|
Tiny Switches
With Big Features
by Tom Lecklider, Senior Technical Editor
Development of MEMS relays has been underway since at least the early ’90s with
many types reported during the intervening years. Anticipated benefits include
extremely small size, long life in excess of 100 billion cycles for military
applications, high operating speed, and excellent RF characteristics.
Unfortunately, the step from a working laboratory device to a reliable
volume-manufactured product has proven to be large.
In addition to technical challenges,
bringing any new technology into an existing market can be difficult. As John
McKillop, chief technical officer at TeraVicta Technologies, explained, "The
market generally requires an entry-level product that meets or exceeds the
reliability, cost, and performance of the incumbent product. This creates a
particularly difficult barrier for the entry of MEMS into the market because the
relays most vulnerable to replacement are, on average, fairly reliable and low
cost and have low actuating voltages."1
Today, only a few companies have proven MEMS switches or
relays. Among them are TeraVicta, Radant MEMS, MEMtronics, and XCOM Wireless.
These companies have developed MEMS products, but other devices may exist that
have not been publicized. In addition, some MEMS relays have been introduced and
subsequently withdrawn, such as the MagFusion products.
A few microminiaturized relays also have been developed.
Advantest produces electrothermally activated relays for the company’s own use
in ATE. The advantages of the device include a very small 4-mm x 5-mm footprint,
25-dB minimum isolation up to 20 GHz, and an expected life of greater than 20
million cycles. In addition, the relay actuation requires 50 mA at 5 V rather
than the much higher voltages associated with electrostatic switches. As
Advantest has implemented electrothermal actuation, it provides sufficient speed
for the intended ATE environment.
The Panasonic ME-X Relay also is a
microminiature device, in this case electromagnetically actuated. A very small
wound coil tips an armature pivoted in its center so that contact pairs at both
ends of the relay are simultaneously actuated, the NO pair closing and the NC
pair opening. Compatibility with common logic circuitry voltages is maintained
with a 3-V rating for the 90-Ω
coil. This relay includes a latching mechanism so that
current is not drawn after the state has changed. If you want to buy a MEMS relay, the
field narrows considerably. Both XCOM Wireless and Radant are in the early
stages of commercial sales. Much of Radant’s funding has been linked to defense
advanced research projects agency (DARPA) and military projects, so the first
uses of the company’s relays are in military radar applications. XCOM Wireless
has sold early devices to select customers and now is transferring backend
assembly from an in-house line to a volume manufacturer.
MEMtronics is concentrating on lower power microwave and
millimeter-wave devices for use up to 110 GHz. A single MEMS relay may never be
sold by MEMtronics; instead, the devices are being integrated into monolithic
microwave integrated circuits (MMICs).
At present, TeraVicta and Radant MEMS are the only brands of
MEMS relays readily available.
Nevertheless, MEMS relays and switches are a reality today
thanks to several critical factors. One is the emergence of MEMS contract
manufacturers, such as Innovative Micro Technology (IMT), which makes parts for
both XCOM Wireless and MEMtronics. MEMS manufacturing is related to other types
of semiconductor production, but it’s also very distinct from them.
John Foster, chairman and CEO at IMT, explained that his
company has many standard capabilities and processes that are offered to MEMS
designers; however, flexibility is key. Comparing only the XCOM Wireless and
MEMtronics parts, he postulated that if his company could not offer a wide range
of materials and processes, it might be capable of building one or the other
device, but not both because they differ that much from each other. And, relays
are only a segment of IMT’s MEMS business, which includes microfluidics,
magnetic actuators, IR, and optical telecommunications devices.
A good way to understand the current
group of MEMS relays is to consider why so many previous developments failed. At
the top of the list is actuation force. One view of the interaction among the
many characteristics of a MEMS relay is shown in
Figure 1. This matrix, presented by TeraVicta,
supports the company’s emphasis on a high-force actuation method.
|
Figure 1. Relationships Among Relay
Characteristics
Courtesy of TeraVicta
|
Of course, for an electrostatic actuator, the operating
voltage increases to generate a higher force. The type of metal used for the
contacts affects eventual degradation and on-resistance and, to a degree,
stiction. High force helps to reduce contact resistance but may accelerate
contact wear. High-force actuation as well as restoration lead to faster
operation. A high actuating force allows the restoring force also to be greater,
which largely overcomes stiction, the tendency for the contacts to stick
together.
Providing Sufficient Push and Pull
MEMS relays mimic the
mechanical operation of their larger relatives. When actuated, one or more
contacts are either brought close to or forced against a stationary contact. The
materials used in the relays, if properly designed, make almost perfect springs.
For example, most of the successful MEMS relays have demonstrated an expected
life in excess of 100 million cycles. Typically, end of life is reached because
contact resistance has risen above a nominal value, not because the relay’s leaf
spring mechanism has broken.
Because MEMS devices are so small, forces that can be ignored
in larger mechanisms can no longer be. Stiction is caused by electrostatic and
interfacial forces, such as capillary and Van der Waals forces, and a major
cause of unreliability. Also, organic deposits and contaminants such as water
from humid air can hold the contacts together.
Stiction cannot be completely eliminated, so most of today’s
MEMS relays have high-voltage electrostatic actuation so a high restoring spring
force also can be used. It’s more important to generate sufficient force than to
interface conveniently with available 5-V or 12-V power supplies.
Another common factor is hermetic sealing accomplished as soon
as possible after the micromachining process has been completed. Most devices
use a form of wafer-level or 0-level packaging in which the MEMS mechanism is
hermetically capped directly on the wafer. This is accomplished in several ways,
but the important thing is to ensure that the just-completed device is not
contaminated.
How large are the forces that are involved? Manufacturers
generally don’t discuss design details in depth, but if you assume an area of
0.04 mm2 (0.2 x 0.2) for each of the actuator plates, 100-V actuating
voltage, and a 5-µm separation, a reasonable force estimate can be made.
Electrostatic forces are at work on the plates of capacitors
in the electronic devices we use each day. Of course, capacitor plates generally
don’t move, but they still experience forces because of the electrical field
between them. For a simple parallel plate capacitor with air or vacuum between
the plates
where: A = the area of either plate in m2
d = the distance in meter between the plates
V = the voltage difference between the plates
To put this into perspective, tissue paper has a mass of about
18 grams/m2 or a weight of 7 x 10-5 N per 400 mm2
area.
This force is balanced by the spring force of the moving
contact support. If the applied voltage is large enough for the Coulomb force to
overcome the spring force, the contacts begin moving toward each other. However,
as they do, the plate separation reduces, causing the force to increase. This
process progresses quickly, causing the movable contact to snap down to the
fixed contact.
The problem now becomes one of making sure the moving contact
support spring is strong enough to overcome stiction when the actuating voltage
is removed. Early relay designs erred on the side of lower actuating voltages to
improve compatibility with other circuitry. However, for some of these designs,
the restoring spring force was only marginally sufficient to overcome stiction,
and reliability in real-life applications was poor.
Product Examples
TeraVicta
The calculated 7 x 10-5
N force is only a rough estimate. Adding a dielectric in the gap between the
plates will increase the force generated. Also, it should be made clear that the
capacitance of the actuation mechanism is very different from the capacitance of
the switched contacts. This can be seen clearly in Figure 2, a photomicrograph
of the patented high force disk actuator used in many TeraVicta switches.
|
Figure 2. Photo of a High-Force Disk Actuator Relay
Courtesy of TeraVicta
|
The large disk measures a bit over 300 µm in diameter so for
the assumed 5-µm gap, it would provide about twice the calculated force with a
100-V drive signal. TeraVicta specifies a maximum 70-V drive, canceling the gain
in force due to area and leaving the original 7 x 10-5 N estimate.
The capacitance of a parallel plate capacitor is given by
or C
= 0.13 pF for the disk actuator with a 5-µm separation. Because capacitance
directly relates to area, the capacitance of the contacts is very much smaller.
It’s difficult to scale the contact area from the figure, but assuming a 50-µm dia for the slightly domed contacts, the capacitance is a factor of 36 smaller
than that of the actuator or about 3.5 fF.
Making this estimate gives an idea of the off-state
capacitance of the switch. TeraVicta quotes the isolation of an open switch at
31 dB up to 1 GHz, which implies a capacitance of about 2.5 fF. On the other
hand, a maximum 43-dB isolation figure also is stated, corresponding to about
0.17 fF. So it’s reasonable to assume an off-state capacitance of only a few fF
for this switch whether based on its physical size or electrical specification.
Insertion loss is the other major
specification quoted for switches and relays. It’s a measure of the amount of
power lost from input to output and, for the TeraVicta switches, is 0.08 dB
maximum up to 1 GHz. The company reports a total contact and interconnection
on-state resistance of about 0.32 Ω
However, some loss also is due to shunt capacitance and
series inductance, and these effects contribute to an increase in insertion loss
at higher frequencies.
XCOM Wireless
Currently, XCOM is ramping up the availability of the type
XW-3100 SPDT relay. Select customers are using it now, but the company’s goal is
for the relay to become a high-volume commercial product. Applications that can
benefit most from using such a small device include ATE testhead switching.
Dave DeRoo, vice president of product
development at XCOM, said this relay is aimed at higher frequency switching up
to at least 20 GHz. In common with other MEMS developers, he stressed the
importance of packaging. Not only are the devices very small, they have critical
electrical parameters as well. Providing a robust, protective package is of no
use if it compromises the insertion loss or linearity, for example. Typical
insertion loss and isolation curves are shown in
Figure 3.
|
Figure 3. Typical Insertion Loss and Isolation
for XW-3100 Relay
Courtesy of XCOM Wireless
|
Although the forces involved in MEMS relays are small, so too
are the contact areas. Mr. DeRoo commented that in the XW-3100, gold contacts
were used because the contact pressure was much smaller than
in conventional relays. Viewed from the perspective of Figure 1, lower pressure
helps prevent contact degradation, but with silver and silver oxide contacts
that are used in much larger relays, it would result in higher on-resistance.
Good, low-resistance contacts are
necessary for both the TeraVicta and XCOM relays given their continuous current
ratings. These relays may be small, but they are tough. TeraVicta specifies 15-W
continuous RF power, which for a 50- Ω system
translates to 550 mA. XCOM quotes continuous RF current as 400 mA.
Radant MEMS
Radant has been developing a switch for five or six years,
primarily with DARPA and military funding and with phased array radar in mind.
The switches have been used by select customers for approximately two years.
Recently, a large number of Radant switches were
part of a successful phased array radar demonstration.
In this application, individual elements of the antenna are
steered electrically by altering their phase. Although such antennas were
feasible in the past, MEMS relays can be integrated with the phase-shifting MMIC
to improve performance and reduce size and weight. Also, the additional biasing
and control circuitry required for other types of
switching are reduced or eliminated.
This switch is considerably smaller than
the TeraVicta structure, being only 100-µm long overall. It has the form of a
classical cantilevered beam and requires from 40 V to 120 V for actuation. FET
terminal names are used, the drain and source being input/output and the gate
the control element
(Figure 4).
|
Figure 4. Photo of Cantilever Beam Ohmic Switch
Courtesy of Radant MEMS
|
According to details in a paper by J.
Maciel et al of Radant, A Packaged, High-Lifetime
Ohmic MEMS RF Switch, "The contact material is a
thin layer of a platinum group alloy deposited on the underside of the beam and
on the drain….Devices are usually operated at a contact force of 200 µN and have
a single contact resistance of 3
Ω. Switches
typically have four to eight contacts in parallel to yield a total on-resistance
of less than 1 Ω at DC and low frequencies."
 The
paper also describes the wafer-level capping process used to provide a hermetic
seal. A silicon wafer with etched cavities is bonded to the switch wafer to
create the package. Typical lifetime is at least 1011 cycles, and switches have
been tested to beyond 9 x 1011 by the Department of Defense. Hot
switching significantly reduces the lifetime of this and other Ohmic MEMS
switches and relays to 107 cycles or less.
MEMtronics
In contrast to other devices that have Ohmic contacts, the MEMtronics switch is
capacitive. Although the contacts may touch, only a capacitive RF shunt path to
ground exists when the switch is activated. Charles Goldsmith, the company’s
president, said that switch characterization was progressing well with an
emphasis on pull-in and release voltage as well as capacitance ratio
measurement. These parameters are key to achieving high performance and
reliability.
The device uses electrostatic actuation,
but the restoring force is provided by a bridge membrane rather than a
cantilever arm. As shown in Figure 5,
looking directly down on the switch, the membrane with curved sides is anchored
at either end and positioned above two parallel striplines with the actuating
contact between them. A dielectric layer beneath the metallic bridge and above
the striplines and actuating contact increases the effective capacitance.
|
Figure 5. Photo of a Capacitive Shunt Switch
Courtesy of MEMtronics
|
When the switch is off with the bridge in
the undeflected up position, the capacitance to either stripline is 30 fF to 40
fF. When the bridge is pulled down to the actuating contact, the shunt
capacitance to ground increases to 1 pF for both striplines. Mr. Goldsmith
explained that the DC control voltages and RF signals are simultaneously present
on the switch conductors. Because the two types of signals are so far apart in
frequency, there’s no problem multiplexing or demultiplexing them onto the
switches. Generally, resistive biasing is used rather than inductive biasing.
Because so little current flows, using 20-kΩ
resistors provides lower loss and higher bandwidth.
In a paper written with David Forehand,
also at MEMtronics, Mr. Goldsmith described how some of the design trade-offs
were evaluated. The company applied transient current spectroscopy to better
understand charge trapping in the switch dielectric. A very long time constant
was observed and the corresponding transient currents measured. From these
parameters, an accurate model of the phenomena was developed. With it, various
insulating materials could be evaluated quickly without the necessity of
building complete MEMS switch devices.
From the detailed test results, some guidelines emerged.
First, keeping the actuating voltage as low as possible, at least less than 40
V, minimizes dielectric charging, a major component of stiction leading to
unreliable operation. Second, reducing the amount of dielectric trades a lower
capacitance ratio for longer life by reducing charge trapping.
 For electrostatically actuated
MEMS switches, once the movable contact has snapped down, very little voltage is
required to keep it down. So, MEMtronics has developed a switch drive waveform
that initially rises to 40 V but quickly reduces to a much lower holding level.
Keeping the control voltage low for as much of the time as possible further
minimizes dielectric charging.2
A second paper described a mcro-encapsulation process used to
package the switch. The applications for this shunt capacitance switch range up
to 110 GHz, so packaging becomes even more critical at these very high
frequencies.
A cage structure made from Si3N4
is formed above the actual switch structure. The cage has many very small holes
in it through which the various processing steps to remove sacrificial layers
must be performed. However, the holes are not large enough for a suitable liquid
encapsulant to flow through them. In this manner, the switch is manufactured
together with its low-cost packaging directly on the wafer. The packaging has
been shown to contribute not more than 0.06-dB insertion loss at 35 GHz.3
Summary
Types of actuators and
packaging and contact materials aside, how many brands of MEMS relays can you
buy today? If you answered one, you’re right. If you’re a defense contractor and
answered two, you’re right as well. Certainly, one or two is better than zero,
and perhaps in a few months or years, the number will be three or even five, but
today it’s one or two.
That’s not to say that there aren’t at least four types of
relays that work reliably and probably more that we don’t know about. So, expect
to see these devices turning up in products for which MEMS relays and switches
can make a big performance difference. The commercial application most
manufacturers feature is high-frequency ATE. For the military, it’s phased array
radar.
Making reliable MEMS switches and relays simply is hard work.
Many of the people directly involved at the companies included in this article
have been researching this topic for 15 or 20 years. And, the relays starting to
appear on the market are all very different.
For example, when discussing the MEMtronics capacitive switch,
Mr. Goldsmith commented that he needed to stay far away from silicon at the very
high frequencies where the switch is targeted because of material losses.
Similarly, although all the manufacturers acknowledge the need for hermetic
sealing at the wafer level, they each do it differently.
Do you see applications for MEMS relays and switches in the
work you do? Why not contact the companies active in this field to discuss your
requirements? And, if you do, tell us about it. This is an innovative new
technology that will develop more quickly given interest from potential users.
References
1.
McKillop, J., Challenges in the Commercialization of
MEMS Switches, Tera-Victa Technologies.
2. Goldsmith, C., et al, Tailoring Capacitive
Switch Technology for
Reliable Operation, MEMtronics.
3. Goldsmith, C., et al, Zero-Level
Packaging for RF MEMS Switches, MEMtronics.
|
