Microsensors, MEMS and Smart Devices - Gardner Varadhan and Awadelkarim Part 5 pdf

Here, weconsider the three main kinds of organic PCBs - solid, flexible, and moulded; the ceramic PCB is known as a thick film hybrid circuit board and is discussed in Section 4.6.1.. 4.

Figure 4.39 Tape-automated bonding technique

hot thermode produces a faster throughput than wire bonding Moreover, the reducedinductance of a probe means that the devices can be AC-tested

The disadvantages of TAB include the relatively high cost of the process and the needfor a large device footprint This problem is overcome in flip-chip mounting

4.4.3 Flip TAB Bonding

In flip TAB bonding, the die is mounted upside down on the substrate, as shown inFigure 4.40 The major advantage of flip TAB over regular TAB mounting is that the diecan be subsequently attached to a metal lid for better thermal management

4.4.4 Flip-Chip Mounting

Finally, flip-chip mounting of the die has a number of key advantages It provides anexcellent contact between the die and substrate by eliminating the wire or beam lead

Chip Flip TAB lead Support material Substrate

Figure 4.40 Flip TAB technique

— Solder bumps

" " - - Substrate

Figure 4.41 Flip-chip mounting technique

entirely (see Figure 4.41) Solder bumps are placed on the substrate and then the die ismounted facedown, and the solder is melted to make the connection The small footprintand pitch, coupled with short interconnect of about 50 urn, and hence low inductance,make this a very attractive technology at a relatively low cost

Full details of these bonding methods may be found in textbooks such as Doane andFranzon (1993)

4.5 PRINTED CIRCUIT BOARD TECHNOLOGIES

Once electronic components have been made and packaged, such as the monolithic ICsdescribed in Sections 4.3 and 4.4, they need to be connected with other components

to form a circuit board The most common way to do this is to make a PCB, which

is also known as a printed wiring board (PWB) There are a number of different PCBtechnologies based on different dielectric materials and their fabrication process Here, weconsider the three main kinds of organic PCBs - solid, flexible, and moulded; the ceramic

PCB is known as a thick film hybrid circuit board and is discussed in Section 4.6.1.

compo-is prepared and, if required, a protective layer compo-is patterned, leaving just the solder areasexposed

In a solid organic PCB, the dielectric material consists of an organic resin reinforcedwith fibres The fibres are either chopped or woven into the fabric, and the liquidresin is added and processed using heat and pressure to form a solid sheet The most

PRINTED CIRCUIT BOARD TECHNOLOGIES 105

- Copper interconnect

Dielectric -Plated through hole

(b) L—Dielectric Copper interconnect

Blind via —\ /-Buried via

Dielectric-/ Copper (c) Plated through hole

interconnect-Figure 4.42 Schematic cross section of three types of organic PCBs: (a) single-sided; (b)

double-sided; and (c) multilayered

Table 4.8 Material properties of some common fibres used in organic PCBs

-10 –3

%

°cg/cm 3

J/g.°ckg/mm W/m.°C kg/mm

e-glass

5.05.81.14.8840

2.54 0.827

350

0.89 7400

s-glass

2.8

4.52

2.65.5975

2.49 0.735

4750.9

8600

Quartz 0.54

3.50.25.0

1420 2.20 0.966

2001.1

7450

Aramid –5.0a

4.01.04.53001.401.0924000.5

13000

aAlong axis of fibre; radial is 60 ppm/°C

commonly used fibres are paper, e-glass, s-glass, quartz, and aramid The precise choice

of the dielectric material depends on the technical demands presented by the device andapplication proposed, and the properties, such as the permittivity and loss factor, arefrequently the most important Table 4.8 gives some of the properties of the fibres thatare commonly used in organic PCBs

4.5.2 Flexible Board

In flexible PCBs, the resin used to make a solid dielectric material is replaced by a thinflexible dielectric material and the metal is replaced by a ductile copper foil Again, a

Etch Laminate

Cover film Adhesive | Etch Copper

Adhesive Base film Cover film Adhesive Copper Adhesive Base film Adhesive Copper Adhesive Cover film Cover film"

Adhesive Copper Adhesive Base film Adhesive Copper Adhesive Cover film Adhesive Copper Adhesive Base film

firm firm mm nrmSingle-sided flex-printed wiring

mm rmn mm nrm

nrm mrn rmn mm

mm mm mm mm

y/////////////////////////zm

Multilayer flex-printed wiring

Figure 4.43 Schematic cross section of three types of flexible PCBs: (a) single-sided; (b)

double-sided; and (c) multilayered

Table 4.9 Material properties of some resins used in organic PCBs

°CW/m.°KGPa

Epoxy

584.50.351300.33.4

Polyimide

494.30.332600.34.1

Cyanateester553.80.352600.33.4

PTFE

992.60.46-0.30.03

aApproximate values

number of different organic materials can be used to make a flexible wiring board such

as polyimide (Kapton), polyester terephthalate (Mylar), random fibre aramid (Nomex),Teflon, and polyvinyl chloride (PVC) The copper foil is processed as before by opticallithography, and layers can be joined together to form multilayer laminates The layersare usually bonded together using an adhesive such as acrylic, epoxy, polyester, and

PRINTED CIRCUIT BOARD TECHNOLOGIES 107

Table 4.10 Material properties of some dielectric films used in flexible organic PCBs

Initial tear strength

Tensile strength, min

Volume resistivity (damp

heat)

Units Polyimide

g/cm3 1.40

4.00kV/mm 79

% 0.15

10–3 12

% 40

g 500MPa 165Q-cm 106

FEP

2.152.30790.30.720020017

107

Polyester

1.383.40790.257.090800138-

Epoxypolyester1.53-5.90.20015170034

105

Aramidpaper0.653.0015.40.30104-28

Flexible PCB dielectric and adhesive films are now manufactured to a standard, andTable 4.10 shows the Class 3 properties of some dielectric films according to the standardIPC-FC-231 Accordingly, organic PCB laminates can now be constructed with increasedconfidence in their performance

4.5.3 Plastic Moulded

The most common forms of PCB - the organic PCB and the ceramic PCB (see nextsection) - are planar, that is, the metal interconnects are formed in two dimensions withplated through holes joining one layer to another However, it is possible to make athree-dimensional PCB by the moulding of a suitable plastic A three-dimensional PCBcan be made from extruded or injection-moulded thermoplastic resins with a conductivelayer that is selectively applied on its surface However, high-temperature thermoplasticsare required to withstand the soldering process, and commonly used materials arepolyethersulfone, polyetherimide, and polysulfone Plastic moulded PCBs have severaladvantages over organic PCBs, such as superior electrical and thermal properties andthe ability to include in the design, noncircular holes, connectors, spacers, bosses, and

so on More often than not, a moulded PCB is in essence an IC chip carrier package.Plastic moulded PCBs may prove to be advantageous in microtransducers and MEMSapplications, in particular, when the assembled microstructure has an irregular structure

or needs special clips or connectors The plastic moulded IC package may also be used

as part of a hybrid MEMS before full integration is realised Future Micro-moulds may

be fabricated using microstereolithography (see Chapter 7)

4.6 HYBRID AND MCM TECHNOLOGIES

4.6.1 Thick Film

PCBs can also be formed on a ceramic board, and these may be referred to as

ceramic PCBs A ceramic board, such as alumina, offers a number of advantages over

organic PCBs, because a ceramic board is much more rigid, tends to be flatter, has alower dielectric loss, and can withstand higher process temperatures In addition, alumina

is a very inert material and hence is less prone to chemical attack than an organic PCB.Ceramic PCBs can be processed in a number of different ways, such as thick-film, thin-film, co-fired, and direct-bond copper The most important technology is probably the thickfilm Circuit boards have been made for more than twenty years using this technology

and are usually referred to as hybrid circuits.

In thick-film technology, a number of different pastes have been developed (known asinks), and these pastes can be screen-printed onto a ceramic base to produce interconnects,resistors, inductors, and capacitors

3 Substrate is cleaned using a sandblaster, rinsed in hot isopropyl alcohols, and heated

to 800 to 925 °C to drive off organic contaminants

4 Each layer is then in turn screen-printed to form the multilayer structure Each paste

is first dried at 85 to 150 °C to remove volatiles and then fired at 400 to 1000 °C

5 The last high-temperature process performed is the resistive layer (800 to 1000 °C)

6 A low-temperature glass can be printed and fired at 425 to 525 °C to form a protectiveoverlayer or solder mask

Thick-film technology has some useful advantages over other types of PCB manufacture.The process is relatively simple - it does not require expensive vacuum equipment (likethin-film deposition) - and hence is an inexpensive method of making circuit boards.Figure 4.44 shows a photograph of a thick-film PCB used to mount an ion-selectivesensor and the associated discrete electronic circuitry (Atkinson 2001) The thick-filmprocess is useful here not only because it is inexpensive but also because it forms a robustand chemically inert substrate for the chemical sensor The principal disadvantage of thick-film technology is that the packing density is limited by the masking accuracy - somehundreds of microns Photolithographically patterned thin-film layers can overcome thisproblem but require more sophisticated equipment

4.6.2 Multichip Modules

Increasingly, PCB technologies are being used to make multichip modules (MCMs) Amultichip module is a series of monolithic chips (often silicon) that are connected and

HYBRID AND MCM TECHNOLOGIES 109

Figure 4.44 ISFET sensor and associated circuitry mounted on a ceramic (hybrid) PCB From

Figure 4.45 Silicon efficiency rating and line width of different interconnection and substrate

technologies After Ginsberg (1992)

packaged to make a self-contained unit This module can then be either connected directly

to peripheral ports for communication or plugged into another PCB One important reasonfor using MCM instead of a conventional die-packaging approach is that the active siliconefficiency rating is improved (see Figure 4.45) In other words, the total area of thesemiconductor die is comparable to the MCM substrate area As can be seen from thefigure, conventional PCB technologies and even SMT and hybrid are much poorer thanthe high-density MCM methods

The ceramic-based technology is referred to as an MCM-C structure; other MCM-C

technologies include high-temperature fired ceramic (HTCC) and low-temperature fired ceramic (LTCC) Table 4.11 lists the relative merits of different MCM-C technologies

co-Table 4.11 Relative merits of MCM-C technologies,

with one being the best

Adapted from Doane and Franzon (1993),

HTCC333113112

LTCC211331111

BenefitImproved wire-bond,assembly yield stabilityImproved high-frequencyperformance

Smaller line and spacedesigns

More rugged packageGood thermal characteristicsCapability of assemblyDevelopment of packagesMore consistent electricalperformance

Better high-frequencyperformance

Table 4.12 Properties of some commonly used MCM-C materials Adapted from Doane and

AI2O3

99.5White7.63.879.90.124400

1014

20-35

-AI2O3

96White7.13.79.50.426250

1014

20-35

-BeO99.5White9.03.016.50.49.5170-240

1015

250-260

-A1N98-99.8Dark grey4.43.2558.8-8.90.7-2.010-14280-320

>1013

0.7480-260

The choice of ceramic substrate is important and the >99 percent alumina

has a low microwave loss, good strength and thermal conductivity, and good flatness.However, it is expensive and 96 percent alumina can be used in most applications Incases in which a high thermal conductivity is required (e.g power devices), beryllia (BeO)

or aluminum nitride (A1N) can be used, although these involve a higher cost Table 4.12summarises the key properties of the ceramic substrates

In addition, modules wherein interconnections are made by thin films are classified asMCM-D and those made by plastic (organic) laminate-based technologies are classified

as MCM-L Table 4.13 shows a comparison of the typical properties of the three maintypes of MCM interconnection technologies

HYBRID AND MCM TECHNOLOGIES 111 Table 4.13 Comparison of MCM interconnection technologies.

Line width (um)

Line pitch (um)

Bond pad pitch (um)

6-9

35-65

200

Cu (Au)15

100-150 250-350 250-350

5 to 10+

0.2-0.3

3.0904.3

HTCC MCM-C Alumina

9.5

100-750 100-200

W(Co)15

100-125 250-625 200-300

50+

0.8-110.01022.1

Adapted from Doane and Franzon Thin film

MCM-D Polyimide

3.52525

Cu (Al, Au)5

10-25

50-125

1004-to 10

1.3-3.4

3.4621.25

Laminate MCM-L Epoxy-glass

4.8120300Cu25

75-125 150-250

20040+

0.06-0.09

0.7721.46

MCM technology has several advantages for integrating arrays of microtransducers andeven MEMS (Jones and Harsanyi 1995) First, the semiconductor dies can be fabricated

by a different process, with some dies being precision analogue (bipolar) components andothers being digital (CMOS) logic components Second, the cost of fabricating the MCMsubstrate is often less expensive than using a silicon process, and the lower die complexityimproves the yield Finally, the design and fabrication of a custom ASIC chip is a time-consuming and expensive business For most sensing technologies, there is a need for newsilicon microstructures, precision analogue circuitry, and digital readout Therefore, fabri-cating a BiCMOS ASIC chip that includes bulk- or surface-micromachining techniques

is an expensive option and prohibitive for many applications

Figure 4.46 shows the layout of a multichip module (MCM-L) with the TAB patterns

shown to make the interconnections (Joly et al 1995) This MCM-L has been designed

for a high-speed telecommunications automatic teller machine (ATM) switching module,which, with a power budget of 150 W, is a demanding application

4.6.3 Ball Grid Array

There are a number of other specialised packaging technologies that can be used as analternative to the conventional PCB or MCM The main drive for these technologies is

to reduce the size of the device and maximise the number of I/Os For example, thereare three types of ball grid array (BGA) packages Figure 4.47 shows these three types:the plastic BGA, ceramic BGA, and tape BGA The general advantages of BGA arethe smaller package size, low system cost, and ease of assembly The relative merits of

Figure 4.46 Example of a high-density MCM-L substrate with TAB patterns From Joly et al.

(1995)

plastic and ceramic PGA packages are similar to those already discussed for PCBs andMCMs The tape EGA uses a TAB-like frame that connects the die with the next layerboard

4.7 PROGRAMMABLE DEVICES AND ASICs

The microtechnologies described in this chapter are used to make a variety of differentmicroelectronic components Figure 4.48 shows the sort of devices that can be madetoday These are subdivided into two classes - standard components, which are designedfor a fixed application or those that can be programmed, and application-specific ICs(ASICs), which are further subdivided The standard components that may be regarded as

having fixed application are discrete devices (e.g n-p-n transistors), linear devices (e.g.

operational amplifiers), and IC logic families of TTL and CMOS (e.g logic gates andbinary counters, random access memory) The other types of standard component may beclassified as having the application defined by hardware or software programming

In hardware programming, the application is defined by masks in the process, andexamples of these devices include programmable logic arrays (PLAs) and read-onlymemory (ROM) chips There has been a move in recent years to make softwareprogrammable components The most familiar ones are the microprocessors (such as theMotorola 68 000 series or Intel Pentium) that form the heart of a microcomputer and its

PROGRAMMABLE DEVICES AND ASICs 113

Figure 4.47 Three main types of ball grid array packages: (a) plastic; (b) ceramic; and (c) tape-ball

nonvolatile memory For example, erasable programmable read-only memory (EPROM)

in which the memory is erased by UV light and its easier-to-use successor, electronicallyerasable programmable read-only memory (EEPROM) In recent years, there has been astrong move toward software programmable array devices; these components do not havethe mathematical capability of a microprocessor but are able to perform simple logicalactions As such, they can be high-speed stand-alone chips or glue chips - that is, chipsthat interface an analogue device with a microcontroller or communication chip Examples

of these are programmable logic devices (PLDs) and PGAs that may have several thousand

gates to define The newest type of component is the programmable analogue array (PAA)device, and these may well become increasingly important in sensing applications in whichthe design of an operational amplifier circuit can be set and reset through I/O ports.

It should be noted that this classification of devices into hardware and softwareprogramming is not universally accepted Sometimes, it may be more useful to distinguish

1

"L PLA Microprocessor MOS ROM EPROM, EEPROM

Increasing density

Increasing time to market

Increasing versatility

Increasing development cost V

Decreasing cost/gate

Figure 4.49 Diagram showing the various trade-offs between the different technologies adopted

to make an ASIC chip

PROGRAMMABLE DEVICES AND ASICs 115

between devices according to where they were programmed In this case, devices mayall be regarded as 'electrically' programmed and then they can be subdivided into thoseprogrammed by the manufacturer (as in mask-programmed parts) and those programmed

by the user (as in field-programmable gate array)

The second class of components are those called application-specific integrated circuit ICs (ASICs) (see Figure 4.48) There are several types of ASIC and these are referred to

as full-custom, semicustom, and silicon compilation Full-custom ASICs are those that aredefined down to the silicon level and, therefore, there is great scope for the optimisation

of the device layout, reduction in silicon die area, and speed of operation However,

a full-custom design can be an expensive option and is only useful for large volumes.Silicon compilation is the exact opposite; hence, it is rather wasteful of silicon and pushes

up the process costs while minimising the design cost The more common approach, andmore relevant for the manufacturing of microtransducers, is that of a semicustom ASICchip This has four subdivisions In gate arrays, the device has been partly processedand the designer simply defines the interconnection of the digital logic devices by one

or two customised masks Thus, most of the process is common to a number of endusers, and hence the costs are greatly reduced In analogue arrays, the same principle isapplied, except that this time a range of analogue components are being connected and

an analogue circuit is formed In master slice, the wafer run can be split at a later stageinto different subprocesses The last type of semicustom approach is the standard cell inwhich the designer selects standard logic or analogue circuit functions from a softwarelibrary and then connects them together on the silicon die The design time is reduced byusing standard cells with a standard process

The various trade-offs of the ASIC technologies are illustrated in Figure 4.49 such

as risk, cost, density, and flexibility Strictly speaking, PLDs are not ASICs but theyhave been included here because they are often the main competitors to an ASIC chip.Although the number of equivalent gates per chip in PLDs is only 500 to 3000, the costadvantage is often attractive

When deciding upon which ASIC technology to use, it is important to weigh the relativecosts involved, such as the development time and the nonrecurring engineering costs (maskmaking etc.), and design consideration such as the architecture required and the number ofgates In the final analysis, it is usually the volume that dictates the cost to manufacture thechips; the production charges per 1000 gates are shown against total volume in Table 4.14.For example, modern microprocessor and memory chips are manufactured in enormousvolume (millions of chips per year) and so the cost is dominated by the time to processand, hence, the size of wafer processed Current microelectronic plants use wafers of adiameter of 8" or more, and companies have to build new plants that cost nearly onebillion dollars as larger diameter wafers become available This situation is usually notapplicable to the manufacture of microsensors because of the much reduced volume andhigher added value

However, all of these production costs per kgate are low compared with the cost offabricating a nonstandard component For instance, when integrating a microtransducer

or MEMS with a standard IC, it is nearly always necessary to develop nonstandard

pre-or postprocessing steps, such as surface pre-or bulk micromachining (see next chapter) Thiscost issue is critical for the eventual success of a component on the market and therefore,

we will return to it later on in Chapter 8, having first described the different fabricationmethods and technologies associated with microtransducers and MEMS

Table 4.14 Typical costs of different ASIC (and programmable device) technologies Adapted

Analogue

RAM,ROM,

AnalogueLogiconlyFixedlogic

Fixedlogic

Density(kgates)

(k€)

50

15a

-5015-100

<5

5-20

Productionvolume(1000s)

5-10

3-42-3N/A3-42-3876

10-207-155-12

aCosts shown for PC-based and workstation-based design, respectively

REFERENCES

Atkinson, J (2001) University of Southampton, UK Personal communication

Colclaser, R A (1980) Microelectronics Processing and Device Design, Wiley & Sons, New York,

p 333

Doane, D A and Franzon, eds (1993) Multichip Module Technologies and Alternatives, Van

Nostrand Reinhold, New York, p 875

Furakawa, S (1985) Silicon-on-insulator: Its Technology and Applications, D Reidel Publishing

Joly J., Kurzweil, K and Lambert, D (1995) "MCMs for computers and telecom in CHIPPAC

programme," in W K Jones and G Harsanyi, eds., Multichip Modules with Integrated Sensors,

NATO ASI series, Kluwer Academic Publishers, Dordrecht, p 324

Jones, W K and Harsanyi, G., eds (1995) Multichip Modules with Integrated Sensors, NATO ASI

series, Kluwer Academic Publishers, Dordrecht, p 324

Sze, S M (1985) Semiconductor Devices, Physics and Technology, Wiley & Sons, New York,

p 523

Udrea, F and Gardner, J W (1998) UK Patent GB 2321336A,"Smart MOSFET gas sensor,"Published 22.7.98, Date of filing 15.1.97 International Publication Number: WO 98/32009, 23July 1998, Gas-sensing semiconductor devices

5.1 INTRODUCTION

The emergence of silicon micromachining has enabled the rapid progress in the field ofmicroelectromechanical systems (MEMS), as discussed previously in Chapter 1 Siliconmicromachining is the process of fashioning microscopic mechanical parts out of a siliconsubstrate or, indeed, on top of a silicon substrate It is used to fabricate a variety ofmechanical microstructures including beams, diaphragms, grooves, orifices, springs, gears,suspensions, and a great diversity of other complex mechanical structures These mechan-ical structures have been used successfully to realise a wide range of microsensors1

and microactuators Silicon micromachining comprises two technologies: bulk machining and surface micromachining The topic of surface micromachining is covered in

micro-the next chapter Furmicro-ther details can be found in micro-the two-volume Handbook of raphy, Micromachining, and Microfabrication (Rai-Choudhury 1997).

Microlithog-Bulk micromachining is the most used of the two principal silicon micromachiningtechnologies It emerged in the early 1960s and has been used since then in the fabrication

of many different microstructures Bulk micromachining is utilised in the manufacture ofthe majority of commercial devices - almost all pressure sensors and silicon valves and 90

percent of silicon acceleration sensors The term bulk micromachining expresses the fact

that this type of micromachining is used to realise micromechanical structures within thebulk of a single-crystal silicon (SCS) wafer by selectively removing the wafer material.The microstructures fabricated using bulk micromachining may cover the thickness rangefrom submicrons to the thickness of the full wafer (200 to 500 um) and the lateral sizeranges from microns to the full diameter of a wafer (75 to 200 mm)

Etching is the key technological step for bulk micromachining The etch processemployed in bulk micromachining comprises one or several of the following techniques:

1 Wet isotropic etching

2 Wet anisotropic etching

3 Plasma isotropic etching

4 Reactive ion etching (RIE)