Ghodssi R., Lin P., MEMS Materials and Processes Handbook

Подождите немного. Документ загружается.

14 T.L. Lamers and B.L. Pruitt

Fig. 1.6 QFD Phase I application of piezoelectric acoustic sensor to laptops (Modiﬁed ﬁgure,

used with permission of ASME. First appearing in Lamers and Fazzio [50])

Concept Screening (CS)

The output of QFD Phase I is used as input for CS. The intent of CS is to screen

design ideas, providing the development team insight on which concepts are worth

pursuing. CS is structured as a matrix, with design concepts comprising the rows

and the most critical engineering metrics from Phase I in the columns. Each concept

is evaluated in comparison to the high-importance engineering metrics.

The concept screening mechanics are detailed for application to readers’ projects.

Refer to Figs. 1.8, 1.9, and 1.10 for the Avago examples of the CS matrix format.

To perform CS, follow these steps.

1. Analyze the QFD Phase I results to select the three, four, or ﬁve engineering

metrics with the highest weightings. Only the three to ﬁve most important engi-

neering metrics from Phase I are utilized in CS to make the tool fast and easy

to use. The decision of whether to use three, four, or ﬁve engineering metrics is

driven by the location of a natural breakpoint in the importance weightings of the

engineering metrics. For example, if the percentages for the top ﬁve engineering

metrics proceeded in decreasing order: 19, 15, 13, 12, and 8%, there is a natural

breakpoint between the fourth and ﬁfth most important metrics. In this case, the

1 The MEMS Design Process 15

Fig. 1.7 QFD Phase I application of piezoelectric acoustic sensor to automotive (Modiﬁed ﬁgure,

used with permission of ASME. First appearing in Lamers and Fazzio [50])

top four engineering metrics would be used in CS. If no such breakpoint is evi-

dent, use the top ﬁve engineering metrics. Put the chosen engineering metrics in

the top row of the CS matrix.

2. Use design and manufacturing team brainstorming, competitor product informa-

tion, literature, knowledge of process capabilities, or whatever other means are

appropriate to create design concepts. Concepts should include physical form

as well as materials and manufacturing process, but should not be to the level

of detailed design. In this respect, the concepts are illustrated similarly to those

used in Pugh’s Concept selection [48]. Five to seven design concepts is a good

target. Record the name of each design concept in the leftmost column of the CS

matrix.

3. Evaluate each concept in comparison to the engineering metrics in the top row

of the CS matrix, recording the score in the box where the concept and engineer-

ing metric intersect. A concept receives a “–1” if it is likely unable to meet the

metric, “0” if it will likely meet the metric, and “1” if it is likely to exceed the

metric.

4. Sum the scores for the concepts across the engineering metrics, and record the

total score in a column on the right of the CS matrix. Rank concepts according

to their total score.

16 T.L. Lamers and B.L. Pruitt

Fig. 1.8 CS for piezoelectric acoustic sensor in cell phones (Modiﬁed ﬁgure, used with permission

of ASME. First appearing in Lamers and Fazzio [50])

Fig. 1.9 CS for piezoelectric acoustic sensor applied to laptop (Modiﬁed ﬁgure, used with

permission of ASME. First appearing in Lamers and Fazzio [50])

1 The MEMS Design Process 17

Fig. 1.10 CS for piezoelectric acoustic sensor applied to automotive (Modiﬁed ﬁgure, used with

permission of ASME. First appearing in Lamers and Fazzio [50])

5. Analyze the results. The highest ranking concept or concepts are those most

likely to be viable and most worthwhile to prototype. Overall scores indicate

how well the design concepts ﬁt the intended application, with scores above zero

indicating a better ﬁt than negative scores.

6. If there is more than one potential application, repeat QFD Phase I and CS for

each application, keeping the design concepts constant. A comparison of total

scores across applications will highlight how well the technology and design

concepts ﬁt the respective applications.

7. Utilize the CS results in determining the viability and direction of the project.

The interested reader is referred elsewhere [43] (ASME IMECE on scent

dispenser) for more examples of the application of CS.

The CS results for the application of piezoelectric mics to cellular handsets, lap-

tops, and automotive sensors are shown in Figs. 1.8, 1.9, and 1.10. The same six

design concepts were used in each of the application areas. The concepts are the

following.

1. Piezoelectric mic with Deep Reactive Ion Etching (DRIE) backside cavity and

annular electrodes

18 T.L. Lamers and B.L. Pruitt

2. Piezoelectric mic with Potassium Hydroxide (KOH) backside cavity and annular

electrodes

3. Piezoelectric mic with DRIE backside cavity and continuous electrodes

4. Piezoelectric mic with KOH backside cavity and continuous electrodes

5. Piezoelectric mic with shallow cavity and annular electrodes

6. Piezoelectric mic with shallow cavity and continuous electrodes

These concepts were developed by the project technical team through consider-

ation of such factors as ease in leveraging the current FBAR process capabilities,

available in-house processing options, literature results for similar devices, and

currently successful product structures.

QFD and CS were performed after the ﬁrst round of prototyping, so the team was

able to use initial test results in creating potentially viable design concepts. The con-

cepts vary in aspects of the manufacturing process, such as using DRIE versus KOH

versus RIE (Reactive Ion Etching) to create a cavity beneath the mic membrane.

The concepts also differ in physical structure through the comparison of contin-

uous versus annular electrodes. In this manner, important aspects of the physical

design layout and function as well as critical manufacturing process options are rep-

resented in the concepts. The assessment of how likely a given concept was to meet

a particular engineering spec was based upon the judgment of members of the tech-

nical team as well as initial test data and published results from other organizations.

In analyzing the outcomes of CS, it is noted that design concept 1, “Piezoelectric

mic with DRIE backside cavity and annular electrodes,” ranked ﬁrst in all three of

the application areas, although it was tied for ﬁrst with concept 3, “Piezoelectric

mic with DRIE backside cavity and continuous electrodes,” in the laptop and auto-

motive areas. The relative ranking of the other concepts shifted depending upon the

application. This result demonstrates that CS can be useful in distinguishing the via-

bility of concepts in different application spaces. A second, even more enlightening

result can be observed by considering the total scores the design concepts received.

The total scores varied between the application areas, as the engineering metrics

and targets changed. For example, “Sensitivity” was critical for cellular handsets

and laptops but not as important for automotive applications. All the piezoelectric

design concepts struggle to meet sensitivity targets, making the total scores gen-

erally lower in cell phone and laptop applications. The higher total scores in the

automotive application suggest the piezoelectric acoustic sensor is better suited for

automotive applications. This insight on technology ﬁt to market is important to

recognize and utilize.

The outcomes of the QFD I and CS offer clarity regarding which design con-

cepts are most likely to succeed so that resources may be assigned commensurately.

Reviewing the scores for each design can be a worthwhile exercise for assessing

how many of the concepts have the potential to be successful. A score greater than

or equal to zero shows a fair likelihood of the concept meeting the design require-

ments, whereas a negative score shows the concept will probably fail to meet critical

design requirements. Designs with negative scores are probably not worthwhile to

pursue further or require modiﬁcation to become viable to pursue.

1 The MEMS Design Process 19

Results

Performance results of the Avago microphones are published elsewhere [50]. Over

the course of the ﬁrst 8 months of the project, hundreds of different designs were

prototyped. These included circular membranes with electrodes covering the entire

membrane surface, circular membranes with annular electrodes, circular membranes

with electrodes restricted to the center of the membrane, circular membranes with

combinations of annular and central electrodes connected in series and parallel

combinations, and a variety of cantilevers.

Figure 1.11a–c provides images of representative devices showing several of

these conﬁgurations. Other projects which diverged less substantially from exist-

ing technology have historically taken between 6 months and 2 years of iterations to

progress as far as this much more difﬁcult project did in 8 months through utilizing

concurrent design and product deﬁnition phase design methodologies.

Extensive leverage of the FBAR process and infrastructure enabled fast fabrica-

tion of initial prototypes and allowed many iterations of prototyping in a short time

span. Five complete mask turns and 17 distinct experimental runs were fabricated

and tested within the ﬁrst 8 months of the program. As devices were produced, they

were tested for sensitivity. The results were then used in guiding future experiments

and device layouts. Phase I of QFD provided the team with technical targets and

guidance on which engineering metrics were most critical to achieve. “Dynamic

range” and “Total packaged linear dimensions” were very important in all t hree

application areas, whereas “Noise ﬂoor,” “Shock resistance,” “Moisture tolerance,”

and “Sensitivity” were key in at least one application.

CS gave the project team direction on which design concepts were most likely to

meet the technical requirements for a given application space. In all three cases, the

design concept of creating a piezoelectric microphone with a DRIE backside cavity

and annular electrodes came with the highest rank. This concept was prototyped,

and the test results supported the CS ﬁndings. Even more important than suggesting

which design concepts to pursue, CS indicated which application space the technol-

ogy might best ﬁt through analysis of the total scores. In the cellular handset and

laptop applications, the concept total scores tended to be zero or negative, showing

(a) (b)

(c)

Fig. 1.11 Representative mic design options. (a) Annular electrodes, (b) cantilever, and (c)nested

electrodes (Modiﬁed ﬁgure, used with permission of ASME. First appearing in Lamers and Fazzio

[50])

20 T.L. Lamers and B.L. Pruitt

the piezoelectric microphone design ideas would have difﬁculty meeting critical

technical speciﬁcations. In automotive sensors, the total scores ranged from –2 to

+4. The design concepts with positive scores are more likely to meet the important

technical requirements. This result indicates, at least for the design concepts con-

sidered, that the technology is better ﬁt to application in automotive sensors than in

cellular handsets or laptops. Early prototype test results aligned with this CS out-

come. The CS results drove realignment of the program to markets that matched the

technology capabilities. Technical challenges remain in piezoelectric microphone

development. As development continues, QFD Phase I and CS should be repeated

with reﬁned customer requirements, engineering metrics, and design concepts. In

this manner, the tools will provide continued guidance throughout the course of the

project.

The general methodology of QFD I and CS may be utilized for rapid develop-

ment of other products and technologies. QFD Phase I can be applied to MEMS

in the standard format, and gives insight on the most critical engineering metrics

on which to f ocus design efforts. The results of Phase I are made more robust

through validating the customer requirements and engineering metrics via feed-

back from customers and competitive product data. Use of engineering judgment

in evaluating the tool results is also important. The concept screening tool, used to

evaluate design concepts versus critical engineering metrics, can easily be applied to

other MEMS development efforts. CS provides a screening mechanism for selecting

design concepts to pursue further, and gives a ﬁrst brush look at the ﬁt of technol-

ogy and design concepts to applications. CS was utilized in the development of

a piezoelectric MEMS microphone, enabling fast iterations of assessing the ﬁt of

the technology and design concepts with various applications. Used in combination

with concurrent engineering techniques, CS could speed the development of other

MEMS products and increase proﬁtability through addressing applications that best

ﬁt the technology and design.

The Avago story demonstrated a case of extending an existing technology to new

applications, which is in essence a technology-driven project. In comparison, the

Knowles story, discussed next, demonstrates a technology that evolved through the

forces of market pull.

1.4.3 The Knowles Story

Knowles was started more than 60 years ago and has been successful in a number

of product areas, including hearing aids. Knowles began their program to develop

a MEMS mic as a defensive strategy to protect their hearing aid business. Dr. Pete

Loeppert led the development effort, assisted by a few process engineers initially,

and later by three test, design, and modeling engineers. This effort ultimately

resulted in the Knowles SiSonic mic. Information for this case study was primarily

obtained through an interview with Dr. Loeppert [51].

Initially the team pursued a two-piece design, where the capacitive plates were

fabricated on separate wafers and later assembled together, as this was the design of

1 The MEMS Design Process 21

choice in the literature at the time. This approach did not provide very good results,

so the team migrated to a one-piece design, fabricating both capacitive plates on

the same wafer. Film stress was the largest variable in performance. As ﬁlm stress

is very hard to control consistently from batch to batch, performance varied widely

across wafers and lots. As the main designer, Loeppert knew it was important to

design a product robust to process variations such as stress. His knowledge and

understanding of product design and process interactions led to a free-plate design

that eliminated dependence on as-deposited ﬁlm stress and reduced the effects of

package stress, a critical breakthrough. Loeppert also knew to choose materials and

processes that would give the mics the desired characteristics, resulting in a migra-

tion from PECVD nitride to LPCVD nitride to allow better control over the stress

gradient.

The team utilized ﬁnite element modeling and crude electrostatic modeling to

create designs that would prevent electrostatic collapse. The modeling served as a

jumping-off point, with building and testing product the primary means of learning.

Rapid prototyping led to the addition of corrugation to the backplate to reduce

membrane bowing. A representative geometry is shown in Fig. 1.12.

By 1996, Knowles had achieved a working design, but it was relatively expen-

sive and the hearing aid market was too small to justify t he product and development

costs. Knowles decided to switch applications from hearing aids to consumer prod-

ucts, ﬁrst targeting laptops and eventually cellular phones. The switch in application

necessitated design changes for cost reduction and process simpliﬁcation so that the

Fig. 1.12 Knowles

Sisonic

Microphone

packaged with signal

conditioning chip and

cross-section showing the

ﬂoating diaphragm,

antistiction posts, and back

plate. Original ﬁgure, used

with permission of Knowles

Electronics

22 T.L. Lamers and B.L. Pruitt

process could be run on standard equipment and would be easily scalable. By 1999,

the team had good prototypes but struggled to ﬁnd production fabrication and pack-

aging partners. In 2002, they began working with Sony for wafer fabrication, and

ultimately built their own packaging facility. In 2003, Motorola put the SiSonic in

what was anticipated to be a low-volume phone platform. SiSonic sales to Motorola

exceeded 30 million by 2009.

Although the Knowles team did not use design methods such as QFD and FMEA

formally, they kept the ﬁt of the technology to the market application in mind

throughout the development process. As the project leader, Loeppert kept a running

Pareto list of critical needs. A Pareto list is a means to determine priorities for pro-

cess improvement via identifying the biggest improvement opportunities. As each

opportunity is addressed, the list must be updated with the next largest improve-

ment areas. Loeppert continuously guided the team on working down the Pareto list

in much the same way as the outcomes of QFD would be used. Loeppert also kept

a focus on developing the right product for the application at the right market time,

rather than just building something that was technically outstanding but did not meet

customer and market needs. QFD and CS can both help with this process if the team

lacks the experience to maintain the focus on market application without formal

methods. Loeppert suggests that all MEMS engineers obtain exposure to design

and process, even if the ultimate job aspiration is one or the other. The hands-on

learning provides deep understanding that enables an engineer to do a much better,

more efﬁcient job in developing new products. It is also necessary to have a balance

between technical knowhow in the relevant areas and market understanding on a

project team, whether those skills come together in one person or through multiple

team members.

Given Knowles position at the top of the MEMS mic market and the fact that they

shipped over one billion units in the ﬁrst 6 years of production [52], the success of

the SiSonic development project is indisputable. SiSonic provides a prime example

of developing MEMS technology to ﬁt the requirements of a speciﬁc target market.

SiSonic’s market dominance shows the team hit the mark.

1.4.4 Summary of Key Concepts

The design process is nonlinear and requires short loops and systematic evaluation

of process and materials in parallel with transducer physics analysis and geo-

metric design. Structured design methodologies minimize the time and resources

spent on design, rigorously employing and weighting the numerous parameters and

voices essential to a good design outcome. The methods ﬁnd use in helping com-

panies with patented, well-characterized, or otherwise advantageous materials or

technology knowhow to evaluate new markets and applications (technology-push)

and also help companies sift through a wider range of technologies and design

options in pursuing a new product or market (market-pull). The rest of the chap-

ter is devoted to advice for structuring steps of the product design process outlined

in Fig. 1.3.

1 The MEMS Design Process 23

1.5 Materials and Process Selection

1.5.1 Materials Selection

The materials list for MEMS continues to grow, while CMOS compatible materials

and silicon still comprise a large fraction of commercial products for their obvi-

ous compatibilities with electronics and attributes for micromachining. Srikar and

Spearing [53] classiﬁed ﬁve materials indices to aid in materials selection. For their

resonator case study these are based on attributes including mass, stiffness, inertial

load, deﬂection, and frequency and are related to materials properties. As the design

process progresses, increasing detail and quality of data are needed to accurately

predict device performance, as shown in Fig. 1.13. To obtain detailed materials data,

test structures fabricated in the same process as the device should be evaluated and

results fed back to a design iteration.

Furthermore, materials selection plots for classes of devices and desirable mate-

rials attributes can be consulted as the design embodiment progresses. For example,

micromechanical ﬂexures are a common design element in MEMS and reliable

robust ﬂexures require large displacements at small forces and without fracture. The

materials selection space may be represented thus as in Fig. 1.14. Material proper-

ties may also be affected by their processing history, such as pressure, temperature,

deposition method, etchant exposure, and the like.

1.5.2 Process Selection

MEMS devices consist of primary (structural) materials and secondary (dielectric,

interconnect) materials [54]. MEMS processes frequently also utilize secondary

Fig. 1.13 The quality of

materials data required for

design increases as the design

process progresses (Modiﬁed

ﬁgure, used with permission

of IEEE. First appearing in

Srikar and Spearing [53].

Reprinted with permission.