Application-Specific

Optical Design

Introduction the patent literature or some other

database of refractive, reflective, or
Optical design software capabilities catadioptric forms to find configura-
have advanced considerably from the tions that might be successful in
late 1950s and early 1960s when com- meeting the design specifications for
puter tools first became available. Ini- their specific application. One tech-
tially, the main purpose of the software nique that can be used successfully
was to geometrically trace rays and in developing a starting point for a
perform limited analyses. However, the complex system is to break the sys-
introduction of automated optimiza- tem functionality into sub-systems
tion, generally using a damped least of recognizable forms. For example,
squares algorithm, is what has made a scanning system may consist of a
software an indispensable tool for the pre-scan objective lens configuration
optical engineer. Figure 1. CODE V Models for Different with a post-scan eyepiece configura-
Applications
Today, optics are used in a variety of tion. In recent years, the introduction
applications, and unique features and of “global optimization,” such as
capabilities are often required to model, The Optical Design CODE V’s proprietary Global Syn-
optimize, and analyze systems designed Process thesis algorithm, have allowed
for a specific application. In this paper, optical engineers to more fully
Before discussing specific applications, search the solution space for useful
we will look at several different applica-
tion areas and discuss some of the soft- it is important to consider the optical starting configurations. Typically,
design process, since the general process designers find one or more promis-
ware modeling, design, and analysis
is common across all applications. The ing forms from prior experience or
features important for those applica-
tions. The applications include: process can be summarized as follows: the patent literature; and use these as
• Commercial Imaging Systems 1. Develop a design specification. starting points for global optimiza-
• Visual Systems (i.e., optical systems This includes 1st order properties tion. The purpose at this stage of the
(Effective Focal Length, Field of process is not to generate a final
that use the human eye as the
View, F/number, spectral band), size complete design; rather, it is to find
detector)
and weight limitations, material and multiple starting forms with unique
• Off-axis tilted and decentered characteristics.
surface shape limitations, nominal
Systems
performance metrics and “as-built” 3. Analyze the starting design(s). The
• Telecommunications Systems requirements (i.e., the performance types of analyses performed depend
• Astronomical applications of an actual system with fabrication on the application and design speci-
• Non-visible Systems (e.g., UV, IR, and assembly tolerances), along with fication. The purpose of this step is
etc.) any other restrictions or issues that to determine how close the initial
• Microlithographic (Optical Stepper) will impact the final optical design. design(s) is to meeting the final
Systems Design specifications are strongly design specification.
Figure 1 shows a montage of several driven by the specific application. 4. Optimize the starting design(s).
CODE V models used in different appli- 2. Determine one or more starting This step includes determining the
cation areas. points. Optical engineers rarely possible variable parameters in the
develop a starting design from system model that can be modified
scratch. Based on the design to improve performance, generating
specification, engineers typically use the appropriate merit function and
 constraints based on the design spec- 7. Perform a global optimization • Verify the availability of optical
ification, and then optimizing the check on the finished design. Since materials in the finished design. In
design forms. the constraints, variables, and speci- recent years, glass manufacturers
5. Analyzing the optimized design(s). fications have likely evolved during have been trying to keep costs low
This step is similar to Step 3, but the optical design process, it is often by maintaining only small inven-
should include an initial tolerance useful to perform another global tories of infrequently ordered
analysis to determine if the current optimization (Global Synthesis) with glasses. When this happens, a
form will be sensitive to manufactur- the finalized parameters. If global glass that is offered in the manu-
ing and alignment errors. CODE V optimization does not generate a facturer’s catalog is sometimes out
includes a tolerance analysis capa- form better than the current design, of stock. This can often require a
bility based on very fast and accurate this gives designers some confidence complete redesign to use different
wavefront differential algorithms. that their final design is the best optical materials for time-critical
This feature includes the impact of a solution. Occasionally, when starting designs. The designer should ver-
single or multiple compensators from the "final" configuration, glo- ify material availability as soon as
used to recover lost performance, bal optimization will find a configu- they have a reasonable idea of
and will determine tolerances ration that does a better job of which materials are required.
(within user-defined limits) in order meeting the performance goals. If • If possible, make bi-convex and
to minimize total performance loss. the optical materials were allowed to
bi-concave lenses with similar
This capability allows for very fast be variable during Step 4, this should
radii equi-convex or equi-concave
initial assessments of design sensi- also be allowed during this step, and to prevent them from being assem-
tivity to fabrication tolerances and real glass substitution will be
bled backwards. This happens
assembly errors. This type of analy- required on the resulting designs.
more often than fabrication shops
sis helps the designer determine the 8. Perform a tolerancing check on would like to admit.
best design form(s) for their specific the finished design. In this step, the
requirements. Other required analy- designer should determine the final • If possible, make spherical sur-
ses may include determining the per- tolerances and compensators for the faces with long radii planar.
formance loss due to environmental design. The designer should already • If possible, round lens thickness
conditions such as temperature or be working with the opto-mechani- values to a limited number of sig-
pressure. At this point, designers can cal engineer to understand what nificant digits (this will make the
begin to choose the best candidates adjustments are possible. Once fabrication shop happy).
from among competing design again, using the wavefront differen-
• Adjust the radii in the design to
forms. tial method of tolerancing allows
match the test plates available to
6. Repeat Steps 4 and 5 as necessary. different compensation schemes to
the fabrication shop. This step will
Optical design is an iterative pro- be tested very quickly. In addition
save fabrication cost if test plates
cess. It is rare that the initial optimi- there are features that allow the soft-
are being used to verify the manu-
zation of a starting form results in ware model to match the opto-
factured radii. CODE V optimiza-
the final configuration. Typically, mechanical implementation (i.e.,
tion supports automatic test
designers need to modify the model “labeled” and “grouped” tolerances
plating to an input test plate list.
by allowing additional variables and compensators). An aspect of this
(e.g., adding lenses, using aspheric step that is often overlooked is to • After all these changes have been
surfaces, etc.), or by allowing more always include an adjustment toler- implemented, perform a final opti-
freedom during optimization by ance on the compensator. In the real mization and performance evalua-
relaxing constraints. If the optical world, no compensation (such as tion(s) (both nominal and as-
materials are allowed to vary, the refocus) can be achieved perfectly; built).
designer will need to substitute there is always some error in the • Create and check lens and compo-
obtainable materials for any theoreti- adjustment (e.g., due to thread toler- nent drawings for the fabrication
cal material definitions and perform ances on a focus barrel). facility.
a quick re-optimization before eval- 9. Modify the finished design for fab-
uating the system. Sometimes, if the rication and cost considerations. 10. Post-fabrication analyses. After the
design has been fabricated, there
resulting design does not meet the Once a final design form is deter-
may be some additional steps neces-
design specifications, the designer mined, there are additional steps that
must revisit the specifications to will facilitate the fabrication of the sary for precision applications:
determine if they can be modified design and potentially lower life- • The glass model for the design
with regards to size and weight lim- cycle costs. These steps may may need to be altered to match
its, or performance goals. include: measured index data from samples
of the real material to be used, and
the airspaces adjusted slightly to

2
 compensate for performance
changes due to the refractive index
differences. This process is called
“melt fitting,” and CODE V
includes some special features to
aid in the modeling of measured
glass data.
• Apply surface deformation data,
measured with an interferometer Starting Form For Global Synthesis
directly to the surfaces in the
model. Most commercial interfer-
ometer manufacturers can export
the measurement data directly into
CODE V’s interferogram file
format.
• Apply measured lens thicknesses
to the model.
• After these changes have been
implemented, analyze the system
“as-built” performance based on
the information.
• If needed, you can perform auto-
mated alignment on the design.
CODE V contains an “Alignment
Optimization” feature that can be Figure 2. Several distinct solutions for a CRT camera lens, generated by Global Synthesis
used to determine the correct
alignment adjustments based on
optical system measurements cation is a 200 mm EFL, F/1.25, all- diffraction-based MTF at specified fre-
using an interferometer. spherical, 8-element, camera lens for quencies) and MTF tolerancing.
With the optical design process outlined, photographing a CRT display. The start-
we can examine how the specifics of the ing configuration is indicated.
process can vary for different applica- Early imaging system performance met-
tions. rics were based on geometrical ray trac-
ing and included ray aberrations,
geometrical Modulation Transfer Func-
Commercial tion (MTF), 3rd and higher-order aberra-
Imaging Systems tions. Later, as diffraction computations
were added to software, the standard
Imaging systems for commercial appli-
performance metrics migrated to Dif-
cations (such as camera objectives and
fraction MTF, Root-mean-square (RMS)
projector lenses) were one of the earliest
wavefront error, Point Spread Functions
application areas that benefited from
(PSF), and Strehl ratios. Figure 3 shows
optical design and analysis software. Figure 3. Several Analyses for a
a mosaic of performance analysis results
These systems cover a large range of Petzval Lens
for a Petzval lens system.
f-numbers (F/#) and Fields of View
(FOV). Since diffraction-based MTF is a lead-
Commercial imaging systems com-
ing performance metric for commercial Visual Systems
imaging systems, optimization and tol-
monly use centered, rotationally sym- The effective design and analysis of
erance metrics that compute diffraction-
metric refractive systems. A starting visual systems can benefit from special-
based MTF can be very beneficial.
point could be chosen from one of the ized handling of the light emerging from
CODE V’s fast and accurate wavefront
2400 patents in CODE V’s built-in the system. In a visual system, the
differential algorithm is used for both
patent database, or by using global opti- human eye is the detector; and it has a
MTF optimization (i.e., an optimization
mization. The applicability of global built-in capability to quickly refocus as
merit function that directly optimizes
optimization for starting point genera- it scans a field of view through an opti-
tion can be seen in Figure 2. The appli-

3
cal system. This suggests that optical or cycles/angular unit in the case of plays (HUDs) can often be fit into a
software needs to model independent MTF. Figure 4 shows a 5x40 visual tele- much smaller space if the components
focus across field. Typically, the focus- scope (i.e., 5x magnification, 40 mm are tilted and decentered relative to a
ing ability of the eye is measured in entrance pupil diameter) and the common axis.
units of diopters of accommodation. A CODE V controls for angular perfor- For these types of systems, flexibility in
diopter is a reciprocal unit of distance, mance units and accommodation across how the tilts and decenters can be
corresponding to 1/(distance in meters) the field of view (which can be variable defined is an important software model-
to where an object appears to be. A per- for optimization). ing feature. For example, CODE V sup-
son with good vision (i.e., emmetropic Notice that each field has a separate ports six different methods for tilting
vision if corrective eye glasses are not accommodation ranging from 0.0 diopt- and decentering a surface, and defining
required, or possibly myopic or hyper- ers at the edge of the field (i.e., objects the coordinate system that follows the
opic vision if corrective lenses are at the edge of the field in the telescope surface. One of the most useful is the
needed) can focus very comfortably at appear to be at infinity) to +0.63 diopt- ability to define all the surfaces relative
infinity. Focusing on an object that ers on-axis (axial objects appear to be to a global coordinate system.
appears to be at infinity corresponds to 1/0.63, or about 1.6 meters in front of
an accommodation of 0.0 diopters For optimization, control of beam and
the observer. This range of accommoda- component clearances becomes very
(i.e., 1/∞ meters). tion meets the typical field curvature
Since the average human has a visual specification < 1 diopter.
resolution that corresponds to about one
arc-minute, performance metrics in
terms of angular units are natural for
evaluating visual systems. Some typical
visual system performance metrics and
suggested goals are:
• Field curvature < 1 diopter (always
positive, i.e., the image is formed in
front of the observer since the human
eye has limited ability to focus light
converging towards it)
• Astigmatism < 0.25 diopter
• Lateral color < 5 arc minutes
(assumes modeling of a full photonic Figure 5. Analysis Curves with units of
Diopters, Arc Minutes, or Cycles/Arc Minute
spectral band, ~ 465 to 645 nm)
System Layout and Full Field Spot Diagram -
• 0.25 MTF at 0.5 cycles/arc minute Before Optimization
(corresponds to a “line width” of one Figure 5 shows a mosaic of analysis out-
arc minute) put with units in arc minutes, or diopters
CODE V supports a “true afocal model- of accommodation.
ing” feature that allows independent
focus across field, with focus values
defined in diopters of accommodation. Off-axis, Tilted, And
The feature also supports aberration Decentered Systems
analysis output in terms of angular units,
Rotational symmetry
has a number of benefits
in an optical design,
including symmetry of
the aberration field and
generally easier fabrica-
tion and alignment.
However, many applica-
tions require the symme-
try to be violated, System Layout and Full Field Spot Diagram -
especially to meet pack- After Optimization (same scale)
aging constraints. Figure 6. 4-Mirror, Off-axis System,
Figure 4. 5x40 Telescope & True Afocal Modeling Controls Space-borne reflective Before and After Optimization
in CODE V optics and heads up dis-

4
important, since the natural tendency is
for the design to become more symmet-
ric to reduce aberrations, and this often
leads to interference problems between
the ray bundles and components.
Figure 6 illustrates the evolution from a
starting system with multiple beam and
component interference problems to a Figure 8a. Optical Isolator System
final system with improved performance
and no beam and component interfer-
ence problems. This design evolution
was accomplished in a single optimiza-
tion run. The key to the successful opti-
mization of this system was the ability
to easily define optimization constraints
to automatically prevent interference
problems. In CODE V, these constraints
are typically handled as Lagrange multi-
pliers, which means they are separate
from the optimization error function (in
this case, the default RMS blur size
error function). This allows the optimi-
zation to converge on the best perfor- Figure 8b. Field Intensity at Output Fiber Figure 8c. Field Intensity at Input Fiber
from Reflection at Output Fiber
mance in the least constrained mode,
while maintaining packaging clearances. field map outputs for the system above, This system has very little insertion loss
Another important aspect for the design one showing astigmatic focal lines and in the desired direction (the intensity of
of systems utilizing off-axis tilted and the other a plot of the magnitude and the optical field at the output fiber is
decentered components, is that tradi- orientation of wavefront Zernike coeffi- shown in Figure 8b). By using birefrin-
tional diagnostic analyses, such as trans- cients for 3rd order coma. gent crystal materials and a Faraday
verse ray aberration curves or field rotator element to rotate the electric
curves, can be deceiving. These tradi- field, any light that is reflected back
tional tools were developed for systems Telecommunication towards the input fiber is split into two
with rotational symmetry, and rely on Systems components, such that very little energy
this symmetry to be meaningful across couples back into the input fiber (Figure
the entire field of view. To address this In recent years, there has been an 8c).
limitation, CODE V includes a field increase in the optical design and analy-
sis of components for telecommunica- Typical analysis models for traditional
map diagnostic option that can plot vari-
tion systems. These components are imaging systems treat all diffraction as
ous performance metrics across the full
generally very small. The governing occurring at the optical system exit
field of view. The information in these
performance metric is how much energy pupil. This may be inadequate for
plots can provide great insight on what
is gathered from an input fiber, laser, or telecommunication systems, since the
steps to take to improve the system per-
waveguide and coupled into an output beams are typically only a few hundred
formance (see Thompson 1996:2 and
fiber. As a percentage this is called the wavelengths in diameter and often
Rogers 1999:286). Figure 7 shows two
fiber coupling efficiency, but it is typi- propagate several thousand wavelengths
cally defined in terms of energy loss in in distance between components. The
decibels (i.e., insertion loss). physics of this arrangement cause the
beam diameter to grow due to
The fibers are very small and typically diffraction spreading, and cannot be
can propagate a single mode, or only a accurately modeled with geometric ray
few modes. Gradient index elements are tracing techniques. In this case, general
common and polarization effects are beam propagation algorithms must be
often important. Figure 8a shows an used throughout the entire optical
optical isolator. The input fiber is on the system to adequately account for
left and the output fiber is on the right. diffraction.
For this system, it is desired that light
reflected from the output fiber face will CODE V’s general beam propagation
Figure 7. Full Field Maps of Aberration
Fields not propagate back into the input fiber. feature can be used to determine the

5
 beam remains coherent and interference
results. The angular size of the object
under observation can be inferred by
adjusting the separation between the two
apertures until the fringe structure disap-
pears. Figure 10b shows the PSF in a
well-aligned system. Since the mirrors
are modeled as separate non-sequential
components, they can be tilted with
respect to each other. The effect on the
LP01 Mode LP011a Mode LP011b Mode PSF of a 1/3 arc-second tilt on one of the
mirrors is shown in Figure 10c.
Figure 9. Supported Modes for Corning SMF-28 Fiber used at 850 nm

Non-Visible (UV, IR, etc.)

complex field throughout the system, modeling defines systems sequentially
and, most importantly, at the output for ray tracing. That is, all rays must go Systems
fiber face. CODE V’s fiber coupling from an object to surface 1, to surface 2, The optical design process outlined on
efficiency feature calculates the overlap to surface 3, and so forth. When seg- page 1 works equally well for systems
integral between the complex field and mented apertures are used, different rays that operate at visible wavelengths and
the mode profile of the single mode will intersect different segments, but not for systems that operate outside the visi-
fiber to determine the insertion loss into intersect the others. In CODE V, this ble spectrum. However, some new
the fiber. Multimode fiber analysis is type of system is handled by using non- issues arise when operating outside of
also supported, by computing the cou- sequential surfaces (NSS). Figure 10a the visible spectral band. For example,
pling efficiency into each supported shows a stellar interferometer system systems that operate in the ultraviolet
mode of the fiber. Figure 9 illustrates the used for ultra-high resolution studies. will require better optics to achieve
amplitude profile for three modes sup- The path for the light from the two sepa- “diffraction limited” performance, since
ported by a Corning SMF28 fiber oper- rated apertures is carefully maintained diffraction is a function of wavelength
ating at a wavelength of 850 nm. to be approximately equal so that the
Typically, this fiber is used at 1310 nm
or 1550 nm, and only the fundamental
mode (LP01) will propagate. However,
when used at 850 nm, the modes shown
in Figure 9 are supported. The data used
to create these plots is used to define the
mode structure for the multi-mode fiber
coupling efficiency calculation.
CODE V supports both a fiber coupling
efficiency merit function for optimiza-
tion and a coupling efficiency toleranc-
ing performance metric. This allows the Figure 10a. Stellar Interferometer Systems
optical design process outlined on
page 1 to be directly applied to these
types of systems.

Astronomical
Applications
Astronomical applications generally
require small fields of view, fast (small
F-number) optics, and point image per-
formance metrics, since the objects of
interest are small and faint. In addition,
many large telescopes and planned
space-borne optics are using segmented
apertures. Traditional optical system Figure 10b. PSF for Aligned System Figure 10c. PSF for Misaligned System

6
and shorter wavelengths allow better Microlithographic
resolution.
Systems
Thermal infrared systems, designed to
detect heat emissions from thermal bod- One of the most demanding optical
ies, typically use cooled detectors. design applications is microlithographic
Often, the detector array instantaneously systems. These are systems used to fab-
images only a portion of the desired ricate integrated circuits. While the opti-
field of view. Using scan mirrors or cal design process is the same, these
other devices, the image of the detector systems require diffraction limited as-
array is scanned across the full field of built performance with essentially zero
view. These systems are subject to some distortion. Successive generations of
unique image defects such as narcissus, Figure 12. Plot of P vs. V for Optical these systems use shorter operational
which must be considered in the design Glasses in the 1550 nm – 1610 nm wavelengths and faster optics in order to
Spectral Band achieve the improved theoretical resolu-
specification.
tion limit. Subtle image degradation due
In general, material considerations have sus Abbe number (V) for the Schott to material stress birefringence, intrinsic
the most significant impact when work- glass catalog, in the visible region. material birefringence, thermal effects,
ing outside the visible spectral region.
Fictitious glasses, i.e., glass models that and other sources must be analyzed
Optical glasses eventually become
can be variable for optimization, are typ- (Li, Ota, and Murakami 2003:127-129).
opaque in the ultraviolet and infrared.
ically based on this normal dispersion In addition, the characteristics of the
Often other crystalline materials must be
model. However, outside the visible illumination beam must be considered.
used. Reflective optics have the benefit
spectral region, the concept of a “normal For example, CODE V can analyze one-
of working across a larger spectral band
line” may not exist. Consider the P vs. V dimensional and two-dimensional image
but sometimes their use is restricted due
plot for the same glasses operating in a structure assuming illumination that
to packaging considerations or fabrica-
typical telecommunications wavelength ranges from fully coherent to incoher-
tion limitations.
band of 1550 to 1610 nm. The result is ent. Figure 13 shows an Offner reflec-
shown in Figure 12. tive microlithographic system and
A fictitious glass dispersion model image structure analysis.
based on the “normal line” in the visible The demands of microlithographic
would not do a very good job of model- optics require that all possible steps be
ing real glass dispersion behavior in the taken to maximize performance. The
infrared. In CODE V, users can redefine post-fabrication steps of the design pro-
the fictitious glass model wavelengths cess become critically important. Manu-
and dispersion characteristics to match facturers have successfully used
those of real glasses used in that wave- CODE V’s alignment optimization fea-
length region. This makes it much easier ture to reduce cost and improve
to substitute real optical materials for performance (Sugisake, et al.,
variable optical materials. 2000:751-758).
Figure 11. Plot of P vs. V for Optical
Glasses in the Visible Spectral Band

In the near ultraviolet and infrared, opti-

cal glasses will transmit, but their opti-
cal properties will be different than
when they are used in the visible. For
example, in the visible spectral region,
the 2nd derivative of the index vs. wave-
length curve, represented by the partial
dispersion (P), varies nearly linearly
with Abbe number for most commonly
available optical glasses. This represen-
tation is sometimes called “the normal
dispersion line” and is useful for under-
standing glass selection to correct the
aberration of secondary color. Figure 11 Figure 13. Offner 1:1 Microlithographic Projection system & 2-D Partial
is a plot of the partial dispersion (P) ver- Coherence Analysis

7
 mination systems is very should be defined. The specification will
different from that of image- affect the approach used in executing
forming applications. In some the optical design process, and help to
respects, it is much more diffi- determine what design features and
cult to develop an optical sys- analysis functions should be applied to
tem that distributes light from a the problem.
condensed source evenly across The benefits of optical design software
a large area (or with some spe- encompass all phases and applications
cific distribution). An attempt
of optical engineering in the 21st cen-
to merge the capabilities
tury. The return on investment for out-
required for both image-form-
standing software pays for itself many
ing and illumination applica-
times over in terms of superior perfor-
tions into a single software
mance, lower manufacturing costs, and
package, necessitates signifi-
in its ability to facilitate technological
cant added complexity to the
innovation.
software interface, or a signifi-
cantly reduced feature set.
ORA has developed another References
software product, LightTools,
to address the specific needs of Li,Yangqiu, Ota, Kazuya, and
the illumination engineering Murakami, Katsuhiko. “Thermal and
field by providing a full set of structural deformation and its impact on
features with an easy-to-use optical performance of projection optics
Figure 14 interface. for extreme ultraviolet lithography,”
Journal of Vacuum Science and Technol-
ogy B, Jan/Feb 2003, pp. 127-129.
Other Considerations Conclusions
Rogers, John R. “Design techniques for
Engineering is an international The optical design process is relatively systems containing tilted components,”
endeavor, but different locales often consistent across various applications. SPIE, Vol. 3737, 1999, p. 286.
require specialized tools. An example However, many applications require
Sugisake, K., et. al., “Assembly and
within CODE V is support of the PRC some specialized design or analysis
alignment of three aspherical mirror
National Standard optical element draw- techniques and features. Successful opti-
optics for extreme ultraviolet projection
ing format. Figure 14 shows an optical cal design requires that the designer
lithography,” Proc. SPIE, Vol. 3997,
element drawing for a singlet. always consider the design specifica-
2000, pp. 751-758.
To this point we have not discussed tion, which is the first and often the
most important step in the optical design Thompson, Kevin. “Practical methods
optics for illumination design and analy-
process. It is here that all the consider- for the optical design of systems without
sis. The optical design problem for illu-
ations and limitations of the design symmetry,” SPIE, Vol. 2774, 1996, p. 2.

3280 East Foothill Boulevard, Pasadena, California 91107

Telephone: (626) 795-9101 FAX: (626) 795-0184
E-mail: service@[Link] Web site: [Link]