Minimizing Error and Maximizing Precision in Microassembly
To optimize positioning systems for photonic component manufacturing and assembly, it helps to understand the errors that can whittle away at precision part placement.
Precision assembly solutions are as unique as the products they are built to automate, especially in
the photonics world. Progressive assembly operations such as guiding, gluing and
bonding sometimes take place in a local region around a static package in three-dimensional
space. Integrating the different mechanical systems so that all of the active components
are available within that small space is a challenge.
To determine system capability for photonic applications,
it helps to understand the types of errors associated with the mechanical system
and the techniques that can help minimize them. This should be second nature to
all parties involved, including the mechanical and control engineers on the integration
side and the designers of the product.
First, consider linear errors. Resolution is
the minimum distance measured by a closed-loop feedback system. This is the resolution
of the system, and it does not imply the ability to actually move that distance.
Also important is minimum bidirectional step size, or the smallest repeatable step
size a system can achieve. The bidirectional nature is important because mechanical
errors and stiction are taken into consideration.
Next, examine position repeatability,
which is the ability of a system to return repeatedly to a given point in space.
Also evaluate the system’s linear accuracy. Derived from repeatability and
resolution, this is the difference between the actual position and the desired position
along a line, compared with an absolute reference. Companies usually specify stage
accuracy, for example, as a percentage of full stroke.
In this spatial axis-specific error
category, flatness and straightness specify the path deviation from the best-fit
straight line along the axis of motion of a stage. Flatness is the vertical runout,
and straightness is the horizontal runout. However, factory specifications assume
that the stage mounts to a lapped surface, which is often not feasible for every
stage in a system.
Figure 1. When designing positioning systems for photonic component
manufacturing and assembly, it helps to understand potential error
for instance, is the angular deviation of the axis over one revolution.
is the translation of the center over one revolution.
For rotary stages, eccentricity, the
deviation of the center over one revolution, and wobble, the angular deviation of
the axis over one revolution, are key (Figure 1).
The process of stacking components basically combines
axes to form a positioning system, with each stage spatially related to and dependent
on the previous stage. Also of note are Abbe errors, linear errors caused by underlying
angular errors. Small angular errors along the stage axis produce translation errors
at the work surface, and the farther the work area is from the stage, the larger
the linear error (Figure 2).
Figure 2. Small angular errors at the stage surface produce translation errors at the work surface. The Abbe error (dx) equals the angle (u) times the offset (h).
Orthogonality error should be added
to the designer’s checklist. This is the degree of perpendicularity error
between two axes in a plane. Over its travel, a pair of axes for a high-quality,
factory-assembled system with a range of motion of 200 x 200 mm can translate to
10 μm in X and Y because of such errors.
Now examine the position, path and
placement accuracy that the system will require:
• Position accuracy is the difference
between the actual location and the desired location, compared with an absolute
reference (Figure 3).
• Path accuracy pertains to a
mechanism in motion and its deviation in position from a planned path. Precision
dispensing of an epoxy ring is an example of where this would be required.
• Placement accuracy usually
means the result of placing one object in some orientation relative to another with
a degree of certainty. While loosely based on the mechanical precision of a system,
this specification is really the outcome of the hardware, the software and calibration
working together to perform an operation.
Figure 3. These figures represent two-dimensional position repeatability and accuracy of
a tool tip relative to a target point; for example, (0,0) at the intersection of
the X- and Y-axes.
All of these errors are compounded
by the thermal expansion of each axis, by tooling plates and by machine bases due
to ambient temperature and duty-cycle variations. Systems also can experience substantial
errors when stages bearing a load are cantilevered. When stages are stacked, the
total error will differ at each location within a work envelope.
Also review the coordinate system that
is used to identify positioning in 3-D space. The native coordinate system for most
platforms is the Cartesian (Figure 4). In some assembly applications, however, maneuvering
around the center of a gripped object is necessary. This center point is sometimes
referred to as the virtual pivot point. By defining a tool coordinate system and
using a motion control code to facilitate it, the object can be rotated without
translating in X, Y or Z, as well as translated in X, Y or Z without rotating.
Figure 4. The Cartesian coordinate system provides a framework to define a location in six degrees of space relative to a datum (0,0,0,0,0,0).
This technique will require spatially
relating the tool tip, which is attached to the tooling surface (on the Rz axis),
to the mechanism’s Cartesian coordinate system (Figure 5). Using real-world
units in one common reference frame, the six degrees of space position and orientation
of each link must be described. And describing how each axis couples with and relates
to the prior axis in the stack is necessary to accurately define the tool-tip location.
Each link is represented as a vector, and multiple vectors are related to define
Figure 5. The tool coordinate system helps define a location in six degrees of space relative to a tool tip (0,0,0,0,0,0).
In the product and platform design phases, a variety
of factors to reduce placement error of the positioning stages and complexity of
the control system can be considered. For example, the number of axes of motion
(degrees of freedom) must be minimized for an assembly process. Top-down assembly
is preferred. Having symmetrical parts reduces pick strategy issues. Also consider
that an angularly oriented part immediately increases complexity by adding a rotational
axis of motion, which then adds complexity to the mathematical description of tool-tip
Other ways to reduce positioning-system complexity include:
• Minimization of planarity and perpendicularity requirements.
• Use of constant mechanical
features such as geometry and height to enable repeatable picking by an end effector.
• Emphasis on repeatable, high-contrast
locating features when using vision guidance.
• Exploration of alternative
picking choices. The ability to vacuum pick a part, for example, reduces the need
for multiple end-effector types and minimizes rotational moves.
Next, examine part presentation techniques.
To reduce the magnitude of pick flexibility required, repeatable presentation in
X, Y, Z and Θz (related to tool-tip location) is important. If vision-guided picking
or placing must be used, repeatable part presentation within the camera’s
field of view reduces the need to move the camera to find the object. And, to minimize
Z accuracy requirements, perform all operations at the same Z height, if possible.
Also evaluate end-effector flexibility;
e.g., if a part is sturdy enough to take a load, it may be possible to allow the
end effector to preload into the part. Force sensing can be used for precision Z
picking and placing. If a part has a repeatable geometry, an end effector might
be designed to mechanically center the part upon picking. Devising a manipulator
for multiple tools by rotating each into an active position or by using a tool-exchanging
scheme also might be an option. For a precision application, locating the tip relative
to the tooling surface after each change may be necessary.
Eye on accuracy
For pick-and-place operations, placement tolerance
will be no better than the pick repeatability unless the position of the gripped
part is refined after gripping. Whenever possible, engineers should design the microassembly
system to rely on resolution and repeatability. If accuracy is truly needed, the
emphasis should be placed on local rather than on global accuracy, which is more
difficult to attain.
If machine vision is employed, consider
the camera magnification needed to achieve the placement tolerance; e.g., how many
magnifications are needed? Other design issues include positioning the cameras and
calibrating the frames to the motion reference frames.
One way to significantly reduce complexity
is to rely on software-driven calibration and mapping techniques, which, on all
but the nonlinear errors, might improve X and Y accuracy by 10 to 20 times. The
integrity of the calibration will depend on accurate, repeatable mechanics and on
quality vision images with good contrast. Mapped relationships remain constant only
when ambient and operating temperatures do not change. A steel surface 200 mm wide,
for example, will grow 2 μm for every degree Fahrenheit the temperature increases.
The goal should be to keep the design
simple. Accuracy needed for mathematically derived locations is not a concern if
one can teach the system to move to a location by first moving the equipment manually.
It also may be possible to avoid the need for absolute placement within a large
area of the work envelope as well as to minimize the required accurate work area.
Visual active alignment is an option
when the placement of a device relative to another in the same field of view can
be monitored. The place ment accuracy then depends on mechanical resolution and
on finding the objects relative to one another within the camera’s field of
view, rather than on total system accuracy. The need for total accuracy also diminishes
to local if one device can be placed relative to another.
In an emitter/detector situation, where
the emitter is activated while the detector signal is monitored, signal-based active
alignment will work. The signal strength provides relative position feedback. Positioning
tolerance then relies on resolution, not on repeatability or accuracy.
One simple rule is to work with as
few degrees of freedom as possible in an assembly process because calibration techniques
for Z, Θx, Θy and Θz are difficult to implement. Minimizing the differences between
work heights and eliminating angular movements also will reduce accuracy requirements.
If a downward-looking camera is designed
to move with the active tool tip, it helps to reduce the distance between the tool
tip and the camera centerlines. Designers also should examine whether an upward-looking
camera can be used to locate the part relative to the tool tip after picking, a
step that negates pick inaccuracies.
When visually locating a gripped part
with an upward-looking camera, move the part into the center of the field of view
before capturing the final image. This reduces reliance on the accuracy of the camera
calibration. Calibrating at a point rather than at the whole camera field reduces
the need to relate location to a point rather than to the area and orientation of
the camera frame.
Development of a successful positioning strategy
for photonic component and other microassembly processes should involve a detailed
collaboration among product designers and mechanical and control engineers. Each
must fully understand all aspects of the system, where positioning errors come from
and how to minimize them. The result is a more simple, cost-effective precision
Meet the author
Bruce Fiala is a senior software engineer in the Robotics and Vision Group at RTS-Wright LLC in Nashville, Tenn.
MORE FROM PHOTONICS MEDIA