MEMS Technology

 

Index - MEMS

Introduction

DiCon Fiberoptics has been developing and perfecting our core MEMS technology since 1997, and we introduced our first commercial MEMS product in 2000, a MEMS VOA (Variable Optical Attenuator). Unlike most of the other Optical MEMS companies, who focused on large port-count Optical Cross- Connects (OXCs), DiCon chose to focus initially on higher-volume individual components, such as VOAs and MEMS Switches. This allowed us to perfect our processes for producing MEMS devices in high volumes, with high quality and reliability, as well as high manufacturing yield.

In-House Wafer Processing

In contrast to other sellers of optical MEMS components who use foundries for the production of MEMS chips, DiCon does its own wafer processing at our inhouse MEMS fab and clean room, located at our Richmond, CA headquarters facility.
DiCon wafer processing facility
This allows us to maintain very tight control of our wafer fabrication and testing processes, and also enables a very tight coupling between design and manufacturing. This in turn allows DiCon to continuously improve the performance and quality of our MEMS-based products, through the combination of design and process improvements.

Why MEMS?

DiCon’s customers are demanding smaller, more efficient, and more durable optical switches and VOAs at lower cost.

DiCon’s MEMS technology provides durability and reliability:

Mechanical materials are either ductile, brittle, or both. Since most metals are ductile, when they are subjected to large amounts of stress they will deform plastically, developing micro cracks that grow when subject to cyclic loading. In contrast, DiCon’s MEMS switches and VOAs are based on single crystalline silicon, a brittle material. It does not deform, fatigue, or wear out over time, and its dimensions and mechanical properties are immune to stress until a critical fracture stress-level is reached. Data results from testing show that DiCon's MEMS switches and VOAs still work within specifications after one billion cycles.

DiCon’s MEMS technology enables reduced size, power, and cost:

Using a single silicon chip, DiCon’s 1x8 MEMS switch with integrated digital control has a size of 25 mm x 16 mm x 8.50 mm. It also costs significantly less than traditional mechanical switches. The MEMS chip itself dissipates almost no electrical power, and the use of low-power control electronics keeps overall power dissipation to a minimum.
MEMS wdm add drop optical switch and variable optical attenuator

DiCon’s Approach to MEMS

DiCon’s approach to the design and manufacture of our MEMS components and integrated solutions is both disciplined and interdisciplinary. By their very nature, optical MEMS devices require interdisciplinary expertise in mechanical, electrical, and optical engineering. DiCon has assembled a strong technical team that is able to work across these three disciplines.

Before introducing our first MEMS product, DiCon spent four years in developing the core technology, our design and analysis capabilities, and processes that are essential to producing leading-edge products with high reliability and good yields. In doing so, we have adhered to the following design guidelines and principles, that are applicable to all of our MEMS products:
  • 100% in-house wafer fabrication in our dedicated MEMS fab and clean room, for tight process control and rapid innovation
  • All single-crystalline silicon construction for stable, reliable performance (essentially no wear-out mechanism at the silicon device level)
  • Stiction-free designs with no touching parts, using silicon torsion beams that suspend the moving parts of the device structure in free space
  • Highly-accurate mass balancing around the axes of rotation, for shock and vibration immunity
  • Innovative and proprietary damping techniques, for shock and vibration immunity and fast switching times
  • Focus on a relatively small number of MEMS chip designs, which can then be produced as high-volume “building blocks”
  • Designs that are insensitive to normal levels of process variation, to improve yields at the wafer and chip level
  • Focus on continuous improvement, through the manufacturing of the thousands of wafers and millions of MEMS mirrors that we’ve produced to date

Single-Axis Tilt-Mirror

DiCon’s highest-volume MEMS products, including our MEMS VOA, MEMS On/Off Switch, MEMS 1x2, and MEMS Add-Drop 2x2, are based on our Single-Axis Tilt Mirror design, as shown in the Scanning Electron Microscope photo below:

MEMS mirror, single axis

The mirror (the octagonal area in the middle of the chip) is tilted by applying a voltage to the electrostatic comb-finger actuators that extend to either side of the mirror. The moving parts of the structure, including the mirror, and half of the comb-fingers in the actuators, are connected to the fixed parts of the structure via silicon torsion beams. Thus, the moving parts of the structure are effectively “suspended in space”, and never come into physical contact with the fixed parts of the structure.

The mirror tilts over a continuous range of motion, with a highly repeatable “tilt angle versus applied voltage” characteristic. For most applications, sufficient tilt angle can be achieved with just 5 volts. Higher voltages can be used for applications requiring higher tilt angles. There is essentially no steady-state electrical power dissipation at the MEMS chip level, as the actuators have the electrical characteristics of small capacitors. The tilt mirror is essentially an analog device. Additional control electronics are typically used to enable digital control of the mirror, as described below.

Dual-Axis (“3D”) Tilt Mirror

DiCon’s MEMS 1xN Switches make use of a more complex dual-axis tilt mirror. This type of MEMS mirror is sometimes referred to as a “3D” MEMS mirror, since light that hits the mirror can be redirected in multiple directions, in three-dimensional space. DiCon’s Dual-Axis Tilt Mirror uses a mirror and gimbal design, as shown below:

MEMS mirror, dual axis
In this design, the mirror is tilted within the gimbal by one set of comb-finger actuators, and then the gimbal itself is tilted by the second set of actuators. A pair of control voltages is applied to independently tilt the mirror and gimbal. Voltages up to 16 volts are used, in order to achieve the somewhat higher tilt angles needed for our higher-fanout MEMS 1xN switches. DiCon is currently producing single-mirror MEMS 1xN switch components with fan-outs as high as 1x24, for use in our MEMS Switch Matrices.

Component Packaging

The packaging of DiCon’s MEMS products at the component level is shown in simplified form below. Using a GRIN lens as a common collimating element, light coming in on the input fiber is aimed at the MEMS tilt-mirror. Depending on the position of the mirror, the light is either aimed directly at the core of one of the output fibers (for switches), or is partially mis-aligned with the core of a single output fiber to achieve variable attenuation (for VOAs).

As stated above, our basic MEMS components in cylindrical packages are essentially analog devices, in which the tilt mirrors move over a continuous range of motion, with either one or two axes of rotation. Thus, for every desired state of the VOA or switch, a specific, precisely controlled input voltage (or a pair of voltages, in the case of 1xN switches) must be applied.

MEMS optical switch cross section diagram

The details of our cylindrical MEMS device packaging are shown in the exploded view below:

MEMS component in cylindrical packaging, expanded view

Control Electronics

DiCon’s MEMS components are analog devices, controlled by applying a precise voltage (or voltages) to the control pin(s) of the device. External control electronics are used to generate the precise voltage(s) needed, and can also be used to provide a digital control interface.

In a typical implantation, a precision DAC and a transconductance amplifier are used to generate each control voltage that is required (one voltage for each VOA, On/Off Switch, or 1x2 Switch, and two voltages for each 1xN Switch). A small non-volatile memory is used to store the digital DAC codes that correspond to switch states, or, in the case of a VOA, to store a digital representation of a VOA’s voltage versus attenuation characteristic. A small MCU is typically used to implement the user interface, using either a common serial bus interface (e.g. I2C), or in some cases a simple TTL-level parallel interface.

For individual MEMS components, the control interface is built on a flex-PWB that wraps around the MEMS device’s cylindrical package. The combined flexcircuit and MEMS cylindrical package are then packaged into a laser-DIP package, as shown below:

inside of MEMS optical switch
Because the tilt-angle versus voltage characteristic of DiCon’s MEMS devices is so repeatable, our MEMS components are typically operated “open-loop”, without the use of closed-loop feedback control. A calibration table is created that defines the tilt-angle versus voltage characteristic, and then simple table-lookup algorithms are used to call up switch states, or VOA attenuation levels. Although not necessary for component operation, closed-loop feedback control can be used to enable system functions. For example, if a MEMS VOA component is being used to maintain a constant absolute optical power level (as opposed to a constant level of attenuation), then a closed-loop feedback system can be implemented, driven off of a tap-detector that is used to monitor the resulting optical power level at the VOA’s output.

From Components to Integrated Solutions

In module or chassis-level applications where multiple MEMS components are used, it is more common to implement the control electronics for the entire module or chassis on a single, larger PWB, supporting all of the MEMS components, which are left in their cylindrical packages. This allows the sharing of the MCU and memory over multiple MEMS components, as well as the use of quad and octal DACs and amplifiers for greater packaging density.

MEMS MxN Matrix Switches

DiCon’s high-fanout MEMS 1xN Switches can be easily integrated into module or chassis-level MEMS MxN Matrix Switches. With a maximum fanout of 1x24, MEMS Matrix Switches in sizes up to 24x24 are achieved by cross-coupling input and output 1xN switches, without additional cascading. The figures below show 8x8, 16x16, and 24x24 MEMS Matrix Switches.

MEMS matrix optical switches DiCon’s modular approach to MxN MEMS Matrix Switches provides a number of significant advantages:
  • Configuration flexibility – using a small set of building-block components, we can offer MEMS Matrix Switches of any MxN dimension
  • Architecture flexibility – we can support such features as broadcast/multicast, and “FlexPort” (any port to any port) designs.

MEMS matrix optical switch diagram

  • Packaging flexibility – everything from individual components to rackmounted chassis designs.
MEMS product line, from components to rack-mount chassis systems
  • High performance – including low insertion loss in the range of 1.5 dB to 1.7 dB for MEMS MxN Matrix Switches up to 24x24, with high levels of cross-talk isolation
  • Use of high-volume building block for reliability and maintainability

Future Directions

DiCon is continuing to develop our MEMS technology, in three major directions:

1. Continuous improvements in performance and reliability DiCon’s in-house MEMS fab and clean room allows us to investigate design and process changes that serve to continually raise the bar on product performance and reliability, as well as improved product yields and cost. One focus of this ongoing work is the development of “hardened” products for extreme operating conditions, including defense and aerospace applications.

2. Higher-fanout 1xN MEMS Switches DiCon’s MEMS 1xN Switch design scales well to even higher fanouts, using the same basic Dual-Axis “3D” MEMS chip design, thereby enabling an evolutionary path toward development of single-mirror components up to 1x48 and beyond. This in turn allows steady increases to the maximum size of a two-stage, minimum insertion loss MEMS MxN Matrix Switch, to 48x48.

3. Development of Integrated MEMS Arrays for larger MEMS Matrix Switches Although DiCon’s modular approach to MEMS Matrix Switches offers significant advantages for smaller and medium -size matrices, above some size limit the advantage swings to the so-called free-space MEMS matrix switch, in which arrays of MEMS mirrors are aimed at each other. For larger MEMS matrix switch applications, DiCon is developing Integrated MEMS Arrays.