.jpg)
by Professor Harry Stephanou
Abstract
Assembly, packaging, and testing activities account for 85% of the cost of
many microsystems. This is primarily due to the lack of backend standards or
general methodology. This presentation focuses on concurrent microengineering
and the need for designing for micromanufacturability. Specifically, how to
design a microsystem so that it can be assembled, packaged, and tested with
high yield, low cost, and short cycle time with low volume automation. The goal
is for the product to be designed concurrently with the fabrication process
and the assembly work cell. Special attention is devoted to tolerance analysis,
error propagation, and their impact on product performance. We present a rigorous
mathematical framework and its implementation into a software tool that allows
the designer to rapidly evaluate tradeoffs among cost, cycle time, and yield.
We will illustrate how The Texas Microfactory™ is using this unique proprietary
tool for pilot production of complex microsystems.
Introduction
Recently, increasing market inclination towards low-cost portable systems
with complex functionality has opened up great avenues for miniaturization technology.
Unlike the past surface-micromachined and monolithically fabricated MEMS products,
newer microsystems have significantly grown in design complexity as well as
material heterogeneity. In order to address these issues, alternative production
technologies are being investigated rigorously. Assembly in micro-domain is
a pioneering concept that sets new paradigms for manufacturing of robust, low-cost,
and mass-producible microsystems. This is achieved through simple designing
of heterogeneous micro-components followed by precise manipulation and assembly
of these components to construct the envisaged complex system.
However, assembly based manufacturing of microsystems poses several unique
challenges due to certain intrinsic reasons such as unavailability of standards
for component design and fabrication, stringent tolerance budgets, workspace
constraints, and surface effects due to scaling. Consequently, selectivity of
manipulation systems, sensors, control schemes, and automation can result in
a significantly large number of iterations in establishing the production cycle
for an acceptable yield, which renders the manufacturing process hugely expensive
and time consuming.
At ARRI's Texas Microfactory™, we exercise a holistic approach
towards microscale manufacturing where, using a complex mathematical framework,
different aspects in pilot production are evaluated concurrently to estimate
the cost functions such as yield, throughput, cost, and performance. Using an
indigenously developed iterative software tool, an acceptable combination for
these cost functions is searched for and the corresponding input parameters
are selected for the manufacturing process.
Results
.jpg)
|
Figure 1.
ARRI's microspectrometer consisting of Silicon MEMS parts, glass
ball lenses and beamsplitter cube, IC laser source and detector integrated
on 1cmx1cm silicon die with device resolution of 5nm for visible wavelength
and 25nm for near infrared wavelength. |
Table 1: Cost function comparison for microspectrometer
manufacturing
| Parameters |
Projections for a pure
open loop control based assembly |
Projections for a pure
closed loop control based assembly |
Projections for a custom
designed hybrid control scheme |
| Overall Yield |
20% |
99.9% |
92.5% |
| Cycle time |
6 - 10 minutes |
50 - 80 minutes |
20 - 35 minutes |
| Sensors requirement |
0 |
4 |
2 |
.jpg)
.jpg)
|
Figure 2.
Snapshots of the software tools for design for micromanufacturability:
(top) manufacturing cost function estimator, (bottom) process simulator
in virtual 3D. |
Copyright AZoNano.com, MANCEF.org