Automated assembly of camera modules using Active Alignment with

up to six degrees of freedom

K. Bräuniger a, D. Stickler a, D. Winters a, C. Volmer a, M. Jahn, S. Kreya
a
TRIOPTICS GmbH, Hafenstr. 35-39, 22880 Wedel, Germany

ABSTRACT

With the upcoming Ultra High Definition (UHD) cameras, the accurate alignment of optical systems with respect to the
UHD image sensor becomes increasingly important. Even with a perfect objective lens, the image quality will deteriorate
when it is poorly aligned to the sensor. For evaluating the imaging quality the Modulation Transfer Function (MTF) is
used as the most accepted test. In the first part it is described how the alignment errors that lead to a low imaging quality
can be measured. Collimators with crosshair at defined field positions or a test chart are used as object generators for
infinite-finite or respectively finite-finite conjugation. The process how to align the image sensor accurately to the
optical system will be described. The focus position, shift, tilt and rotation of the image sensor are automatically
corrected to obtain an optimized MTF for all field positions including the center. The software algorithm to grab images,
calculate the MTF and adjust the image sensor in six degrees of freedom within less than 30 seconds per UHD camera
module is described. The resulting accuracy of the image sensor rotation is better than 2 arcmin and the accuracy
position alignment in x,y,z is better 2 µm. Finally, the process of gluing and UV-curing is described and how it is
managed in the integrated process.
Keywords: active alignment, MTF, camera, objective, sensor, camera module, six degrees of freedom

1. INTRODUCTION
The world of camera module has a wide variety like video and photo cameras for the tip of endoscopes, surveillance,
mobile phone, consumer, security, game console, and the automotive sector. All types of camera module are available
nowadays not only performing 2D imaging but also 3D imaging.
In optical manufacturing the accurate positioning of the optical elements is an essential requirement to reach best optical
quality. For objective lenses this is valid for each lens element. After the assembly, it’s common to use a mechanical
mount, such as screwing, to fix the lens to the sensor unit. The camera module has then only one degree of freedom to
bring the on-axis focus to the specified object distance. Manufacturing error like misalignment of single elements lead
not only to a lower image quality and slightly changes of focal length but also to a tilt in the image plane. A tilt in the
image plane is equal to objects that are in front focus on one side and in back focus on its opposite side. For camera
modules which are on-axis focus-adjusted to infinity, back-focus has no real object distance and such sensor areas are
lost. Not only simple one axis alignment but also 3 axis translation alignment isn’t able to overcome that issue. To reach
best performance and a high production yield of a high quality camera module, active alignment in 5 to 6 degrees of
freedom is necessary. The 5-axis manipulation aligns the sensor in both tilt-directions, centers the on-axis focus and
adjusts the focus to the correct object distance. The last degree (the sixth) of freedom is aligned by rotating of the sensor,
which is necessary for applications, where the sensor orientation has to fit to a reference of the lens unit.

2. MEASUREMENT PRINCIPLE

2.1 Measurement of Modulation Transfer Function (MTF)

The Modulation Transfer Function (MTF) is an important aid to objective evaluation of the image-forming capability of
optical systems[1,2]. Not only that the MTF provides a means of expressing the imaging quality of optical systems
objectively and quantitatively, but it can be calculated from the lens design data.

Photonic Instrumentation Engineering, edited by Yakov G. Soskind, Craig Olson,

Proc. of SPIE Vol. 8992, 89920F · © 2014 SPIE · CCC code: 0277-786X/14/$18
doi: 10.1117/12.2041754

Proc. of SPIE Vol. 8992 89920F-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 The Modulation Transfer Function (MTF), describing the resolution and performance of an optical system is the ratio of
relative image contrast divided by relative object contrast.
MTF = Relative Image Contrast/ Relative Object Contrast
The MTF is a function of a spatial frequency like “line pairs per mm” (lp/mm) or for camera modules “per pixel pitch
(lp/pixel). When an object (illuminated test chart or reticle) is observed with an optical system, the resulting image will
be somewhat degraded due to inevitable aberrations and diffraction phenomena. In addition, a real lens will not fully
conform to the design data. Manufacturing errors, assembly and alignment errors in the optics will deteriorate the overall
imaging performance of the system. As a result, in the image, bright highlights will not appear as bright as they do in the
object, and dark or shadowed areas will not be as black as those observed in the original patterns. In general an
illuminated target can be defined by its spatial frequency (number of bright and dark areas per millimeter) and the
contrast (the apparent difference in brightness between bright and dark areas of the image).
By convention, the modulation transfer function is normalized to unity at zero spatial frequency. For low spatial
frequencies, the modulation transfer function is close to 1 (or 100%) and generally falls as the spatial frequency increases
until it reaches zero. The contrast values are lower for higher spatial frequencies as shown above. As spatial frequency
increases, the MTF curve falls until it reaches zero. This is the limit of resolution for a given optical system or the so
called cut off frequency (see figure below). When the contrast value reaches zero, the image becomes a uniform shade of
grey.

Figure 1. Change of the Modulation by an imaging lens. The object’s modulation has the full (black to white) amplitude,
while at the image plane the modulation is a function of spatial frequency.

The grids shown in Figure 1 are actually no longer used in order to measure the MTF. Modern MTF-instruments use a
single illuminated slit on opaque background as the object. From a mathematical point of view a single slit can be
regarded as the sum over all spatial frequencies (Fourier synthesis). All frequencies contribute with the same amplitude
(=1) to this slit not taking the finite slit width into account for this description. This single slit will be imaged into the
image plane of the sample. Due to diffraction and aberrations there will be no perfect slit image in this plane, instead the
slit image is broadened. It represents the so called Line Spread Function (LSF). On the basis of Fourier analysis the
contribution of each spatial frequency to the LSF can be calculated. Actually the amplitude of each spatial frequency is
equal to the contrast at this frequency. The Fourier analysis of the Lines Spread Function corresponds to the MTF of the
sample. Taking a single image of the LSF unveils the complete MTF. Alternatively it is also possible to use a cross (i.e.
two perpendicular slits) for the target. This enables the instrument to measure the MTF in two image directions
simultaneously. The basic terms and mathematical relations used for MTF are described in the ISO 9334 standard.
The modulation transfer function varies not only related to the spatial frequency but also with the position in the field of
view. The MTF measurement along the axis of symmetry of the optical system is known as on-axis measurement.
To completely characterize the imaging performance of an optical system, the MTF must be measured at different
positions within the field of view. The MTF measurement within the field of view is known as off-axis measurement. In
order to achieve an off-axis measurement, the target is placed in the field of view at the desired object position.
In the so-called finite-finite imaging condition the illuminated slit or crosshair target is directly moved in the object plane
of the sample. In the more common infinite-finite imaging condition, the illuminated slit or crosshair is part of a

Proc. of SPIE Vol. 8992 89920F-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 collimator projecting the target to infinity. The collimators are then oriented at different off-axis angles for characterizing
the MTF at the corresponding image fields.

2.2 Measurement of Image Plane-Tilt

Centering errors or mechanical tolerances lead to focus deviation and a tilted image plane. The result is a non-centered
image with variation in object distances from opposite field position. To determine the image plane or the angle between
the image plane and the sensor (see Figure 2 left), a through focus-scan by vary the distance between lens and sensor
while looking to a fixed object is the common technique. After the measurement, the stage aligns the angle between lens
and sensor to have a flat image plane on the sensor (see Figure 2 right).

collimated light
or test chart

Tilt and Shift

corrections

Figure 2. Image plane measurement via Through-Focus-Scan of the lens, tilt and shift correction and final measurement of
the aligned lens.

Figure 3 (a) shows an example for a sensor image, while the camera is looking to an object test chart or toward a
collimator system. The square surroundings give the searching area (regions of interest) for the crosses. During the scan
the system has to calculate the MTF. Looking to the variation of the MFT at a specified spatial frequency (e.g. 0.2
lp/pixel) for each field position, maxima will appear for the fields best focus positions. Figure 3 (b) shows an example
for the MTF variation versus the focusing (Z) position (distance between lens and sensor). In this example the best focus
for left, the center and right are on different heights, while upper and lower border vary simultaneously. Compared to the
sensor image of Figure 3 (a) this is the result for a tilt in the horizontal direction. A perfect sensor lens alignment is
given, when the through focus curves have their maxima at the same height. The center (on-axis) focus-position can be
different due to slight field curvature. In case of field curvature, the best focus position is calculated by weighting the
center and the field position.
In the given example, 5 positions and 5 curves are shown, but in the real measurement, every curve is separated into a
sagittal and tangential MTF through focus-curve or MTF in horizontal and vertical sensor direction. To avoid aliasing
effects, the object crosses are slightly rotated.

3. INSTRUMENT
3.1 Element for alignment and object distance
The first basic question of active camera module alignment is to align the lens in front of a fixed sensor unit or the sensor
behind a fixed lens. For 2D camera module, this question is answered by the element that has the reference for later
integration of the camera module. This can be a reference surface of the lens package or of the sensor unit. 3D camera
modules consist of a sensor body with two separated sensors or one sensor element with two CCD pixel arrays. In this

Proc. of SPIE Vol. 8992 89920F-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 case it is necessary to fix the sensor package and to align the lens above. Otherwise a simultaneous image plane-tilt
alignment is impossible or both lenses are looking into different direction.
The second basic question is the type of object generator and the object distance. For infinity object distance a collimator
system of five or more collimators with cross hair reticle is used. The advantages of a collimator system are the real
infinite object distance and the easy handling for wide angle camera modules by positioning the collimators at specific
field angles. The focal length for the collimator and the reticle slit width has to be optimized for focal length and the
pixel pitch of the camera module. For finite distances a test chart reticle with multiple cross hairs on a larger glass plate
with a back light is a solution. This is the optimum for short distances like the object distance of endoscopes. Such test
charts are easy to mount on a linear stage to vary the object distance also after fixing sensor and lens together. To vary
the object distance with a collimator system, focusing collimators are an option.

MTF @ 0.2 Ip/pixel

1
rn.

(D
d
n
(a) (b)
Figure 3. (a) sensor image of a test chart and searching areas for crosses at the on-axis and 4 off-axis positions, (b) example
of a through-focus-scan of non-aligned sample.

3.2 Instrumental setup

Figure 4 show the stand alone instrument for small and intermediate camera modules. The sequence of dispensing,
alignment and bonding will be described in the next paragraph. The main components of the system are schematically
drawn in Figure 5 (a). As mentioned before, one component (lens or sensor package) is fixed by the sample reference
holder. The second part is gripped and moved by a 6 axis stage. The 6-axis stage is able to carry a payload of up to 3 kg.
Its bi-directional repeatability is better than 0.5 µm for all translations and better than 0.02 mrad (4 arcsec) for tilt and
rotation.
Above the camera module the instrument has the object generator like a collimator system (see Figure 5 (b)) or a test
chart. While a collimator is easy to adjust to wide field angles, the field range of a test chart is limited by its size and the
maximum size of the back-light. Typical sizes are up to 200 mm x 200 mm. Otherwise the target for to be spitted its
different fractions with individual light sources.

Proc. of SPIE Vol. 8992 89920F-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 Figure 4. Stand-alone instrument for automated assembly of camera modules with active alignment, dispensing and curing.

Further components are a dispenser unit, a UV light source, sample transportation, an alignment tool, process camera and
in-door light. The alignment tool is an autocollimator to align and center the reference of the camera module and the
target system one to another. Fine-adjustment screws for sample holder tilt-adjustment and tilt and centration adjustment
of the test chart exist.
All motorized axes and the pneumatic sub-system are controlled via a single 8-axis controller.
The sample holder and gripper have fast-lock interfaces to use the instrument for different module types.

4.oWiramr

objective lens
semple M1Oltler with Objective)

pk holder with sen

(a) (b)
Figure 5. (a) schematic drawning of the main components for active alignment with fixed sensor unit, gripped lens, 6-axis
stage and collimator system.

Proc. of SPIE Vol. 8992 89920F-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 The active alignment system in a modular design has options to test the aligned and bonded camera modules. The
available quality check test functions for camera performance-verification include tests like, color reproduction, Opto
Electric Conversion Function (OECF), exposure and distortion. The system is able to test these functions with a specific
test chart. With a homogeneous pull-in back-light or an Ulbricht-Sphere, functions like vignetting and dead/hot pixel are
available.

4. SOFTWARE AND SEQUENCE

The software for active alignment has several features for image processing, MTF calculation, image plane calculation,
definition of the regions of interests and stage control etc. The process control for the whole sequence is script
controlled. This script base controlling has the benefit that the process is easy to change to a more complex version, or
the costumer is able to implement additional steps for quality check or interactions in sequence breaks.

After inserting sensor and lens package, the process sequence gets started. In a first step, the slider for sample
transportation moves the sample to the dispensing position.

(a) (b)
Figure 6. (a) sample at dispensing position, (b) sample at alignment position. Gripper and dispenser mounted on 6-axis
stage.

4.1 Dispensing
Depending on the type of glue and its viscosity a volume dispenser or an electric pneumatic dispenser is used. Figure 6
(a) shows a high precision volume dispenser for low viscosity glue. Via a bridge, the dispenser is mounted on the same
6-axes stage like the gripper for alignment. Therefore the dispenser is able to make smooth circular movement, even with
slightly tilted pointing to the sample in the range of up to 5°. In a modular version, the dispensing is done by a separate
stage system and tilting is realized by additional hardware. The dispensing motion can be programmed to point, line,
and/or segment dispensing. After dispensing the dispenser moves back and the sample gets transported to the alignment
position.

Proc. of SPIE Vol. 8992 89920F-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 4.2 Alignment, curing and testing
After reaching the alignment position, the already inserted lens gets gripped and lifted or the variable element gets
gripped from a position close-by and is moved to the fixed element (see Figure 6 (b)).
The alignment sequence starts with a rough centering, followed be a through-focus scan to determine the angle between
sensor and image-plane. The angle gets corrected without a strong focus shift, because the 6-axis stage is able to set the
pivot point to the gripper position. A second scan is used to check the image plane-alignment and to determine the new
best-focus position after alignment. After the alignment process the glue becomes cured via high power UV LEDs with
focusing condenser lenses, while the lens is still gripped and hold in position by the 6-axis stage. In a post-sequence, the
quality of the module can be tested, or the camera module gets transported to the load position.
After the dispensing, alignment and bonding sequence has time consumption in the range of 30 second including 5
seconds for UV curing.

5. MEASUREMENT RESULTS
For the image alignment and the reliability of the instrument, the measurement accuracy of the image plane angle is most
important.
To determine the measurement repeatability and the alignment accuracy of the first alignment step, statistical records
have been analyzed. All tests have been performed with a 1/3’’ camera and an EFL 5mm lens, f/2.2 and a 5-collimator
system with EFL 50mm and a 20 µm slid width.
The distribution of 100 continuous image-plane angle measurements has a standard deviation of 0.16 mrad. The small
measured angle was 1.27 mrad.

160 80
1 1

120 60
,A
4-,
80 40
o o Tilt X o o Tilt X
U U
40 o Tilt Y 20 o Tilt Y

o III] d
_I,,,,,T-rrl I I
II,
I 0 . . rI I1...... .

-1 2 -0.8 -0.4 0 0.4 0.8 1.2 -1.2 -0.8 -0.4 0 0.4 0.8 1.2
Image Plane Tilt / mrad Image Plane Tilt / mrad

(a) (b)
Figure 7. (a) distribution of alignment result (image plane tilt) after first alignment step, (b) after second alignment step.

To check the reliability, 500 alignment procedures has been saved to have a detailed look on the result after just one
correction cycle, started from a misalignment of around 0.5°. Figure 7 (a) shows the result for X and Y tilt measured
after the first correction step. It is obvious that the tilt in X has a distribution around zero with (0.01 +/- 0.14) mrad,
while Y is shifted by (0.63 +/- 0.12) mrad. This could be attributed to the fact, that the sensor has a 17% larger pixel
pitch in Y-direction. This has not been considered in the software at the time of testing.
The slight misalignment in Y is than corrected by a second alignment step, but this is also effecting the alignment in X
(see Figure 7 (b)). The alignment after the second correction step is (0.14 +/- 0.2) mrad. Therefore, with 2-sigma (stdev)
it can be expected that 95.5% of all sample are in an alignment state of 0.54 mrad (< 2 arcmin) after the second
alignment step.

Proc. of SPIE Vol. 8992 89920F-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms

 In future we expect the same accuracy for both values like the one for Tilt X in Figure 7 (a) and a tilt alignment in the
range of 0.28 mrad (< 1arcmin) with the first alignment step.

6. CONCLUSION
The importance of the accurate alignment of the optical system in respect to the sensor was discussed by describing the
MTF measurement principle and the influence of a tilted image plane and focus deviation. Especially the image plane-tilt
can only be corrected by an automated alignment in 5 or 6 degrees of freedom. The process of the correction using the
information of the through focus curves has been described. Besides the theoretical concept, the instrument with the
necessary components was presented, showing the integration of the dispensing, alignment and curing process. Besides,
the two different illuminations with test chart or collimators were introduced, showing especially the advantages of the
innovative collimator based system. All the different steps are only possible with advanced software performing the
different sequences and providing the measurement result also for a possible and more complex quality check of the
aligned camera modules. In the end, measurement results show the current accuracy of the system and give a forecast of
further improvements.

REFERENCES

[1] Boreman, Glenn D., “Modulation Transfer Function in Optical and Electro-Optical Systems”, SPIE Press
[2] Williams, T., “The Optical Transfer Function of Imaging Systems”, IOP Publishing Ltd (1999).

Proc. of SPIE Vol. 8992 89920F-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 05/20/2015 Terms of Use: http://spiedl.org/terms