1. Introduction

Figure 1 shows the system architecture of the single panel projectors used in this study. A mechanical layout is given just below the system raytrace for clarity. The raytrace is also divided into two systems: an illumination system and an imaging system. The illumination system raytrace features lenslet homogenization where the lamp arc is conjugate to the second array which is conjugate to the pupil of the imaging lens. Thus the imaging lens pupil will contain many images of the arc. There is also a polarization conversion system at the second lenslet array to improve throughput. The column of light emitting from a lamp with parabolic reflector is modeled by an slightly extended point source that is collimated by a paraxial lens. The imaging system is represented by two paraxial elements and a stop. This represents a telecentric projection lens with efl=32.5 mm. and bfl=59.5 mm.



Figure 1. System Architecture

The ad hoc procedure used to align the first prototype projector was as follows:

1) perform rough mechanical alignment of system components, 

2) focus condensor and adjust magnification of light body at panel,

3) shift panel to light body,

4) steer panel tilts toward center of PBS,

5) mount lens so projected image is reasonably sharp and uniform, check centration of light on front element, check centration of illumination in lens pupil, and

6) tilt panel so that pixels on screen are sharp.

A system aligned in this manner is said to be aligned to "breadboard" tolerances. These are hypothesized to be +/- 1mm. and +/-1 degree, and will be used in future error budgets.

 

2. Performance Comparison

Figure 2 shows contour maps of lux on a 50" by 30" screen. The projector with the optimized alignment has much greater luminous flux. Integrating over area gives 82 lumens, a 28% improvement in throughput. Contrast improved 36% to 102:1. Defocus due to field tilt and blurring due to laterally displaced ghost images was eliminated. Although absolute performance of these systems is limited by the use of mismatched off-the shelf components, this data illustrates dramatic relative performance improvements due to alignment alone.

Figure 2: Performance Comparison

3. Performance Budget - Sharpness

Figure 3 shows typical through focus MTF performance of a high performance projection lens ( design form shown above).



Figure 3: Typical Through Focus MTF at 38 lpm

The figure shows that in order to preserve at least 50% modulation for all field points, the display panel must reside in an envelope only 53um deep. This depth can be transformed into a tilt budget for rotations about the short axis of the display panel. For a panel 1280 pixels wide and pixel pitch of 13.2 um, the maximum tilt is given by:



Since the current architecture folds the imaging system with the PBS, we consider angular errors of the display panel () and of the PBS ()



Allocating just 1 mrad angular error to the panel and PBS consumes the entire sharpness budget. Since the ad hoc procedure cannot produce angular alignments this precise, it is no surprise that pixels are not well modulated on the screen. Moreover, a ghost image, displaced laterally by 5 pixels indicates much more severe angular errors.

4. Performance Budget – Throughput

The throughput budget of Table 1 tests the hypothesis that "breadboard" tolerances of +/- 1mm and +/- 1 degree can actually affect throughput by as much as 28% in the current system.

A number of factors can cause angular errors to accumulate and, consequently, cause vignetting and/or coating losses to occur. Using raytrace techniques, a sensitivity for each factor has been determined. A linear model is used to accurately estimate throughput losses.

Vignetting factors considered include lateral shifts of condensor elements and tilts of the display panel and PBS. Also considered are vignetting factors that cause magnification changes such as focus shifts of the condensors. Vignetting may occur in the projection lens, at the panel clear aperture (or mask ) and, in poorly aligned systems, at internal baffles.

Throughput losses also occur because of PBS coating performance. Manufacturer’s data sheets indicate even high performance PBS’s have a throughput loss of about 3.4% per degree of skew. Angle of incidence related coating losses are combined with vignetting losses to estimate total throughput losses in Table 1.

Table 1. Throughput Budget



The total throughput loss, corresponding to "breadboard" tolerances of +/- 1mm and +/- 1 degree, does indeed budget to just over the 28% seen in the system with the ad hoc alignment.

5. Performance Budget – Contrast

In the current architecture (Fig 1), dark pixels reflect P-polarized light off the PBS. Although much of this light is extinguished by a clean-up polarizer, some of this light leaks through and reduces contrast.

Using PBS manufacturer’s transmission vs. angle data, we can determine the value of R_p or P-leak and estimate contrast.

In the ad hoc system, (with "breadboard" tolerances of Table 1) incident light on PBS (2nd pass) is skewed 3.85 deg, giving R_p (P-leak) of 5% of 64 lm. The cleanup polarizer has 75% extinction so 0.85 lm reach the screen, yielding a contrast of 75:1.

In the optimized system, R_p (P-leak) is 4% of 82 lm or 3.2 lm, After cleanup, 0.8 lm reach the screen, yielding a contrast of 102:1

 



6. 10 Step Alignment Procedure

The alignment procedure employed in optimization utilizes: a cone-mounted alignment telescope, a laser pointer, and an optical rail as depicted in Figure 4.

 

 

Figure 4. Alignment Setup

 

Figure 5 gives images as seen through the alignment telescope during different steps, as follow:

1. Set up Laser Reference Axis, parallel to rail. Adjust beamsplitter so return spot from reference mirror is always coincident with 1^st pass spot, even during carriage translation.

2. Align Telescope to Reference Axis (Fig. 5a).

3. Remove projection lens, establish projector PBS perpendicularity.

4. Prep front of projection lens with reticle wire, mount lens and bring lens axis into coincidence with reference axes (Fig 5b). Also bring multiple return spots into coincidence (Fig 5a).

5. Remove projection lens and detilt display panel (lightvalve off). Detilt by bringing laser return spots into coincidence (Fig. 5a). Turn lightvalve on and center condensor lenses. Examine for centration by taking scope through focus and examining laser spots on condensor and lenslet surfaces.

6. Turn on lamp, focus and set magnification of condensor system. Examine light box directly with scope (Fig 5c). Shim lamp as necessary to center lightbox.

7. Center panel with respect to reference axis by using a display panel target and bringing it into coincidence with scope crosshairs.

8. Re-Mount lens and double check centration by examining display panel target.

9. Check conjugacy, magnification, and centration of lenslets in lens pupil (Fig. 5d).

10. Reverse scope in cone mounts. Mark center of screen. Check that mark and fiducial conicide in projected image.

 

7. Conclusions

A rigorous alignment procedure such as the one given here can dramatically improve performance of a breadboarded projector simply by addressing alignment errors. Throughput and contrast

improvements of 30% or more are possible. Image sharpness will also benefit greatly.

 

 

 

8. Acknowledgements

Special thanks to Jeff Frisk and Displaytech, Inc. for use of the single panel projectors and assistance evaluating performance. Thanks also to Dave Jenkins of Radiant Imaging, Inc. for his assistance with the Prometric CCD Photometer used in these evaluations.