Feature Article Think Small! issue 4, Volume 2:
RF MEMS switches deliver on early promise
by Jérémie Bouchaud & Bernardo Knoblich
Few MEMS components have created so
much excitement, and then such disappointment in so short a time.
Is this component finally in production? Will there be an interesting
market? And for whom? We have been following RF MEMS switches since
2000 and have just updated our technical and market analysis. Although
the market was just $6 million last year, it will increase to $210
million in 2011.
Up and down
The hype curve was first proposed by analysts at Gartner as a way to place technologies in their particular stage of evolution. This curve follows a distinct trend. We abstract this principle to the case of RF switches (see figure 1). Switches reached their peak visibility - also called the “Peak of Inflated Expectations” - in 2003, when Magfusion and Teravicta announced the first samples. Unfortunately, these two small start-ups could not process the demands for samples that suddenly poured in from around the world. A number of potential adopters did not receive their samples and were frustrated. Confusion also plagued the announced specifications. As a result, those who were fortunate enough to receive samples were unimpressed by the performance. This is represented by the “Trough of Disillusionment” phase.
Happily, it appears that
RF MEMS switches has now left the latter and emerged shining into
the “Slope of Enlightenment”, an indicator that the industry and
technology are both maturing. Start-ups like Teravicta have completely
changed management and sales teams and now tend to adopt a more
modest or realistic marketing approach. Good funding rounds were
recently finalized, for example, at Teravicta and Wispry. Meanwhile
switches from companies like Radant and MEMtronics have surpassed
the 100 billion cycles mark, further quelling reliability doubts.
In addition, major improvements in low-cost in situ packaging and
CMOS integration have been achieved.
First commercial products implementing MEMS
switches, i.e. Automated Test Equipment (ATE) systems from Advantest,
started shipping at the end of 2005. And finally, there is encouraging
news from Wispry and NXP, which plan ramp-up production of MEMS
switches (more precisely switched capacitors) for cell phones.
The last level, the so-called “Plateau of
Productivity”, is expected to be conquered by 2010 when the benefits
of RF MEMS switches are demonstrated and accepted throughout a number
of applications in test, telecom and defence applications.
The final height of the plateau will depend on whether MEMS switches will
make it into cell phones, or be implemented only in niche markets.
Can we buy them
yet?
A handful companies have started commercialising
RF MEMS switches.
Such devices are in small production at Teravicta, Radant (in the
US only for the moment, due to export restrictions), Advantest and
Matsushita.
Advantest has started serial implementation of proprietary MEMS
switches in its ATE and production of these MEMS switches is captive.
However, the company just founded a branch “Advantest Component,
Inc.” to make switches available to the free market.
Additional companies have started sampling
switches for selected customers including WiSpry for mobile handset
applications and MEMTronics and XCOM for high end applications e.g.
in defence. Omron recently joined the pack and is expected to start
serial production by mid-2008.
NXP is currently in the “industrialisation phase”
of switched capacitors for cell phones and is focussing its efforts
on process and manufacturability. Several foundries have also launched
serial production of DC switches for automated distribution frames
(ADF) for telecommunications including IMT and APM.
As of September
2007, we estimate that around 50,000 switches are being shipped
every month and that all in all 500,000 to 600,000 switches have
already shipped (not considering ADF DC switches produced for qualification
phase).
The selling price varies according to the specifications
and volumes. Teravicta ships its 7 GHz SPDT (single pole double
through) for $ 35 and 26,5 GHz SPDT for $92 in small volumes. Teravicta
also announced a new 5 GHz SPDT switch for $ 8 by the end of the
year. At Radant, SPST (Single Pole Single Through) can be bought
for $25 and the SPDT for $40 in 2000 unit quantities. Matsushita
positions itself at the high end ($50 and more) so that the MEMS
solution does not cannibalise its line of conventional EMR relays.
Switches for ADF are already shipping at less than $1 due to the
high demand even in the qualification phase.
Potential applications
RF MEMS switches are currently being shipped or are in development
for a very wide spectrum of applications, from high-value niches
such as satellites up to mobile phones. We distinguish four application
fields: test equipment, telecom infrastructures, aerospace and defence
and finally, mobile phones.
Test and instrumentation
Automated
Test Equipment (ATE) for semiconductors: this
was the first commercial application for RF MEMS switches. Implementation
of MEMS is quick and easy since it meets technical requirements
and conventional relays can be directly replaced with MEMS without
changing the system. ATE market leader Advantest developed its own
MEMS switch, which it started to implement in its own ATE systems
at the end of 2005. Most major ATE system suppliers including Agilent,
Advantest, Teradyne, Credence and Verigy are currently evaluating
or have started replacing conventional relays with MEMS.
RF Instrumentation
Manufacturers of RF instrumentation equipment,
i.e. network analysers, spectrum analysers and oscilloscopes like
Agilent, Rohde & Schwarz and Tektronix are evaluating MEMS switches
and would implement them “better yesterday than tomorrow”. However,
power handling issues must be solved prior to implementation in
RF instruments since hot switching is required, contrary to ATE.
Serial implementation is expected in 2009.
Telecom infrastructure
Automated Distribution
Frames:
Although this application has often been neglected, MEMS switches
will start to deploy for wireline telecom switching matrices early
in 2008. These switches are no longer classified as RF (as they
operate in MHz range) but we include these devices in our analysis
because each can be considered to transmit a signal. Production
is ramping up at foundries like IMT, APM and tMt. Simpler Networks
is a key player in this area and signed an agreement with Alcatel-Lucent
at the end of 2006. Worth mentioning are also Telepath Networks,
Norcada and MEMScap, which owns key IP in this area.
Base stations:
RF MEMS switches are also under development for base stations in
the 0.5 to 6 GHz range (see interview with Alcatel-Lucent on page
13). RF MEMS is a real enabler for reconfigurable architectures
due to high linearity properties. 
Additional switch applications in telecom infrastructure are currently
being explored and include microwave infrastructure (at 20 to 60
GHz) and even RFID readers.
Aerospace &
defence: more than a niche
Defence applications have historically
driven the development of RF MEMS switches in the US and still does.
We estimate that the US Department of Defence invested over $30
million in RF MEMS in 2006, mostly for switches. The volume application
will be phased array antennas for communication and radar, e.g.
for missiles, helicopters, aircrafts, drones, ships and so forth.
A particular effort is also dedicated to switches and tunable capacitors
for agile filters used in tactical radio. It comes as no surprise
that the US leads in this field.
The pioneers - Raytheon and Rockwell –
now play the role of system integrators, leaving most of the work
on switches to start-ups Radant, MEMTronics or XCOM. In Europe,
since ITAR issues bar access to these US-sourced components, system
companies like EADS, BAE Systems and Thales have started to develop
their own switches. In Asia, the major effort in the aerospace sector
is performed by Mitsubishi.
Outside defence, NASA and ESA are sponsoring research for satellite
communication payload. EADS and Thales are currently cooperating
for the development of MEMS based reflect array antenna for internet
access in civilian aircrafts (EC project RETINA).
Automotive: not
mid-term
We do not see any opportunity for
RF MEMS switches in the automotive sector by 2012, contrary to other
recent market reports. In this case, the competitive technology
is not at the component but at the system level. For automotive
radar, RF MEMS switches are very promising for the beam steering
approach investigated in the European project MIPA. Our interviews
with the automotive industry show however that a digital beam forming
approach offers more promise, since digital processing can make
it cheaper. In this context, RF MEMS is irrelevant since a short
switching time of some nanoseconds is required for digital beam
forming systems [1].
In a similar fashion, we do not believe in the success of automotive
roof antennas. The reason is the presence of 3G+ systems that could
provide the same service for lower costs. Terrestrial infrastructure
(base stations) are considerably cheaper than satellites.
Will RF MEMS hit
cell phones?
Cell phones have brought the most
excitement for switches since coming to light in 2000 with the looming
perspective of a billion unit market. A number of applications have
been investigated including T/R switch, multiband filters and filter
banks, and impedance matching networks, and so on. Infineon, Motorola/Freescale
and ADI have abandoned the idea for several reasons, besides the
well-known reliability and packaging difficulties:
- Extreme price pressure. Even if MEMS
switches enable lower insertion loss, higher isolation… these improvement
are nice to have, but in most cases no customer will be ready to
spend a cent more.
- Alternative technologies kept improving!
In particular, SoS CMOS switches from Peregrine offer excellent
linearity and insertion loss. These devices have bridged a major
part of the gap between MEMS and other technologies in term of performance,
making the need for MEMS less obvious.
What’s left for MEMS switches in cell phones?
From all the possible cell phone applications mentioned above, WTC
believes that impedance matching networks for the PA (Power Amplifier)
or the antenna module offer the best (and only) prospects for RF
MEMS switches—or more precisely, switched capacitors—in the next
5 years. NXP, RFMD and WiSpry are currently developing switches
for this function. NXP plans to commercialise antenna matching circuits
based on switched capacitors as early as 2009 or 2010. NXP leverages
its PASSI process to integrate MEMS with other passive at a very
low additional cost. A prototype will be presented for the first
time at Semicon Europe in October.
RFMD is very strongly positioned in the PA business and plans to
commercialise MEMS-based reconfigurable PAs, again by the end of
the decade.
The
RF MEMS switch market
All in all, the RF MEMS switch market
reached a close-to-insignificant $6 million last year, mainly for
ATE and automated distribution frames (see figure). We expect the
total market to reach $210 million in 2011. The main applications
will continue to be ATE. The figure for cell phone refers only to
the additional cost of RF MEMS switches in antenna or PA modules
(mainly packaging cost). If one considers the RF MEMS based modules
which NXP or RFMD will ship, cell phone will dominate with more
than $ 100 million in 2011.
In terms of the supply chain, who is best
positioned on this emerging market? We have referenced more than
40 companies actively developing or manufacturing switches. Interesting
is the prominent number of system companies with an own switch development
effort, e.g. Raytheon, Rockwell, Thales, EADS, BAE Systems, for
defense and aerospace, Advantest for test, and NXP, RFMD, NTT DoCoMo,
LG and Samsung for cell phones. There are two reasons why system
companies do this:
- The high impact of MEMS switches at system
level (enabling new architecture), where system companies need a
hands-on approach to understand the potential impact at this level.
- The limited availability of RF MEMS switches
so far, especially at microwave frequencies as a result of ITAR
restrictions. Therefore, system companies in Europe or Asia need
to develop their own technology.
This high involvement of system companies
at the MEMS component level should not be taken as a sign of a closed
market. Most of these companies prefer to buy from the free market.
In fact most defence and aerospace companies – even pioneers like
Raytheon and Rockwell – will either:
- buy the switches from components from
the start-ups such as Radant, MemTronics, XCOM…
- or transfer their design at MEMS foundries
such as Tronics, IMT…
Also worthy of note, some system companies capitalize on their MEMS
switch development – initially for their own needs – and commercialise
them on the free market. This is the case for Advantest, the leading
ATE company. Additional system companies are currently investigating
this opportunity.
Conclusion
Returning to the hype curve, we see
that RF MEMS switches have finally left the “Trough of disillusionment”
for the much more enjoyable “slope of enlightenment”.
There are interesting opportunity for suppliers
of the following:
- Single components for test and instrumentation
- Modules with higher value, e.g. phase
shifters or tunable filters for base station or defence and aerospace.
There are also great chances for module
suppliers to cell phones. Since better performance is not sufficient
per se, we believe that integration will be the key for success.
Therefore, the best opportunities are for those organizations that
can integrate MEMS into their existing module business, like RFMD
or NXP. Start-ups in this field will probably have to licensee the
technology or take the more risky route of developing their own
modules.
Important lessons, alternative technologies also improve; new competing
technologies emerge such as SoS FET or ferroelectric devices. Therefore
MEMS switches need to continuously keep improving where they already
are superior e.g. insertion loss, linearity… And last but not least,
sometimes, the alternative is not at the component level but at
the system level like shown in automotive radar and roof antennas.
References:
[1] ARRRO