Optomechanical design of multiscale gigapixel digital camera

 
CONTINUE READING
Optomechanical design of multiscale gigapixel digital camera
Optomechanical design of multiscale
                                gigapixel digital camera

     Hui S. Son,1 Adam Johnson,2 Ronald A. Stack,2 Jeffrey M. Shaw,3 Paul McLaughlin,3
                   Daniel L. Marks,1 David J. Brady,1 and Jungsang Kim1,*
                  1
                      Electrical and Computer Engineering Department, Duke University, Duke–129 Hudson Hall,
                                                Durham, North Carolina 27708, USA
           2
               Distant Focus Corporation, Distant Focus–4114B Fieldstone Road, Champaign, Illinois 61822, USA
                           3
                            RPC Photonics, Inc., RPC–330 Clay Road, Rochester, New York 14623, USA
                                            *Corresponding author: jungsang@duke.edu

                         Received 6 December 2012; revised 25 January 2013; accepted 27 January 2013;
                                posted 28 January 2013 (Doc. ID 179787); published 4 March 2013

                 Recent developments in multiscale imaging systems have opened up the possibility for commercially
                 viable wide-field gigapixel cameras. While multiscale design principles allow tremendous simplification
                 of the optical design, they place increased emphasis on optomechanics and system level integration of the
                 camera as a whole. In this paper we present the optomechanical design of a prototype two-gigapixel
                 system (AWARE-2) that has been constructed and tested. © 2013 Optical Society of America
                    OCIS codes: 120.4880, 110.0110.

1. Introduction                                                        for surveillance instruments and astronomical tele-
In traditional single aperture imaging systems, the                    scopes [2,3], but to date an economical ground-based
maximum attainable resolution is determined by                         camera that can be scaled beyond a few gigapixels
either the geometric aberrations or the diffraction                    has not been feasible. The availability of such a cam-
limit of the optics. An efficiently designed camera                    era would have a significant impact on our ability to
matches this resolution to the pixel-limited resolu-                   acquire visual data on both astronomical and terres-
tion of the sensor. A multigigapixel camera therefore                  trial objects.
requires optics that can effectively resolve billions of                  A multiscale camera circumvents these difficulties
image points with corresponding sensors capable of                     by splitting the work of imaging the field over several
acquiring these image points. However, in conven-                      small-scale optics and using digital image processing
tional imaging systems, increasing the resolution                      to form the composite image [4,5]. Figure 1 illus-
typically necessitates an increase in the size of the                  trates the basic elements of a multiscale design. A
optics, in turn increasing geometric aberrations,                      main objective lens, shown as a sphere, captures the
which scale with system size. Traditional approaches                   total field and produces an intermediate image. This
to correct these aberrations across the entire field                   image is then further corrected and relayed through
require a large number of optical elements, leading                    a set of smaller optics (called “microcamera” optics)
to excessive levels of complexity, weight, and size                    to produce partial images at corresponding focal
[1]. In addition, individual digital sensors with pixel                planes. These partial images are designed to contain
counts numbering in the billions are currently                         the total field of interest and can be manipulated in
unavailable. A limited number of highly specialized                    postprocessing to produce a continuous image. The
gigapixel imagers have been developed, specifically                    array of the microcamera effectively behaves as a
                                                                       curved focal surface to capture the image produced
                                                                       by the main objective. An appropriately designed
 1559-128X/13/081541-09$15.00/0                                        shared objective lens can correct a large portion of
 © 2013 Optical Society of America                                     the aberrations and increase the aperture size for

                                                                             10 March 2013 / Vol. 52, No. 8 / APPLIED OPTICS   1541
Optomechanical design of multiscale gigapixel digital camera
thermal simulations and their predicted effects on
                                                                     AWARE-2 are presented in Section 4, and the conclu-
                                                                     sions are summarized in Section 5.

                                                                     2. Design of AWARE-2 Camera
                                                                     In this work, we will investigate a monocentric multi-
                                                                     scale design. Monocentricity refers to the spherical
                                                                     symmetry of the objective lens, in this case a ball
                                                                     lens, which allows identical microcameras to be
                                                                     arrayed radially from the center of the objective. A
                                                                     more complex, nonspherical structure can be de-
Fig. 1. (Color online) Schematic illustration of a multiscale ima-   signed if needed, but for most applications like wide-
ging system.                                                         field imaging a spherical distribution is preferred.
                                                                        For a monocentric camera, the main support struc-
                                                                     ture is a spherical dome which precisely places each
each set of microcamera optics. Thus the microca-                    microcamera in the correct position. A microcamera
mera optics need only correct the residual aberra-                   is essentially a finite conjugate relay lens. Because
tions over a small field, which can readily be                       the objective is monocentric, its aberrations are field
achieved using optical components with manageable                    independent and thus the same microcamera can be
levels of complexity. Here “multiscale” refers to the                used for all field angles. The main objective lens is
use of a large objective lens combined with smaller                  held in place by either a passively or actively aligned
micro-optics to achieve both small-scale aberrations                 mechanical frame, depending on the level of accuracy
and increased clear aperture diameter.                               required.
   Using an array of lenses to relay an image is a well-                The designs illustrated in the following sections
known technique for applications such as document                    are for the AWARE-2 wide-field camera. AWARE-2
copying and three-dimensional imaging [6,7], and                     was designed to realize a platform capable of captur-
arrays of individual cameras have been used for a                    ing up to 120° field of view (FOV) in all directions
variety of interesting applications such as synthetic                with diffraction-limited resolution of over two giga-
apertures and high-speed photography [8]. The key                    pixels and the ability to focus from infinity to 30 m
difference between these systems and a multiscale                    in object space, achieving an instantaneous FOV
design lies in the use of optical components of differ-              (iFOV) of 40 μrad with 1.4 μm pixel pitch CMOS
ent scales.                                                          sensors. This is the first of our full-scale gigapixel
   The division of the imaging work in multiscale                    systems and serves as a testbed for optimization of
systems allows each set of microcamera optics to                     future cameras.
be much simpler than a single monolithic design                         A ZEMAX [9] optical design of AWARE-2 is shown
and creates the potential for massively parallelizable               in Fig. 2. The main objective lens is a two-layer glass
image acquisition. However, precisely holding and                    ball and the micro-optics consist of three plastic
aligning each of these microcameras to form proper                   optical groups. This design has a focal length of
images and overlap requires an innovative optome-                    34.2 mm and an f -number of 2.2 at infinity focus.
chanical approach. Furthermore, this increase in                     Table 1 shows the alignment tolerance budget dic-
array density is only possible if the physical compo-                tated by the optical design to achieve the target
nents can be tightly packed, thus presenting a signif-               image quality of above 20% modulation transfer
icant engineering challenge not present in previous                  function at the Nyquist frequency (357 cycles∕mm).
systems.                                                             The tolerance budget is split into two groups, the
   The overall goal of our optomechanical design is                  micro-optics and the subassembly. Micro-optics toler-
the same as that of traditional optical systems: build               ance specifies the alignment required between the
a stable platform to position the optical components                 components of the microcamera with respect to each
according to the optical design in a way that can be                 other in order to maintain image quality at the
readily manufactured and assembled. However, the                     sensor. Subassembly tolerance specifies the align-
details unique to multiscale design require the reso-                ment required between the main objective and the
lution of several mechanical design, assembly, and                   microcamera assembly. As shown, the subassembly
tolerancing issues to successfully construct a multi-
scale gigapixel scale digital camera. We address
these issues here via descriptions of the designs we
implemented in our first prototype. Section 2 de-
scribes the general design considerations and compo-
nent design through an example of a two-gigapixel
prototype system called AWARE-2 (here 2 refers to
the resolution capability of the camera in gigapixels).
Section 3 describes assembly and alignment issues in                 Fig. 2. (Color online) ZEMAX optical design of AWARE-2. The
building AWARE-2. Results from mechanical and                        chief ray and two marginal rays of 0° field angle are shown.

1542     APPLIED OPTICS / Vol. 52, No. 8 / 10 March 2013
Optomechanical design of multiscale gigapixel digital camera
Table 1.   Tolerance Budget for AWARE-2a

             (a) Microcamera                  (b) Subassembly
     Surface decenter          25 μm       Lateral         300 μm
     Surface tilt              0.05°        Axial          150 μm
     Thickness                 25 μm         Tilt           0.12°
     a
    (a) Optical alignment tolerances of micro-optics with respect
 to one another. (b) Optical alignment tolerances between
 objective lens and microcamera assembly.

                                                                    Fig. 3. (Color online) Illustration of how having an image that
tolerances are much looser in comparison to those                   slightly underfills the sensor requires careful alignment to prevent
of the micro-optics due to the fact that the working                loss of important image areas.
f -number of the main objective is large (∼3.5) com-
pared to that of the microcamera (∼2.2).
                                                                    maximize light collection when close packed. Optical
A.       Micro-Camera Design                                        components for each microcamera are assembled
In a monocentric camera, the logistics of system as-                into machined aluminum lens barrels as shown in
sembly and repair become straightforward if the mi-                 Fig. 4(b). The plastic optics are positioned and
crocameras are designed to be mass produced and                     aligned in machined seats in the barrels and cemen-
interchangeable. These criteria influenced the de-                  ted down, which hold axial and lateral misalign-
sign of the system toward making each microcamera                   ments to within 13 μm of the design, well within the
an independent unit that can be easily installed and                micro-optics tolerance indicated in Table 1. The outer
replaced. This strategy allows us to take advantage                 diameter of the barrel serves as a datum for center-
of economies of scale, similar to the use of modular                ing, and a flange serves as a datum for tip/tilt and
integral field spectrograph units planned for use in                alignment along the axis.
the VIRUS integral spectrometer array [10]. In addi-                  One of the most powerful features of the multiscale
tion, as was seen in Table 1, the micro-optics align-               design is the ability to provide each microcamera
ment tolerance is much tighter than the subassembly                 with an independent focus. This is especially challen-
tolerance. This further justifies independently build-              ging in AWARE-2 due to the limited lateral space
ing each microcamera to ensure the micro-optics are                 available in a close-packed configuration, and most
aligned precisely in each barrel. Each microcamera                  focus adjusting technologies, such as voice coils
consists of a barrel containing the optics for imaging              and deformable optics, proved inadequate for this
and a sensor module containing the digital CMOS                     application. Focus in AWARE-2 is thus achieved by
sensor. Optical design principles for the micro-optics              translating the CMOS sensor with respect to the
are described in detail by Marks et al. [11]. The mi-               image plane. The sensor module, shown in Fig. 5,
crocamera design process is a complex interplay be-                 consists of the sensor and the associated circuit
tween optical properties (e.g., demagnification, FOV,               boards for receiving and transmitting data. The sen-
and associated tolerances), mechanical constraints                  sor package itself is mounted to a flexible circuit
(e.g., physical size of the cameras, manufacturability,             board, which can be translated along the optical axis
and assembly), and the impact of the electronics (e.g.,
thermal management and cabling).
   Microcamera designs require special considera-
tions uncommon to most imaging systems. Due to
the need to tightly pack the microcameras side by
side in the array, the optics and mechanical supports
must be laterally compact. This forces the cameras to
be long along the optical axis to fit necessary compo-
nents and take advantage of the divergence between
cameras. Typical single aperture cameras tend to
overfill the sensor so that all the pixels are utilized
in forming an image. In contrast, the image in our
system tends to underfill the sensor in the microca-
mera to ensure that most of the usable field points
are captured while ensuring sufficient overlap be-
tween adjacent cameras. This requires careful align-
ment of the sensor to each microcamera barrel so
that overlap regions are not clipped off, as illustrated
in Fig. 3.
   The micro-optics for the AWARE-2 camera are                      Fig. 4. (Color online) (a) Solid drawing of micro-optics for
shown in Fig. 4 and consist of three groups of plastic              AWARE-2. (b) Solid drawing of micro-optics assembled into a
elements. The front optic has a hexagonal profile to                cutout barrel.

                                                                         10 March 2013 / Vol. 52, No. 8 / APPLIED OPTICS          1543
Optomechanical design of multiscale gigapixel digital camera
array. Determining this envelope is discussed in
                                                                    detail by Marks et al. [11].
                                                                    B. Dome Structure
                                                                    Naturally, once the microcameras are independently
                                                                    assembled and tested, a structure is needed to posi-
                                                                    tion them correctly relative to the main objective lens
                                                                    in order to form an array of cohesive images. Active
                                                                    alignment of each micro-camera to the objective is
                                                                    impractical, both in terms of cost and complexity, and
Fig. 5. (Color online) Solid drawing of complete sensor module
                                                                    unnecessary since small lateral and longitudinal
assembly.                                                           misalignments can be compensated for by shifting
                                                                    the image in postprocessing and adjusting focus, re-
                                                                    spectively. We concluded that for the AWARE-2 sys-
via a fine-pitched screw for focus. Sensor position can             tem, a passive alignment method based on using a
be set by manual rotation of the focal screw or by                  structure machined on a five-axis mill is sufficient.
an automated servo motor. To achieve the object side                Specialty machine shops can easily obtain tolerances
focal range of infinity to 30 m, 100 μm of sensor travel            on the order of 25–50 μm and 0.05°, which are well
were required. Alignment of the sensor face to the                  within the subassembly tolerances in Table 1.
barrel axis is achieved via adjustable pin and dog-                    For the monocentric case, the passive alignment
bone bushings, which line up with pins on the barrel                structure follows a spherical dome. This dome fixture
face. More details on the alignment procedure be-                   must be able to hold each microcamera within a cer-
tween the barrel and the sensor module are given                    tain tolerance dictated by either image quality or im-
in Section 4.                                                       age overlap. The dome should also have mechanisms
   Attachment of the two components was accom-                      in place for easily installing and replacing any micro-
plished by flexible wire clips, which provide compres-              camera without disturbing others.
sion of the sensor face against the barrel. A fully                    The AWARE-2 support dome was machined out of
assembled microcamera is shown in Fig. 6.                           aluminum, as shown in Fig. 7. Each microcamera has
   Having one microcamera design to cover the entire                a corresponding hole in the dome. The microcameras
field is desirable from a design, assembly, and cost                sit in counterbore holes and are locked down with
perspective. The conical envelope inside which the                  press fit pins on the sides. For a wide-field applica-
microcameras must fit in order to pack them in                      tion where the dome needs to cover a large section
the array must be determined to prevent physical
interference during assembly. There is currently no
known way of packing an arbitrary number of equal
circles on the surface of a sphere while eliminating
the variation in nearest-neighbor distances between
circles. Therefore, if only one microcamera design can
be used, it must fit within the envelope determined
by the smallest nearest-neighbor distance in the

Fig. 6. (Color online) (a) Solid drawing of microcamera assembly.
(b) Photo of actual microcamera with a US quarter coin for size     Fig. 7. (Color online) (a) Solid drawing of dome. (b) Photo of
comparison.                                                         actual machined aluminum dome.

1544     APPLIED OPTICS / Vol. 52, No. 8 / 10 March 2013
Optomechanical design of multiscale gigapixel digital camera
of a sphere, it is typically difficult to tightly pack the
cameras while maintaining low variations in interca-
mera distances. A packing strategy based on a dis-
torted icosahedral geodesic was developed for this
purpose [12]. Son et al. concluded that a spatial pack-
ing density of around 0.85, which is near the theore-
tical maximum of a geodesic configuration, can be
achieved by using hexagonal and pentagonal shaped
front micro-optics,. This arrangement was used for
AWARE-2, where the 120° FOV is covered by 226 mi-
crocameras. The packing geometry of the dome can be
adjusted for different applications. For example, for a
wide, rectangular FOV, simpler schemes based on
slightly perturbed hexagonal packing can be applied.
C.   Objective Lens Mount
Depending on the tolerances of the optical design,
active alignment of the main objective lens could
be required in order to obtain acceptable image qual-
ity and overlap. For the monocentric case where the
objective is spherically symmetric, rotational align-
ments (tip/tilt) are unnecessary and only transla-
tional positioning is needed. Traditional precision
machining can readily align an objective lens to             Fig. 8. (Color online) (a) Solid drawing of flexure mount for main
around 100 μm relative to the center of the dome             objective lens. (b) Photo of machined flexure.
using passive alignment features. This assumes that
parts are machined to 25 μm and several components
connect the objective to the microcameras, resulting         the axis of the barrel [Fig. 9(a)]. The sensor module
in a tolerance stack-up. If higher accuracy is re-           has corresponding bushings on its matching side
quired, an active alignment scheme utilizing a flex-         [Fig. 9(b)], which control the lateral and rotational
ure structure can be used to position the objective.         position of the sensor.
The alignment range should reflect the machining                Alignment is achieved by adjusting the position of
tolerances it needs to compensate.                           the bushings so that they will center the active area
   An aluminum flexure frame capable of translating          of the sensor to the axis of the barrel. The schematic
the position of the objective lens in all three directions   for this method is illustrated in Fig. 10. A shadow
was constructed in case passive alignment strategies         mask is back illuminated and displays a precision
proved inadequate (Fig. 8). The flexure is actuated by
precision set screws that are able to translate the lens
with a resolution of tens of micrometers. Lateral
actuation is achieved via two set screws and a third
screw to lock the position. Axial translation is actu-
ated by three screws placed directly over the lateral
actuation screws. For AWARE-2, we found passive
alignment to be sufficient, but for future applications
where higher precision is required the proposed flex-
ure mount is a likely option.

3. Assembly and Alignment of AWARE-2
Most of the components in AWARE-2 were designed
to be passively aligned using machined features for
ease of assembly and low assembly costs. However,
an active alignment step was required to center the
sensor to the microcamera barrel axis in order to en-
sure proper image overlap. This alignment procedure
and the subsequent installation of the microcameras
into the dome are described below.
A.   Sensor-to-Barrel Alignment
The sensor is aligned to the barrel by adjustable
bushings. The back of the optics barrel has two align-       Fig. 9. (Color online) (a) Back of optics barrel showing alignment
ment pins, which are precisely positioned relative to        pins. (b) Front of sensor module showing pin and dogbone bushings.

                                                                  10 March 2013 / Vol. 52, No. 8 / APPLIED OPTICS         1545
Optomechanical design of multiscale gigapixel digital camera
Fig. 11. (Color online) (a) View from behind dome of arrangement
Fig. 10. (Color online) Steps for aligning bushings on the sensor.   of microcameras in AWARE-2 system. (b) View from inside dome of
(a) Align bushings relative to a known shadow mask. (b) Illuminate   stacked micro-optics.
mask to cast shadow on sensor. (c) Move sensor into position and
power on to capture shadow position. Adjust position and angle
until mask pattern is aligned with predetermined position.
(d) Photo of actual setup.

pattern on the sensor face. The shadow mask is held
in a jig that holds the bushings in a known position
relative to the mask. Alignment of the sensor to the
bushings is performed by powering up the sensor and
adjusting the lateral and rotational position until the
mask pattern is in the proper position. Once they are
in place, the bushings are bonded to the sensor head
with UV cure epoxy. A small sample group of sensors
were checked after the epoxy was cured to ensure
shrinkage did not shift the sensor more than 50 μm.
B.   Camera Assembly
Once the microcameras have been assembled, they
can be used to populate the dome to cover the tar-
geted FOV. For a ground-based camera, it is reason-
able to only populate the cameras along a horizontal
strip where most of the interesting features are
located. This type of arrangement was initially used
for AWARE-2 and is shown in Fig. 11. Ninety-eight
microcameras capture a 1 gigapixel image with a
FOV of approximately 120° × 40°.
  Insertion of the microcamera into the dome is illu-
strated in Fig. 12. A clocking pin is used to properly               Fig. 12. (Color online) (a) View from sensor perspective of micro-
align the orientation of the sensor face for maximum                 camera retention and clocking mechanisms. (b) Cross-sectional
overlap. The microcameras are held in the dome by                    view of microcamera in dome showing how the camera seats into
press fit roll pins. The face of the counterbore hole in             the counterbore. (c) Photo of assembly step.

1546     APPLIED OPTICS / Vol. 52, No. 8 / 10 March 2013
Optomechanical design of multiscale gigapixel digital camera
Fig. 14. (Color online) (a) Change in BFL as a function of uniform
                                                                         temperature of dome. (b) Change in spot size at edge of field as a
                                                                         function of uniform temperature of dome.
Fig. 13. (Color online) (a) Photo of assembly of camera to front
faceplate of enclosure. (b) Full camera in enclosure with electronics.
                                                                         this rigidity is achieved due to the inherently sturdy
                                                                         nature of the honeycomb-like, hexagonally distribu-
the dome serves as the datum that sets the pointing                      ted holes employed in the design.
angle and axial position of each microcamera, while                      B. Thermal Simulations
the sidewalls of the body of the hole set the lateral
position. The faces of the counterbore holes were                        The optical performance of a complex and large
machined to within 0.05° tip/tilt tolerance, which                       optical system such as AWARE-2 can be highly
is well within the 0.12° subassembly tolerance speci-                    susceptible to temperature variations and gradients.
fied in Table 1, by a five-axis mill.                                    Thermal loads can come from external sources,
   Figure 13 shows the assembly process of the                           such as ambient temperature changes, or internal
AWARE-2 camera system. Figure 13(a) shows the                            sources, such as the heat generated from the sensor
front plate of the camera enclosure that holds                           and other electronic components within the system.
the dome that houses the microcameras and some                           Internal thermal effects on system performance can
of the electronics. Figure 13(b) shows the full enclo-                   be significant with each microcamera producing up
sure completely assembled.                                               to 1 W of heat at full frame rate.
                                                                           To investigate changes in environmental tempera-
                                                                         ture, a first-order thermal analysis was performed
4. Mechanical Stress and Thermal Considerations                          where the system temperature was varied as a
A.   Mechanical Simulations
To ensure that the weight of the microcameras does
not excessively deform the dome, load simulations
were performed in SolidWorks [13]. Each microca-
mera was modeled as a 41 g weight evenly distribu-
ted on the inside surface of the counterbore holes in
the dome. This ignores any moments or uneven
weight concentrations, but the results should provide
a good estimate of the mechanical strain applied to
the system. The direction of gravity was varied to
simulate the camera facing up, down, and to the side.
Material displacements of less than 1 μm are pre-                        Fig. 15. (Color online) Temperature gradient induced in dome
dicted, indicating that the dome is structurally very                    by internally generated heat from microcameras. 1 W of heat per
rigid. Even though most of the dome is hollowed out,                     microcamera at 20°C room temperature.

                                                                              10 March 2013 / Vol. 52, No. 8 / APPLIED OPTICS         1547
Optomechanical design of multiscale gigapixel digital camera
Table 2.   Change in Pointing Angles and Positions of Microcameras due   actual AWARE-2 system the microcameras are air
                        to Thermal Deformations
                                                                         cooled via several fans so that significant amounts
                   N                  T           Pointing Angle         of heat are pumped out through forced convection
Camera        Displacement      Displacement      Displacement           rather than conductively through the dome, so these
No.               (μm)              (μm)             (degrees)           simulations should represent a worst-case scenario.
1                  89.9              0.3               0.010
                                                                         5. Conclusion
2                  85.4              0.5               0.018
3                  78.5              0.8               0.026             In this work, we have presented general mechanical
4                  69.3              1.5               0.033             guidelines for designing a multiscale, gigapixel cam-
5                  58.0              2.6               0.040             era and details on a working prototype based on
6                  45.1              4.1               0.044             these principles. Although this first system design
7                  31.5              5.4               0.046             and construction effort leaves room for further design
8                  18.8              5.6               0.040
                                                                         improvements, we believe that this work has laid a
9                   9.8              4.2               0.022
                                                                         solid foundation for future development of gigapixel
                                                                         systems.
                                                                             Some key points that are currently being in-
                                                                         vestigated for next-generation cameras are the
                                                                         mechanisms for installation and assembly of the mi-
                                                                         crocameras, microcamera focus, and thermal manage-
                                                                         ment. The current method of using roll pins to provide
                                                                         lateral compression that holds the microcameras in
                                                                         the dome does not fully utilize the alignment face
                                                                         on the counterbores and could result in pointing er-
                                                                         rors. In addition, the alignment of the bushings
                                                                         and assembly of the sensors to the barrels are
                                                                         labor-intensive procedures and should be simplified
                                                                         in order to make manufacturing more scalable. We
                                                                         have also found that while moving the sensor face
                                                                         for focus was possible, the design could be simplified
uniform thermal bath between 0 °C and 40 °C from                         if the focus was adjusted by moving the optics instead.
an initial temperature of 20 °C. With its spherical                      A focal mechanism that translates an optical compo-
symmetry and uniform use of T6061 aluminum                               nent is currently being developed. Thermal loads for
throughout the structural components, all the micro-                     the first-generation system were easily managed by
cameras in AWARE-2 expand or contract about the                          forced convection. However, as we scale up the resolu-
center of the dome under uniform temperature var-                        tion by incorporating a larger number of microca-
iation. An optical analysis through ZEMAX shows                          meras, internal heat generation will scale roughly
that these types of uniform temperature variations                       with the square of the radius of the dome, while heat
can be compensated for by an adjustment in the back                      dissipation by conduction through the dome only
focal length (BFL) of the microcameras. The change                       scales with the radius of the dome. This indicates that
in BFL for various temperature values and changes                        heat sinking through the dome will eventually be
in optimized image spot size near the edge of the field                  inadequate and extracting the heat directly out of
for AWARE-2 are shown in Fig. 14.                                        the back of the microcameras will be necessary in or-
   Internally generated heat from the microcameras                       der to make this a truly scalable system. While these
can cause significant temperature gradients in the                       challenges remain for higher pixel-count systems, the
dome. Figure 15 shows the simulated temperature                          current 1 gigapixel AWARE-2 is fully operational and
distribution of a symmetric section of the dome in                       has been collecting images at various sites [14].
SolidWorks when each microcamera dumps 1 W of                              We gratefully acknowledge David Kittle and
heat directly into each hole of the dome. The front                      Andrew Holmgren for their assistance and the
of the dome is constrained to be at room temperature                     Defense Advanced Research Projects Agency (DARPA)
(20 °C).                                                                 for financial support via award HR-0011-10-C-007.
   This temperature gradient generates thermal de-
formations of the dome and leads to changes in point-                    References
ing angle and position of each microcamera. The                           1. A. W. Lohmann, “Scaling laws for lens systems,” Appl. Opt. 28,
changes in a sample set of nine microcameras from                            4996–4998 (1989).
                                                                          2. K. W. Hodapp, U. Laux, W. A. Siegmund, N. Kaiser, K. C.
the center of the dome to the outer perimeter were                           Chambers, U. Laux, J. Morgan, and E. Mannery, “Optical
tracked and the results are shown in Table 2. Point-                         design of the Pan-STARRS telescopes,” Proc. SPIE 5249,
ing angle deviations of less than 0.05°, lateral displa-                     165–174 (2004).
cements (T direction) of less than 6 μm, and axial                        3. B. Leininger, J. Edwards, J. Antoniades, D. Chester, D. Haas,
                                                                             E. Liu, M. Stevens, C. Gershfield, M. Braun, J. D. Targove, S.
displacements (N direction) of less than 100 μm                              Wein, P. Brewer, D. G. Madden, and K. H. Shafique, “Auton-
are observed, all of which are within the tolerances                         omous real-time ground ubiquitous surveillance—imaging
listed in Table 1 to maintain image quality. In the                          system (ARGUS-IS),” Proc. SPIE 6981, 69810H (2008).

1548       APPLIED OPTICS / Vol. 52, No. 8 / 10 March 2013
Optomechanical design of multiscale gigapixel digital camera
4. D. J. Brady and N. Hagen, “Multiscale lens design,” Opt.                 J. D. Murphy, T. Prochaska, R. D. Allen, R. Bender, G. Blanc,
    Express 17, 10659–10674 (2009).                                          T. Chonis, G. Dalton, K. Gebhardt, J. Good, D. Haynes,
 5. D. L. Marks and D. J. Brady, “Close-up imaging using micro-              T. Jahn, P. J. MacQueen, M. D. Rafal, M. M. Roth, R. D.
    camera arrays for focal plane synthesis,” Opt. Eng. 50, 033205           Savage, and J. Snigula, “VIRUS: production of a massively
    (2011).                                                                  replicated 33 k fiber integral field spectrograph for the
 6. R. H. Anderson, “Close-up imaging of documents and displays              upgraded Hobby–Eberly telescope,” Proc. SPIE 8446,
    with lens arrays,” Appl. Opt. 18, 477–484 (1979).                        84460N (2012).
 7. R. F. Stevens and T. G. Harvey, “Lens arrays for a three-          11.   D. L. Marks, H. S. Son, J. Kim, and D. J. Brady, “Engineering a
    dimensional imaging system,” J. Opt. Pure Appl. Opt. 4,                  gigapixel monocentric multiscale camera,” Opt. Eng. 51,
    S17–S21 (2002).                                                          083202 (2012).
 8. B. Wilburn, N. Joshi, V. Vaish, E. V. Talvala, E. Antunez, A.      12.   H. S. Son, D. L. Marks, J. Hahn, J. Kim, and D. J. Brady,
    Barth, A. Adams, M. Horowitz, and M. Levoy, “High perfor-                “Design of a spherical focal surface using close-packed relay
    mance imaging using large camera arrays,” ACM Trans.                     optics,” Opt. Express 19, 16132–16138 (2011).
    Graph. 24, 765–776 (2005).                                         13.   SolidWorks 2012, Dassault Systemes SolidWorks Corp.
 9. Zemax 12, Radiant Zemax LLC (Redmond, Washington,                        (Waltham, Massachusetts, 2012). http://www.solidworks.com.
    2012). http://www.radiantzemax.com.                                14.   D. J. Brady, M. E. Gehm, R. A. Stack, D. L. Marks, D. S. Kittle,
10. G. J. Hill, S. E. Tuttle, H. Lee, B. L. Vattiat, M. E. Cornell,          D. R. Golish, E. M. Vera, and S. D. Feller, “Multiscale gigapixel
    D. L. DePoy, N. Drory, M. H. Fabricius, A. Kelz, J. L. Marshall,         photography,” Nature 486, 386–389 (2012).

                                                                             10 March 2013 / Vol. 52, No. 8 / APPLIED OPTICS            1549
Optomechanical design of multiscale gigapixel digital camera
You can also read