Sensing microgripper with pid control systems

Figure 1.13: Solid model of the electrostatic microgripper with integrated force sensor: one comb capacitor is used as actuator and the other one is the sensing part [7].. In this chapte

VIETNAM NATIONAL UNIVERSITY, HANOI

COLLEGE OF TECHNOLOGY

Phan Huu Phu

SENSING MICROGRIPPER WITH PID CONTROL SYSTEMS

MASTER THESIS

Hanoi - 2008

VIETNAM NATIONAL UNIVERSITY, HANOI

COLLEGE OF TECHNLOGY

Phan Huu Phu

SENSING MICROGRIPPER WITH PID CONTROL SYSTEMS

Major: Electronics and Telecommunications Technology

Concentration: Electronics Engineering

TABLE OF CONTENTS

DECLARATION 1

ACKNOWLEDGEMENTS 4

ABSTRACT 5

LIST OF ABBREVIATIONS 6

CHAPTER 1 INTRODUCTION 7

1.1 M ANIPULATION IN MICRO - WORLD 7

1.2 M ICRO - GRIPER FOR MICRO - MANIPULATION 7

1.2.1 Electrostatic microgripper 7

1.2.2 Piezoelectric microgripper 8

1.2.3 Electrothermal microgripper 8

1.2.4 Polymeric electrothermal microgripper 10

1.3 M ICRO - MANIPULATION WITH A FEEDBACK SYSTEM 11

1.3.1 Force sensor 11

1.3.2 Sensing microgripper 14

CHAPTER 2 SENSING MICROGRIPPER 17

2.1 I NTRODUCTION 17

2.2 F ORCE - SENSING CANTILEVER 17

2.3 S ILICON - POLYMER ELECTROTHERMAL MICROGRIPPER 18

2.4 S ENSING MICROGRIPPER 20

2.5 T HE SENSING MICROGRIPPER CHARACTERISTICS 22

2.5.1 Electrothermal actuator characteristics 22

2.5.2 Sensing cantilever beam characteristics 25

2.5.3 Response frequency of the sensing microgripper 27

CHAPTER 3 BUILDING PID CONTROL FUNCTION 29

3.1 F EEDBACK LOOP CONTROL 29

3.2 B UILDING A PID TRANSFER FUNCTION FOR THE SENSING MICROGRIPPER SYSTEM 30

3.2.1 Transfer function of sensing microgripper 30

3.2.2 Transfer function of driver circuit 30

3.2.3 Open-loop control 31

3.2.4 Proportional control 32

3.2.5 Proportional – Integral control 32

3.2.6 Proportional – Derivative control 33

3.2.7 Proportional – Derivative – Integral control 34

CHAPTER 4 ELECTRICAL DESIGN 35

4.1 I NTRODUCTION 35

4.2 P ROCESS SELECTION AND SIMULATION 35

4.2.1 Process selection 35

4.2.2 Device modeling 36

4.2.3 Silicon-level simulation 36

4.2.4 Analog-only simulation 38

4.2.5 Mixed Analog/Digital simulation 38

4.3 S YSTEM BLOCK DIAGRAM 39

4.4 C ELLS DESIGN 40

4.4.1 Voltage reference generator 40

4.4.2 Internal regulator 45

4.4.2.1 The regulator: 45

4.4.2.2 The high temperature detector: 46

4.4.2.3 The UVLO 47

4.4.2.4 Bias current generator 47

4.4.3 Digital to Analog converter (DAC) 52

4.4.4 Buffer 56

4.4.5 PID Controller 56

4.4.6 Other cells 60

4.5 F ULLY SCHEMATIC OF SYSTEM AND SIMULATION RESULTS 62

4.5.1 Schematic 62

4.5.2 Simulation results 62

CHAPTER 5 PHYSICAL DESIGN 65

5.1 I NTRODUCTION OF L AYOUT 65

5.1.1 Matching concepts 65

5.1.2 MOS transistor layout 66

5.1.3 Resistor layout 68

5.1.4 Capacitor layout 69

5.1.5 Layout rules 70

5.2 S YSTEM FLOOR PLAN 71

5.3 S YSTEM LAYOUT 72

5.3.1 Sensing microgripper 72

5.3.2 Electrical circuits 72

CHAPTER 6 CONCLUSION & FUTURE WORKS 74

6.1 C ONCLUSION 74

6.2 F UTURE WORKS 74

6.2.1 Finishing the system layout 74

6.2.2 Process establishment 74

6.2.3 Layout verification 75

6.2.4 Sample fabrication 75

6.2.5 Characterization 75

BIBLIOGRAPHY 76

List of Abbreviations

AFM: Atomic Force Microscope

AlN: Aluminum Nitride

Bi-CMOS: Bipolar junction transistors and CMOS technology

CAD: Computer Aid Design

CMOS: Complementary Metal-Oxide-Semiconductor

CTE: Thermal expansion coefficient

DAC: Digital-to-Analog Converter

DC: Direct Current

DIMES: Delft Institute of Microsystems and Nanoelectronics

DOFs: Degrees(s) of Freedom

DRC: Design Rules Checking

ERC: Electrical Rules Check

ESD: Electro Static Discharge

FIB-cut: Focused ion beam-cut

GDSII: Graphic Data System II

IC: Integrated Circuits

LSB: Least Significant Bit

LSI: Large-Scale Integration

LVS: Layout versus Schematic

MEMS: Micro-Electro-Mechanical systems

MIS: Minimally Invasive Surgery

MOSFET: Metal-Oxide Semiconductor Field-Effect Transistor

NMOS: N-channel Metal-Oxide Semiconductor Field-Effect Transistor PID: Proportional Integral Derivative

PLI: Photolithographic Invariance

PMOS: P-channel Metal-Oxide Semiconductor Field-Effect Transistor PSRR: Power supply rejection ratio

PTAT: Proportional to Absolute Temperature

PVDF: Polyvinylidene Flouride

PZT: Lead Zirconate Titanate

SEM: Scanning Electron Microscope

SOI: Silicon on Insulator

SPICE: Simulation Program Integrated Circuit Emphasis

TC: Temperature Coefficients

UVLO: Under Voltage Lock-Out

ZnO 2: Zinc peroxide

Chapter 1 Introduction

1.1 Manipulation in micro-world

The dominant physical principles in the micro-world can be quite difference from those of the macro-world When the size of the object is less than 1mm, adhesive forces between the manipulation tool and the object such as surface tension, electrostatic and van der Waals forces can be significant compared to the gravitational force [23]

Manipulation of micro-particles can be done using several physical principles and methods For manipulation of a micro-structure under specific ambient conditions or liquid; suction, cryogenic, electrostatic, and friction are the most often considered methods

Friction principle is chosen for addressing the micro-manipulation of small objects presented in this thesis Like the human hand, friction manipulation uses at least two fingers applies on two sides of a clamped object A friction manipulator can old an object due to the friction force between the tools and the object surfaces This friction method is the most widely used in micro-manipulation because of size, cost, reliability, and fabrication aspect in particular

1.2 Micro-griper for micro-manipulation

In the recent years, microgrippers have been widely researched as they are in great demand in many research and application areas, such as advanced micro-assembly, micromanipulation, micro-robotic, minimally invasive and living cell surgery For the development of microgrippers fabricated using integrated circuits (IC) or IC compatible technology, electrostatic, piezoelectric and electro-thermal actuation are generally used

1.2.1 Electrostatic microgripper

The electrostatic principle is based on the distance change between a fixed electrode and a suspended one when the voltage applied to these two electrodes changes The first successful electrostatic microgripper based on bulk and surface silicon micro-machining techniques was presented in 1992 (see Fig 1.1) [3] The 12 µm thick and

1500 µm long polysilicon microgripper is overhanging from a supporting silicon cantilever The microgripper jaw displacement is 10 µm at an applied voltage of 45 V, with a basic frequency of 5 kHz A monolithically fabricated electrostatic microgripper has been recently presented [7] A lateral comb drive has been chosen to actuate this gripper This microgripper can manipulate glass or copolymer spheres of size ranging from 20 to 90 µm with an applied force up to 380 µ at an applied voltage

of 140 V The main limitation of this device is the high voltage, the large size and the complicated electronic circuit typical of the electrostatic method

Figure 1.1: The schematic design of a polysilicon electrostatic microgripper (adapted from [3])

Figure 1.2: Schematic drawing of a piezoelectric microgripper [22]

1.2.3 Electrothermal microgripper

A thermostat consist a bi-metallic strip, which is made of two thin metallic pieces of difference materials that are bonded together As the temperature of the strip changes, two pieces change length at difference rates, forcing the strip to bend Based on the

Timoshenko’s bi-metal thermostat theory, an electrothermal microgripper (see Fig 1.3) is fabricated using doped silicon and a special bonding technique [9] This structure consists of a silicon cantilever beam with a doped layer on top Out of plane bending is obtained when a current is induced though the doped layer

Figure 1.3: Schematic drawing of a typical bi-material cantilever actuator The doped-silicon expands

when applying a current, therefore the cantilever bends downwards (adapted from [9])

Figure 1.4: Schematic drawing of a typical flexure actuator: (a) single hot arm configuration and (b)

two hot arms configuration

Figure 1.5: A microgripper product of Zyvex company: (a) the entire device is about 650 µm long,

270 µm wide, 50 µm thick The initial gap between the two jaws is 36 µm and the maximum opening is

80 µm; [(b) and (c)] the grippers being used to manipulate a FIB-cut coupon [13]

Since the well-known flexure thermal actuator (see Fig 1.4(a)) was introduced in

1992 [10], the actuator has received wide interest as they can produce a large displacement, a large force and use IC compatible fabrication process The flexure thermal actuator consists of a thin arm with higher electrical resistance than its thick arm The thin arm (hot arm) gets more heat than the thick one and consequently elongates more than the thick arm and in-plane bending occurs A configuration more efficient in terms of power consumption uses two hot arm thermal actuators (see Fig 1.4(b)) The electrical current just passes through two hot arms As there is no electrical current in cold arm and flexure, the efficiency of power consumption is improved compared with the single hot arm structure This principle is also used in the Zyvex gripper shown in Fig 1.5 The limitations of these devices are the extremely high operating temperature and high power consumption

1.2.4 Polymeric electrothermal microgripper

Recently, polymeric electrothermal microgrippers have been extensively researched as they are capable of producing large displacements at a lower drive voltage and operating temperature [19] Based on the above-mentioned flexure thermal actuator, polymeric microgrippers are developed using a polymer layer with a thin metal heater

on top [19] The structures have a large displacement at low operational temperature and low power consumption due to the large thermal expansion coefficient (CTE) of the polymer Fig 1.6 shows the schematic design of a developed polymeric microgripper using SU8 with a thin metal layer (Cr/Au) on top as a heater [19] The microgripper jaw displacement is 12 µm at an applied voltage of about 2 V and an operational temperature of less than 100 0C This microgripper has been designed to manipulate cells in air and also in fluid solutions

In these polymeric microgrippers the metal heater is deposited on top of a high thermal expansion coefficient polymer layer The interface between the heat source and the polymer layer is limited by the surface area of the metal layer, and the heat transfer along the vertical dimension is not effective Since the polymer layers have low thermal conductivity, these devices can generate limited movement Moreover, the unintentional vertical movement couples and interferes with the desired lateral movement [19].

Figure 1.7: Schematic drawing of a fabricated SU8 microgripper (adapted from [19])

1.3 Micro-manipulation with a feedback system

When manipulating micro-objects, operating on living cells or in minimally invasive surgery (MIS), enhanced dexterity, accuracy and speed are considerably improved when the force on the objects can be sensed and controlled in real-time

The development of miniaturized manipulators with force control is also of great interest in micro-robotics and micro-assembly

Manipulation of micro-objects with traditional microgrippers without a built in force sensor normally requires a camera inserted into the system to obtain visual feedback This approach results in a two-dimensional image Depth perception of the contact between the manipulating tool and the object being manipulated is lost, making it difficult to identify the position of the tool [16] Moreover only displacement, and not force, can be detected A microgripper with a built-in force sensor can address this limitation and is therefore suitable for holding objects firmly, whilst avoiding any squeezing of delicate objects

1.3.1 Force sensor

The contact forces between living cells in a laboratory or between micro-particles and

a manipulator are generally in the nano-Newton to mili-Newton range Cantilever force sensors are generally used to measure force in this range

The bending of the cantilever is related to the applied force By monitoring the deflection of the beam, the amplitude of the applied force can be detected Several force-sensing methods, such as capacitive, piezoelectric, optical laser detection, and piezoresistive can be used [30]

The capacitive method is based on the capacitance change which occurs when the structure is deformed, and is widely used in micro-accelerometers and harsh environment sensors An example of a capacitive force sensor (see Fig 1.8) is shown

in [29] It has two degrees of freedom (2-DOF) and uses silicon on insulator (SOI) wafers This sensor is capable of measuring forces up to 490 µN with a resolution of

0.01 µN along the x -axis and up to 900 µN with resolution of 0.24 µN along the

y-axis Complete isolation between the two electrodes is a limitation of capacitive force sensors, so normally SOI wafers are used Alternatively, trench isolation can be used However, it is difficult to control the etching area to obtain a completely isolated structure Moreover, the capacitive method requires a complicated fabrication process and complex electronic circuitry

Figure 1.8: Schematic drawing of a capacitive force sensor in two dimensions (adapted from [29])

A two-dimensional piezoelectric force sensor is presented in [28] (see Fig 1.9) It consists of two perpendicular pieces of polyvinylidene fluoride (PVDF) material This structure is symmetric in the vertical and lateral dimensions, with resolution and sensitivity in the µN range However, the PVDF cannot be patterned optically The two pieces have to be glued perpendicularly to each other, resulting in a rather large sensor structure With this approach, the sensor cannot be miniaturized and the fabrication process is not IC-compatible The piezoelectric method also requires complicated electronic circuits for processing the signal

Cantilevers based on optical force measurement are often used in atomic force microscopy (AFM) and high-resolution measurement With this principle, the deflection of a cantilever is amplified by a laser beam reflected on to the cantilever tip, and monitored using a split photodiode detector This principle is very powerful for

measuring small displacements but requires high accuracy of optical alignment and adjustment The cantilever surface must be reflective and larger than the laser beam spot It is not easy to create a lateral and multidimensional force sensor using an optical method Moreover, the lasers required are relatively large, and so it is impossible to miniaturize the entire force-sensing system down to the micrometer size range

Piezoresistive transducers translate a force into a change in the value of a resistor They are widely used as sensing elements in pressure sensors, accelerometers, and AFM cantilevers [15] Recently, developments in piezoresistive cantilever fabrication have led to submicrometer cantilevers with a resolution of pN and even fN Most of the previously developed high-sensitivity force sensors use SOI wafers and vertical structures

The sidewall-doping technique is normally used in lateral force-sensing Independent detection of a vertical and lateral forces sensor is shown in [25] (see Fig 1.10) Separate piezoresistor circuits serve the triangular probe and the two inner ribs, enabling independent detection of vertical and lateral forces However, oblique implantation, a rather special technique, is required in order to produce resistors on the vertical sidewalls of the cantilever

Figure 1.9: (a) Force cantilever using piezoelectric polymer PVDF and (b) a two dimensional force

sensor is based on the two perpendicular glued pieces [28]

Figure 1.10: Dual-axis AFM cantilevers with orthogonal axes of compliance Oblique ion implants are

used to form electrical elements on vertical sidewalls and horizontal surfaces simultaneously [25]

1.3.2 Sensing microgripper

In recent years, several designs of microgripper with force feedback have been demonstrated A force-sensing microgripper for MIS application is shown in [18] (see Fig 1.11) This device uses piezoelectric actuation with a strain gauge sensor on the sidewall of the structure It is capable of actuating at high frequency (hundreds Hz) with very high drive voltage In [5], a similar device is shown

Figure 1.11: Scheme of the microgripper based on PZT actuator with the location of the strain gauge

sensors [18]

It uses electromagnetic actuation and piezoelectric force sensing It generates large displacement at low voltage and a linear sensing output However, the main limitations of the above-mentioned devices are a fabrication process that is not compatible with CMOS technology and the rather large dimensions which are not suitable for the manipulation of micro-objects

In Fig 1.12, another type of sensing microgripper is shown [12] The actuator design

is based on three doped silicon beams connected by an end bar The two inner beams are electrically connected in parallel When a voltage is applied between the outer beams and the two connected inner beams, the current in the outer beam is twice the current in the inner one Therefore, the outer beam functions as the hot arm of an electrothermal actuator The displacement of the actuator is detected by monitoring the resistance change of the two inner beams The microgripper displacement is due to the thermal expansion of silicon, a material with a low CTE Therefore, the microgripper displacement and sensor output is quite small

Figure 1.12: Piezoresistive feedback microgripper: (a) optical image of the device; (b) a typical

Wheatstone bridge resistor configuration; (c) gripper made with two actuators in the three-beam configuration, which are connected as a Wheatstone bridge circuit; and (d) gripper configuration with three actuators consisting of the three-beam structure connected to avoid thermal stress influencing

the force detection [12]

Figure 1.13: Solid model of the electrostatic microgripper with integrated force sensor: one comb

capacitor is used as actuator and the other one is the sensing part [7]

An electrostatic microgripper with an integrated capacitive force sensor is shown in [18] (see Fig 1.13) It consists of a lateral comb drive for actuating the gripper and another for force sensing This device is capable of motion of up to 100 µm with force sensitivity of 4.41 kV/m, and corresponding 70 nN force-sensing resolution However, this device requires high drive voltage, large dimensions and a complicated electronic circuit

Another sensing microgripper is shown in [17] (see Fig 1.14) This device combines a CMOS photo-detector and a polymeric electrothermal actuator based on the hot-and-cold-arm design [19] By using an external light source, this device is capable of detecting the object present between two gripper jaws However, this method does not offer the possibility of monitoring the contact force between the manipulator tip and the object grasped

Figure 1.14: (a) Schematic of a simple 2 x 2 CMOS micro-machined gripper array with on-chip optical

detection and (b) gripper with a polystyrene bead [17]

Chapter 2 Sensing Microgripper

2.1 Introduction

For many of the application areas of micromanipulators force control is highly desirable as dexterity, accuracy and speed are considerably improved by a feedback system [16] A microgripper with a built-in force sensor is suitable for holding objects firmly, whilst avoiding any squeezing of delicate objects

As discussed in chapter 1, several designs of microgrippers with force feedback have been demonstrated in recent years [5, 7, 9, 12, 18] However, these devices suffer from some limitations: either they are large in size and thus not suitable for micromanipulation; or require a fabrication process that is not CMOS-compatible and/or need a complicated electronic circuit In this chapter a novel sensing microgripper based on silicon-polymer electrothermal actuators and a piezoresistive force-sensing cantilever beam is introduced

2.2 Force-sensing cantilever

A novel two-dimensional nano-Newton force-sensing piezoresistive cantilever is presented with providing nano-Newton sensitivity, a wide force measurement range, and the possibility of being combined with handling tools Instead of using sidewall implantation, two separate sensing piezoresistors are located on each sides of the cantilever By using two switches to change the Wheatstone bridge configuration, the piezoresistive cantilever is capable of detecting both the lateral and vertical bending independently Consequently this force-sensing cantilever can detect the applied forces in both parallel and perpendicular direction to the wafer surface The cantilever

is made on regular silicon wafers with a fabrication process compatible with CMOS technology

In Fig 2.1 a schematic drawing of the two-dimensional piezoresistive force sensing cantilever is shown The four piezoresistors are located on the surface of the structure The piezoresistors are aligned along the [110] direction of the (001) crystal plane of a silicon wafer The resistor pairs located on the cantilever are stress-sensing resistors (the resistance changes when the cantilever is deflected) Two other resistors are outside the cantilever They are not subjected to stress and therefore they are used for common-mode signal compensation in a Wheatstone bridge arrangement [4]

Figure 2.1: Schematic drawing of the two-dimensional piezoresistive force-sensing cantilever with

geometric symbols and orientations used [4]

When a force F is applied to the tip of the cantilever the cantilever bends On the yz plane, perpendicular to the longitudinal x -axis of the cantilever, the applied force F can be determined as two components: F z is the lateral and F y is the vertical

-component (see Fig 2.1) By monitoring F z and F y we can obtain both amplitude and direction of the applied force [4]

2.3 Silicon-polymer electrothermal microgripper

A novel silicon-polymer laterally stacked electrothermal in-plane microactuator is proposed The device is composed of three materials: a metal heating layer, a silicon frame with high heat conductivity and a polymer with a high thermal expansion coefficient (CTE) During actuation, heat is efficiently transferred from the heater to the polymer thanks to the high thermal conduction of the deep silicon structures which have a large interface with the surrounding polymer Moreover, the polymer layer is constrained between two silicon plates The thermal expansion of the constrained polymer is three times larger than the no constraint one

Fig 2.2 shows the sketch of a silicon polymer block It consists of laterally stacked silicon-polymer segments Each segment is formed by a rectangular plate of polymer which is bonded between two silicon plates The metal heater is on the top of the silicon Instead of transferring heat directly from the heater to the polymer as in the conventional developed designs [19], the efficient distribution of the heat to the polymer occurs through the large interface areas between the silicon plate and the polymer Polymer layers are bonded between two silicon plates which constrain the expansion of the polymer layer in the direction parallel to the plate surfaces Therefore, the thermal expansion of the polymer plate is enhanced in the perpendicular direction

Sketches of the microgripper based on a silicon comb structure are shown in Fig 2.3(a) The microgripper is designed for normal open operation mode Each actuator

has a silicon comb finger structure with the aluminum heater on the top A thin layer

of silicon nitride is used as the electrical isolation between aluminum structure and silicon substrate The gaps between comb fingers are filled with SU8 polymer Each actuator consists of 41 silicon fingers and SU8 layers in between The polymer comb fingers are 3 µm wide, 75 µm long and 30 µm thick The ratio between the width of the polymer layer and the dimensions of its bonded surface with silicon rigid plate are

25 and 10, respectively These ratio values satisfy the prerequisite for the maximum constraint effect

When the heater is activated, the generated heat is efficiently transferred to the surrounding polymer though the deep silicon comb finger structures which have a large interface area with the polymer layer The polymer layers expand along the lateral direction causing bending displacement of the actuator arm [4]

Figure 2.2: (a) Scheme of the silicon polymer laterally stacked segments; (b) Side cross-section of a

single segment depicting the constrained polymer between two rigid silicon plates [4]

Figure 2.3: a) Sketch of the silicon-polymer electrothermal microgripper based on the silicon comb

structure, b) top and cross-section view of the silicon-polymer electrothermal microgripper based on

the silicon comb structure [4]

2.4 Sensing microgripper

The sensing microgripper is based on the combination of a silicon-polymer electrothermal microactuator and a piezoresistive lateral force-sensing cantilever beam A schematic drawing is shown in Fig 2.4 When the electrothermal actuator is activated, the microgripper’s arm and also the sensing cantilever are bent This causes

a difference in the longitudinal stress on the opposite sides of the cantilever This changes the resistance value of the sensing piezoresistors on the cantilever changes The displacement of the microgripper jaws can be monitored by the output voltage of the Wheatstone bridge of the piezoresistive sensing cantilever beam The contact force between the microgripper jaws and clamped object is then determines based on displacement and stiffness of microgripper arm [4]

Figure 2.4: a) Schematic drawing of the sensing microgripper, b) Top and cross-section view of a

sensing microgripper arm with geometry symbols and parameters The Wheatstone bridge

configuration is also shown [4]

The sensing microgripper is designed for normal open operating mode, using the silicon-polymer electrothermal microgripper based on the silicon comb structure This configuration is more rigid than the one based on the silicon serpentine and therefore preferred

The force sensor is designed based on the lateral force-sensing piezoresistive cantilever beam as presented in 2.2 paragraphs The four piezoresistors are located on the cantilever beam structure and connected to create a Wheatstone bridge The piezoresistors are aligned along the [110] direction in the (001) crystal plane of the silicon wafer The resistor pairs located on the cantilever are the stress sensing resistors When the electrothermal actuator is activated, the cantilever beam is bent parallel to the wafer surface Therefore, the differential change of resistance occurs on

the two resistors R S1 and R S2

Figure 2.5: SEM picture of (a) the sensing microgripper and close-ups of (b) the piezo-resistors; (c)

the jaws, and (d) a section of the thermal actuator [4]

The Wheatstone bridge reduces the temperature influence on the output voltage from a first-order to a second-order effect, because both sensing resistors on a beam undergo the same temperature shift The two additional resistors outside the sensing cantilever are not subjected to stress They form a matched reference pair that makes the sensor signal more insensitive to common-mode external error sources, such as variations of the environmental temperature

The fabricated sensing microgripper is shown in Fig 2.5 The device is 490 µm long,

350 µm wide, 30 µm thick and with a 40 µm gap between the two jaws The piezoresistive force-sensing cantilever is 390 µm long, 10 µm wide with four piezoresistors on the surface The fabrication process is based on the DIMES bipolar process and the silicon-polymer actuator process The processes separately developed for the sensor and the actuator part of the system are CMOS-compatible and thus can

be easily combined

2.5 The sensing microgripper characteristics

2.5.1 Electrothermal actuator characteristics

Figure 2.6: The device operation: (a) initial position of the sensing microgripper jaws; (b) when

applied 4.5 V to both arms; (c) just before clamping; and (d) with clamped object (bonding wire) [4]

Fig 2.6 shows images of several typical positions of the microgripper jaws In Fig 2.6(a) the initial position is where the 40 µm gap between the two jaws can be seen The distance between the two jaws is closed to 8 µm when applying a voltage of 4.5 V

to both arms [see Fig 2.6(b)] Fig 2.6(c) and (d) illustrate the manipulation of a

23 µm bonding wire

Fig 2.7 shows the displacement response of the microgripper jaws in air when a DC voltage is applied to the electrothermal actuator This measured movement is the total change between the two microgripper jaw positions when both arms are activated The

measurement error is estimated to be ±1.5 µm The measured results come within

7.5 % of the simulated value A maximum movement of 32 µm is measured at applied voltage of 4.5 V Therefore, this microgripper is capable of manipulating a micro-object with a diameter between 8 and 40 µm

Figure 2.7: The simulated and measured sensing microgripper jaw displacement versus applied

voltage The maximum measured displacement is 32 µm at 4.5 V [4]

The power consumption is calculated by the applied voltage and the corresponding current on the electrothermal microactuators Fig 2.8 shows the measured of the jaw displacement with linear fitted and simulated values versus power consumption The experimental results come within 9 % the simulated values, indicating that the fabrication process and the simulated model behave as expected On average the device needs around 5 mW for a 1 µm displacement of the microgripper jaws

Figure 2.8: The simulated and measured sensing microgripper jaw displacement versus power

consumption [4]

The average increasing temperature in the electrothermal actuator ∆T ave can be

estimated by monitoring the change of the resistance of the aluminum heater

It is given by

Al T

R

T R T R T

act

act res act

1(

)()(

) 0

where αAl is temperature coefficient of resistance of aluminum film, R act (T 0) is the

resistance of the electrothermal actuator (205 Ω at room temperature), R act (∆T res) is the resistance of the actuator when the average temperature on the actuator is changed by

4.5 V, resulting in a maximum average temperature change of 176 0C Fig 2.9 shows the jaw displacement versus the average working temperature The experimental values meet the simulated ones within 7 %

The results of the thermal characterisation are also shown in Fig 2.9 The values obtained with the external heat mode come within 7 % and 5 % of the electrical and simulated ones, respectively It indicates once again that the aluminum depositing process behaves as expected and the average working temperature of the actuator can

be well estimated from the resistance change of the aluminum heater

As shown in Fig 2.9, this proposed device works on both the glassy and rubbery

plateau regions with the glass transition temperature T g of 120 0C [93] It explains the non-linear characteristic of the displacements versus power consumption and working temperature (see Figs 2.8 and 2.9)

Figure 2.9: The simulated and measured sensing microgripper jaw displacement versus average

working temperature [4]

2.5.2 Sensing cantilever beam characteristics

Figure 2.10: The output voltage of the force-sensing cantilever versus the applied voltage on the

electrothermal microactuator The inset shows the microgripper jaws with the clamped object [4]

Fig 2.10 shows the output signal of the Wheatstone bridge versus the voltage applied

to the electrothermal microactuator The measured zero-stress value of the piezoresistors at room temperature is 39 kΩ The bias voltage is 1V dc The maximum

output voltage is 49 mV at the applied voltage of 4.5 V The relation between the output voltage and the sensing microgripper jaw displacement is shown in Fig 2.11 The sensitivity of the sensing microgripper derived from this curve is 1.5 kV/m

Fig 2.10 also shows the output voltage of the piezoresistive force sensing cantilever when the microgripper grips a 23 µm diameter bonded wire as an object with an image

of the clamped object The sensing microgripper jaws close gradually until it grips the object

Figure 2.11: The output voltage of the force-sensing cantilever versus the displacement of the

sensing microgripper jaws [4]

The contact force between the microgripper jaws and the clamped object can be estimated by the jaw displacement in Fig 2.11 and by considering the simulated gripper arm stiffness of 1.8 kN/m Fig 2.12 shows the calculated contact force of this proposed microgripper The contact force is zero until the two gripper jaws reach the object at applied voltage of about 3.75 V The contact force then increases up to

135 mN at 4.5 V applied voltage Combining the measured results in Fig 2.10 and the calculated force results, the sensitivity of this built-in force-sensing microgripper is estimated to be 1.7 V/N

Figure 2.12: The contact force between microgripper jaws and the objects versus the applied voltage

2.5.3 Response frequency of the sensing microgripper

Fig 2.13 shows the measured voltage gain and phase shift as a function of frequency

of this sensing microgripper The cut-off frequency is defined as the frequency at which the voltage gain is attenuated by a factor of 0.707 (or -3 dB) and the phase shift

is 45 degrees Therefore, the measured cut-off frequency of this sensing microgripper

is 29 Hz (see Fig 2.13) The transient response of the full range displacement of this sensing microgripper is also characterised in Fig 2.14 The rise time (the time required for the response to rise from 10 % to 90 % of the final value) and the settling time (the time required for the response to reach and stay within 2 % of its final value)

is about 13 and 18 ms, respectively

Figure 2.13: The response frequency of the sensing microgripper [4]

Figure 2.14: The transient response of the sensing microgripper [4]

Chapter 3 Building PID control function

3.1 Feedback loop control

There are several schemes to be chosen for a control system It was start from initial idea for the microgripper that compatible with CMOS process, the control system is designed and integrated on chip with the microgripper of great advantages to manufacturing cost, setting displacement, transient response, voltage supply range, etc A unity feedback control system is needed as presented in Fig 3.1 In this figure, plant is a physical system to be controlled which consists of a driver circuit and the sensing microgripper; and controller provides an excitation for the plant which is designed to control overall system behavior

Figure 3.1: Block diagram of a unity feedback control [27]

The transfer function of a PID controller is given by:

s

K s K s K s K s

K

D

I P

++

=+

where: K P = proportional gain, K I = integral gain, and K D = derivative gain

Variable e represents the tracking error, the difference between desired input value X (desired position or status of a physical system) and actual output Y This error signal

e will be sent to the PID controller, and the controller computes both derivative and integral of this error signal Signal u just past the controller is now equal to the

proportional gain times the magnitude of the error plus the integral gain times the integral of the error plus the derivative gain times the derivative of the error, i.e., [27]

Applying this model to our case, the variable e is the difference between the voltage

setting and the feedback voltage and is sensed by the piezoresistors

3.2 Building a PID transfer function for the sensing microgripper system

3.2.1 Transfer function of sensing microgripper

Fig 2.13 shows the measured voltage gain and phase shift as a function of frequency

of this sensing microgripper using the lock-in amplifier The large-signal cutoff frequency of this sensing microgripper is measured as 29 Hz (ωP =200 rad/s) Thus,

the transfers function of the sensing microgripper in s domain:

1 005 0

1 1

1

1 )

3.2.2 Transfer function of driver circuit

Figure 3.2: Driver circuit

In order to control this microgripper a voltage source supply the actuator will be generated and controlled by the output of the PID controller There are many ways to get a voltage source controlled and depending on what is the level of input supply, possibilities are buck/boost for step down/up, fly-back, etc In our case, we only need

a controllable active device connected in serial to pass current from power supply The controllable device may be a MOSFET or bipolar transistor, but the process to fabricate our sensing microgripper was designed to be CMOS compatible and thus a MOSFET is chosen With MOSFET devices, the output voltage to supply the actuator

is controlled by the Gate voltage There are 2 kinds of MOSFETs: NMOS and PMOS

If a NMOS is used, the maximum output voltage is limited by the supply voltage minus VT (threshold voltage, about 0.7V) when the maximum gate voltage equals the

supply voltage With PMOS, the output voltage can go up to the supply voltage when using a single supply The final circuit (called Driver circuit) to generate voltage for the actuator is shown at Fig 3.2

The behavior of the driver circuit is like a RC circuit (the PMOS is a “dynamic” resistor) and the transfer function of the drive circuit is:

1

1 )

is changed from 0 to the maximum value depending on what is the setting point) The driver circuit requires higher current capability, but the more current capability the more area of the PMOS takes.Typically,R DS =50 Ω is reasonable For the capacitor C

which play the part of a filter, and just take a value equal 1 µF (in this case, the transient response time due to the value of the filter capacitor is ignored) Thus, we have the transfer function for the driver circuit:

1 00005 0

1 1

).

( 10 1 ).

( 50

1 )

0

1 1

00005 0

1

1 005 0

1 )

( ).

( )

+ +

= + +

=

=

s s

s s

s H s H s

H

Figure 3.3: Open-loop step response of the sensing microgripper and the driver circuit

Let's first view the open-loop step response of system shown at Fig 3.3 The DC gain

of the plant transfer function is 1/1 (at s=0), so 1 is the final value of the output to a unit step input Furthermore, the rise time is about 10 ms (from 10 % to 90% of rising edge), and the setting time is about 25 ms

3.2.4 Proportional control

The closed-loop transfer function of the above system with a proportional controller is:

) 1 ( 00505 0 00000025

0 )

P

P P

K s

s

K s

H

+ + +

=

With several trial values of proportional gain ( ) we have the step response of the control system shown at Fig 3.4 The plots show that the proportional controller reduced both the rise time and the steady-state error, increased the overshoot and decreased the settling time in comparison with the open-loop response When

, the rising and setting time reduces to about 2 ms and have no overshoot; the

DC gain is shifted down a bit to 10/11 When

3.2.5 Proportional – Integral control

For the given system, the close-loop transfer function with a PI control is:

I P

I P PI

K s K s

s

K s K s

H

+ +

+ +

+

=

) 1 ( 00505 0 00000025

0 )

Let's use the K P =200 and do some trial with the integral gain We will get the plots as shown at Fig 3.5 The integral gain also reduces the rise time and increases the overshoot as the proportional gain does, but it needs to have a huge integral gain ( ) to get a considerable changes In conclusion for this system there is no need to use the integral gain and we can keep the system control as simple as possible

I

K

I

K

Figure 3.5: Step response of closed-loop feedback proportional-integral controller in comparison with

proportional only controller

3.2.6 Proportional – Derivative control

The closed-loop transfer function of the given system with a PD controller is:

) 1 ( ) 00505 0 ( 00000025

0 )

P D

P D PD

K s

K s

K s K s

H

+ + +

+

+

=

Figure 3.6: Step response of closed-loop feedback proportional-derivative controller in comparison

with proportional only

The results for and with several values of are shown at Fig 3.6 Let’s compare the various cases one by one with the proportional control system The derivative gain will reduce the overshoot and rising time and make the system more stable (reduce number of ringing), but the more value of the more setting time need The optimum values for this system are

time = 20 ns, setting time =50 ns and there is no overshoot at all

03.0

=

D

K

3.2.7 Proportional – Derivative – Integral control

Now, let's take a look at a PID controller The closed-loop transfer function of the given system with a PID controller is:

I P D

I P D

K s K s K s

H

++

++

+

++

=

)1()00505.0(00000025

0)

But from the analysis before, the integral gain does not need to be a concern, and our system is only proportional and derivative to keep the system as simple as possible Let’s use K P =500 and K D =0.03 and the final transfer function is:

50103505

.000000025

0

50003

.0)

1()00505.0(00000025

0

)

2 2

2

++

+

=+++

+

+

=

s s

s K

s K s

K s K s

H

P D

P D

PD

The response of the close-loop control system in comparison with the open-loop is shown at Fig 3.8

Figure 3.8: Step response of close-loop feedback proportional-derivative controller in comparison

with open-loop response

By using the PD close-loop control scheme, we improve both rising time and setting time to about 500 times faster (rising time from 10 ms down to 20 ns, setting time from 25 ms down to 50 ns), and still keep the system stable

Chapter 4 Electrical Design

4.1 Introduction

This chapter presents the circuit of the PID controller and a description of the design strategy and simulation of each sub-circuit that composes the controller The design strategy is based on the functionality of each sub-circuit face to tradeoffs such as low current consumption, low supply, fast transient response, stability, etc

The design procedure of the electrical circuits which contained in our system will be demonstrated in Fig 4.1

The best advantage of the sensing micro gripper is CMOS compatible process so that

we can add sensing microgripper’s fabrication steps into a prior CMOS process and then that process is called CMOS-MEMS process The integration of MEMS and integrated circuits is a big step forward for the MEMS industry The CMOS-MEMS process can be divided into 2 categories, one is processes allowing MEMS before CMOS and the other is processes allowing CMOS before MEMS

These processes are evolving advantages and disadvantages but in this thesis will not take account of this field The process selection will be the CMOS selection because the microgripper is ready and compatible with any CMOS process

Back to our electric system, circuits are including analog and digital signal signal), high voltage and current The standard high voltage Bi-CMOS 1.2 µm process

(mix-with twin well (n-well), double poly, double metal was chosen Beside of hand calculating for each parameter of devices when design electrical circuit of the system

we will need to verified by simulation software Due to the need of simulation software, the model of every devices of the process was obtained which can give the results of simulation as close as happened on fact

Figure 4.2: The cross view of a device from CMOS-MEMS process [14]

4.2.2 Device modeling

Before one can design a circuit to be integrated in CMOS technology, one must first have a model describing the behavior of all the components available for use in the design A model can take the form of mathematical equations, circuit representations,

or tables

It should be stressed at the outset that a model is just that and no more—it is not the real thing! In an ideal world, we would have a model that accurately describes the behavior of a device under all possible conditions Realistically, we are happy to have

a model that predicts simulated performance to within a few percent of measured performance There is no clear agreement as to which model comes closest to meeting this “ideal” model By the fact that, HSPICE offers the user 43 different MOS transistor models from which to choose! In this thesis, the models of CMOS devices are provided by the foundry

4.2.3 Silicon-level simulation

The almost universally used program for silicon-level simulation is SPICE (‘Simulation Program, Integrated Circuit Emphasis’), and particularly many commercial derivatives such as PSPICE, HSPICE, etc SPICE was first developed at

the University of California at Berkeley in the early 1970s, and is based upon the small-signal AC equivalent representations of active and passive electronic devices These, in turn, are based upon the physical properties and terminal characteristics of each device, with particular emphasis on the nonlinear characteristics of the active devices (transistors) It should be appreciated that as the operating point (current and voltage) in a nonlinear device changes, so the small-signal equivalent circuit values also change, and hence it is necessary for a simulation program such as SPICE to perform a large number of simulation calculations with different operating points as the circuit currents and voltages vary

SPICE requires the following information to be given:

• the circuit interconnection details;

• R, C, L values of all linear devices and perhaps interconnections;

• details of all nonlinear devices (transistors, diodes and perhaps monolithic capacitors) and their parameter values at some chosen operating point;

• the DC supply voltage(s);

• the input signal(s) to the circuit;

• which output(s) or other details the program is to calculate and display

A particular feature of the SPICE program, and an indication of its power, is that it is not necessary to specify the operating points of each nonlinear device, nor is it necessary to give the small-signal parameter values over a range of operating points Instead, the program will first calculate the DC operating point of each device from the DC equations it knows for the device and from the circuit information that it has been given, and then will work out the small-signal parameters at this point from the known values at some other point In order to perform such calculations, the equations relating the small-signal equivalent values at one operating point to those at other operating points are built into the program, which will perform this revaluation of parameter values many times in a simulation run as circuit currents and voltages change

Because the full electrical performance of the active devices, together with all other circuit parameters, can be built into the SPICE simulation, very detailed simulation results can be obtained, the accuracy depending solely upon the accuracy of the model and the parameter values Every node of the circuit is evaluated for changing conditions at each simulation interval, which means a great deal of unnecessary calculation occurs in digital circuits which have a high degree of latency, for example, where a long counter is involved with only one or two stages changing state, and hence there is the need for the higher but less detailed gate-level or switch-level

simulation for most digital circuits of LSI complexity and above, as we have

previously considered [20]

4.2.4 Analog-only simulation

For analog-only circuits, the almost universal simulation tool is SPICE and its commercial derivatives The simulation accuracy of SPICE over a wide range of voltage, current, frequency and temperature makes it supremely relevant Its principal disadvantage, namely its inability to handle circuits of VLSI complexity due to the detailed calculations involved, does not arise in analog circuits, since it is rarely the case that more than a few hundred transistors at the most are involved

Most commercial SPICE-based tools provide a graphical as well as a textual output, with the capacity to display frequency characteristics, etc., in the conventional manner with which a designer is familiar The ease of changing component values or tolerances in a simulation and observing the results makes analog design a very much faster process than having to construct a physical breadboard layout, but this does not,

of course, relieve the designer from having to provide all the creative input

4.2.5 Mixed Analog/Digital simulation

For mixed analog/digital simulation it is exceedingly difficult to provide a single comprehensive CAD tool, largely because of the fundamentally different requirements

of fine detail but low component count of the analog parts and the coarser detail but high component count of the digital parts This is sometimes referred to as the ‘little-

A, big-D’ situation The present solution is either (i) to have separate analog and digital simulators and manually transfer data between them, or (ii) to have one system containing two separate simulators within one simulation environment, with the user programming which one to apply to the different parts of the circuit For example, a SPICE derivative may be used for the analog parts and switch-level or gate-level simulation for the digital parts However, for ‘little-A, little-D’ situations, for example uncommitted analog arrays which also have some limited digital capability, SPICE may be relevant for simulating the complete circuit, provided sufficiently comprehensive CAD computing power is available

In some commercial packages, the ability to move from one simulation tool to the other, for example to initialize circuit conditions before switch-level or gate-level simulation is started, or to analyze individual digital macros in greater detail than is provided by the digital simulator, may be particularly useful, capitalizing upon the feature that the circuit simulator analyzes every individual node in a simulation step, whereas the digital simulator processes only event changes

However, it is the method of organizing the interface between the two simulators that distinguishes different vendors’ approaches to this area This is made more complex if there is feedback between the digital and analog parts, for example in phase-locked

loops, than in the case where there is only a one-way transfer of data, for example in D-to-A and A-to-D conversion The five principal methods which have been adopted

to handle mixed simulation are as follows:

• unidirectional coupling of separate analog and digital simulators, running one

• similar to above but with the core simulator being a digital simulator with analog behavioral models for the analog parts

In the first three of the above categories, an appropriate boundary interface must be provided to link the two quite separate types of simulator This data link tends to be vendor-specific, as are the behavior models used in the last two of the above categories

Among the difficulties in linking the analog and digital simulations is the dissimilarity

of information in the two domains For example, an analog circuit driving a digital gate requires knowledge of the gate input impedance characteristics before the analog output signal can be accurately determined, but this is not usually available in digital circuit modeling Conversely, in the digital simulation, discrete information on signal driving strengths and on unknown conditions is required, which is foreign to analog simulators

A single mixed-mode simulator with one unified modeling language which can handle both analog descriptions and digital descriptions, and with hierarchical ability to cover primitive to architectural level, would clearly be advantageous

Considerable research work is continuing in this area, but the fundamental difference between analog and digital simulation will always remain

The graphics available for mixed analog/digital simulation, however, are already very useful [20]

4.3 System block diagram

The sensing microgripper with controller containing all the active function required for an automatic micro-manipulation system This device contains a sensing microgripper and its control circuit that was designed on the same die of a MEMS-