E146 Vol. 54, No.

28 / October 1 2015 / Applied Optics Research Article

Freeform lens design for light-emitting diode

uniform illumination by using a method of
source–target luminous intensity mapping
JIN-JIA CHEN,1,* ZE-YU HUANG,1 TE-SHU LIU,1 MING-DA TSAI,1 AND KUANG-LUNG HUANG2
1
Department of Electrical Engineering, National Changhua University of Education, Changhua 50074, Taiwan
2
Department of Materials Science and Energy Engineering, Mingdao University, Changhua 52345, Taiwan
*Corresponding author: jjchen@cc.ncue.edu.tw

Received 30 April 2015; revised 10 July 2015; accepted 7 August 2015; posted 11 August 2015 (Doc. ID 240103); published 31 August 2015

We present a freeform lens for application to LED uniform illumination. This lens, which is designed with a
method of simple source–target luminous intensity mapping, can produce irradiance uniformity of greater than
0.8 and optical efficiency above 90% with an arbitrary half-beam angle greater than 45 deg. Typically, as com-
pared with a conventional source–target energy mapping method, this design method can achieve better optical
performance of lenses for general LED lighting. When a non-Lambertian-type light source is employed, for ex-
ample, the chip on board LED, the use of the method can result in a compact LED lens without losing the optical
performances of high irradiance uniformity and high optical efficiency as yielded by lenses for Lambertian-type
LED light sources. © 2015 Optical Society of America
OCIS codes: (220.4298) Nonimaging optics; (220.2945) Illumination design; (230.3670) Light-emitting diodes.

http://dx.doi.org/10.1364/AO.54.00E146

1. INTRODUCTION method [9]. This method is typically based on simple source–

In recent years, owing to the progress of LED fabrication
technology and advantages of energy-saving, long lifespan,
compact volume, high color-rendering index, and environmen-
tal benefit, high-power white-light LEDs have gradually taken
the place of conventional lamps to be the main light source for
various illumination applications. However, the semi-spherical
radiation pattern of LEDs renders them as hardly used directly
for uniform or collimating illumination, and a nonimaging op-
tic based on various conic surfaces or freeform surfaces is always
required to redistribute light rays from LED light sources to
satisfy the illumination request.
Owing to freeform optics capable of achieving high effi-
ciency and compact fixtures for LED lighting, many related
methods have been proposed in the past years, for example,
the tailored freeform surface method [1], the source–target
energy mapping method [2–5], the feedback modification
method [6], and the simultaneous multiple surface (SMS)
method [7]. However, most methods involve iterative calcula-
tion or complex Monge–Ampère equations, and this results in
complicated LED lens designs. To solve this problem, an initial
design with L2 Monge–Kantorovich theory was given by
Wu et al. [8]. With this initial design, the Monge–Ampère
approach can converge more stably and faster.
To obtain a highly efficient lens design, we propose a design
method with an approach similar to that of the functional
target luminous intensity mapping while without involving iter-
ative calculations and solving differential equations. Particularly,
when applied to Lambertian-type LED light sources, this
method can achieve an analytic form for the ray-angle relation
between the incident light ray and the output refracted ray with
an arbitrary output beam angle. Therefore, its algorithm is
simple and straightforward.
The primary design methodology was presented at the
International Optical Design Conference [10], and in
Section 2, we only give a brief description of the method. Its
applications to design freeform secondary lenses for uniform
illumination with Lambertian- and non-Lambertian-type LED
light sources are given in Section 3. The relative optical sim-
ulation results and optical performance of these two types of
LED light sources are also discussed in Section 3. To compare
the efficacy of this method with that of the conventional
source–target energy mapping method, a typical energy map-
ping method is given in Section 4, and the optical performances
of lenses designed with these two methods are compared.
Finally, a brief conclusion is given in Section 5.
2. METHODOLOGY
The proposed freeform lens design method is based on lumi-
nous intensity mapping between the LED light source and
the prescribed target intensity distribution. Relative geometric

1559-128X/15/28E146-07$15/0$15.00 © 2015 Optical Society of America

 Research Article
analysis and freeform-surface construction methodology are
briefly explained in this section. With a geometric analysis, the
tangential vector at each refractive point is calculated, and
the associated freeform lens can be constructed by using the
method [11] given by the authors.
A. Geometric Analysis
The basic geometric analysis for an arbitrary refractive surface,
which is used to redirect a light ray emitted from a light source
S with a ray-angle of θI to a specific azimuth angle θO , is
depicted in Fig. 1. From this figure, we have the following
relations:
θt
θT − θO
and
θi
θT − θI ;
where θi is the incident angle, θt is the refractive angle, and θT
is the slope angle of the tangential vector T̄ at the refractive
point P. According to Snell’s law, θi and θt should satisfy the
following relation:
n1 sin θi
n2 sin θt ;
where n1 and n2 are refractive indices in the incident and the
refractive spaces, respectively.
Substituting Eqs. (1) and (2) into Eq. (3) and rearranging
the resultant equation, we obtain the tangential slope angle at
the refractive point P as


−1 n1 sin θI − n2 sin θO
θT
tan : (4)
n1 cos θI − n2 cos θO
Therefore, the tangential vector at point P can further be ex-
pressed as
 
n sin θI − n2 sin θO
T̄
1; tan θT 
1; 1 ; (5)
n1 cos θI − n2 cos θO
where the incident ray-angle θI is arbitrarily given, and the
refracted ray-angle θO is calculated based on source–target
luminous intensity mapping. The detailed calculation of the
refracted ray-angle θO is given in the following subsection.
B. Freeform Surface Construction for Constant
Illumination
According to the basic photometry principle, a light source,
which produces constant illumination on a target plane of
Fig. 2, should satisfy the following relation [12]:
Fig. 1. Geometric analysis for an arbitrary refractive surface.
Vol. 54, No. 28 / October 1 2015 / Applied Optics E147
(1)
Fig. 2. Geometric relation of a light source and a target plane with
(2) constant irradiance.
Eθ Iθ

cos3 θ; (6)
E0 I0
(3) where E θ is the irradiance produced by the light source at an
azimuth angle θ to the normal axis, E 0 is the irradiance at the
center of the target plane, and I 0 is the intensity of the radiation
in the axis.
To make the irradiance the same for all points on the target
plane, that is, E θ
E 0 for arbitrary θ, we must have
I0
Iθ
: (7)
cos3 θ
However, the luminous intensity distribution for a high-power
LED light source is generally a Lambertian type, that is,
I S θ
I 0 cos θ: (8)
Therefore, to obtain constant irradiance on the target plane, we
should design a lens (or a luminaire) such that the target
intensity distribution I T θ can fulfill Eq. (7).
In addition, if the lens is axial symmetric, then according to
the energy conservation law, the flux emitted from the LED
light source with an azimuth angle interval of 0 ≤ θ ≤ θI
should be equal to that radiated from the lens with an azimuth
angle interval of 0 ≤ θ ≤ θO , where the Fresnel’s loss is ne-
glected. In other words, the following mathematical relation
should be satisfied:
Z θ Z θ
I O
2π I S θ sin θdθ
2π I T θ 0  sin θ 0 dθ 0 : (9)
0 0
Substituting Eqs. (7) and (8), respectively, for I T and I S , and
then integrating on both sides of Eq. (9), we obtain
sin θI
tan θO : (10)
Equation (10) gives the ray-angle relation between the incident
and the refracted light rays. The ranges of the incident and re-
fracted ray-angles are, respectively, given by 0 ≤ θI ≤ 90° and
0 ≤ θO ≤ θO;max , where θO;max is the maximum ray-angle (also
defined as the half-beam angle) of the refracted light rays when
θI
90° occurs. By employing Eq. (10), we can calculate the
refracted ray-angle θO for a given incident ray-angle θI . Then,
by substituting both values of θI and θO into Eq. (5), we obtain
the tangential vector T̄ at the corresponding refractive point.
 E148 Vol. 54, No. 28 / October 1 2015 / Applied Optics Research Article

The maximum ray-angle of the refracted light rays can fur-

ther be calculated by setting θI
90° in Eq. (10), that is,
θO;max
θO jθI
90°
tan−1 1
45°: (11)
However, for a target plane with an area not equal to that of
the original one such that the irradiance is altered to K times
the original one with a half-beam angle of 45°, the ray-angle
relation between the incident and the refracted rays can be
rewritten as
pffiffiffiffi
sin θI
K tan θO : (12)
By using Eq. (12), the maximum ray-angle for the refracted Fig. 4. Simulated irradiance distribution and the associated cross-
light ray is calculated as irradiance profile with K
1 in a target plane 2.5 m away.
 pffiffiffiffi
θO;max
θO jθI
90°
tan−1 1∕ K : (13)
Once the tangential vector T̄ at each refractive point is
calculated, a 3D freeform surface can easily be constructed by
using the method [11] given by the authors. For convenience, a
schematic construction procedure of a 2D freeform contour is
given in Appendix A. Finally, the associated 3D freeform sur-
face can thus be obtained by revolving the 2D contour around
the optical axis for an axially symmetric lens.

3. APPLICATIONS TO FREEFORM LENS

DESIGN FOR LED UNIFORM ILLUMINATION
Based on the proposed method, we can design a freeform lens
with an arbitrary half-beam angle for uniform illumination
with a Lambertian- or non-Lambertian-type LED light source.
The computer simulation for lenses with different half-beam
angles is conducted without taking Fresnel’s loss into account,
and the simulated results are given in the following.
A. Application to a Lambertian-type LED Light
Source
In this case, the ray-angle relation between the incident and the
refracted light rays obeys Eq. (10), where K is set to be unitary,
and the maximum ray-angle of refracted light rays (or the half-
beam angle) is 45°. By using the freeform surface construction
method proposed by the authors, the 2D outline and 3D pro-
file of the freeform lens are, respectively, given in Fig. 3 with a
thickness of 10 mm and an aperture diameter of 10.8 mm for
the acrylic lens material.
The simulated irradiance distribution as well as the associ-
ated cross-irradiance profile in a target plane of 2.5 m away are
given in Fig. 4. In addition, the luminous intensity distribution
of the output beam is also shown in Fig. 5. The computer
Fig. 5. Luminous intensity distribution of the output light beam.
simulation results show that the optical efficiency, which is de-
fined as the ratio of the flux within the prescribed target zone to
the output flux of the LED light source, of the lens is about
93%, the irradiance uniformity (defined as E ave ∕E max ) is 0.81,
and the half-beam angle is 45°.
To test the performance for lenses with different half-beam
angles, we also take different values of K (K ≠ 1) into the sim-
ulation. The associated simulation parameters are given in
Table 1, and the simulated half-beam angle, lens thickness, and
lens aperture for different K values are shown in Table 2. In
addition, the associated variation tendencies of the optical ef-
ficiency and irradiance uniformity with the K value are given in
Fig. 6. From the results of Table 2, we can see that both the
half-beam angle and the lens aperture increase with the decease
of the K value for a fixed lens thickness of 10 mm. Also, from
the variation tendencies in Fig. 6, we can see that the optical
efficiency increases with the K value first and then decreases
with K when K becomes greater than a unit, while the irradi-
ance uniformity decreases with the increase of the K value due
to the decrease of the lens aperture. To illustrate the descent of
uniformity for a larger K value, the irradiance distribution and
the associated cross-irradiance profile for K
1.6 are also

Table 1. Associated Simulation Parameters

Light Source 1 mm × 1 mm LEDChip
Total light-source flux 100 lm
Fig. 3. Constructed freeform lens with K
1. (a) The 2D lens out- Radiation pattern Lambertian type
line, and (b) the 3D lens profile. Lens material Acrylic
 Research Article Vol. 54, No. 28 / October 1 2015 / Applied Optics E149

Table 2. Simulated Half-Beam Angle, Lens Thickness, ing method, called chip on board (COB) technology, has been
and Lens Aperture for Different K Values developed. With this packaging technology, the luminous flux
Half-Beam Angle Lens Thickness Lens Aperture
emitted from a single LED device can be increased by ten to
one hundred times as compared with LEDs packaged with the
K (deg) (mm) (mm)
conventional method; however, the associated luminous inten-
0.2 65.9 10 25.8 sity distribution is, as shown in Fig. 8, no longer of Lambertian
0.4 57.7 10 17.4
0.6 52.2 10 13.8
distribution. In this situation, the source intensity I S θ should
0.8 48.2 10 12.0 be modeled or approximated with high-degree polynomials or
1.0 45.0 10 10.8 spline functions [13], and thus the integration on the left side
1.2 42.4 10 10.0 of Eq. (9) should be numerically calculated.
1.4 40.2 10 9.4 To explain the effect of the luminous intensity distribution
1.6 38.3 10 9.0 of LED light sources on the irradiance uniformity yielded by
the freeform lens, we applied the freeform lens of Fig. 3 to a
given in Fig. 7. From this figure, it can be seen that the irra- COB LED with a chip size of 1 mm × 1 mm and obtained an
diance uniformity decreased as compared with the case in Fig. 4 irradiance distribution, as shown in Fig. 9. This figure shows
when K
1. To obtain a better optical efficiency and irradi- that the irradiance uniformity is 0.78 and apparently looks
ance uniformity, for example, the optical efficiency greater than poor. To improve the uniformity, the intensity distribution
90% and irradiance uniformity greater than 0.8, the K value of the COB LED shown in Fig. 8 was modeled with cubic
should be less than a unit; in other words, the output half-beam splines, and then was substituted into Eq. (9) for the source
angle of the lens should be greater than 45°. intensity I S θ so that a new freeform lens was constructed.
After computer simulation, the new irradiance distribution was
B. Application to a Non-Lambertian-type LED Light obtained, as shown in Fig. 10. This figure shows that the uni-
Source formity becomes 0.88 and is improved greatly.
As mentioned in Section 2, the algorithm of the lens design In addition to the improvement of the irradiance uniform-
method for uniform illumination with Lambertian-type LED ity, to further realize the effects of the beam angle on the optical
light sources is simple and analytic. However, to get more emit- performance of the lens designed for the COB LED, the
ted luminous flux per LED device and simultaneously solve the half-beam angle, the lens thickness, and the lens aperture for
heat dissipation problem, a new large-sized integrated packag- different K values are also shown in Table 3, and the associated
variation tendencies of the optical efficiency and irradiance

Fig. 8. Luminous intensity distributions of the Lambertian-type

Fig. 6. Variation tendencies of the optical efficiency and irradiance LED and COB LED light sources.
uniformity versus the K value.

Fig. 9. Simulated irradiance distribution and the associated cross-

Fig. 7. Simulated irradiance distribution and the associated cross- irradiance profile with a lens designed for a Lambertian-type LED
irradiance profile with K
1.6 in a target plane 2.5 m away. and applied to a COB LED.
 E150 Vol. 54, No. 28 / October 1 2015 / Applied Optics Research Article

Fig. 10. Simulated irradiance distribution and the associated cross-

irradiance profile with a lens designed for a non-Lambertian-type LED
and applied to a COB LED. Fig. 11. Variation tendencies of the optical efficiency and irradiance
uniformity versus the K value.

uniformity are given in Fig. 11. These simulation results show

that, similar to results of the lens designed for Lambertian-type
LEDs, both the half-beam angle and the lens aperture increase
with the decease of the K value, as shown in Table 3. However,
as shown in Fig. 10, the optical efficiency typically decreases
with the increase of the K value, but the irradiance uniformity
increases with the K value first and then decreases with K when
K becomes greater than 0.6. In order to obtain a satisfying op-
tical efficiency and irradiance uniformity, the value of K should
be chosen to be less than 1.4, which corresponds to a half-beam
angle of 42.1°.

4. COMPARISON WITH THE METHOD OF

SOURCE–TARGET ENERGY MAPPING
In order to compare the efficacy of the proposed method with
that of the source–target energy mapping method, in the follow-
ing we give a brief description of a typical source–target energy
mapping method [3,14]. By employing both the source–target
energy mapping method and the proposed method, their respec-
tive simulated optical efficiency and uniformity are compared Fig. 12. Geometric relation of flux mapping between an ideal point
with the same simulation parameters given in Table 1. source S and a target plane with a radius of R and a distance of H for a
freeform and symmetrical lens.
A. Brief Description of the Source–Target Energy
Mapping Method
Figure 12 shows a geometric relation of energy mapping be- source is divided into n annuli with an equal azimuth interval
tween an ideal point source S and a target plane with a radius Δθ and the flux within an annulus of θi ≤ θ ≤ θi1 is pro-
of R and a distance of H through a symmetrical freeform jected into an annulus through the lens with r i ≤ r ≤ r i1 in
lens. Suppose that the radiation pattern of the light source is the target plane, then, according to the flux conservation law
Lambertian type and the totally radiated flux is given by ΦS. and the uniform illumination purpose, the average irradiance
Meanwhile, if the semi-spherical radiation space of the light on each annulus in the target plane should be equal to the
global average irradiance, that is,
Table 3. Simulated Half-Beam Angle, Lens Thickness, R 2π R θi1 R 2π R π∕2
and Lens Aperture for Different K Values 0 θi I 0 cos θ sin θdθdϕ I 0 cos θ sin θdθdϕ

0 0 :
Half-Beam Angle Lens Thickness Lens Aperture πr 2i1 − πr 2i πR 2
K (deg) (mm) (mm) (14)
0.2 67.3 10 26.8 From Eq. (14), if the values of θi , θi1 , r i , and R are known,
0.4 59.4 10 18.2 the radius r i1 (corresponding to an azimuth angle of θi1 ) can
0.6 54.0 10 14.5 therefore be calculated as
0.8 50.1 10 12.5 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1.0 46.9 10 11.2 r i1
r 2i  1∕2cos 2θi − cos 2θi1 R 2 ; (15)
1.2 44.3 10 10.3
1.4 42.1 10 9.8 where θi1
θi  Δθ, and the initial values of θi and r i are
1.6 40.2 10 9.3 given by θ0
0 and r 0
0, respectively.
 Research Article Vol. 54, No. 28 / October 1 2015 / Applied Optics E151

For an incident ray with a ray-angle of θi , the corresponding

refracted ray-angle θi0 , from the geometry in Fig. 12, can be
expressed by


0 −1 r i − PQ
θi
tan : (16)
OQ
By employing Snell’s law at point P and using the geometric
relation in Fig. 12, we can calculate the slope angle of the tan-
gential vector T̄ as
 
sinθi0 − θi 
θT
θi  tan−1 ; (17)
n − cosθi0 − θi 
where n is the refractive index of the lens, and the tangential
vector T̄ is expressed as
T̄
1; tan θT : (18)
Once the tangential vector T̄ at each control point is calculated,
a freeform surface and thus the 3D freeform lens can be con-
structed by using the method mentioned in Section 2.
B. Simulated Results Comparison between Both
Methods Fig. 15. Irradiance distribution and the associated cross-irradiance
Based on the proposed method and the source–target energy profile for a lens of 60° half-beam angle with a method of (a) energy
mapping and (b) intensity mapping.
mapping method stated in the previous subsection, we simulate
the individual optical efficiency and irradiance uniformity for
both design methods with different half-beam angles. The as-
sociated simulation parameters are same as given in Table 1, and the simulated optical efficiency versus different half-beam
angles is shown in Fig. 13, while the simulated irradiance uni-
formity versus different half-beam angles is shown in Fig. 14.
From results in Fig. 13, it can be seen that the freeform lens
designed by using the proposed method can achieve better op-
tical efficiency than that by using the source–target energy
mapping method if the half-beam angle becomes greater than
50°. The same tendency also occurs in the irradiance uniform-
ity, as shown in Fig. 14. To illustrate the difference of irradiance
distributions when the half-beam angle is greater than 50°, the
irradiance distributions for both methods with a half-beam

Fig. 13. Optical efficiency versus half-beam angle for both design
methods.

Fig. 14. Irradiance uniformity versus half-beam angle for both de-
sign methods. Fig. 16. Construction method of the 2D freeform locus.
 E152 Vol. 54, No. 28 / October 1 2015 / Applied Optics
angle of 60°are also given in Fig. 15. From Fig. 15(a), it can be
seen that the irradiance aroud the axis becomes larger than that
near the edge of the prescribed target zone by using the source–
target energy mapping method. However, from Fig. 15(b), it
can be seen that an even irradiance distribution, no matter if
near the axis or the edge, is achieved by using the proposed
intensity mapping method. In other words, the use of the pro-
posed source–target intensity mapping method in freeform lens
design for LED uniform illumination will achieve better per-
formance, both in optical efficiency and irradiance uniformity,
for large output half-beam angles.
5. CONCLUSIONS
A freeform lens design method based on source–target lumi-
nous intensity mapping and a simple freeform surface construc-
tion procedure is presented in this paper. By employing this
method, a freeform LED lens of arbitrary half-beam angle
greater than 45 deg can be constructed with irradiance uni-
formity greater than 0.8 and optical efficiency above 90%.
The computer simulation results also demonstrate that the
use of the method can achieve better performance than those
by using general source–target energy mapping methods in
LED illumination with large beam angles. Other than the
Lambertian-type LED light source, this method also works well
for designing a lens for the non-Lambertian-type LED light
source for uniform illumination. Though the cited examples
in this paper are particularly for LED uniform illumination,
the proposed method is also suitable for various indoor/outdoor
luminaire designs for a specific light source intensity distribu-
tion and a prescribed target intensity distribution.
APPENDIX A
As shown in Fig. 16, red lines marked with i 0 , i 1 , i 2 , and i 3
stand for incident light rays emitted from the light source S,
and then hit the lens surface at points of P 0 , P 1 , P 2 , and
P 3 , respectively. Point P 0 , whose position depends on the pre-
scribed lens thickness, is the initial point of the 2D freeform
locus to be constructed, and is given by the designer. The blue
dashed lines T̄ 0 , T̄ 1 , T̄ 2 and T̄ 3 , respectively, stand for the
tangential vector at point P 0 , P 1 , P 2 , and P 3 , which are used
to construct the 2D locus. The construction procedure is to
first give the coordinates of the initial point P 0 and to calculate
the tangential vector T̄ 0 at point P 0 by using Eq. (5), where θO
is calculated with Eqs. (10) or (12), and then to calculate point
P 1 , which is the intersection of the incident light ray i 1 and the
tangential vector T̄ 0, next by using Eq. (5) again to calculate
Research Article
the tangential vector T̄ 1 at point P 1 , and then to calculate point
P 2 , which is the intersection of the incident light ray i 2 and the
tangential vector T̄ 1. Duplicate the above process until θI
equals 90°, and then the whole 2D locus can be constructed
with these P-points and the help of CAD programs.
Furthermore, to ensure that the tangential slope at each P-point
is unchanged, the curve between any two neighbor P-points
should be fit by a high-degree polynomial [11].
Funding. Ministry of Science and Technology, Taiwan
(MOST) (102-2221-E-018-029).
REFERENCES
1. H. Ries and J. Muschaweck, “Tailored freeform optical surfaces,”
J. Opt. Soc. Am. A 19, 590–595 (2002).
2. Y. Ding, X. Liu, Z. Zheng, and P. Gu, “Freeform LED lens for uniform
illumination,” Opt. Express 16, 12958–12966 (2008).
3. W.-C. Chen and H.-Y. Lin, “Freeform lens design for LED illumination
with high uniformity and efficiency,” Proc. SPIE 8123, 81230K (2011).
4. G. Wang, L. Wang, L. Li, D. Wang, and Y. Zhang, “Secondary optical
lens designed in the method of source–target mapping,” Appl. Opt. 50,
4031–4036 (2011).
5. F. R. Fournier, W. J. Cassarly, and J. P. Rolland, “Fast freeform
reflector generation using source–target maps,” Opt. Express 18,
5295–5304 (2010).
6. Y. Luo, Z. Feng, Y. Han, and H. Li, “Design of compact and smooth
free-form optical system with uniform illuminance for LED source,”
Opt. Express 18, 9055–9063 (2010).
7. J. C. Miñano, P. Benítez, J. Y. Liu, J. Infante, J. Chaves, and W. Lin,
“Applications of the SMS method to design of compact optics,” Proc.
SPIE 7717, 77170I (2010).
8. R. Wu, Y. Zhang, M. M. Sulman, Z. Zheng, P. Benítez, and J. C.
Miñano, “Initial design with L2 Monge–Kantorovich theory for the
Monge–Ampère equation method in freeform surface illumination
design,” Opt. Express 22, 16161–16177 (2014).
9. J. Bortz and N. Shatz, “Generalized functional method of nonimaging
optical design,” Proc. SPIE 6338, 32–47 (2006).
10. J.-J. Chen, Z.-Y. Huang, T.-S. Liu, and K.-L. Huang, “Freeform LED
lens design for uniform illumination with prescribed candela distribu-
tion,” in International Optical Design Conference (IODC), Classical
Optics 2014, OSA Technical Digest (online) (Optical Society of
America, 2014), paper JTu5A.28.
11. J.-J. Chen, T.-Y. Wang, K.-L. Huang, T.-S. Liu, M.-D. Tsai, and C.-T.
Lin, “Freeform lens design for LED collimating illumination,” Opt.
Express 20, 10984–10995 (2012).
12. J. Chaves, Introduction to Nonimaging Optics (CRC Press, 2008),
Chap. 16.
13. R. L. Burden and J. D. Faires, Numerical Analysis (Thomson Brooks/
Cole, 2005).
14. M.-D. Tsai, Z.-Y. Huang, T.-S. Liu, J.-J. Chen, and K.-L. Huang,
“Freeform lens design for uniform illumination with high power LEDs,”
in Optics & Photonics Taiwan International Conference, Chung-Li,
Taiwan, 5–7 Dec. 2013.