sought. In this particular example, transparent spacer layer 165 is positioned directly above the emitting apertures 177 to provide the exact spacing needed between the emitting apertures and prism sheet layers 58 and 60 (G1′ as in FIG. 7 ). Prism sheet 58 may be optically coupled (laminated) to transparent spacer layer 165 to minimize any unrecoverable power losses due to total internal reflection within the spacer. The collapsed multi-layer illuminator is shown in detail 143 .

[0093] Light is emitted over uniformly over the full aperture of multi-layer illuminator 143 , which for the illustrative 3 mm packages is 36 mm by 36 mm.

[0094] The same conventional packaging approach may be used for just a single row of packaged LEDs as illustrated by FIG. 10 in details 145 and 147 . Exploded detail 147 shows the same vertical layout applied to 6 equally spaced LED packages 161 . In this case, the full aperture size is the width of 2 packages and the length of 12 packages. Hence, using the illustrative 3 mm packages 161 , and their diffusive output layers 177 , output light would emit through layers 58 and 60 over a 6 mm by 36 mm region. Using prism sheets 58 and 60 with 90-degree prisms, the output light would be spread over substantially a +/−22.5-degree angular cone.

[0095] Array illuminators 143 and 145 can be used in a variety of lighting and backlight applications, including the red high mount stop lamp on the rear deck of automobiles. For such use, the size of the array and the type of diffusive layers added are adjusted to meet the visual needs of the market. Other applications of these arrays will be described shortly.

[0096] The main practical limitations associated with conventional packaging described in FIG. 10 is the physical limit they impose on the number of lumens that can be delivered per unit area density and the wasteful redundancies of discrete packaging leading to higher than necessary manufacturing costs.

[0097] These limitations are addressed by introducing a more preferable packaging arrangement, one in which the constituent LED chips are contained in what becomes a single extended package.

[0098] 2.2.2 Monolithically Packaged LED Arrays

[0099] Best use of the present inventions (FIGS. 3 , 7 - 9 ) occurs when constituent LED chips are arranged in a monolithically laminated multi-layer package.

[0100] A distributed manufacturing approach is adopted wherein there is but a single continuous package structure accommodating every LED arranged in a super-array, containing many sub-arrays. This approach is more inefficient than using discrete printed circuit boards 163 and discretely packaged LEDs 157 , as has become common practice, or even extended electronic circuit boards with individually die-bonded LED chips, discrete conventional optics glued above them. The multi-layer invention of FIG. 7 , for example, can be implemented using very large (if not continuous) sheets or panels for each and every layer, with no need for the inefficiency of handling discrete elements, other than the LED chips 70 , themselves. This distributed multi-layer packaging approach is shown schematically in FIG. 11 , with multi-layer composite panel 181 much larger in physical extent than any constituent sub-panel that is to be used as a yielded product. Unlike the discrete circuit boards 163 and packages 157 of FIG. 10 , the approach of FIG. 11 is more akin to the multi-layer planar processing used in silicon microelectronics, wherein the distributed multi-layer microelectronic wafers are later diced into individually yielded devices with advantageous economies of scale. With similarity, overall multi-layer composite panel 181 is later cut or sliced into separate sub-panels 196 (along preplanned slicing lines 191 , which may in turn be reduced to even smaller illuminating entities such as bars 183 and plates 179 . Layers 163 , 167 , 165 and 58 are ruggedly laminated.

[0101] Similar attachment of layer 60 above layer 58 is complicated by the need to maintain an air (or low refractive index) gap between them over the output aperture. One solution is to apply a bonding agent between layers 58 and 60 only in the dead regions surrounding the effective sub-array apertures, with these same dead regions exceeding the width of cut lines 191 . Another solution is to add pre-cut pieces of layer 60 , and any output diffusing layer 28 , as a post process prior to use. Yet another solution is to choose prism refractive index and geometry in layer 58 , spacing G1′, and the space allowed between the prisms of layers 58 and 60 anticipating a transparent low refractive index media adhesive or epoxy filling the gap between layers 58 and 60 , rather than air. Fluorinated polymeric liquids manufactured, for example, by Addison Clear Wave LLC or DSM Desotech, can be polymerized with refractive indices as low as 1.42. Prism elements can be formed in acrylates and other polymer materials with refractive indices as high as about 1.7.

[0102] The distributed manufacturing approach symbolized by the multi-layered panels or sheets of FIG. 11 only pre-suppose a practical method for distributing and incorporating large numbers of LED chips efficiently within them. Although conventional pick-and-place methods are compatible with this approach, it would be preferable to place the LED chips in the extended arrays in a collective rather than individual manner. Collective attachment methods are enabled by recent advancements in LED technology creating availability of LEDs with transparent substrates having both electrical contacts on the same side of the chip (allowing so-called flip chip mounting). Such one-sided LEDs can be soldered to metallic circuit elements en masse by heating generally to re-flow deposited solder contacts for all LEDs at the same time. Collective LED placement is enabled by the continuous packaging structure envisioned herein, and introduced further below.

[0103] Practical applications vary with the density of illuminating pixel apertures ( 1 . 8 in FIG. 11 ; 102 in FIG. 7 ), the number of lumens provided by each pixel aperture, and the size and shape of the resulting panel. Some general lighting applications are offered by the present invention used with discrete LED packages, as illustrated by way of two examples that follow. Yet, there is a much wider variety of lighting applications made possible by the distributed packaging approach that will be addresses through additional discussions and examples.

[0104] 3.0 General Lighting Applications with Pre-Packaged LEDs

[0105] Mono-colored light emitting diodes (LEDs) are usually 0.5 mm to 1.0 mm square chips cut from 2″ diameter wafers, 6-10 thousandths of an inch thick (0.010″=0.254 mm). Although the diode itself is formed by epitaxial-layers grown on top of the square substrate's surface, light is emitted from the entire chip, which is preferably transparent. While such a chip makes an ideal emitting region 70 , manufacturers prepackage it with wires attached, either in a clear plastic bullet-shaped molding, or as contained on a small ceramic circuit board. In either case, the discretely packaged LED can be arranged to emit through a square emitting aperture, and organized with companion LEDs into a planar array that would be favorably treated by the present invention. As such, an array of pre-packaged LEDs implemented as in FIG. 7 or FIG. 8 could be used, at least in principal, in a variety of practical general lighting application.

[0106] As one of the many general lighting applications possible for an illuminator of the form of FIG. 7 , consider the case where each conventional package element 161 ( FIG. 10 ) contains one each of a state-of-the-art red, green and blue LED 70 , and that the array of pixels is arranged as in FIG. 10 , details 141 (exploded) and 143 (collapsed). Suppose each LED group has an output aperture 24 that is made square and 3 mm on a side, with spacing W′ between all emitting squares 24 also 3 mm. Total thickness of multi-layer 143 is approximately 3-3.5 mm, including the 1 mm thickness of LED packages 157 , spacer thickness G1′ between emitting apertures 177 and prism sheets 58 and 60 , and the combined thickness of layers 58 and 60 .

[0107] Spacer thickness G1′ for the contiguous output images of FIG. 4 is about 0.625W (or 1.875 mm). High-performance semiconductor LEDs, such as those manufactured by LumiLeds, yield approximately 20 lumens per die at drive powers in the range of 0.25-0.35 watts dc. Assuming adequate heat sinking, and an approximate optical transfer efficiency of 50% from output apertures 218 , means that approximately 30 lumens of mixed red, green and blue light could be yielded from each pixel's output aperture. As industry advancements in the number of lumens per die, L _d , are made over time, n dies are used per pixel, and as the optical transfer efficiency, η, is optimized, the number of yieldable lumens per pixel, n L _d η, may become significantly greater than 30.

[0108] 3.1 LED Equivalent of 100-Watt Light Bulb

[0109] Yet, with 30 RGB lumens per 6 mm by 6 mm illuminating pixel 218 , the luminous equivalent ( 1690 lumens) of a 100-watt General Electric SoftWhite™ light bulb can be achieved with only 56 discrete LED packages 157 and a total of 168 light emitting diodes. If arranged in a nearly square 7 pixel by 8 pixel array, the resulting panel would be 42 mm×48 mm, and less than 4 mm in overall thickness T′, depending on the various layer thickness and offsets used. Such a compact lighting element 143 , represented schematically in FIG. 10 , would have many practical uses, as its life expectancy alone exceeds that of incandescent lamps like the General Electric SoftWhite™ light bulb by more than 100 times. With its 168 diodes driven at 0.25 watts, the total electrical power dissipated would be 42 watts. In addition, the color temperature of the white light emitted by the mixture of red, green and blue LEDs is adjustable electronically, allowing user selectable panel color.

[0110] 3.2 LED Equivalent of 75 Watt PAR-30 Flood Lamp

[0111] As a related example, consider GE's industry standard 75 watt, wide-angle halogen floodlight PAR-30, which delivers 1050 lumens over a useful life of 2000 hours. Using the same configuration and dimensions as just above, equivalent performance can be achieved with the 6-element by 6-element array 143 illustrated in FIG. 10 . Outside dimensions are 36 mm by 36 mm, and electrical power, 27 watts.

[0112] The current worldwide market for all light bulbs is over 1 billion units per year. For solid-state lighting structures of any kind to serve even a small share of this market, manufacturing costs must be reduced towards comparable levels with existing light bulb technologies. Not only does the distributed multi-layer packaging envisioned in FIG. 11 address this need, but facilitates panel combinations such as back-to-back arrangement 187 in FIG. 10 and five-sided lighting cube 189 .

[0113] 4.0 High Lumen-Density Light Sources Panels with Monolithic LED Packaging

[0114] The distributed packaging of LED chips within the context of the present invention enables a new class of high lumen-density light sources that potentially replace high-lumen light bulbs currently in use within many commercial application, including video projectors, spot and flood lighting luminaires, automotive head and taillights, to mention just a few.

[0115] 4.1 LED Light Sources for LCD and DMD Video Projectors

[0116] The most demanding application example for monolithically formed LED light source panels formed by the present inventions involves replacing the 90 to 150 watt halogen arc discharge lamps used in all LCD and DMD front and rear image projectors with comparably performing LED light source panels anticipated, for example, by FIGS. 7 and 11 . Applying the present invention to LCD and DMD projectors, however, requires a denser packing of LEDs per unit area than any imagined general lighting or illumination need. The reason for this is that the total illumination provided by the LEDs in a projector must pass through an image aperture that is typically less than about 18.3 mm×