Professional Documents
Culture Documents
terminology to create design tools for a wider variety alistic numerical models or on measurement-based
of illumination systems. models.15–19 Although these time-consuming models can be
4. We approximate the irradiance variation with dis- useful for trial and error designs of light sources consisting
tance with the inverse square law for a point source of multiple LEDs, a practical model is required to analyti-
because the emitting region of LEDs is typically less cally derive quick estimation formulas. We model each
than 1 mm on a side. This approximation is true even single LED as an imperfect Lambertian emitter, which
for the midfield region and for some part of the near
means the radiant intensity distribution is a power law of
zone.
5. The distance to the target is large enough compared the viewing-angle cosine function.5,13,20
to the spherical array radius to warrant the far-field Because we assume that the illuminated object is a flat
approximation. surface, we write the irradiance pattern for a LED displaced
6. All LEDs of each array have equal values of radiant to position 共x0 = R sin cos , y 0 = R sin sin , z0
flux and equal distributions in space and wavelength. = R cos 兲 over the outer surface of a sphere with radius R
However, because distributions of irradiance and in terms of Cartesian coordinates 共x , y , z兲 共see Fig. 1兲. The
color of assembled LEDs are not exactly equal, even irradiance E 共W m−2兲 over every point 共x , y兲 on a flat
among LEDs of the same type,14 this should be con-
screen at distance z from the LED array may be expressed
sidered for an exact design.
as
Traditionally, a single LED is optically modeled on re-
关共x − R sin cos 兲sin cos + 共y − R sin sin 兲sin sin + 共z − R cos 兲cos 兴m
E共x,y,z;R, , 兲 = ALEDLLED , 共1兲
关共x − R sin cos 兲2 + 共y − R sin sin 兲2 + 共z − R cos 兲2兴共m+2兲/2
where LLED is the radiance 共W m−2 sr−1兲 of the LED chip ance twice and setting it to zero at the target center 共x = 0,
and ALED is the LED emitting area 共m2兲. The number m is y = 0兲 eventually yields the maximally flat condition for the
given by the angle 1/2 共a value typically provided by the angular spacing.
manufacturer, defined as the view angle when radiant inten-
sity is half of the value at 0 deg兲 3.1 Array of Two LEDs
For this array 关Fig. 2共a兲兴, the irradiance E is given by the
− ln 2 sum of the irradiances for two LEDs 共the angle between
m= , 共2兲 LEDs is 2兲.
ln共cos 1/2兲
where m depends on the relative position of the LED emit- E共x,y,z兲 = E共x,y,z;R, ,0兲 + E共x,y,z;R,− ,0兲. 共3兲
ting region from the curvature center of the spherical Differentiating E twice and setting E / x = 0 at x = 0 and
2 2
encapsulant.21 If the chip position coincides with the cur- y = 0 yields the maximally flat condition for
vature center, the number m is nearly 1, and the source is
nearly a perfect Lambertian 共e.g., some Lumileds® and m共m − 1兲共m + 2兲−1关R2 sin2 + 共z − R cos 兲2兴2
Lamina® LEDs兲. Usual LEDs often have a larger value of
m, and thus become more directional emitters 共e.g., some ⫻关z cos − R兴−2 + 2mR关R2 sin2 + 共z − R cos 兲2兴
Nichia® LEDs兲. ⫻关z cos − R兴−1 − sin−2关R2 sin2 + 共z − R cos 兲2兴
+ 共m + 4兲R2 = 0, 共4兲
3 Designs for Uniform Far-Field Irradiance If numerical values for z and R are provided, Eq. 共4兲 may
We apply the theory of the optimal lamp placement to pro- be easily solved for the optimum value of for both near-
duce uniform illumination.22 This theory consists of setting and far-field cases. For the far field z ⬎ ⬎ R, Eq. 共4兲 can be
several terms in the two-dimensional Taylor series expan- significantly reduced giving the condition for uniform irra-
sion of the resulting irradiance function to zero. Arranging diance
冉冑 冊
the LEDs symmetrically, only one series term is sufficient,
which is the equivalent to applying the Sparrow’s criterion m+2
共used in image resolution兲.5 Using this criteria, the angle o = arctan . 共5兲
m共m − 1兲
between each pair of LEDs can be adjusted so that the
combined irradiance distribution is uniform, that is, the in- This design condition is independent of z 共the distance from
dividual irradiance patterns are separated optimally to the center of the array to the target center兲. Equation 共5兲
eliminate the minimum between the maxima from each pair and the following equations are approximately valid for
of distribution. In other words, differentiating total irradi- panel-target distances ⱖ20R.
冉 冊 冉冑 冊
N
2n
E共x,y,z兲 = 兺 E x,y,z;R, , , 共6兲 o = arctan
2共m + 2兲
. 共7兲
n=1 N m共m − 1兲
where 2 is the angle between each pair of LEDs. The This condition is again independent of z and curiously in-
symmetry makes the problem one dimensional so that with- dependent of the number N of LEDs that assemble the ring.
out loss of generality we can calculate the maximally flat Therefore, the number of LEDs that assemble a ring does
condition along any radial axis on screen, such as the x axis not appreciably affect the uniformity of the illumination. A
at y = 0. Differentiating E twice and setting 2E / x2 = 0 at similar conclusion for both point22 and LED5 sources, uni-
x = 0 and y = 0, yields the maximally flat condition. For the formly spaced along a coplanar circle, was previously
far field, the optimal angle o is proven. Figure 3 shows the design of a circular ring array
o = arctan 冉冑 2共2m + 1兲
共m − 2兲共m − 1兲
冊
. 共8兲
⌽o = N 冉 m2 + m + 4
m2 − 4
冊
cosmo . 共9兲
= 2. We proved several intuitive functions for n = f共n , 0兲,
Figure 4 illustrates the uniform irradiance pattern for a ring where 0 is the angle between the optical axis and the near-
array with one LED in the center. We included the curve est LED. However, due to the asymmetry between the ge-
共for m = 30兲 of Fig. 3共c兲 in Fig. 4 for comparison purposes. ometries of array and target, the functions that yielded a flat
We can appreciate that for the ring with a centered LED, pattern over the largest target region were
the irradiance distribution is uniform over a larger region
than the pattern produced by a simple ring. n = arctan关共N + 1 − 2n兲tan 0兴, N even,
兺
i=1
关m共m − 1兲tan2i − 2共m + 2兲兴Ni cosmi = 0, 共14兲
再
E共x,y,z兲 = 兺 兺 ⌽iE x,y,z;R, i,
i=1 n=1
2n
Ni
.冎 共16兲
兺
i=1
关m共m − 1兲tan2i − 2共m + 2兲兴⌽iNi cosmi = 0, 共17兲
M
d
兺
i=1
关2共2m + 1兲 − 共m − 1兲共m − 2兲tan2i兴
d
i
再
E共x,y,z兲 = 兺 兺 E x,y,z;R, i,
i=1 n=1
2n
Ni
.冎 共13兲
We performed a simple demonstrative experiment with a
two-LED spherical array assembled with R = 3.3 cm 共radius
is measured from the origin of coordinates to the chip-
image position, which in our LEDs coincides with the en-
Setting 2E / x2 = 0 at 共x = 0, y = 0兲 gives the design condi- capsulant base兲. These LEDs 共Steren® 5/Ultra White兲 emit
tion for the far field white light with m = 64.66 共measured with respect to the
4 Conclusions
LEDs emit light into only one hemisphere with some de-
gree of directionality; therefore, a compact LED array must
be assembled upon a convex surface to distribute light over
a large area. We have analyzed the optimum LED-to-LED
angular spacing of light sources consisting of multiple
LEDs, assembled upon a spherical surface, to uniformly
illuminate a distant target. Practical equations and formulas
were derived for four representative array configurations.
These design tools offer an easy way to estimate the per-
formance of LED spherical lamps due to the explicit depen-
Fig. 7 Uniform irradiance pattern 共along the x direction at y = 0兲 for a dence on 1/2 共typically provided by the manufacturer兲, the
multiple-ring array of LEDs with different relative radiant flux for configuration geometry, and the number of LEDs that as-
each ring. The design parameters are M = 2 共two rings兲, N1 = 4, N2
= 8, m = 5, z = 300 cm, R = 7 cm. Using Eqs. 共17兲 and 共18兲, curve 共a兲 semble the array.
shows the resulting distribution for 1 = 0 = 15 deg, 2 = 56.3 deg, The derived analytical expressions can be a practical
⌽2 = 2.2, and k = 2.5. By using the values 1 and ⌽2 of 共a兲 as starting tool for both quick estimations 共first-order designs兲 and as
values, curve 共b兲 shows the resulting pattern after a quick process of starting points 共to reduce the computation time兲 for exact
trial and error for 1 = 0 = 18.9 deg, 2 = 62.7 deg, ⌽2 = 2.71, and k designs that must use a realistic LED model.15–19 Depend-
= 2.5. Curve 共c兲 shows the resulting distribution for an array with
equal values of flux for all LEDs, for 1 = 0 = 34.3 deg, 2 = 44 deg, ing on the application, our analysis can be extended to
and k = 0.5. other array configurations to generate uniform irradiance
over distant targets.
15. W. J. Cassarly, “LED modeling: Pros and cons of common methods,” midfield region,” Opt. Lett. 31, 2193–2195 共2006兲.
Photonics Tech Briefs 共Special Supplement in NASA Tech Briefs兲, 20. Y. Uchida and T. Taguchi, “Lighting theory and luminous character-
pp. IIa–2a 共April 2002兲. istics of white light-emitting diodes,” Opt. Eng. 44, 124003 共2005兲.
16. S. J. Lee, “Analysis of light-emitting diodes by Monte Carlo photon 21. I. Moreno, “Spatial distribution of LED radiation,” in The Interna-
simulation,” Appl. Opt. 40, 1427–1437 共2001兲. tional Optical Design Conference, G. Gregory, J. Howard, and J.
17. F. Hu, K. Y. Qian, and Y. Luo, “Far-field pattern simulation of slip- Koshel, Eds., Proc. SPIE 6342, 634216 共2006兲.
chip bonded power light-emitting diodes by a Monte Carlo photon- 22. M. A. Gennert, N. Wittels, and G. L. Leatherman, “Uniform frontal
tracing method,” Appl. Opt. 44, 2768–2771 共2005兲. illumination of planar surfaces: Where to place the lamps,” Opt. Eng.
18. A. Borbely and S. G. Johnson, “Performance of phosphor-coated 32, 1261–1271 共1993兲.
light-emitting diode optics in ray-trace simulations,” Opt. Eng. 44,
111308 共2005兲.
19. C. C. Sun, T. X. Lee, S. H. Ma, Y. L. Lee, and S. M. Huang, “Precise
optical modeling for LED lighting verified by cross correlation in the Biographies and photographs of the authors not available.