Review of Workshop
IEEE European Systems Packaging, Blarney Park Hotel, Cork Ireland
January 25th -27th, 1999
(from an Interconnection Decision Consulting Report)
Overview
This was the 6th European Workshop of the Systems Packaging Committee
of the IEEE. Again it was co-sponsored by the Computer Society
and TC-14 of the CPMT Society. This was the 6th workshop in Europe
( two Brussels, two previous in Cork, one in Malta ) and represented
a return to Cork.
Attendance was standard for Europe - about 70. Costs were phenomenally
low because of the Cork location. Hotel fees were
bundled into the registration fee, and the total was only $350
(230 Irish Pounds) Cork remains a very low cost place to hold
a workshop.
Session I - Telecom
Technological Solutions for Wireless Communications
Systems was presented by David Pedder of Mitel who talked of the
TW: CS Project of Espirit, with participation involvement by Mitel,
Ericsson, Sindelfingen STP, and Bull. The goal was to get bumping
capability down to 55 micron bump size.
The construction was MCM-D, with the substrate having the
chips and passives. The deposited layers provided the
connection to other mixer circuits. Operating frequency was 1.65
GHz for the VCO module; the balanced mixer operated from 80 to
200 MHz. Hiperlan partitioning design was used.
Plasma etched vias were not good enough ! Laser drilling
was necessary. An alternative would be sequential buildup using
photo-sensitive dielectrics.
This technology takes half the area of a Ceramic BGA
module.
One comment from Ericsson - losses are too great in ceramic at,
or over 20GHz, because signal travels only on the surface and
the surface of ceramic is rough and of poorer conductivity than
copper.
It is better to use copper, and particularly rolled copper, for
its smoother surface. With rolled copper, one can go to 40GHz.
The Copper frequency limit is 6GHz; higher performance substrates
are used only when needed.
Lutz Richter of NOKIA presented a report on Project
Bluetooth, a Low Power RF Module. This was a project to create
a Universal Mobile Communications Chip that would provide different
connectivity for different folks. It can wirelessly serve phone,
PalmPilot and PCs, and sales volume is over 100 million modules
per year, at a total module cost, including chips of only $ 5.00
! This is a 6-layer FR4 substrate with microvias.
Range is now 100-150 meters within a building. This is for
the equivalent of cordless telephone range, not satellite.
Lief Bergstradt of Ericssoon was invited to give his
presentation on Microwave MCMs in PTFE ( Rogers ). This talk was
presented at the Beijing Symposium. but was worth re-playing for
this workshop. Performance of the MCM in ceramic was limited to
20GHz, because of the high conductivity of the metallization in
the top few angstroms, because of the shallow skin depth. Moving
to copper as a conductor improved the conductivity and permitted
performance to 40GHz.
Another alternative was to use a copper/moly central core,
sandwiched in Brass coated with just 1 micron of dielectric.
Again, this was rolled copper. Rogers RO 3003 was the dielectric
- a filled PTFE. If highest performance was not needed, the same
dielectric and copper was used on FR4.
The processor was INTEL, fabrication and assembly was by
Ericsson, and the customer was Toshiba.
Walter De Raedt of IMEC in Belgium was to talk on MEMS,
Microm Electrical Mechanical Systems, but he gave very little
detail or examples. The talk in the evening session by Mathuna
was the definitive talk.
Still, De Raedt did have useful passing comments:
Micromachining precludes the use of chemical technology - he sees
that as good. (How do they etch without chemicals? ED)
He thinks embedded passives degrades performance and are expensive.
(Major disagreement here)
The MEMs technology was used to produce a variable capacitor.
In the NOKIA construction, there are two
interdigitated combs of the end bars of a surface capacitor.
The one comb is movable, and can be driven by voltage on a paddle
plate attached to that comb. Changing the voltage moves the combs
closer together, increasing the capacitance in a manner proportional
to the electrostatic voltage. Capacitance change is 3:1, works
fine at 1 - 2GHz.
Inductances were spirals directly on the silicon.
The Evening Session II - High Density Interconnect and
MEMs
Gerard Nicholas of CEA-LETI talked of their Chip contact to
Flex and other chips using Anisotropic Conducting Films (ACF Technology).
Chips are tested on a glass wafer, then the good chips are transferred
to the substrate.
These were big modules 4" x 4", with 10 micron lines
and spaces. The unit provided 8 x the output of the unit it replaced.
Claude Massit of Matra Space talked of their connections to flex
on chips also, through the use of Anisotropic Conductive films.
This was flipchip technology, REWORKABLE! Mass production is planned
in COS, a Glass Fiber reinforced substrate.
But there are pad spacing limitations, and contact resistance
dispersion.
Cian Sean O'Mathuna presentation "Microsystem Packaging -
Case Studies", was one of the high spots of the meeting -
even though he was precluded from presenting some of the best
work. Small micromachined movable liquid pump chips have been
made for use in dispensing insulin and similar body necessities,
and milk pumps were developed to meter out micro quantities of
milk for chemical analysis, etc. He talked of $14 billion of MEMS
devices in the year 2000.
They developed microsurgery "pills" to be swallowed
whole to patch ulcer sores in the stomach. Mostly the sensor information
is conventional, but the micromechanical devices are etched on
the surface of the silicon as needed.
The technology is Chip-on-Board on PCBs, with globtop with a hole
in the middle ( like a donut ! ). The height of the
globtop protects the chips, but globtop could not be used in the
middle - how could the wheels turn ?
For contacts with food, it must be noted that yeast loves plastic,
so fungicides must be used. He sees a major need for a paradigm
shift - a combination of Silicon Technology, Sensor and Packaging
Techniques. That technology cannot reach its potential if they
are packaged separately.
They did some of the assembly using the 3D technology from
3D corp, ( Christian Val )
Eric Beyne of IMEC talked about their Indented Reflow
Soldering Techniques as applied to a Miniature relay. This was
another high spot.
Many packages in development have not been packages at all, and
those that have been were in ceramic with clear top windows. -
too costly. Better to seal the chip at the chip level than use
a non-hermetic plastic package with a pyrex top lid on the cavity.
Holes have to be provided for the cases where the Micro-pumps
need to access the fluids in question.
He talked about a microswitch developed jointly with Clare
Relays that was operated by a magnetic field, but this seems no-where
near as elegant as the one talked of at the WCC that had electrocapacitative
actuation of a see-saw with paddle ends.
Quite surprisingly the die cavity can be filled with a low viscosity
fluid, pumped into a sealed cavity, with the opening remelted.
This fluid is magnetic and controls the field.
This seems unduly complicated to me, and nowhere near as elegant
as the switch with the capacitive driven see-saw paddles.
Session III, Tuesday, European Union Packaging Projects
The definitive talk of the State of European Union Projects was
given by Mark Hutton of BPA In this talk, Hutton made the following
points:
1.- Systems need CSP and Flip chip
for electrical performance and small size
2.- Only a few manufacturers can now provide microvias -
and microvias are necessary
3.- Wafer level CSP manufacture of the chips with surface
distribution can be cost effective.
Evan Davidson, an attendee, pointed out that multilayer ceramics
are now very cost-competitive; they are 1/5 the cost of
a few years ago. But their high frequency performance is poorer.
J.M.Wilkinson of Technology for Industry reviewed Current
Activities in MST Packaging (MST = Microsystems Technology). This
market is growing at 30% per year, and the talk was mostly giving
examples.
Philips, Cochlear Implants for the Deaf
1,000 per year, MCM-C
Dyconex, substrates for hearing aids, Flex PCB
Hamlin, Bosch, others, Accelerometers, Thin film Hybrid
Airbag controls skid controls
Inkjet controllers,
Smart card & fingerprint sensor
Europe has 22 MCM suppliers, plus the Dyconex activity in
flex.
Eric Beyne of IMEC discussed Chip Scale Packaging, as
examined by a consortium, including Saab Combitech, Siemens and
Philips. The aim was to study 25 different constructions for particular
discovery of whether the redistribution layers would provide adverse
performance if they were designed UNDER the chips. Both ceramic
and organic packages were examined, with the MCM-L units assembled
by Solectron.
The construction was Polymer Stud Grid Array, metallized and laser
ablated to leave only the interconnection circuits in place connecting
the stud bumps. A comparatively easy process with a low failure
rate. One of the advantages is that the studs can be closer together
than the usual bumps.
The charts were overwhelming - maybe 25 types of constructions,
subjected to 12 different tests in 60 weeks. Too long a time.
The delay is intolerable.
They proposed that future programs will have a fast track, by
omitting biasing of the traces, reduce the number of devices under
test and operate at lower voltages.
Rory Doyle of NMRC did his usual overview of Market Led Infrastructure.
He talked about the MCM Consortium, a creation of local service
contractors, 6 service centers, 6 types of MCMs and 25 manufacturers.
The customer is given a choice of all 25 constructions, he picks
10-12 for early information, and 3-4 for final analysis.
Prototypes are made using quick turn technology, because tooling
is a problem for production technology. (Tooling for production
is from 24K to 30k dollars. Various demonstrator substrates were
shown.
The goal is self-sufficient, self-funding commercial source
availability. A One Stop Shop.
O'Reilley of PEI talked of their experience with integrated passives.
Their goal was to examine the process steps for passives and try
to improve. Small value resistors are a particular problem. Films
were either 25 ohms/sq or 100 Ohms/sq; either had a tolerance
of 10%. If tighter tolerance is needed, laser ablation must be
used. The rest of the processes for capacitors and inductors were
standard, with 4@ micron height for the inductors.
Session IV - Advanced Interconection Applications II
Tim Lenihan gave the progress report on the Sheldahl technology
and manufacturing progress.
1.- There is a new thrust to go to 2 mil tape for better
vias and better impedance control. Vias are 60 micron and
25 micron, impedance is 50 Ohms.
2.- They investigated alternatives to the holographic laser
via drilling, and found that Plasma works, but its more
costly,
and using photosensitive dielectric is the most costly of all.
3.- They continue working with HiDEC to use the IMPS
technology and make it available to customers. It, of course,
provides the lowest cost of all.
Johan Liu talked about the work of IVF in developing Smart
Cards assembled using Light-Curable adhesives.
Planarity must be tightly controlled and the dielectric laminate
film over the conductive adhesive pads cannot be too thin -- 30
micron, but must be thicker - 100 micron. Too thin and there will
be gaps or voids adjacent to the deposited conductive pads.
I/O is by Ni/Au coated plastic balls. Pressure MUST be exerted
to deform the particles to get reliability. Subsequent curing
of the dielectric provides the joint stability. After the exposure
to the UV, there is a post cure at 80C.
The work at Marconi Technology Center is now called FLINT - Fine
Feature Built-up PCBS. This is MCM-L work and some SMT & BGA.
Wirebond is still used, but some flip chip also.
Package pin counts are 600 to 1,000, and microvias are getting
smaller - from 150 to 200 mils to 35-50 mils.
Constructions are u to 12 layer.
Most vias are laser drilled - by a new $ 40 million laser drilling
machine. They use it, not only for drilling, but also for laser
ablation of solder mask.
Projects include a Marconi IR Image Processor where 2 SMT boards
were replaced by 1 Microvia Board. They now use no SLT or BGA
packages, but they can mix other packages instead.
These are laminate build-up boards of polyimide glass-reinforced.
They like the performance better than FR-4.
The PEI talk on their HCCAAP Power Electronic Packaging of integrated
photovoltaic converters which are intended to reduce the dependence
on fossil fuel for world electronics consumption.
The technology features many small converters in each sub-panel.
Power collection density is 110 Watts per square meter. The goal
is 8-year lifetime, and their reliability testing indicates greater
life with greater bump heights.
If only 15 good units are fabricated on each sub-panel, they have
a viable project. ( Technology is 50 micron line & space,
but their goal is 25 micron.)
HIPERPRINT A High performance PCB for Telecommunications and RF
Applications presented Johan De Baets, University of Ghent. Partners
in this project included Alcatel, IMEC, University of Berlin and
Solectron. This is 50/50 micron line and space, and 40 micron
via technology. There is a significant cost difference between
the 100 micron work and the 50 micron, so they are taking an intermediate
step of 75 micron technology.
Mostly they are using Rogers RO 4050, though they have looked
at FR-5 ( High Tg ) and Lin Lam. The FR-5 is for the future, and
possibly Lin Lam later.
They have been using photodefinable dielectrics, both
Multideposit from Shipley and Probelec from CIBA-GEIGY. The technology
can be used to pattern the circuits, but pattern the photoresist
also.
They explored 10 types of underfill and reduced the trial to the
best 5. Data on the comparison tests is not available yet. Technology
is 4 built up layers per side.
Didier Sampson did the talk from Matra. Their work is fine line
thick film technology with NO via fill (!) and the usual buried
passives. The only major problem was screen printing the conductive
inks.
The application is for a SPARC application, with 1,000 I/O
(substrate was 64 x 56 mm rectangles.) They are moving to Aluminum
Nitride. They found the thick film cheaper than thin for this
application, which is not high speed.
Steve Breibels of IMEC discussed the possibilities of MCM-D modules
with integrated antennas for Microwave and Millimeter Waveguide
Electonics. The trick to putting the antenna on the chips and
modules is to get no coupling between the antenna and the circuits.
These are circuits operating at 44GHz, and the market is for collision
avoidance radar, for planes and automobiles.
Self-oscillating mixer technology only works on monolithic or
MCM packaging density.
The output beam is shifted in direction target, using steering
pointing through phase changes of the outgoing antenna drive.
Construction used multilayer BCB on Ceramic, with
thermocompression bonding. Thermocompression bonding of chip wires
is difficult on organics, but in this case with the ceramic base,
they were able to get the reliability up.
Digital Modules Interposers were discussed by Michael
Massenat of MATRA. They chose the technology to permit the use
of sockets so that the modules could be easily replaced for upgrades,
etc.
For that reason they avoided Plated Through hole pin parts
and BGA, and they also decided not to use SMT. The technology
of choice was a direct clamp of the parts to the PCB without soldering
or conductive adhesive.
The interposers were made by CIN-APSE and Berg, or even the gold
dot process of Hughes.
Comments:
1.- Berg contacts were not auto-wiping
2.- Elastomeric is not space qualified.
3.- They developed a new connector with a new slide contact
by FRB ( FRB - I don't know the company )
The packages were double-sided and stacked, similar to the
AT&T stacked technology. But this seems to be not work in
progress, but work not yet started.
Conclusions
Europe is moving to high density MCMs, with almost every player
involved in a consortium. Very few major players seem to be going
it alone. Anisotropic Conductive Adhesive is only slightly considered,
and some fully conductive adhesive is possible.
Abridged from an IDC Report by Jack
Balde
(Full reports are available to IDC Clients )
www.Inergy.com/balde/IDC_Consulting.html