More than 900 component/packaging/manufacturing enthusiasts showed
up at the Wyndham Palace
Resort for the 51st ECTC. There were six parallel sessions
at ECTC for all 3 days. So even someone who attended must rely
on the Proceedings to find out about talks they were unable to
hear during the meeting. None-the-less we will list some of the
notes taken by one happy CPMT member. The meeting was abuzz with
people signing in,
purchasing CPMT publications,
and networking.
Session 1: High speed packaging.
Felix Mederer of University of Ulm described how
they fabricated VCSELs with thicker monolithic cavities. This
allows a single mode even with a big enough aperture (7 microns)
to match standard polymer optical fibers. The target was 10 Gbs
information for WAN applications.
Michael Vrazel of Georgia Institute of Technology
discussed an inverted MSM structure they used to detect high speed
information at 1310 nm. The device operated at 3.3V; they have
55fF of capacitance when designed for 2.4 Gbs operation.
Isao Yoneda of NEC showed details of a 4 channel
parallel optical transceiver. Trench isolation was used between
the edge emitters. A fixture on silicon that aligned the 4 fibers
and bounced their signal from etched mirrors up to the flip-chip
photo detector array was displayed.
Mikko Karppinen of VTT Electronics described an
interesting way of obtaining low cost passive alignment of fibers
perpendicular to an LTCC structure. The fibers align by vias designed
into the LTCC material and the transmitter chips are flip chipped
onto the far side of the LTCC structure. No ferule is needed to
align the 5 fibers and UV curable adhesive locks things in place.
Eric Grann of Blaze Network described Optical Transceiver
in use for Course WDM which is much in demand for short range
data transmission (even within a radar system). His system works
with multimode at 850 or 1310 nm but this limits you to 1 Gbs
for 2 km. He uses injection molded optics, standard interference
filters that can be batch made 1000s to a wafer, VCSELs, passive
alignment, and no forced cooling on package. This approach was
designed to need only the precision of standard pick-and-place
machines.
Shimya Abe of Fujikura detailed a very small transceiver
using VCSEL and pin photodector.
Session 10: RF MEMs and Systems
Dave Palmer of Sandia Labs gave a tutorial update
on fabrication of RF MEMs switches for applications such as radar
phase shifter circuitry. The many questions indicated the growing
popularity of mechanical switching.
Nathan Bushyager of Georgia Tech demonstrated a
modeling and design approach to MEMs. It is called MRTD - MultiResolution
Time Domain. Typical simulations on a PC still take 3-4 hours
but used on clusters of PC should make design more interactive.
Joseph Hobbs of Georgia Tech studied the trade
off of expensive on-chip capacitors versus several discretes on
the PWB with their frequency limitations. The compromise was distributed
capacitors within the PWB. They showed suppression of switching
noise by a factor of 7 at the high frequency range (300MHz). Traditional
lumped capacitors handle the low frequency problems.
Y. L. Li detailed work on distributed models for
capacitors. By modeling capacitor-like tapped transmission lines
with mostly C and a little L.
The last presentation from the National Sun Yat-Sen University
described success at using a leadframe CSP for RFICs. The motivation
is that the frequency performance of transistors is severely degraded
by standard packaging.
Session 13: All Optical Networks and MEMS Packaging
Opening remarks by Ron Scotti of Lucent gave an
excellent overview of MEMs technology applied to Optical Networks.
First he talked about the simple designs that have lead to the
big MEMs money makers: Analog Devices accelerometers and Lucent
router mirrors. (I don't think the engineers considered them simple
at the time of development). He reviewed the positive and negative
of the dominant polysilicon MEMS technology. Pro: highly versatile,
components have low mass and high resonance frequencies so they
are immune from normal vibrations, technology is well understood
and can produce robust designs. Con: high temperature processes
cause high stress in films, only limited thickness can be used,
polysilicon has a grainy poor optical surface, and there is reliability
concern in a non-hermetic environment.
He then described a new approach using Deep reactive Ion Etching
(Bosch) combined with SOI and wafer bonding. Pro: very good optics,
bulk silicon, no stress gradients, and thick structures easy.
Con: processing has limited flexibility, hard to get good SOI
starting material, high mass structures give lower resonance frequency,
limited number of layer.
James Walker continued in the overview mode. He
said that MEMS reliability is an unknown since the field is new.
For example, SOI MEMS are only 2 years old and optical MEMS are
6 years. There is little mechanical force available in most implementations
of MEMS, so moving light beams is a natural application matched
to the technology, "photons are very light".
The trend in fiber optics is screaming for MEMS cross bar switching.
Need to keep total power small, optical channel spacing is shrinking
(DWDM), bit period is shrinking, range of wavelength control increases,
carefully engineered fiber links. MEMS could give more dynamic
control of links providing gain equalization and dispersion compensation.
"Everyone" is doing optical crosslink for the rapidly
expanding DWDM market. Worries that too many "49ers"
are chasing too few nuggets. However, 3D MEMS is the only scalable
solution to this crosstalk challenge. The other big application
of MEMs in the telecommunication world is MEMS tunable VCSELS.
Walker worries that the hype of MEMS has long outrun the reality
in this application space. He reminds everyone not to fall in
love with one solution to the telecom problems since alternatives
always exist. Deliver MEMS not hype or an alternative will win
the war.
Steve Robinson of Photons Inc. repeated the theme
that 3D crosslink technology was needed. He emphasized the challenge
of analog control of the mirrors as being much harder than the
digital mirror control of past applications. He saw the crosslink
business growing 44%/year arriving at $4B in 2008. He is an advocate
of bulk MEMS for optics not surface MEMS (half the audience looked
for things to throw at this point). He predicted that within a
few years high capacity MEMS switches with 10msec times would
be common.
Jeff Bennett of Intel discussed automated fiber
alignment regardless of temperature, aging, and shock. The goal
was to produce a 10Gbs transceiver in 1/8 the previous volume.
A hierarchy of fluxless solders was used. The study of wetting
and interactions was a full time effort for some material scientist.
Session 18: Fatigue and Delamination (Figure: Bob Howard talks to co-chair Donna
Noctor as Andrew Tay prepares to talk)
Andrew Tay of National University of Singapore discussed
the affects geometry can have on the delaminations in rectangular
IC packages. Several design rules of thumb were ferreted out of
the computer analysis: a 1 mm border of lead frame pad must be
around the chip to stop crack propagation, also the ratio of encapsulant
above and below the leadframe and the ratio to the chip thickness
made a large difference.
Session 22: Novel Packaging Technology
Kyocera Corporation discussed the use of LTCC (low dielectric
constant) for broadband mobile communication up to 50 GHz. They
analyzed the radiation that came from signals going through vias
between SMT layers.
Session 25: Optical Interconnects
Larry Coldren
of University of California Santa Barbara updated everyone
on the development of long-wave length single-mode VCSELs (for
long haul telecommunication). Aiming at 1.55 micron intended
for 10 Gbs under OC-192 for WDM. He reviewed the GaAs DBR in
the 850-980 nm and all the attempts to lengthen the wavelength.
The index contrast in this material gets diminished at longer
wave lengths. To date the quantum dots and GaAsSb quantum wells
do not have enough gain. There has been success by wafer fusing
mixed material systems such as GaAs/AlAs mirrors on an InGaAsP
active region.
He described the InP based system made using MBE. The system
used metamorphic AsAs/AlAs top mirrors to make the system reliable
and manufacturable; multiple active regions to overcome bad mirrors,
used all n-type material by placing a degenerate tunnel junction
to get bias correct. They use oxidation to control the aperture.
To get enough power in the fiber he used 8 VCSELs like wedges
of pie forming a circle.
Yue Liu of Honeywell Laboratories concentrated
on OE arrays. Their goal is to integrate logic and switches in
2D optical array and therefore minimizing channel crosstalk. By
placing the VCSEL light emitter, MSM detector, and control electronics
in a cell you get a "smart pixel" which allows some
control of crosstalk.
Honeywell has been shipping millions of 850 nm VCSELs/month and
has recently added oxide confinement to their mass manufacturing
process. One unique part of this government consortium effort
was the application of arrays of microlens on a wafer with a jet-printer
head.
Yuzo Ishii from NTT Telecommunications Energy Laboratories
showed flip chip carriers on PWB used to create complex OE systems.
The carriers had light source and detector chips placed in a
downward cavity of a carrier that has ICs on its top. The PWB
has a via hole that allows light to pass from the carrier to optical
waveguides or fiber butt ends on the other side. They have obtained
82% coupling efficiency. They also used microlens dispensed by
ink-jet printers.
A talk by Aegis Semiconductor detailed a transparent
monitor for giving gain control feedback to each individual VCSELs
in an Array. Their technology uses amorphous semiconductor from
solar cell development and transparent conductive ITO oxide...all
put down at less than 300C. They have worked up to 32 x 32.
Jerome Eichenberger of Optobahn Corportation discussed
2.5 Gbs/channel communication. The immediate customer was the
short server rack to rack communication, but they also anticipate
a board to board business. They showed a 4 channel unit using
MT ferule and hermetically sealed. It passed all the environments
industry asks for. It used standard pick and place and automatic
active optical alignment. Their modules can be reflowed onto customer
boards. Alignment in next assembly is not super important because
there is lots of power to spare.
Motorola Fellowship
"Low Loss Deep Glass Waveguides Produced with Dry Silver
Electromigration Process", by Ricky Chuang
(student) and Prof Chin C. Lee from University of California Irvine
became the winning paper for the annual Motorola CPMT graduate
Fellowship.
Session 16 Co-Chairs
Sudipta Ray (IBM) and Joseph Soucy (Draper Labs)