Nuclear Instruments and Methods in Physics Research A 473 (2001) 102106

Multi Chip Modules technologies

P. Gerlach*, C. Linder, K.-H. Becks
Fachbereich Physik, Bergische Universit . aat Wuppertal, Gaussstrasse 20, D-42097 Wuppertal, Germany
Abstract
In the past few years, impressive progresses in electronic devices and their packaging have been achieved. High energy
physicists benet from this, being able to minimise the amount of material in their Vertex-Detectors. Although the
environment of Vertex-Detectors is much more controlled and stabilised than in the industry- or end-user-applications,
the requirements of material-budget and device complexity are higher.
This paper tries to give a short overview of the existing Multi Chip Module technologies used in Vertex-Detectors and
focuses on the prototyping results of an upcoming Multi Chip Module technology, called Multi Chip Module-
Deposited (MCM-D), as it is planned to be used for the innermost part of the ATLAS Pixel detector. r2001 Elsevier
Science B.V. All rights reserved.
Keywords: ATLAS; Bump bonding; Flip chip; MCM-D; Multi Chip Module; Thin lm; Pixel; Vertex-Detector
1. Introduction
There are two developments that help a designer
of a Vertex-Detector to implement an increasing
number of features in the system. In the industry,
the rst is known as System on a Chip (SoC),
meaning that all features are implemented in one
integrated circuit. This is achieved by enhancing
the manufacturing process (mixed signal,
BiCMOS, embedded RAM, etc.).
The other way is to build the dierent function-
alities in separate integrated circuits, using the
appropriate process for each part (Bipolar,
CMOS, PROM, RAM, NVRAM, etc.) and to
build the specic device by mounting the parts
needed in a single but complex package. This
package provides the necessary connections be-
tween the parts and the outer world.
For Vertex-Detectors, the requirements on the
modules are quite dierent to the normal industry
ones. Cooling issues, for example, are well known
in industry, but the temperature and power
consumption ranges and the need for material
reduction leads to completely dierent solutions.
For the planned high energy physics experiments,
additional requirements like radiation hardness
have to be faced, which are more or less unknown
in other elds. Protection against environmental
impacts may be reduced to a minimum in Vertex-
Detectors, causing one to talk about hybridisation
instead of packaging.
Although encapsulation is not needed in Vertex-
Detector environments, the number of hybrids to
be built and installed in the planned high energy
physics experiments are orders of magnitudes
higher than in the existing ones. Here, the
production process and handling issues become
*Corresponding author.
E-mail address: [email protected], gerlach@whep.
uni-wuppertal.de (P. Gerlach).
0168-9002/01/$ - see front matter r 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 8 - 9 0 0 2 ( 0 1 ) 0 1 1 2 8 - 7
crucial for the projects. In addition, the rising frequ-
encies that were used require a dened behaviour of
the connecting lines, meaning that the thickness of
dielectrics and the width of signal lines have to be
very well controlled. This leads to the challenge of
improving all the connections within the module,
the modules intra-connections, to a robust system
with a high quality frequency behaviour.
2. Multi Chip Module techniques
There are two widespread MCM techniques well
known in the industry, the MCM-Ceramic and the
MCM-Laminated. Driven by the rising needs of
telecommunication devices, a third one called
MCM-Deposited is currently reaching a high
volume production state.
2.1. MCM-Ceramic
The MCM-Ceramic (sometimes called MCM-
Coered) provides the most robust modules,
demonstrated by its usage in car-engine environ-
ments. Three types of this technology are used:
Thick Film, High Temperature Coered Ceramic
(HTCC) and Low Temperature Coered Ceramic
(LTCC). As dielectric materials, Al
2
O
3
; AlN or
glass and as conductors, Tg, Cu or Au are used.
The main diculty is the shrinkage of the material
during the ring process, resulting in a dicult
dimensional control and the impossibility of
building ground-planes due to warpage. On the
other hand, more than 50-layer substrates are built
using this technique.
2.2. MCM-Laminated
Here, the layers are built separately and
laminated together. This allows rigid and exible
modules to be built. Enhanced lamination steps
allow through holes, to connect non-adjacent
conductor layers. Copper is used as conductor
and as dielectric materials Kapton
1
and Upilex
2
dominate the market; both are available in a
plurality of mutants. Common to both is the
absorption of moisture in the range of 12 wt%,
resulting in a rising dielectric constant e
r
:
This advanced and exible technique with a
large number of manufacturers will be used in
many planned high energy physics projects, as it
may be found in these proceedings.
2.3. MCM-Deposited
The MCM-Deposited provides the highest con-
nection density by depositing dielectric and con-
ductor materials on a substrate. This is done using
thin lm processing techniques.
Al or Cu is used as conductor material. Copper
may be deposited in two dierent methods. A
subtractive method is to sputter the desired
thickness of the layer. It is structured afterwards
by masked etching. An additive method uses a thin
sputtered seed-layer as plating base. The layer
thickness is achieved by electro-plating the struc-
tures needed. Only the seed-layer has to be etched
afterwards. The nal metallisation of the top layer
may be tuned to the mounting technique. As an
example, for solder bump-bonding it may be Ni:P
with an Au cover.
The three mainly available dielectric materials are
shown in Table 1. Some of them are photo-
sensitive, thus reducing the number of process steps.
As a summary, Table 2 shows a comparison of
the Multi Chip Module techniques.
3. The ATLAS MCM-D Pixel Detector Module
concept
The ATLAS Pixel Detector will consist of two
dierent types of hybrid pixel modules [1].
Following the current plans, the larger part will
Table 1
Dielectric materials used for MCM-Deposited
Material Trade name Manufacturer e
r
Cure temp.
PI Pyralin DuPont 3.3 3504001C
Epoxy Probelec Ciba 4.1 1201C
BCB Cyclotene Dow 2.65 2501C
1
Kapton is a registered trade mark of Dow Chemical.
2
Upilex is a registered trade mark of UBE.
P. Gerlach et al. / Nuclear Instruments and Methods in Physics Research A 473 (2001) 102106 103
consist of an MCM-L based hybrid. Details of this
design may be found in Ref. [2]. The innermost
barrel layer at a radius of 5 cm will be removable,
so handling issues become more important. Taking
this into account, a module concept based on the
MCM-D technology was prototyped.
The main idea of the ATLAS MCM-D Pixel
Detector Module concept is to use the sensor (see
Ref. [11]) as substrate for the thin lm structures.
These structures provide all module intra-connec-
tions. As illustrated in Fig. 1, this allows the
integration of the signal-lines and the power-
supply connections within the module, using
bump-bonding=ip-chip as the only assembly
technique. The signal-lines are placed above a
ground-plane in a micro-strip conguration, guar-
anteeing a dened impedance. All the module
intra-connections are placed on an additional,
insensitive area of the sensor, which is covered
after ip-chipping by the data-buering and
collecting part of the electronic circuits (see Refs.
[3,4,10]). Major requirements for the thin lm
technique are:
*
highest process temperature below 3001C;
*
50 O micro-strip signal lines,
*
low power-supply connection resistance,
*
high inter-connection density.
The rst item is caused by the sensitivity of the
high resistive, oxygen-enriched silicon substrate.
(Above 3001C the generation of thermal donors
starts.) The second and third items are needed to
ensure the proper functionality of the module. The
last item is caused by the need for a connection
between each sensor cell through the thin lms
towards the bump-bond connecting the corre-
sponding electronic cell. These connections, called
feed throughs in the following, have to be built
with a 50 mm pitch in the ATLAS pixel geometry.
Feed throughs give an additional capacitive and
resistive load to the electronics front-end. But they
give the freedom to segmentise the sensor equally,
determined only by the area and the number of
channels provided by the electronics. In addition,
the feed throughs may be used to connect the
readout cells to a sensor segmented in a bricked
way, improving the mean resolution in the long
direction of the pixels.
To achieve the items given above, a process
using Benzocyclobutene (BCB, see Chapter 2:3)
and electro-plated copper was chosen. In colla-
boration with the Fraunhofer Institut f . uur
Zuverl . aassigkeit und Mikrointegration (IZM, Berlin,
Germany) a successful prototype program has
Fig. 1. Schematic cross-section of the four layer bus and signal system. On the right-hand side, the feed through connections from the
detector to the read out chip are illustrated.
Table 2
Exemplary numbers of the dierent Multi Chips Module
techniques
MCM-C MCM-L MCM-D
Max. no. of layers > 50 20 5
Dened waveguides Dicult Possible Easy
Min. line-width=space mm 75=200 75=75 10=20
Min. via-diam.=pitch mm 90=200 50=100 20=50
Max. proc. temp. 1C 8501300 350 250
e
r
710 34 2.65
Moisture absorption (wt%) F 12 0.2
P. Gerlach et al. / Nuclear Instruments and Methods in Physics Research A 473 (2001) 102106 104
been carried out, showing that high performing
modules can be built following this concept.
3.1. Prototyping results
To determine the inuence of the thin lm
structures on the performance of the module, a
comparison has been done with the results of the
measurements of conventional hybrids. Here,
only the results of the testbeam measurements of
the charge collection and the spacial resolution are
shown for the MCM-D hybrids. The results of the
conventional hybrids can be found in Ref. [5] and
detailed information about other measurements
done with MCM-D hybrids in Refs. [69].
3.1.1. Charge collection
The increased coupling capacity between adja-
cent pixels may result in a reduced charge seen by
each electronic cell individually. The charge seen
may be reconstructed using the calibrated pulse
height information given by the pixel unit cell [4].
Selecting the appropriate single and double hit
events of a testbeam measurement, the Landau
distribution can be reconstructed as shown in
Fig. 2. The maxima of these distributions are at
the expected value and not changed compared to
conventional hybrids. The maxima of the single
and the added double hit information is at the
same value, meaning that there is no additional
charge loss for double hits. The ratio between the
number of single and double hits is also not
changed.
3.1.2. Spacial resolution
The position of a hit pixel may be subtracted
from the track position given by the testbeam set-
up. This residual gives the spacial resolution of the
pixel system. In Fig. 3, the resolutions in the short
pixel dimension of 50 mm are shown. The single hit
resolution, shown on the left-hand side, results in a
plateau of 37:6 mm; representing the middle of the
pixel. Here, single hits are most probable. The tails
with a width of 5:2 mm are caused by the
increasing charge sharing between the neighbour-
ing pixels. Therefore, the number of single hits in
this region decreases. For double hits, on the right-
hand side of Fig. 3, the coordinates of the hit
pixels may be weighted with the charge informa-
tion, resulting in a resolution of 5:1 mm: Both the
single and double hit resolutions agree with the
ones measured with conventional hybrids [2,5].
4. Conclusion
The technologies developed in connection with
electronic packaging (as MCM-D) may be adapted
to the needs of Vertex-Detectors. But their specic
requirements are not always covered by industry,
i.e. there is still no exible MCM-L using
aluminium as commercially available conductor.
The upcoming developments driven by the mobile
telecommunication have the potential to integrate
passive components in the modules intra-connec-
tion structures and new dielectric materials, and
large area processing techniques will result in new
ways of building Vertex-Detectors.
Fig. 2. Reconstructed energy deposition seen by an MCM-D Single Chip assembly.
P. Gerlach et al. / Nuclear Instruments and Methods in Physics Research A 473 (2001) 102106 105
Acknowledgements
The prototyping program of MCM-D pixel
detector modules has been done in the context of
the ATLAS Pixel Project (CERN), in close
collaboration with the Fraunhofer Institut f . uur
Zuverl . aassigkeit und Mikrointegration (IZM, Berlin,
Germany).
References
[1] ATLAS Pixel Detector, Technical Design Report,
CERN=LHCC=98-13, (5=1998).
[2] F. Ragusa, Recent Developments in the ATLAS Pixel
Detector, Proceedings of the Eighth International Work-
shop on Vertex Detectors, VERTEX 99, 2025 June 1999,
Texel, Netherlands, Nucl. Instr. and Meth. A 447 (2000)
184.
[3] K. Einsweiler, et al., On the performance and limitations
of a dual threshold discriminator pixel readout circuit for
LHC, Proceedings of the IEEE Nuclear Science Sympo-
sium, Toronto, Canada, 1998.
[4] S.I. Meuser, Pixel Readout Chip for the ATLAS experi-
ment, Proceedings of the IEEE Nuclear Science Sympo-
sium, Toronto, Canada, 1998.
[5] C. Troncon, Detailed studies of the ATLAS Pixel
Detector, Proceedings of the IEEE Nuclear Science
Symposium, October 1999, Seattle, Wash., IEEE Trans.
Nucl. Sci. 47 (2000) 737.
[6] K.-H. Becks, et al., A Multi Chip Module, the basic
building block for large area pixel detectors, Procee-
dings of the IEEE Mutli-Chip Module Conference, 1996,
p. 16.
[7] K.-H. Becks, et al., A MCM-D type module for the
ATLAS Pixel Detector, Proceedings of the IEEE Nuclear
Science Symposium, Toronto, Canada, 1998.
[8] M. T. oopper, et al., Fabrication of a high-density system
using BCB=Cu technology, Proceedings of the Interna-
tional Conference on High Density Packaging and MCMs,
Denver, Colorado, April 1999.
[9] C. Grah, Pixel Detector Modules using MCM-D technol-
ogy, Proceedings of the Pixel 2000, Genova, Italy, 2000
Nucl. Instr. and Meth. A 465 (2001) 211.
[10] R. Beccherle, The Module Controller Chip (MCC) of the
ATLAS Pixel Detector, Proceedings of the IEEE Nuclear
Science Symposium, Toronto, Canada, 1998.
[11] M.S. Alam, et al., The ATLAS silicon sensors, Nucl. Instr.
and Meth. A 456 (2001) 217.
Fig. 3. Residuals expectedFmeasured position of single hits and weighted double hits, with the latter resulting in a resolution of
s 5:1 mm:
P. Gerlach et al. / Nuclear Instruments and Methods in Physics Research A 473 (2001) 102106 106