challenges of the emerging microsystems industry

E L S E V I E R Microelectronics Journal 29 (1998) 587-600

Micro onics

Claallenges of the emerging microsystems industry Per Ohlckers a'b'*, Henrik Jakobsen a

~SensoNor, N-3192, Horten, Norway bUniversity of Oslo, Oslo, Norway

Accepted 18 February 1998

Abstract

The microsystems industry is now a rapid growing industrial sector that has emerged during the three last decades. The onset was mainly a technological spin-off from microelectronicsfintegrated circuit technology. Silicon micromachining gave the main triggering technology push as a very promising process technology with distinctive features. Sensor applications gave the first market opportunities, and batch- organised processing teclhnology adapted from the microelectronics industry the key to high quality at low cost. Today, this has matured into a separate industry sector with their own market and manufacturing infrastructure, also with a limited use of other materials than silicon. The microsystems are used in electronic systems with widespread applications, ranging from low cost, high volume automotive applications to high cost, low volume instrumentation applications. Key technological success and inhibitor factors for this industry are discussed. It is proposed that the most important are market opportunities, microelectronics manufacturing infrastructure and micromachining. In the future, computer-aided design and simulation tools will also gain importance. Examples from research and manufacturing of microsystems, mainly from SensoNor and SINTEF Microsystems in Norway, are used to illustrate these factors. It is proposed that the most important challenge for SensoNor is the transition from being sensor company to becoming a true microsystem company. © 1998 Published by Elsevier Science Ltd.

1. Introduction and motivation: key factors to successful industrial innovation of microsystems

Microsystems are trueborn offsprings of the microelectronics industry, just as its subsector micromachined device technology is a true-born offspring from integrated circuit technology with silicon as the most used material. Earlier, we have discussed key factors to successful innovation micromachined devices [1,2], which are mostly used as sensors and actuators in microelectronics systems, with sensors being the most important. We would like to extend this discussion to microsystems, where most of the same factors are important. Therefore, we suggest the following subjective list:

I. 1. The top 10 success factors of the microsystems industry

1. Market opportunities and market acceptance. Micro- systems are used and can be used in a multitude of applications in many market sectors, ranging from high volume, low cost applications to high-end special products in niche markets.

2. Microelectronics :manufacturing infrastructure offers a wide range of advanced services for the manufacture of

* Author to whom correspondence should be addressed

.

microsystems, e.g. high quality materials, e.g. silicon wafers and photo masks, sophisticated equipment, e.g. aligners and implanters, availability of advanced integrated circuits, and services within electronic packaging and interconnection technology. Without this availability of the microelectronics manufacturing infrastructure, the emergence of the microsystems would probably have been delayed for more than a decade and maybe much more. The most striking example is batch-organised planar processing technology, which offers high quality, low cost, batch-organised ways to manufacture key components of microsystems, e.g. micromachined devices, integrated circuits and hybrid substrates. This includes both subtractive techniques, e.g. etching and additive techniques, e.g. thin film deposition. Lithographics is a very important aspect of this technology as a key to high performance at low cost. Micromachining and wafer bonding. Anisotropic, iso- tropic and selective etching techniques combined with photolithography make high-precision, miniaturised three-dimensional structures for microsystems feasible. This micromachining capability is first of all achievable with single crystal silicon, but both in principle and practice other materials can be used, e.g. gallium arsenide, quarts or polysilicon. To assemble into packaged units, bonding on wafer level or chip level is necessary. The

0026-2692/98/$17.00 © 1998 Published by Elsevier Science Ltd. All rights reserved Pll S0026-2692(98)00022-6

588 P. Ohlckers, 11. Jakobsen/Microelectronics Journal 29 (1998) 587-600

main techniques are silicon-to-silicon bonding, silicon- to-glass bonding and silicon-to-thin film bonding. The techniques can be used for both chips and wafers. Sili- con-to-silicon fusion bonding is a high temperature process giving high strength, used primarily for wafers. Anodic bonding is a low temperature process used both for chips and wafers. Other available processes are glue- ing, soldering and welding.

4. Computer-aided design and simulation tools are making it possible to design new microsystems with a much higher performance and productivity of the design and development work. This has resulted in lower development cost for a given complexity and performance. This has also allowed for development of much more complex microsystems with more predictable and more optimised performances in a shorter time. These tools are so far mostly tools for more general use in the electronic and mechanical domain, but specialised tools for microsystems are now starting to emerge as the suppliers see the market opportunities for their tools. In the future, this factor will be more and more important, justifying the high ranking given here.

5. Foundry services for integrated circuits and microsystems make it possible to enter the market with new microsystems at a lower threshold for the needed in- house manufacturing capabilities. This often cuts the time to market and reduces the cost burden for the manufacture of low volume products. The success of the IC foundries is the obvious model and already in use for the ICs in microsystems, and now true microsystem foundries are emerging offering silicon micromachining, microsystem packaging, etc. Wafer level semiconductor processing, mainly of silicon ICs and silicon micromachined components, is the most obvious example. Electronic packaging and interconnection technology is also a very important aspect of foundry services for microsystems, often giving the main part of the manufacturing costs as well as restricting performance due to packaging artifacts, e.g. packaging stress. Microsystem packaging and interconnection are much more complex and challenging than for ICs, since the packaging scheme must facilitate inter- action with the environment where sensing or actuating operations of the microsystem are taking place.

6. Research results from solid state technology, design of integrated circuits, electronic systems and other related fields of microelectronics, as well as other fields of science and technology, e.g. materials technology, mechanical engineering, chemistry, etc. The most striking example is the accumulated research results of silicon microelectronics, making silicon by far the most investigated electronic material.

7. Sensor and actuator effects. The materials used have a large number of physical effects that can be used for signal conversion for sensors, e.g. piezoresistivity for sensing in the mechanical signal domain and Hall effect for magnetic signal. The technology also facilitates

microscale actuating functions, e.g. micromotors and micropumps.

8. Integrated electronics. Many of the used materials have excellent properties as electronic materials. By combin- ing micromachined devices with on-chip or hybrid integrated signal-conditioning electronics, we get micromachined smart devices. With this approach, improved or new sensing and actuating methods can be used. For instance, silicon capacitive sensors are very hard to make without on-chip or hybrid integrated front-end electronics, because both capacitance and capacitance change are in the picofarad range, which can only be sensed if parasitic capacitances are mini- mised. Often, a two-chip hybrid scheme is prefered as giving a favourable compromise between performance and cost.

9. Mechanical material characteristics. The used materials feature favourable characteristics as mechanical materials. Single crystal silicon is here an excellent example, with its high strength and eminent elasticity, making it an excellent spring element for sensors and actuators.

10. Combination of features. Single crystal silicon is in this respect outstanding, as a sensor or actuator material, and as a material for integrated circuits. E.g. a silicon micromachined smart pressure sensor [3] makes extensive use of these features.

However, microsystems have a much slower development in both performance increase, cost improvements and market impact compared to other fields of microelectronics [4,5]. The example given in Fig. 1, the SINTEF smart capacitive silicon pressure sensing microsystem, was a successful research demonstrator, but was never put into production. However, some of the developed process steps and design principles have been and will probably be used as platforms for new product innovations. This visu- alises that there are many bottlenecks slowing innovation of microsystems. These trends have motivated us to suggest the following list:

1.2. The bottom 10 list of inhibiting factors slowing innovation of microsystems

1. Market reluctance. Microsystems are still not well known in most of the market segments, and users are hesitant to use products with a very short track record which often need to be made application specific. In the last part of the 1990s, this is changing rapidly for the better, resulting in very strong market demands.

2. Slow time-to-market. Microsystems still need extensive time from concept to innovation as products on the market. Less than two years is the exception, while some products have needed more than five years. As technologies mature and more efficient development tools are available, this will improve. However, increased complexity of new products will make short- ening of time-to-market a major obstacle.

P. Ohlckers, H. Jakobsen/Microelectronics

Fig. 1. The SINTEF Smart Capacitive Silicon Pressure Sensing Microsys- tem [3]. The photograph of the backside of the 4 X 4 mm micromachined chip shows the 2 X 2 mm diaphragm of approximately 10 #m thickness, and the centre boss structure of approximately 30/zm thickness. The photograph of the triple stack silicon-to-silicon anodic bonded chip set mounted in a transistor header shows the integrated bipolar signal conditioning circuitry on the top chip. This device is demonstrating the power of a combination of features. This sensor is using many of the listed features. It has a micromachined diaphragm made by batch processing, as shown in the detailed photograph of the diaphragm chip. It uses the variable air gap capacitive sensing principle, has integrated electronics and is assembled by silicon-to-silicon anodic bonding. This demonstrator is not yet commercialised into a product innovation, but some of the developed process steps and design principles have been and will probably be used as platforms for later product developments (photo courtesy of Jan D. Martens, SINTEF).

3. With a few exemptions, most products have low market penetration and are therefore produced in low production volumes, resulting in high costs.

4. Immature industrial infrastructure. Most of the industry is still very young and very fragmented both from a geographical and technological point of view. Strong industrial clusters from other industry, e.g. Silicon Valley in microelectronics, have not yet been established.

5. Complex designs and processes optimised for performance, not cost. Sophisticated designs and high complexity of the production process for most of the

Journal 29 (1998) 587-600 589

microsystems call for large development and manufacturing resources, and long time-to-market from the idea to successful commercial production. Complex products produced in low volumes for a relative short time can also give questionable reliability.

6. Immature processing technologies. Most often, many individual device process steps with an early transition from batch processing are used giving yield problems and high cost.

7. Immature media-compatible packaging and interconnection technologies. Media compatibility, or interfacing of the device with the medium it senses or actuates on, is most often difficult and expensive since several individual device packaging fabrication steps are needed. For example, pressure sensing microsystems for applications in hostile environments are often using a housing with media separation by a welded metal diaphragm, which requires costly parts and expensive, individual device processing.

8. Immature microsystems integration methods to design microsystem with all the components optimised, compatible and working together to achieve the targeted functionality.

9. Limited resources. The industry is still small with limited commercial success. Therefore, resources, human, technological and financial, are inhibiting the growth of the industry. For example, the industry has only a limited number of seasoned first-class experts, and the education system has only a few exemptions yet to recognise to the full extent the need to train students in the field of micromachined devices.

10. High costs. Generally, as a consequence of inhibiting factors, e.g. those mentioned above, the industry is generally still manufacturing high-cost products. This situation gives a negative feedback to those same factors, e.g. slows down the market acceptance of the products.

The performance versus cost is a key issue for market penetration for microsystems. With competitive performance versus cost ratio, microsystems will both replace traditional instrumentation systems and open up new applications.

It is a striking resemblance between the microelectronics industry of the 1970s and the microsystems of the 1990s. Just as the microelectronics industry was able to push its technologies and infrastructure to a high market acceptance, it is most likely that the microsystems industry will have a similar success story, making it one of the fastest growing industry sectors in the next decade. The market is there, with an anticipated size of 35 billion dollars by the year 2002 [6] Fig. 2. This is both a challenge and a huge opportunity, and there is probably much to leam from the success story of the microelectronics industry when the large scale build-up of the microsystems industry has started.

590 P. Ohlckers, H. Jakobsen/Microelectronics Journal 29 (1998) 587-600

World microsgstems market

MS 35000

30000

25000

20000

15000

10000

5000

0

1996

D World emerging microsystems market

[] World existing microsystems markets: Read/write heads for HDD

• World existing microeystems markets: Ink jet heads

• World existing microsystems markets: Pressure sensors

• World existing microsystems market: other devices

Fig. 2. The NEXUS forecast of the microsystems market [6]. (NEXUS: Network of excellence of multifunctional microsystems, an EU/ESPRIT sponsored European collaboration.)

2. Highlight examples of the microsystem research at SINTEF

SINTEF Microsystems is a department of SINTEF, a large Norwegian independent multidisciplinary research institute organisation located in Oslo and Trondheim. SIN- TEF conducts contract research and development for industry and the public sector in technological areas, and in the natural and social sciences. With 2000 employees and a turnover of 170 MECU, the SINTEF Group is Scandina- via's largest independent research organisation. Contracts for industry and the public sector account for more than 90% of operating revenues. The main part of the microsystem research activities took place at SINTEF in Oslo, known as SI (Center for Industrial Research), before a merger with the SINTEF Group in 1993. Since 1960, there has been a research group at SUSINTEF which developed silicon sensors and actuators, and lately also microsystems. The activities within this research group are focused on the following main application areas: Micro Electro-Mechanical Systems (MEMS), Micro Opto-Mechanical Systems (MOMS), Silicon radiation sensors, Application-specific integrated sensors, and packaging and interconnection technology. After a reorganisation in 1995, these activities were split into departments: Department of Microsystems and Department of Microelectronics. Together, these two departments have approximately 15 research scientists and a production staff of five involved with microsystems research and prototyping. The Department of Microsystems has a wafer line for research and prototyping of silicon sensors. Yearly throughput is 3000-5000 wafers.

The best way to describe the story of microsystem research at SFSINTEF is to use highlight examples given in the following chapters.

2.1. The SI silicon cantilever beam sensor element

This was the first demonstrator of the applied research

done on silicon sensor technology at SI in the early 1960s. The work on silicon piezoresistance was started in March 1961 with the scientists Jan Barstad and Halle Sandmo as key members of the team, to be joined by Olaf Stavik and Fin Serck-Hanssen later that year. Their team laid the technology platform using planar technology to make piezoresistive devices. This work stimulated the invention of the SI Silicon Cantilever Beam Sensor Element by another scientist on the team, Odd Eriksen. His patent application [7] for this device was submitted in 1965, and was accepted in several countries in Europe and USA. The patent claim was the combination of mechanical support of the 5 mm x 1 mm x 100/zm silicon cantilever beam and electrical contact to the two piezoresistors integrated in the beam by the use of a glass header with four contact pins. The principle and a photograph are shown in Fig. 3.

This sensor element was later commercialised under the family name 'The AE-800 Series Sensor Elements' by the company Akers Electronics, which was started in 1965 to capitalise on the research results at SI in planar and hybrid technology. Later on the company name was changed several times, and was called AME when their Transducer Division was spun off in 1985 as the new company Senso- Nor. The AE-800 Series Sensor Elements was then a key product, and remarkably it is still in production at SensoNor after more than 30 years since the invention, with only minor changes in the original design. This also is an example that good silicon sensor products can stay in the market for a very long time.

2.2. The S l physiological pressure sensor

The AE-800 Series Sensor Elements could be used in a multitude of sensors, provided the measurand could be converted to a bending of the cantilever beam and corresponding piezoresistive response. One very nice product invented by Odd Eriksen and Fin Serck-Hanssen at SI was the SI Physiological Pressure Sensor, patented

P. Ohlckers, H. JakobserdMicroelectronics Journal 29 (1998) 587-600 591

Fig. 3. The SI Silicon Cantilever Beam Sensor Element: the AE 800 Series Sensor Elements. The drawing shows the principle of operation, and a photograph of the device, wtfich is delivered in a protective glass tube.

in 1972 and immidiately put in production by AME under the name AE 840, later to be changed to the name SensoNor 840 [8]. Pressure is sensed by a lathe-machined nickel- copper alloy diaphragm and converted to a proportional deflection of a few microns at the centre of the diaphragm, which acts on the cantilever of an AE 800 element. However, the main patent claim was the use of a disposable plastic dome with a flexible separation diaphragm, making it possible to biologically separate the sensing media from the multiusable transducer. In this way, a sterile setup was ensured by using pre-sterilised, low cost disposable plastic domes, while the metal-based transducer housing did not need to be sterilised prior to each patient set-up (Fig. 4).It is worth mentionioning that SI with most of their activities on silicon sensors in these pioneering years went all the way from the basics to actually making commercial products either sold directly from SI or produced as subcontractor to Akers Electronics/AME. This is in contradiction to the generally accepted view that insti- tutes should not make products. In hindsight, it looks like the SI tactic in this respect was the right approach to com- mercialise their research results. Most likely, this tactic was chosen because no companies were found that were willing to venture into this new technology. Later on, this tactic has not been used so extensively, but examples still exist at SINTEF today, e.g. they are offering small scale production of silicon radiation sensors based upon the reverse-biased pin diode principle [9], and making a high speed silicon micromachined infrared radiation source [10].

Fig. 4. The AE 840 Physiological Pressure Sensor showing the disposable plastic dome and metal-based electronic sensor housing having a pressure sensitive diaphragm, and using an AE 800 element as transducer element.

2.3. The SI silicon micromachined pressure sensor

In 1974, research scientist Knut Asskildt at SI started up an applied research activity on a piezoresistive silicon micromachined pressure sensor, motivated by research work and products from Fairchild and National Semicon- ductors in the USA. His work was sponsored through a two- year fellowship, and he succeeded in making working prototypes. These successful prototypes motivated AME to start a product development project, which resulted in the AE880 Pressure Sensor, later to be upgraded to SP80/81 Pressure Sensor, which is still in production. More details on the AE880 and SP80/81 are given later.

2.4. Other highlight examples f rom SI/SINTEF

Later on, the research group at SUSINTEF reported several excellent research results with a general high academic quality, and in some cases with a good impact on commercial innovation. The SI Silicon-to-Silicon Anodic Bonding Process [11,12] is a good example of excellent academic work, which still has not been commercialised to any large extent. The work on silicon radiation sensors and microsystems is also excellent research work, which is now partly commercialised in the way that the work on front end electronics is innovated by the start-up company IDE-Integrated Detector and Electronics, mainly for solid state-based X-ray imaging systems. The SI double-sided microstrip radiation sensing microsystem made by the combined efforts by the sensor scientist and the front end electronics was definitely state-of-the-art when published in 1992 [9,10], but with the start up of IDE by key front end electronics scientists at SINTEF, the remaining activity on radition sensor front- end electronics at SINTEF soon withered into more or less extinction (Fig. 5).

In 1994, three SINTEF researchers patented [13] a scheme for high pressure silicon sensors using the Bourdon principle in a micromachined pressure sensor to circumvent


2.5. General comments on the SI/SINTEF activities

Fig. 5. Customer-specified radiation sensing microsystem for measuring high energy elementary particles. The microsystem consists of a microstrip radiation sensor and integrated circuit chips for front-end signal conditioning (photo courtesy of Jan D. Martens, SINTEF).

the limitations of diaphragm pressure sensors at high pressure sensor (several hundred Bars and more, e.g. 2000 Bars full scale pressure). This idea was later improved with a pending patent by two other scientists at SINTEF, and they are now commercialising their concept in the start- up company PreSens.

In 1995, three SINTEF researchers patented [14] a scheme for gas sensors using silicon micromachining to make photoacoustic gas sensors. The principle was followed up by another patent using an enclosed cavity with the gas type to be measured as a kind of reference. It works in the way that if this type of gas is present in the optical path from the chopped infrared radiation source to the silicon micromachined sensing unit with the closed gas cavity, a propor- tionally less photoacoustic response is given due to the transmission loss. Results from a working prototype have been published in Ref. [15], and the principle for the prototype is shown in Fig. 6, with the photoacoustic response shown as the oscilloscope display shown in Fig. 7.

The electronic chopped silicon micromachined IR source is also a SINTEF invention, already in widespread use in the gas sensing systems delivered by the Norwegian instrumentation company Simrad Optotronics, presented in Ref. [9].

Looking back, it is the subjective view of these authors that the microsystems research activities at SINTEF have a very long history with many excellent research results. However, the activities have always struggled with marginal resources with a small staff and very limited equipment resources. This is partly because microsystems research needs a level of funding which is larger than is ordinarily available in Norway as a small country, partly because the Norwegian research community up till now has not fully realised the potential of microsystems for industrial innovation. This also leads to the fact that microsystems as a relative small activity in the SINTEF Group had difficulties in obtaining sufficient committments and backing from top management. The good news is that the backing from SIN- TEF top management and funding agencies has improved during the last years as the microsystems field has come of age in Norway as well as the rest of the world as a growing and promising business field.

In the pioneering years of the 1960s, silicon piezoresistance was ventured with the SI Silicon Cantilever Sensor Element and transducers using this element were successfully developed. The research was motivated by the work of Chuck Smith of Bell Labs, and they very successfully extended the work to innovative application devices later to be commercialised by the spin-off company Akers Elec- tronics, which was founded in 1965. However, the main focus of the research group was integrated circuit technology and hybrid technology, giving an activity level of less than five man-years per year on silicon piezoresistive technology and devices. SINTEF are continuing their technology push, and one should expect further future industria- lisation by existing or start-up companies. For instance, the start up company PreSens is exploiting concepts from a SINTEF patent [13] on silicon micromachined pressure sensing technology for high pressure of more than 1000 bars, and their technology for microstrip radiation sensors [9] based upon the reverse-biased PIN diode principle is very promising for commercial innovation. However, i n hindsight, the microsystems community of Norway should regret that their very strong position as pioneers in

Modula ted I R source

I R ~

Fig. 6. Schematic drawing showing the principle operation of the SINTEF novel photoacoustic sensor concept using silicon micromachined electronic chopped IR source and a silicon micromachined pressure sensor for sensing the photoacoustic signal.

P. Ohlckers, H. Jakobsen/Microelectronics Journal 29 (1998) 587-600 593

> E

e~ E E

a

>

8

4O0

3'20

240

160

80

o

-2'0

I Iq.sctirce ~upla~ vdl~e '

radlation" j \ 11. /' / --PA-~ /' ,,/\ ,//

/ ,t \ ' , / /

6 ~o ~ ~o ~o loo12o14o1~1~o time Ires]

Fig. 7. Supply voltage for IR source, monitored IR-radiation (dashed lines), photoacoustic response (PA) ~md background noise signals (solid lines) for the early prototype photoaconstic sensor. Note that the supply voltage and monitored IR radiation (dashed lines) are not in scale. In this test, N20 was used as the gas to be monitored.

piezoresistive silicon sensor technology only resulted in limited industrial innovation until SensoNor started their strong efforts in the late 1980s.

3. Highlight examples of the microsystem activities at SensoNor

SensoNor is a silicon microsystem manufacturer, located in Borre in Norway. The company is a European market leader in design, development and production of electronic air bag sensors for use in the automotive industry. The company had a turnover in 1996 of 25 MECU. Silicon sensor and microsystem technology with focus on silicon micromachining is the core technology of the company. State-of-the-art in silicon micromachined crash sensors is represented by the SA20 device from SensoNor, which is the undisputed world leader in this field with over 23 million of these devices sold ahnost exclusively into the automobile industry. The success of the SA20 is based on the market- driven strategy of Sen,;oNor. Research on micromachined pressure sensors started in 1976 and on micromachined accelerometers in 1982, and SensoNor was founded as a company in 1985 to commercially exploit this work. In the meantime, SensoNor has grown to 300 people, including 85 graduate engineers, and has developed and manufactured a range of silicon-based sensors and microsystems. About 85% of the production is crash sensors and 10% goes into medical and aerospace application. A strong feature of the company is that it performs its own plastic packaging in Norway and keeps the cost down using a high level of automation. SensoNor has been qualified according to quality assurance certificates of ISO 9001, NATO AQAP 110 and Ford Q 1.

We would like to give some highlight examples of the microsystem innovations at SensoNor. The AE 800 Series Silicon Sensor Element has been mentioned earlier. In hindsight, this product family was indeed a very important

cornerstone for the development of SensoNor as a major microsystem company, as this element was a key success factor through the Akers Electronics/AME years and the first years of SensoNor as a separate company. The product family was and is partly sold as a sensor component to transducer manufacturers, and partly used and in use in SensoNor products, e.g. the AE 840.

3.1. The AE 880 and SP80/81 silicon micromachined pressure sensor

The product development started up at AME, with Per Ohlckers as key scientist [16,17]. The project had a very important technological impact: this was the introduction of silicon micromachining as a technology platform for AME. The processes and prototypes developed by SI were an excellent starting point, but performance was still not good enough and manufacturing schemes needed to be developed. After a slow start up, major breakthroughs were made in 1979, showing that it was possible to make sensors with good linearity, high thermal zero shift stability, and excellent long term stability, and in 1980 the first units were sold. Later on, upgrading of the silicon technology by using buried, high stability piezoresistors was implemented with Henrik Jakobsen as key scientist. This product family is now selling very well in the high end market, with aerospace applications as the dominating market sector (Figs. 8- 10). (At that time, AME was mainly a hybrid and system manufacturer with a major project on a portable radio trans- mitter for defence applications.)

3.2. The SA20 Crash Sensor

The SA20 Crash Sensor is a major innovation with its impact on air bag-based safety systems for automotives. The product was launched in 1992, with a very innovative low cost plastic packaging scheme based upon well known piezoresistive and silicon micromachining technology. The crash sensor is a 50 g full scale range accelerometer,

Implanted piezoresietive resistor Thln film aluminium conductor

Square diapl

Pn

Fig. 8. A cross-sectioned view of the chip set for the AE 880 Pressure Sensor. The SP80/81 is an upgrade of this design and corresponding silicon process technology, with buried piezoresistors as the main improvement.


Reference Pressure Port

Cap

t Transistor Header

Connecting Pins

Pressure

Pressure Connection Tubing

Fig. 9. Cross-sectioned view of the SP80 Pressure Sensor packaged in a transistor header. Please note that the assembly of the chip set on the top of the glass tubing ensures excellent decoupling of packaging stress from the header, as well as good thermal isolation, making chip thermostatting a feature to optimise stability.

with a ceramic mass as inertial mass glued on a silicon micromachined cantilever beam with piezoresistors, mounted in a silicone oil-filled cavity in a thermoplastic package for environmental protection and viscous damping [19]. The SA20 Crash Sensor is shown in Fig. 11.

The first main technological obstacle to make this product was the scheme with a sealed cavity with the silicone oil for viscous damping. The next main obstacle was to establish a high volume packaging and assembly line producing millions per year at low cost and high yield. SensoNor struggled hard in the first years to solve those and other main obstacles and many lesser ones. Today, the success of this work is well known, with an annual production of approximately 6-7 million units, meeting targeted cost and

Fig. 10. A micrograph of the SP80/81 chip, taken during deflection measurements with electronic speckle pattern interferometry at 0.25 bar [18]. The white lines are of equal height, with a height separation of 0.4 #m.

quality goals. At peak production years in 1996 and most of 1997, production was run in five shifts around the clock 24 hours a day seven days a week, with a new crash sensor leaving the line approximately every 3 s. With this product, SensoNor captured approximately 70% of the non-captive European market for electronic crash sensors in 1996.

However, this product is a sensor and not a microsystem. In response to market changes requiring the performance, cost and quality improvement potentials of using microsystems, SensoNor is now developing a new generation of crash sensors, of which the SA30 Crash Sensor will be described later. However, in hindsight, it was remarkable perfect timing for the product towards the emerging air bag safety system market, and SA20 also played its role as an important catalyst for the making of this market.

3.3. The SP13 Tyre Pressure Sensing Microsystem

In the autumn of 1997, SensoNor released their SP13 Tyre Pressure Sensing Microsystem. This product is capi- talising on the technology platform of the SP80/81 Pressure Sensor. The two-chip hybrid integration of the sensor die and the application specific front end electronics were the main design challenges, and lead frame-based low cost epoxy transfer-moulded packaging was the main production technology challenge. SensoNor, in response to several new products that were in development and expected to come into development, decided to build a flexible assembly line with the capability to produce many different products with a short configuration time of less than 1 h when changing the product to be produced on the line. The developed and commissioned assembly line used standard IC assembly machines as much as possible to capitalise on the strong infrastructure of this packaging technology, and, in addition,

P. Ohlckers, H. Jakobsen/Microelectronics Journal 29 (1998) 587-600 595

Fig. 11. The SA20 Crash Sensor in a cutaway view: the interior sensor element shown mounted inside the plastic package.

microsystem-specific assembly and test machines were developed and incorporated in collaboration with equipment manufacturers. The line has the capacity to package several tens of millions of microsystems per year, and bottlenecks can be cleared by duplication of equipment. A photograph of the lead frame mounted chip set is shown in Fig. 12.

Tyre pressure sensing in automotives is still only an option in heavy duty trucks and luxury cars, but SensoNor and their customers are with this product and the corresponding system,; in a very favourable position to capitalise on the market opportunities if a similar growth scenario as for crash sensors comes true in the coming years.

As a part of the development of the leadframe-epoxy mould packaging technology, research is done on the origin and effect of packaging stress on the packaged sensing microsystem. In Ref. [20], it is shown that the epoxy transfer mouldiing is setting up severe packaging stress in SP13 prototypes, with increased thermal zero shift and reduced long term stability as a result. By intro- ducing glob topping of the chip set with low-viscosity silicone mould, which has a decoupling effect from the epoxy, packaging stress was reduced to an acceptable level. The principle and effect are shown in Figs. 13 and 14.

3.4. New microsystem products f rom SensoNor

Among the future microsystems under development by SensoNor, we would like to highlight the SA30 Crash Sen- sing Microsystem [21].

This accelerometer microsystem is optimised to be a low cost crash sensor for airbag systems, with versions for fron- tal impacts and side impacts. SA30 is a two-chip solution with the sensor on one chip and signal conditioning on another chip, packaged in a small outline package for sur- face mount. Mounting on PCB can be normal or horizontal with respect to the sensitive direction. SA30 is completely self-contained with no need for any additional components or trimming (Fig. 15).

The output signal can be either PWM (Pulse-Width- Modulated) for innovative system designs, with respect to noise, EMI or A/D conversion, or analog (ratiometric) for traditional interfacing. A threshold signal is available for designers to improve the system reliability and performance. Due to the intrinsic continuous self-test of the sensor, monitoring of a status signal is all that is needed to check the reliability of the sensor signal. This solution makes it possible to check the sensor at all times, even during a crash without interruptions and loss of information. SA30 utilises a small single crystal silicon resonator, which shifts its frequency due to change in acceleration. The


Fig. 12. A photograph of the lead frame mounted chip set of the SPI 3 Tyre Pressure Sensing Microsystem, showing the pressure sensor chip in the upper left of the picture and the front end chip in the lower right. Interconnection is done by gold wire bonding from chip to chip and chip to lead frame, using no interconnect substrate.

innovative concept for self-test has been made possible due to the development of a new sensing principle. Since all parts of the resonator (both 'springs' and 'mass') are in flexure during operation, the consistency of the function can be monitored continuously.

3.5. Foundry services from SensoNor

SensoNor will, effective from the fourth quarter of 1997, start foundry services in silicon microsystems, offering a manufacturing infrastructure focused on microsystems based upon bulk silicon micromachining technology and low cost plastic transfer-moulded packages [22]. As an important part of this

a)

foundry service, we have teamed up with associated design centres in the Nordic countries and Spain, making up the NORMIC consortium approved and sponsored by the Eur- opean Commission as a Europractice Manufacturing Cluster under the ESPRIT IV Programme ofDG lII. The partners are: SensoNor (N), VTI" Electronics (FIN), MIC (DK), SINTEF (N), CNM/D + T (E) and IMC (S).

The cluster will offer a complete service from design to mass production of microsystems based on the following service tasks to be provided:

1. Feasibility studies; 2. Customer training on the specific technologies offered by

the cluster;

v0

b) Fig. 13. (a) Schematic of pressure sensor placed in moulded package. (b) Wheatstone bridge structure.

P. Ohlckers, H. Jakobsen/Microelectronics Journal 29 (1998) 587-600

Thermal zero shift

597

3

2

1

-4 T (C)

Fig. 14. Thermal zero shift for sample with ($2) and without glob top (S 1). The improved thermal stability with the use of glob top is due to the packaging stress decoupling effect of the low viscosity silicone mould glob top, and this is also confirmed by direct stress measurements with the use of the piezoresistors.

3. Design support for the electronic parts of the microsystems;

4. Microsystems design from customer specifications; 5. Prototyping (including MPW) and volume production in

processes compatible with the core technologies of the cluster;

6. Assembly and packaging; 7. Test and qualification services.

The cluster will focus its services on silicon bulk micromachining and microsystem packaging.

The manufacturing cluster will be established with Sen- soNor as the manufacturing site, and five sites offering design support and customer interface, including principle study, feasibility evaluation, process evaluation/transfer and manufacturability studies. CNM/D + T will develop the customer interface and support. Sintef will develop common design rules for bulk micromachining. MIC will through subcontractor DELTA develop testing and reliability interface and support. IMC will develop and provide MPW interface and support. Assembly, packaging and electronic control circuit design support will be provided by VTT. The effort is divided into six work packages and a number of tasks relate to each work package. The focus of all tasks is to lower the risk and bring down the entry costs of microsystems manufacturing by, e.g. offering standardised design rules and MPW services.

Links to other EUROPRACTICE Basic Services (e.g. IC, MCM and MST manufacturing, Software Support and Training and Best Practice) are proposed to get a consistent service.

Fig. 15. The chip set and exploded view of the assembled transfer moulded packaged SA30 Crash Sensor.

4. Discuss ion

Success stories occur when there is a unique combination of technology push and market pull in combination with


Fig. 16. The SensoNor learning curve. It all started in the early 1960s with the invention of the AE 800 series Transducer Elements, with an evolutionary development through innovative products to the microsystem activities of SensoNor today. In the early pioneering years, the main activities were first at SI as research and prototyping activities, which were gradually transferred to Akers Electronics after the foundation of the company in 1965, later to be renamed AME. At AME, the sensor activities were organised in the Sensor Department, which later on was split in 1985 into the separate company SensoNor. The first years can be characterised by a strong technology push, later to be more and more dominated by a strong market pull, with automotive applications as the main market. In the later years, microsystem products are starting to dominate over sensor products.

individuals with a strong will being able to cope with all the obstacles blocking the road to innovation. SensoNor is now in a position to become a successful company in the field of microsystems. We would like to use the experiences and expectations of SensoNor to show the importance of some of the success factors and bottlenecks given earlier in this

review. In Fig. 16, we have tried to visualise what we call the

SensoNor learning curve, starting from the simple and innovative AE 800 Silicon Transducers Element in the early 1960s, to the complex microsystems under development in the late 1990s, e.g. the SA 30 Crash Sensor and the SP 13 Tyre Pressure Sensor. In between, the AE 80/SP 80 Silicon Micromachined Pressure Sensor family launched in the early 1980s was a major technological breakthrough as the first micromachined products. The SA 20 Crash Sensor put on the market in the early 1980s was probably the most important product as the first to be produced in quantities of millions per year, designed for low cost using highly auto- mated and cost-effective assembly techniques.

In the first years both at SI and AME, the sensor activities

were conducted as a small parallel activity compared to the mainstream microelectronics activities of the research institute and industrialising company. In this way, the sensor activity capitalised on the research results manufacturing infrastructure of microelectronics globally in general and also internally. It is our subjective observation that this was very important in order to lay the foundation for later

activities, even though the roles outside the mainstream priorities sometimes were very frustrating and slowed the progress. At that time, hybrid and monolithic microelectronics were regarded as very promising, and the excellence of silicon as a sensor and actuator material was not yet fully realised. For instance, success factors, e.g. micromachining, and the strength of combination of feaures, which later led to the introduction of microsystems, had not come of age. This may partly explain why the sensor activities were given relatively low attention by top management. However, in hindsight, if the technological and commercial opportunities had been more aggressively innovated, an earlier establish- ment of a strong Norwegian sensor and microsystems cluster could have been the result.

P. Ohlckers, H. Jakobsen/Microelectronics Journal 29 (1998) 587-600

On the other hand, the earlier mentioned bottlenecks, e.g. slow market acceptance, immature industrial infrastructure, immature processing technologies, and immature packaging and interconnection technologies were casting their shadows, bogging down the activities and scaring away entrepreneurs.

With the foundation of SensoNor in 1985, the company could concentrate on the innovation of silicon sensor technology. The entrepreneural spirit of the company put up a better position for the innovations that led to growth of the company in the later years. The strategy was to use the existing technology platform, so far used for small volume products, to venture into high volume, low cost markets. The main market breakthrough was done in the automotive market with the first generation crash sensor, the $64A crash sensor, which was ba,;ed upon the AE 800 element. How- ever, just as the dewflopment and commissioning of the manufacturing line was finished, the main customer pulled out due to problems on their side with the system electronics, leaving SensoNor stranded and unable to capitalise on the product investments. This happened late in the 1980s, and resulted in very troubled times for the company. One way to put it is that some of the previous bottlenecks con- cerning technology and market immaturity played a more dominant role than the success factors.

However, thanks to the team spirit mentioned earlier, the company was able to make a remarkable comeback with the release of the second generation crash sensor in 1992, the SA20 Crash Sensor. The management was able to attract investors and motivate the development team to make the SA20 a successful innovation. This was and is a team effort of the management, owners and all the employees of the company. Without forgetting this team effort, the management should be honoured for their skills and stamina in the making of the successful innovation.

At time of writing in the autumn of 1997, SensoNor is working hard with the development of several new products, of which most of them are microsystems. This is a very challenging time for tile company, as the new microsystems call for added skills as silicon microsystems manufacturer in addition to the established skills as silicon sensor manufacturer. A strong build-up of competence of expertise in the fields focusing on the microsystems aspects is needed and is being implemented. Most noteable is the establish- ment of a new group for the design of the application- specific front-end circuits, and a simulation group to cope with the predictions and modifications of the functionalities of the complete microsystems. The most challenging part is to get the different development and manufacturing teams to put their acts together and blend their expertise into making functional and competitive microsystems for the future. If the company can get the microsystem technologies to work at competitive costs, it is a very strong market worldwide which can give the company a very strong growth with profitability in the future years. Hopefully, the company will be able to capitalise on the previous mentioned success

599

factors and its competitive edge, without being bogged down by the bottlenecks, which are still present.

5. Conclusions

In this paper, we have proposed success factors and bottlenecks for the emerging microsystems industry, built upon an earlier proposed concept for the industry of micromachined devices. It is proposed that the most important are market opportunities, microelectronics manufacturing infrastructure and micromachining. In the future, computer- aided design and simulation tools will also gain importance. These success factors will propably continue to be corner- stones for the microsystems industry in the near future. We have reported on the emergence of the Norwegian microsystems activities, and discussed the relevance of some of the given success factors and bottlenecks for this regional sector of the microsystems industry. Focus has been given to SUSINTEF as the main technological pioneer in Norway, and to SensoNor as the main industrial company with successful innovations, e.g. the AE 800 Transducer Element, the AE 880/SP80/81 Pressure Sensor and the SA20 Crash Sensor. Promising future innovations are also described: the SP13 Tyre Pressure Sensor and the SA30 Crash Sensors, together with the commissioned flexible manufacturing line for microsystems. The transition of the company from being a silicon sensor company to become a true silicon microsystem company is the main challenge for the future.

References

[1] P. Ohlckers, A. Hanneborg, M. Nese, Batch processing for micromachined devices, Technical Digest of the MME '94, Micromecha- nics Europe 1994, Pisa, 5-6 September, pp. 2-16. Appeared as an invited paper in Journal of Micromechanics and Microengineering 5 (1995) 47-56. 2. S. Middelhoek, Quo vadis silicon sensors? Sensors and Actuators 8 (1985) 39-48.

[2] P. Ohlckers, The emerging industry of micromachined devices, Proceedings of 3rd International Conference on Manufacturing Tech- nology, Hong Kong, 13-16 December 1995, pp. 52-70.

[3] A. Hanneborg, T.E. Hansen, P.A. Ohlckers, E. Carlson, B. Dahl, O. Holwech, An integrated capacitive pressure sensor with frequency- modulated output, Sensors and Actuators 9 (1986) 345-351.

[4] S. Middelhoek, A.C. Hoogerwerf, Smart sensors: when and where, Sensors and Actuators A 41-42 (1994) 1-8.

[5] J.E. Brignell, Quo vadis smart sensors? Sensors and Actuators A 37- 38 (1993) 6-8.

[6] R. Wechsung, J.C. Eloy, Market analysis for microsystems--an interim report from the NEXUS task force, Proceedings of Eurosen- sors XI, Warzaw, Poland, 21-24 September 1997, pp. 519-526.

[7] O. Eriksen, Electromechanical transducer for stress, pressure and acceleration measurements. (in Norwegian), Norwegian Patent no. 115502. Submitted 6 June 1965, accepted 1 January 1969.

[8] F. Serck-Hanssen, Sensors--new Norwegian export products? (in Norwegian), Article in the Norwegian periodical named Teknisk Ukeblad 50, 1972.

[9] P. Ohlckers, B. Sundby Avset, A. Bjorneklett, L. Evensen, J. Gakkes- tad, A. Hanneborg, T. Hansen, A. Kjensmo, E. Kristiansen,


H. Kristiansen, H. von der Lippe, M. Nese, E. Nyg~rd, F. Serck- Hanssen, O. S~rhsen, Industrial microelectronics toward the year 2000--a report on objectives and results of a research collaboration at the Center for Industrial Research and the University of Oslo, Invited paper, ISHM--Nordic 30th Conference, Oslo, 20-23 September 1992. Appeared in Hybrid Circuits No. 32, September 1993, pp. 4-11.

[10] P. Ohlckers, T. Gleditsch, A. Hanneborg, H. v.d. Lippe, Microsystems for instrumentation--presentation of SINTEF microelectronics, Oslo, Norway, VDI/VDE mst news,'.Berlin, No. 11 (1994) 3-5.

[11] A. Hanneborg, P. Ohlckers, Silicon-to-silicon bonding using sputtered borosilicate glass, Proceedings, Micro Mechanics Europe 1990, Berlin, 26-27 November.

[12] A. Hanneborg, M. Nese, P. Ohlckers, Silicon-to-silicon anodic bonding, Journal of Micromechanical Engineering 1 (1991) 139-144.

[13] A. Hanneborg, P. Ohlckers, F. Serck-Hanssen, Pressure sensor (Trykksensor), Norwegian patent application. Oslo, June 1994. Accepted December 1994.

[14] R. Bernstein, A. Ferber, P. Ohlckers, Gas sensor (Gassdetektor), Nor- wegian patent application submitted in Oslo, January 1995. Norwe- gian patent approved in 1997.

[15] R.W. Bernstein, A. Ferber, L. Furuberg, S.T. Moe, Development of a wavelength selective silicon infrared detector for gas sensor applications, Presentation at Sensor '97, Nurnberg, Germany, 13-15 May 1997.

[16] P. Ohlckers, Kombinasjon av mikroelektronikk og mikromekanikk-en sterk teknologi (Combination of microelectronics and micromechanics-an excellent technology), Paper in Norwegian in ELEK- TRO 2/1981, IngeniOffoflagets Pressebyr~.

[17] L. Halbo, P. Ohlckers, Electronic Components, Packaging and Production, University of Oslo, December 1995, 330 pp.

[18] Yet unpublished results from work by P. Ohlckers, B. Erik Seeberg, A. Larsen.

[ 19] Data sheet: SA30 Crash Sensor; SensoNor asa, P.O. Box 196, N-3192 Horten, Norway.

[20] J.B. Nys~ether, A. Larsen, B. Livered, P. Ohlckers, Measurement of packaged induced stress and thermal zero shift in transfer molded silicon piezoresistive pressure sensors, Technical Digest, MME '97, Micromechanics Europe 1997, Southampton, 1-2 September, pp. 219-222.

[21] P. Ohlckers, R. Holm, H. Jakobsen, T. Kvisteroy, G. Kitfilsland, M. Nese, S.M. Nilsen, A. Ferber, An integrated resonant accelerometer microsystem for automotive applications, Submitted paper presented at Transducers '97, Chicago, USA, 16-19 June 1997, pp. 843-846.

[22] P. Ohlckers, S.-I. Hansen, H. Jakobsen, The emerging industry of micromachined devices, Invited presentation at the NORCHIP Con- ference, Tallinn, Eustonia, 10-11 November 1997. Included in the proceedings from the conference.

challenges of the emerging microsystems industry

Documents