To evaluate the direct-emitting backlight with a specific LED binning distribution, the commercially mathematical software was utilized to accomplish this purpose. In the software, the remote phosphor sheet direct-emitting backlight system was set up as shown in Fig. 4-1, which composed of blue LED chips, reflector white, diffuser plate, phosphor sheet, diffuser sheet, BEF and DBEF. In the simulation, the module gap (h) and the period of blue LED chips (p) were both 30 mm. The range of blue LED peak wavelengths in the center bin (0th) was from 455.0 nm to 457.5 nm, and the others were every 2.5 nm nearby the center one. Besides, in the same conditions, a small-sized remote phosphor sheet direct-emitting backlight system (180 mm x 180 mm x 30 mm) with 36 blue LED chips (6 x 6) was demonstrated to verify the calculation model, as shown in Fig. 4-2 (a) and (b). According to the target of the thesis, the acceptable color deviation of whole backlight was 0.014. This required condition was strict, so the accuracy of computation was very important. In this section, the errors between
37
simulation and measurement were discussed step by step.
Fig. 4-1 Structures of simulated remote phosphor direct-emitting backlight model.
Fig. 4-2 Ten inch direct-emitting prototype of (a) blue LEDs; and (b) phosphor sheet and optical films with LED illuminating.
4.1.1
Correctness of LSF superposition method
The first step is to exam the correctness of the proposed LSF superposition method.
Fig. 4-3 (a) shows the experiment setup for verification. Considering the 3×3 LED
38
array, the calculation result is derive from the superposition of the nine LSF, as shown in Fig. 4-4. From Table 4-1, the color difference between measurement and calculation is 0.0007. The errors are caused from the limitation of the spectrometer dynamic range.
Here the electronic noise and straight light strongly affect the spectrum distribution of the superposition.
(a) (b)
Fig. 4-3 Spectrums measurement setup with turning on (a) 3×3 LEDs; and (b) 5×5 LEDs.
Fig. 4-4 Simulated spectrums superposition compared with measurement data.
=
39
Table 4-1 Color difference between simulated superposition and measurement.
4.1.2
Tolerance of self-fabricated module
Second, we found out the tolerance of the measurement set up. The deviation between different times measurement is 0.0012. There are two main reasons. One is the unstable of the self-fabricated module, and the small tilt angle (θ) between the panel normal and the axis of objective, as shown in Fig. 4-5. The other is the radiance tolerance of the instrument, where the noise rises as the increased exposure time.
Fig. 4-5 Measurement deviation by small tilt angle between panel normal and detector.
Color Coordinates In Center Place Measurement (3×3) 0.1905,0.4560
Calculation (3×3) 0.1908,0.4554
Difference 0.0007
40
4.1.3
Affected color composition range of LED
Thirdly, we compared the measured 5×5 and 6×6 data with the calculated 3×3 result to verify the approximation of 3×3 LSF superposition. We assume the LEDs of outer ring from 3×3 matrix wouldn’t affect the color composition in the center. The comparisons are shown in Table 4-2. Here the color difference is 0.0071 and 0.0130 respectively. The huge deviation shows that the 3×3 assumption is incomplete for this case. However, the calculated 5×5 case has small color difference from the measurement results. Therefore, the 5×5 approximation is more suitable to simulation the color deviation.
Table 4-2 Color difference between different considered LEDs and measurement.
4.1.4
Built-in tolerance from bin width
The forth step is to exam the calculation tolerance caused from bin width. We consider the situation that the 457.7 nm and 460.0 nm blue LEDs are all in 457.5~460.0 nm bin. The color difference between them by measuring 3×3 LEDs is 0.0034, as
41
Table 4-3 Color difference between two different wavelength LEDs in the same bin.
The real manufacture of 3×3 single bin LEDs, which composing varied peak wavelengths, was constructed for testing the built-in error, as shown in Fig. 4-6. The color coordinates measured from module is compared with the simulated condition, which LEDs peak wavelengths are all in the bin edge, as compared in Table 4-4.
Fig. 4-6 Peak wavelengths of LEDs in the same bin to test the built-in errors.
Peak wavelength (nm) Color coordinate (u’v’)
457.7 0.1901,0.4604
460.0 0.1887,0.4635
Difference 0.0034
Difference
(intensity variable±5%) 0.0043
x y z
Pitch x=30 mm
Pitch y
=30 mm Gap=
30 mm
42
Table 4-4 Built-in color errors by a width of peak wavelength in the same bin.
4.1.5
Boundary effect by light reflection
The last step is discussing the boundary effect. The color uniformity was influenced since the light reflected by the panel frame. A diffuse white frame is to cause the diffuse reflection, and an absorption black frame is to mimic the situation that there is no boundary, as shown in Fig. 4-7, Fig. 4-8 (a) and (b).
By measuring the backlight systems, the color difference between these two types is compared in Table 4-5, where ε is defined as the color error value. According to the results, the ε in the outer rings is larger than the inner locations, which are far away from the boundary. Also, the black boundary model is more similar to the proposed approach, as shown in Table 4-6.
Table 4-5 Color difference between the white boundary and black boundary modules.
Peak wavelength (nm) Color coordinate (u’v’) (1) Mixed LEDs of single bin 0.1886,0.4615 (2) 457.7 0.1901,0.4604
Difference between (1)& (2) 0.0019
(3) 460.0 0.1887,0.4635
Difference between (2)& (3) 0.0020
43
Fig. 4-7 Prototype with white boundary.
Fig. 4-8 Prototypes (a) without boundary; and (b) with black boundary.
(a)
(b)
44
Table 4-6 Color accuracy of simulation and black / white boundary prototypes.
4.1.6
Total difference between calculation and experiment
Fig. 4-9 Total differences between simulation and measurement in LUV color space.
All for all, combing all the steps of verification, we can find that the color deviation value by measuring 10-inch prototype is 0.0118, which is similar to the result of simulation one, 0.0196.
Moreover, by comparing the color difference between calculation and experiment, the average error value is ±0.0019 and the maximum error value is 0.0087, which is small enough to verify the accuracy of calculation model, as shown in Fig. 4-9.
Color errors with measurement Black boundary White boundary
Δu’v’ average 0.0041 0.0287
Δu’v’ maximum 0.0099 0.0338
Δu’v’ minimum 0.0003 0.0260
45
After comparing simulated results and measured data of prototype, the maximum error value is 0.0087 in CIE 1976 LUV color space, which is acceptable but could be reduced. Therefore, the simulated module was classified to many parts to analyze errors step by step. According to the results in this chapter, the color errors are small enough in the steps of correctness of LSF superposition method, and tolerance of self-fabricated module. Moreover, the tolerances of bin width are built-in errors, which could not improve anymore. However, the errors caused from affected color composition range of LED cannot be ignored. All these errors are attributed to reflected light by diffuser plate and reflector white. It means that the distance of emitted light guiding in module cavity is longer than predicted value, as shown in Fig.
4-10. Accordingly, to enhance the accuracy of proposed method, the affected intensity ratio of far LEDs must be considered when calculating the color coordinate in backlight system. Finally, designing the boundary effect, the black frame is to mimic the situation that there is no boundary. In this structure, the results are more similar to the proposed approach since there are without the unpredictable reflected light.
Fig. 4-10 Effected LEDs light by multi-reflection between reflector and diffuser plate.
46
Chapter 5
Results
After the verification of the calculation model, the color uniformity of backlight system could be computed handily in any conditions. The compositions of this chapter could be briefly classified to three parts. The first part is to calculate the color uniformity of backlight system with LED random arrangement. Furthermore, with the same conditions, the optimized best arrangement of different binned-LEDs designed to get the highest color uniformity is analyzed in the second part. Finally, some ideas of reducing gap were proposed.