Advertisement

Article

Shrinking RF Circuits Using MEMS”Toward the Single-Chip RF Solution

With the continuing miniaturization of electronic systems and
advancement of MEMS capabilities, promising new design options in
RF and microwave applications continue to emerge. Integration of
on-chip MEMS devices frequently offers a number of benefits
including lower cost, higher performance, reduced size and weight,
and increased reliability.

In the RF/microwave technology domain, much attention continues
to be given to the development and integration of MEMS-based
components. As shown in the block diagram for a representative RF
telecom system (Figure 1), designers are already using MEMS device
technology for selective (passive/active) components including
inductors, capacitors, switches, and filters. These devices, when
integrated with the RF chip, offer higher value for such
applications as VCO, PLL, and other RF functionality required for
advanced telecom systems.


Figure 1: RF system basic block diagram

Designers are just beginning to see the benefits of such
integrated devices, including enhanced performance (both passive
and active devices) and/or a reduction in cost, size, or weight of
the system. Integration of RF-MEMS devices eventually offers a
means to replace all passive RF chips with on-chip devices for the
long-awaited, single-chip RF product.

The industry is seeing the emergence of a CMOS-compatible
process with very attractive performance in the 2-GHz range. Our
design teams have reduced the ohmic losses in inductors using
large, thick copper metallization layers. MEMCAP has fabricated a
3-nH spiral inductor with a Q factor of 76 at 1.9-GHz. We have also
obtained high quality factors ranging from 30 to 80 at 2-GHz. A
full-wave electromagnetic modeling approach, based on Finite
Element Method (FEM) analysis, shows very good accuracy between
experimental data and theoretical predictions (for more details on
the analysis, please refer to Reference 7).

RF MEMS Component Design Considerations

As shown in Figure 1, there are a number of
targeted integration opportunities for RF MEMS components.
Following are some general guidelines for each component.

Switches

You fabricate MEMS-based switches with low-loss metallic
structures in order to achieve lower insertion losses and higher
linearity. The actuation can be electrostatic or electromagnetic.
The advantage of an electrostatic actuation is that there is no
current consumption. However, the drawback with this approach is
the higher actuation voltage (>10V) that is required. The
advantages of electromagnetic actuation include a lower actuation
voltage. In this case, however, current consumption may be
significantly higher. Electro-magnetic actuation also adds slightly
more processing complexity, due to the use of magnetic thin
films.

Electrostatic switches offer the most promise as configuration
switches (antenna switches or frequency band-selection switches,
referred to as S1 in Figure 1), where the key factor is low power
consumption. Electromagnetic switches are better candidates for the
T/R switch where high switching speed is a key requirement
(referred to as S2).

Inductors

MEMS-based inductor fabrication requires a thick
copper-on-insulator process in order to achieve higher Q-factor and
resonant frequencies. For some specific applications, you can use a
3-D coil geometry with magnetic materials to improve further
inductor performance. There are additional details on the
integration of inductors later in this article.

Variable Capacitors

Similar to the inductor, MEMS-based variable capacitors use a
thick copper-on-insulator process to minimize ohmic losses and to
insulate the capacitors from lossy substrates. The component
achieves tunability by using movable metallic structures with an
actuation process that is very similar to the process used for
switches. The objective remains to improve the quality factor of
the devices and the circuit's tunability range. The combination of
inductor and capacitor capabilities optimized for low loss and
flexible tunability range is very important for the design of VCOs
and filters. Developers have reported MEMS varicap and switch
prototypes.

Process Compatibility Considerations

Compatibility of the MEMS process with CMOS active circuitry
processing is critical for the realization of integrated MEMS. The
ideal MEMS process must have little if any impact on the
wafer/electrical yield of the active circuitry. This capability
allows component suppliers to offer increased performance designs
while maintaining or reducing device cost. Additional benefits are
the inherent reliability of the IC-based processing as well as the
reduced lifecycle cost (when contrasted to discrete passive
components interconnected via a printed circuit board). Leveraging
IC-based, low-cost processes, current estimates for volume
production of integrated MEMS inductors are already below $.10 per
unit as opposed to a minimum of $.15 per unit for traditional
discrete devices.

As manufacturers use on-chip RF MEMS to drive towards a
single-chip RF solution, it will be critical to produce all on-chip
MEMS devices (inductors, capacitors, switches, resonators, and
filters) using the same process in order to streamline design and
manufacturing. For example, a design team may consider using both
an inductor and a varicap on-chip to achieve a superior LC network
for circuit tuning. It would be cost-prohibitive to realize the
devices using two separate MEMS processes (such as thick metal
inductors and polysilicon varicaps), due to increased process
complexity and cost, as well as a very probable yield hit. A thick
copper metallization process can achieve the two device types at
the same time. Likewise, the IC derivative process is inherently
compatible with the fabrication requirements of switches,
resonators, and filters.

Again, a critical feature of the process is to be able to
realize these integrated structures above-the-IC, that is within
the device footprint of the active circuitry. Otherwise, you must
consider a tradeoff between the device cost increase due to
increased silicon area usage and the cost reduction (component
cost, assembly, and reliability) from the passive-component
integration.

An On-Chip High-Q Inductor Illustration

Many efforts have been focused on improving the RF performance
of silicon technology, in part due to its low cost, dielectric
compatibility, and micro-machining properties. The high-Q inductor
offers an excellent example of the cost reduction, high integration
density, and performance improvement benefits of RF MEMS and
microwave devices.

Typically, the quality factor of inductors using standard
silicon technology is less than 10 at 2-GHz, due to the parasitic
effects of low-conductivity metallization as well as lossy
substrates interactions. Therefore, to achieve high-performance,
most RF integrated circuits applications still require the use of
off-chip inductors. However, drawbacks of off-chip inductors
include significant parasitic effects, prohibitive size, and large
losses due to the board-level flip-chip and/or surface-mount
interconnects.

Developers have focused recent efforts on improving basic
silicon processing capabilities to yield high-quality-factor
inductors. In these efforts, the focus has been to leverage the
inherent low cost of silicon-process technology as well as the use
of low permittivity, microwave-compatible dielectric materials, and
thick copper metallization. Currently, integrated MEMS inductor
designs are under way in circuit applications at a number of
well-known telecommunications equipment and telecom semiconductor
manufacturers. Design teams have also begun work on other
integrated passive components as well. The same integrated
(on-chip) components may also be fabricated for discrete (off-chip)
applications. In these cases, the benefits of increased electrical
performance of the device can be realized while maintaining the
same package specification and assembly operations.

Topology of the System

To achieve a high-quality-factor inductor design, design teams
require a CMOS-compatible process technology to minimize parasitic
effects of the substrate and provide high-integration density. The
new process enables definition of a completely new inductor that
can be deployed on RF circuits using a flip-chip “discrete”
technique or a new above-IC approach, where the inductors are
post-processed on top of the wafer and active circuitry.

Such a process typically consists of electroplating copper over
a low loss, low permittivity, insulator material layer that is
placed on the integrated circuit. The inductor is further
interconnected to the IC by vias through the insulator layer. The
above-IC technique is very attractive due to the advantages of
small size, package cost reduction, superior interconnect
performance, applicability for mixed technologies, and overall cost
reduction for high-volume applications.

The innovative aspect of the process is the thick metal
fabricated on a microwave-compatible dielectric material. The
electroplated copper conductors are implemented within a
photoresist mold using conventional UV equipment. This technique
provides high aspect ratios with angles near 90°.


Figure 2: Cross-section and photo of MEMSCAP's high-Q
inductor, offering a quality factor (Q) of up to 80 at 2-GHz. In
the photograph, the intrinsic part of the inductor and the two
coplanar ports (only necessary for the probe test and
characterization) are shown.

The spiral inductors and the different layers are shown in
Figure 2. Two kinds of inductors have been processed, one using
silicon substrates and one on quartz. Different inductor values
have been designed in the 2-GHz range. We have achieve typical
values ranging from 1.5- to 18-nH for geometrical parameters
compatible with RF integrated circuits.

Modeling of the Inductors

We use classical methods to obtain the different electrical
(inductance and resistance) and geometrical (length, width, gap)
parameters of the inductors. Modeling other inductor parameters
requires more of an original approach. To obtain Q factor and a
more accurate value of the inductance, one design team has
developed a technique that uses full wave electromagnetic
simulation based on Finite Element Method (FEM) analysis.


Figure 3: An example of inductor modeling with nominal
meshing. The finite-element method (FEM) model allows design teams
to create a parameterized description of the inductor for defining
high-performance components.

The results demonstrate a high degree of accuracy between
measurement and theoretical predictions. The nominal meshing of an
inductor is shown in Figure 3. This technique allows engineers to
model the complete structure, including the different layers and
properties of the materials used to fabricate the component.

Inductor Characterization

We characterize the MEMS-based inductors on a wafer with
scattering parameters measured from 100-MHz to 16-GHz using an HP
8510 Network Analyzer and a Cascade probe station. Two kinds of
characterizations were used, a classical one relying on TRL
(Through-Reflect-Line) calibration in the probe planes and a
de-embedding response that allows determination of the inductor's
intrinsic parameters.


Figure 4: The results obtained up to 16-GHz for a 3-nH
inductor using a silicon substrate and a TRL (Through-Reflect-Line)
calibration procedure


Table 1: Inductor Q-factor summary using one and two
metal layers on the silicon substrate

Emerging MEMS Solutions

With increasing miniaturization of electronic systems and
advancement of MEMS capabilities, design solutions continue to
emerge based on the integration of MEMS technology. Specific to
RF/Telecom applications, potential integration opportunities exist
for inductors, capacitors, switches, filters, and other integrated
RF functions. The ability to integrate these components within the
RF IC using a singular post-processing recipe provides a clear path
to realizing the long-awaited single-chip RF solution. These same
optimized passive components can also be manufactured in discrete
formats.

The architecture of high-Q inductors designed with the MEMS
process increases the integration of RF chips and monolithic
microwave integrated circuits (MMICs). Furthermore, the
electromagnetic simulation techniques and new process technology
offers high-quality passives for silicon RF chips using thick
copper metallization, low loss and low permittivity dielectric
material, and a new above-IC process.

All these improvements lead to high-quality-factor inductors in
the 2-GHz range. We have seen high Q values ranging from 30 to 80
with the mean value equal to 50 for inductor values ranging from
1.5- to 18-nH. Such high-quality factors for inductors are among
the best results seen in silicon, particularly when using standard
processing technology.

About the Authors

Patrick Albert is director of the wireless
business unit of MEMSCAP in Grenoble, France. He can be reached at
.
John Costello is director of professional
services for MEMSCAP in Raleigh, NC. He can be reached at .
References

[1] M. Aubourg and P. Guillon; “A mixed finite element formulation
for microwave device problems. Application to MIS structures,”
Journal of Electromagnetic Waves and Applications, Vol. 5, pp
371-386, 1991.

[2] J.N. Burghartz, M. Soyuer, and K.A. Jenkins, “Microwave
Inductors and Capacitors in standard Multilevel Interconnect
Silicon Technology,” IEEE Transactions on Microwave Theory and
Techniques, Vol. 44, No. 1, January 1997.

[3] N. Klemmer and J. Hartung, “High Q Inductors for MCM-Si
Technology,” IEEE, 1997

[4] H.M. Greenhouse, “Design of Planar Rectangular Microelectronic
Inductors,” IEEE Transactions on Parts, Hybrids and Packaging, Vol.
PHP-10, No. 2, pp. 101-109, June 1974.

[5] J.N. Burghartz, D.C. Edelstein, K.A. Jenkins, and Y.H. Kwark,
“Spiral Inductors and Transmission Lines in Silicon Technology
Using Copper-Damascene Interconnect and Low-Loss Substrates,” IEEE
Transactions on MTT, Vol. 45, No. 10, October 1997.

[6] D. Lovelace, N. Camilleri, and G. Kannell; “Silicon MIC
Inductor Modeling for High Volume Low Cost Applications,” Microwave
Journal, Vol. 15, pp. 173-180, 1994.

[7] L. Bastères, P.
Blondy, B. Guillon, F. Valentin, and A. Mhani “High-Q Inductors for
RF and Microwave Applications in CMOS Technology,” Workshop on New
Technologies for RF Circuits, Bracknell, UK, March 28, 2000.

0 comments on “Shrinking RF Circuits Using MEMS”Toward the Single-Chip RF Solution

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.