The online journal of IDEMA,

the global trade association for the disk drive industry Q2 2005 Features:

• GMR and TMR Corrosion Issues and Solutions for Magnetic Read-Write Devices, by Donald C. Zipperian, Ph.D, Pace Technologies, and Professor Srini Raghavan, University of Arizona

• Inertial Sensing for Hard Disk Drive Drop Protection, by Gregory J.

Galvin, Ph.D., President and CEO, Kionix, Inc. • DISKCON USA, September 19-22, Santa Clara, CA Charity for Youth Golf Tournament, September 19

Conference and Tradeshow, September 20-21 Keynote Dinner featuring Mike Wingert, Maxtor President & COO, Sept. 20 Global Standards Meetings, September 22

ARTICLE: GMR and TMR Corrosion Issues and Solutions for

Magnetic Read-Write Devices By Donald C. Zipperian, Ph.D., Pace Technologies, and Professor Srini Raghavan, University of Arizona

INTRODUCTION Magnetic read-write storage devices, in addition to computer hard drives and

magnetic tape, are now being incorporated into many commercial multimedia

information and communication devices. Driving this commercialization has been

the continued 60-80% growth rate per year in the areal recording density. The

technological developments utilized to reach the present areal density levels of 100+

Gbits/in2 have placed stringent demands on the manufacturing processes for these

devices. The primary driving force for the present technology and for the future

growth has been the continued development of giant magnetoresistive (GMR) spin-

valve heads and the next generation tunneling magnetoresistive (TMR) heads.

- 1 -

Figure 1a shows the air bearing surface (ABS) for a magnetic hard drive read-

write head or magnetic slider. The air bearing side is designed in order to produce a

stable and minimal air cushion which positions the magnetic head above the spinning

magnetic disk. Figure 1b shows the profile of an MR magnetic tape head. The true

“business end” of the magnetic head is the pole-tip; this is where the actual data

reading and writing occurs.

Figure 1a: Disk Drive Magnetic Head (slider) Air Bearing Surface (ABS)

Figure 1b: Tape Magnetic Head

Figure 2a shows an SEM image of a typical hard drive pole-tip and Figure 2b

shows the pole tip for a magnetic tape head. The pole-tip consists of a reading and

writing element. The writing element is an inductive magnet and writes data to the

- 2 -

media disk by converting an electrical signal into magnetic flux, which produces a

magnetized orientation change on the media disk. Manufacturing of the inductive

head and the shields is accomplished by electroplating processes during wafer

manufacturing.

Figure 2a: Disk Drive Magnetic Pole Tip

Figure 2b: Tape Head Magnetic Pole Tip

The reading sensor measures the change in the magnetic orientation on the

media disk by measuring the change in magnetism or its change in resistance by the

magnetoresistive (MR) effect. The efficiency of the magnetoresistive effect is

characterized by the magnetoresistive ratio, which is defined as the maximum

resistance change divided by the resistance at magnetic saturation. For GMR heads,

the GMR ratio for a 100 Gbit/in2 spin-valve head is less then 0.20 and typically about

0.15 (1). For TMR heads, the TMR ratio can be as high as 0.70 (2,3). The current

technological advances for the magnetic hard drive head are focusing on the

- 3 -

- 4 -

conversion from GMR to TMR devices, whereas the tape head industry is presently

converting from MR to GMR technology. The other significant difference between the

hard drive head and tape head is the size of the writing device. For tape heads the

dimensions of the writer pole are much larger.

The GMR spin-valve film layer consists of four laminated metal layers: an

anti-ferromagnetic pinning layer, a pinned layer whose magnetization is fixed by an

exchange-coupling field from the anti-ferromagnetic layer, a copper interlayer, and a

free layer whose magnetization rotates according to the signal field from the disk

media (1). The TMR structure consists basically of two ferromagnetic thin film layers

separated by a thin tunnel barrier. The TMR effect is very dependent upon the

barrier layer thickness and decreases rapidly with increasing thickness. For example,

the TMR ratio is 0.70 for GaMnAs/AlAs/GaMnAs tunnel junctions having a barrier

layer thickness of 1.5 nm, but decreased to 0.20 as the barrier thickness increased

to 1.8 nm (3).

The evolution of the GMR spin-valve head and the corresponding increase in

the GMR ratio is a result of thinner film layers with tighter tolerances, an increased

number of film layers, and the incorporation of a wider range of magnetic and non-

magnetic metals into the stack layers. Figure 3 shows one of the more common

GMR structures (4). The thickness of these layers can range from angstroms to

several hundred angstroms, requiring angstrom level tolerances. These GMR and

TMR thin films are manufactured with vacuum ion beam deposition and ion beam

etching techniques. Future generation GMR and TMR heads will consist of various

metal and alloy combinations of iron, nickel, manganese, tantalum, iridium, copper,

chromium, ruthenium, zirconium and cobalt. The deposition thickness of these

metals will also continue to shrink.

Figure 3: Common GMR Stack Materials

Magnetic Head Corrosion Issues

Corrosion has always been an issue for the complex manufacturing of

magnetic heads; however, as the number of metal layers and their reduction in

thickness continues to evolve, the corrosion issue will become more complex and

critical. Not only does corrosion affect the drive performance, it also reduces yields

and therefore increases manufacturing costs. To make matters more difficult,

corrosion in the past has been easy to see by visual inspection of the writing pole

and shields; however the dimensions of the magnetoresistive stack make it

impossible to visually observe and therefore characterize or measure. Corrosion has

been determined to occur primarily during the slider manufacturing process, where

the materials are exposed to a number of manufacturing solutions and chemistries

via the cutting, lapping, cut-rate enhancing, cleaning and rinsing operations.

Figure 4 shows the potential manufacturing steps where corrosion can occur.

Typical in-line quality control for corrosion at the lapping, transfer, and cleaning

operations are minimal and involve rejecting affected heads rather than focusing on

corrosion prevention or reduction. As a first pass, optical inspection of the writer has

- 5 -

been used to discard corroded heads. In addition, electrical and magnetic testing

eliminates the magnetoresistive heads adversely affected by corrosion.

Figure 4: Potential Areas for Magnetic Head Corrosion

Figure 5 shows the SEM micrographs of corrosion pits on the write pole,

shared pole and shield metals as a result of accelerated corrosion testing. The

metals used for the write head are typically alloys of nickel, iron and cobalt.

Intuitively, it makes sense that any observable corrosion on the larger writer, shared

and shields cannot be very promising for the more sensitive magnetoresistive stack

metals, which require TEM analysis in order to observe the sensor metal layers.

- 6 -

Figure 5: SEM Micrograph of Corrosion Pits in the Share Pole (original magnification 10,000X)

Figure 6a shows both the SEM and TEM image of a good slider as compared

to the corroded slider shown in Figure 6b. The TEM images show that the corroded

slider has a missing layer in the GMR stack.

Figure 6a: TEM Micrograph of a Good Slider

- 7 -

Figure 6b: TEM micrograph of a corroded slider

Dissimilar Metals Corrosion

The type and number of metals and alloys being incorporated into the

GMR/TMR stack offers a plethora of opportunities for galvanic and electrolytic

corrosion for the magnetic hard drive and tape read-write heads. Galvanic corrosion

occurs when two metals having different electrochemical characteristics contact each

other in an aqueous or semi-aqueous solution. Table IA shows the reduction

potential difference between the metals increases potential of the metals used in the

GMR or TMR stack at a metal ion activity of 10-6 and a pH value of 5.7 (distilled

water). The metals with a more negative reduction potential will corrode

preferentially to the more noble or higher reduction potential metals when in contact

with each other. Galvanic corrosion is also more prevalent as the reduction potential

difference between the metals increases. - 8 -

Table 1A: Reduction Potentials in Distilled Water Based on the reduction potentials shown in Table IA, the newer metal

additions of manganese, ruthenium, zirconium and chromium into the GMR and TMR

stack significantly increase the susceptibility for galvanic corrosion over the more

traditional iron, nickel and cobalt metals, which incidentally have comparable

reduction potentials to each other. Alloying these metals also complicates the

galvanic corrosion issue even more. Additional factors that affect galvanic corrosion

include: the area ratio of the metals and the bulk solution properties such as oxygen

content, pH, and solution conductivity.

The variations in surface area ratios for the thin films are very small in

comparison to the size of the lapping plates used for the finish and kiss lapping

steps. Ideally the best corrosion scenario would be to have a large anodic mass for a

lapping surface, which would provide a level of galvanic corrosion protection

(sacrificial anode). The most undesirable scenario is to have a small anodic surface

(thin film stack metals) in contact with a large cathodic lapping plate. The primary - 9 -

lapping plate materials used to lap magnetic sliders are bismuth/ tin, tin/ antimony

(EBismuth = 0.30 volts, Etin = -0.31 volts Eantimony = -0.19 volts), zinc (E = -0.94 volts),

and copper (E = 0.16 volts) (Table IB). Zinc is a much better choice of lapping

plate metals because it has a reduction potential closer to the more reactive metals

of zirconium, manganese and ruthenium. Zinc lapping plates, although less

common, have been used for both tape head and magnetic head lapping because

they have very similar hardness and recrystallization characteristics to tin lapping

plates (5).

Table IB: Reduction Potentials of Lapping Plates in Distilled Water

Pourbaix Diagrams

A useful way to study the relationship or susceptibility for metal corrosion

based on pH and oxidation potential is with electrochemical equilibrium diagrams

(Pourbaix Diagrams). Pourbaix diagrams are thermodynamic plots of solution

potential vs. pH. Figures 7a-7d show the Pourbaix diagrams for the magnetic metals

of iron, cobalt, nickel and manganese. Cobalt, nickel and manganese are shown to

form ionic species or corrosion products at pH values typically below 9. At pH values

above 9-10 passive oxide species are more characteristic of cobalt, nickel and

manganese. For the iron-water system, the Pourbaix diagram indicates that the

more stable ionic species for iron at lower pH values is Fe+2, with Fe2O3 iron oxide

more stable at high pH values. Based on these diagrams one might be led to believe

that the corrosion of these magnetic materials could be eliminated by simply

maintaining processing conditions at high pH values. This is somewhat deceiving

because the iron oxide species do not form continuous or passivated oxide coatings.

- 10 -

In fact, at higher pH values, iron corrosion may be more aggressive because of

localized pitting corrosion.

Figure 7a: Pourbaix Diagram for Iron-water System (metal ion activity 10-6)

Figure 7b: Pourbaix Diagram for Nickel-water System (metal ion activity 10-6)

- 11 -

Figure 7c: Pourbaix Diagram for Cobalt-water System (metal ion activity 10-6)

Figure 7d: Pourbaix Diagram for Manganese-water System (metal ion activity 10-6)

Figures 8a-8e show the Pourbaix diagrams for tantalum, iridium, chromium,

ruthenium and zirconium. These metals are typically covered with an oxide layer in

the presence of water. It would not be anticipated that these metals would have an

electrochemical corrosion issue in the presence of water; however, when subjected

- 12 -

to grinding or lapping the passivated surface is continually removed and can actually

result in accelerated corrosion of these metals. As shown by Fang and Raghavan

(6), the galvanic corrosion potential can increase significantly during the abrasion

process. They demonstrated an accelerated corrosion rate for the aluminum and

titanium galvanic cell when subjected to abrasion.

Figure 8a: Pourbaix Diagram for Tantalum-Water System (metal ion activity 10-6)

Figure 8b: Pourbaix Diagram for Iridium-water System (metal ion activity 10-6)

- 13 -

Figure 8c: Pourbaix Diagram for Chromium-water System (metal ion activity 10-6)

Figure 8d: Pourbaix Diagram for Ruthenium-water system (metal ion activity 10-6)

- 14 -

Figure 8e: Pourbaix Diagram for Zirconium-water System (metal ion activity 10-6)

Figure 8f is the Pourbiax diagram for the aluminum water system. As seen,

aluminum, or perhaps the AlxOy undercoat and/or overcoats, form a stable passive

oxide in the pH range of 4-10. However, at higher pH values greater than 10, the

ionic aluminum species of AlO2- becomes more predominate. Consequently, high pH

cleaning solutions for alumina coatings formed by oxidizing aluminum targets can

result in significant overcoat recession, especially if a true alumina is not formed.

Figure 8f: Pourbaix Diagram for Copper-water System (metal ion activity 10-6)

- 15 -

Figure 9 shows the Pourbaix diagram for copper. At basic pH values, copper

can form stable passivated oxides. At lower pH values, especially at higher oxidation

potentials, copper can corrode.

Figure 9: Pourbaix Diagram for Copper-water System (metal ion activity 10-6)

For all of the Pourbaix diagrams shown in Figures 7-9, the pH and oxidation

potential stability region for distilled water has been added. The typical pH for

distilled water ranges from 5.5 to 6 depending upon the quality of the distilled water,

which naturally absorbs carbon dioxide from the air to form carbonic acid. This

absorption is what gives distilled water an acidic pH. Table II lists the feasibility of

the head materials to corrode in the presence of distilled water. Thus, distilled

water can have a very significant and negative impact on many of the magnetic

read-write head metals (Table II).

- 16 -

Table II. Corrosion rate summary in distilled water without corrosion inhibitor

Pourbaix diagrams are useful for analyzing a metal’s thermodynamic

corrosion potential in water; however, they do not provide a means to determine the

corrosion rate. Pourbaix diagrams thus, provide a sound theoretical start; however,

predictive corrosion rates need to be simulated with the use of electrochemical

polarization curves. Polarization curves are derived by plotting current density

verses the degree of oxidation or potential change. The most common technique for

determining the corrosion rate is based on the Tafel equation.

Corrosion Rates of Metals used in MR Stacks

The corrosion rate of metals can be estimated using electrochemical

polarization technique. Figures 10-14 show the polarization curves for manganese,

cobalt, iron copper, and nickel in a glycol/ surfactant based lubricant at a pH 7.5 with

and without the addition of a corrosion inhibitor.

- 17 -

Figure 10a. Tafel Plot for Mn without inhibitor

Figure 10b. Tafel Plot for Mn with inhibitor

- 18 -

Figure 11a. Tafel Plot for Co without inhibitor

Figure 11b: Tafel Plot for Co with inhibitor

- 19 -

Figure 12a. Tafel Plot for Fe without inhibitor

Figure 12b: Tafel Plot for Fe with inhibitor

- 20 -

Figure 13a. Tafel Plot for Cu without inhibitor

Figure 13b: Tafel Plot for Cu with inhibitor

- 21 -

Figure 14a: Tafel Plot for Ni without inhibitor

Figure 14b: Tafel Plot for Ni with inhibitor

Table III summarizes the estimated corrosion rates for these metals with and

without the inhibitor. The corrosion rate for iron, nickel, cobalt and copper are all

less than 10 angstroms per minute (0.04 µinches/minute). Clearly, the most

reactive metal is manganese, with a corrosion rate of 124 Angstroms per minute (0.5

µinches/minute) without the addition of the corrosion inhibitor. To complicate

matters, present fly heights are approximately 0.5 µinches and PTR specifications are

approximately 0.01-0.02 µinches and are measured from the larger and less

corrosion sensitive shared and write pole metals of iron, nickel or cobalt. These

metals are 50-100X less reactive than manganese, corroding at a rate of 0.04

- 22 -

µinches/minute. In addition, magnetic heads are typically rejected for corrosion

based on visible inspection (pitting) or discoloration of the shared and write poles. It

makes sense then, that any observable corrosion at visual inspection cannot be very

promising for the more sensitive magnetoresistive stack metals.

Table III. Corrosion rate comparison with and without organic corrosion inhibitor

With the incorporation of an azole and/or thiazole corrosion inhibitor additive

(PACE Technologies patent pending), the effective corrosion rates for the metals

tested were reduced by more than 90%. In particular, the corrosion rate of

manganese was decreased to 2 angstroms per minute with the corrosion inhibitor

verses 124 angstroms per minute without inhibitor. This is very significant, because

this corrosion rate is about the same as observed for the traditional iron, nickel and

cobalt metals used in the successful manufacturing of past generation MR and GMR

heads.

The advantages of using organic corrosion inhibitors such as the azole and/or

thiazole compounds is that they can be added to the present manufacturing

chemistries and/or and as an intermediate or transfer dip (Table IV). The inhibitors

act by adsorbing onto the metal surface, thus providing a protective barrier to the

corrosive environment. The advantages of corrosion inhibitors include:

♦ Presence of film prevents uniform corrosion attack

♦ Increases the activation energy on the metal surface (passivation)

♦ Eliminates corrosion over a wide range of pH and solution oxidation values

- 23 -

Table IV. Corrosion rate summary with and without inhibitor

CONCLUSIONS

Corrosion of GMR and TMR devices is complicated because it can occur by a

number of known corrosion mechanisms. The two discussed in the scope of this

article include galvanic and uniform corrosion. In order to minimize galvanic

corrosion during lapping, where the surface of the reactive metals are continually

being exposed, the use of zinc lapping plates provides a large sacrificial anode area

for corrosion protection. Zinc also has a similar hardness and recrystallization

temperature to the traditional tin alloy lapping plates.

Azole and thiazole organic corrosion inhibitors (patent pending) are very

effective for reducing the corrosion of magnetoresistive metals. These corrosion

inhibitors can be incorporated into the various cutting, lapping and cleaning

processes or applied immediately after the more corrosion susceptible processing

steps. The inclusion of the corrosion additive to the individual processing chemicals

and solutions in contact with the reactive metals and alloys offer the greatest level of

protection for both galvanic and electrochemical corrosion. The addition of the

organic corrosion inhibitor provides a constantly regenerating protective coating on

the reactive metals. A secondary benefit of the organic inhibitors is that they may

equalize the sensor metal recession created by galvanic or electrochemical metal

- 24 -

- 25 -

recession. Manganese is a good example of a metal where the corrosion rate can be

comparable or greater than the mechanical removal rate without the addition of a

corrosion inhibitor. Subsequently, the more reactive metals, such as manganese

would have at greater sensor recession and thus alter the magnetic head’s reading

sensitivity. Additionally, because of the inability to easily, accurately, and reliably

measure this recession, the addition of the corrosion inhibitor to the diamond

slurries, lubricants and cleaning solutions is certainly warranted. For added

protection, a dip, spray or immersion soak in the inhibitor after each process step

would provide an additional measure of corrosion protection for the in-between

processing steps.

ACKNOWLEDGEMENTS

The authors would like to thank Dr. Dallas Meyer, Paul Tyler and Ray Moudry from

Seagate Technology; Chakkrit Supavasuthi and Ittipon Cheowanish from Western

Digital, Thailand; Li Jian from LAFE Group Hong Kong; Chip Ragan from Veeco

Instruments; and Ashok Kumar Muthukumaran from the University of Arizona for

their contributions to this publication.

REFERENCES

1. “Advanced Spin-Valve GMR Head.” Hitoshi Kanai, Kenji Noma and Jognill

Hong, Fujitsu Sci. Tech. J, 37, 2 (December 2001), p. 174-182.

2. “Cell Shape and Patterning Considerations for Magnetic Random Access

Memory (MRAM) Fabrication.” Robert Ditizio, Paul Werbaneth and Jian-Gang

Zhu, Semiconductor Manufacturing Magazine, January 2004.

3. “Large Tunneling Magnetoresistance in GaMnAs/AlAs/GaMnAs Ferromagnetic

Semiconductor Tunnel Junctions.” Masaaki Tanaka and Yutaka Higo, RIKEN

Review, No. 33 (March 2001) pp 35-38.

4. “Advanced Sensor Fabrication Using Integrated Ion Beam Etch and Ion Beam

Deposition Processes.” John Londono, Kurt Williams, and Adrian

Devasahayam, Veeco technical note,

http://www.veeco.com/appnotes/AdvanSensorApplNotes12_00.pdf

5. “Zinc A-10 Lapping Plates.” PACE Technologies,

http://www.metallographic.com/adobe/zinc.pdf.

6. “Electrochemical Investigations during the Abrasion of Aluminum/Titanium

Thin-Film Stacks in Iodate-Based Slurry.” Yan Fang and Srini Raghavan

Journal of the Electrochemical Society, 151 (12) G878-881 (2004).

About the Authors:

Donald C. Zipperian, Ph.D. is the Chief Technical Officer for PACE Technologies. Dr. Zipperian’s background has been in the areas of precision surface finishing and electrochemistry. At PACE Technologies he overseas the lapping and surface finishing developments for the data storage and industrial precision surface finishing markets, which include the development of lapping plates, corrosion inhibitors, lapping and cutting lubricants, diamond slurries, etc. His previous industry experience includes first hand knowledge of the slider lapping process as manager of the Advanced Surface Finishing Group at Seagate Technology and as the Director of Data Storage Marketing for Veeco Instruments, where he was involved with the development of metrology tools for 100% slider inspection

following the lapping process.

(donz@metallographic.com) (www.metallographic.com)

Dr. Srini Raghavan is a professor of materials science and engineering at the University of Arizona in Tucson. His research interests are in the areas of wet chemical processing in IC manufacturing, chemical mechanical planarization and corrosion and degradation of materials. He is a principal investigator in the NSF-SRC Center for Environmentally Benign Semiconductor Manufacturing at the University of Arizona. He teaches classes in Semiconductor Processing,

Applied Surface Chemistry and Corrosion and Degradation.

- 26 -

- 27 -

ARTICLE: Inertial Sensing for Hard Disk Drive Drop Protection, by Gregory J. Galvin, Ph.D., President and CEO, Kionix, Inc. HDD Protection — A Market Imperative

The worldwide growth rate for portable electronic products, which are more

and more likely to contain a hard disk drive (HDD), is impressive. Earlier this year,

the IT market research specialists at IDC reported that worldwide mobile phone

shipments totaled 194.3 million units in the fourth quarter of 2004; that 46.8 million

digital music players were sold in 2004, with a projection of 132 million in 2009; and

that worldwide PDA shipments totaled 3.4 million units in the first quarter of 2005.

USA Today, in an April 14 special report, indicated that nearly 49 million laptops

were sold worldwide in 2004, almost double the number sold in CY 2000.

These numbers alone add up to a compelling requirement for an active hard

disk drive drop-protection feature in highly portable, “always-on” products. Data

loss, and its resulting impact on productivity, adds another persuasive argument to

the critical need for HDD drop protection.

Objectives and Complexities

The objective in HDD drop detection is to reliably detect, by means of an

inertial sensor (accelerometer), an object in free fall and to signal the drive head to

park safely prior to impact. Very simply, free fall is the descending motion of an

object subject only to gravity.

Drop detection is a simple notion that involves a complicated detection

process, largely because portable electronic products are subject to complex motion

during use. For example, we want to signal free fall when a device is at risk and

tumbling toward the ground. But, we do not want to signal free fall when the device

is responding to typical-use events such as running, where low-g periods are long

enough to look like free-fall, and dancing, where high-g periods can reach 4.2g.

Thus, the process must be capable of distinguishing between typical-use motion and

a genuine fall, so as to not trigger a false positive. Consumers grow tired quickly of

- 28 -

a protection feature that disables a portable computer when the user adjusts its

tabletop position.

The process is further complicated by irregular real-life motions such as

tapping or shaking, the requirements for positioning the motion sensor, the

complexities associated with threshold detection techniques, and the changes in the

center of mass on products such as flip phones and laptops when open or closed.

The Physics of Free Fall

In a simple free fall event, since a = 1 g (9.8 m/s2) and initial velocity is 0 (v0

= 0), the equation of motion can be used to determine the position of an object at

any time after it is dropped.

Equation of Motion z = z0 + v0t + ½ at 2

If the initial height from the ground (z0) is known and the initial velocity (v0) is

zero, then the time it takes for an object to impact the ground is t = (2 z0/g)½.

Therefore, from a height of one meter, an object will take 0.45 seconds (450

milliseconds) to impact the ground.

Unfortunately, free fall is not always simple because hand-held devices rarely

fall without incurring a tumbling spin, thus complicating the physics. A spinning

object is subject to acceleration that can be significant in its centripetal and

centrifugal forces. Centripetal force is that which holds the object in the center of

the spin; centrifugal force is Newton’s “equal and opposite reaction,” displacing the

object from the center of the spin.

Accelerometers for Drop Detection

Silicon micromechanical technology has enabled the production of low-cost,

small form-factor accelerometers capable of detecting linear acceleration on one, two

or three axes. These devices are etched from single-crystal silicon and function on

the principle of differential capacitance. Essentially, it is a mass on a spring. The

mass is capable of motion relative to the substrate and moves in response to

acceleration. This movement results in a change of capacitance that is detected and

transformed into an electrical signal by an Application Specific Integrated Circuit

(ASIC).

An accelerometer at rest measures 1g of acceleration and an accelerometer in

simple free fall measures 0g of acceleration, no matter their orientation. Simple free

fall suggests a very simple drop detection algorithm: When total acceleration equals

0, the object is in free fall.

The “signature” of a simple drop is shown in Graphic 1. The signature depicts

the fall from the object’s pre-drop (1g) condition to its post-drop (1g) condition.

0

0.5

1

1.5

2

2.5

3

3.5

4

0.00 0.50 1.00 1.50 2.00 2.50

Time (seconds)

Acc

eler

atio

n (g

)

Impact

Held in Hand DropFirst BounceBouncing Resting on Floor

Accelerometer registers 0g during drop.

0

0.5

1

1.5

2

2.5

3

3.5

4

0.00 0.50 1.00 1.50 2.00 2.50

Time (seconds)

Acc

eler

atio

n (g

)

Impact

Held in Hand DropFirst BounceBouncing Resting on Floor

Accelerometer registers 0g during drop.

Graphic 1. Simple free fall of a cell phone.

In the above case, a Kionix KXM52-1050 tri-axis accelerometer was placed in

a cell phone approximately 3 cm from the center of mass and flat-dropped to a hard

surface from approximately one meter. The accelerometer, as expected, measured

at or near 0g acceleration prior to impact in this simple free-fall experiment.

This same phone was dropped again to a hard surface but was subjected to a

spin of approximately 4 revolutions per second, a more accurate real-life scenario.

As can be seen in this drop signature (Graphic 2), the accelerometer never

registered 0g during the entire fall. Rather, it measured over 3.5g during the drop

- 29 -

with spin. A similarly-mounted tri-axis accelerometer, located 5 cm from the center

of mass on a popular MP3 player mock-up, produced the same impact profile,

registering over 3g of acceleration as it spun out of control prior to impact.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0.0 0.5 1.0 1.5 2.0 2.5

Time (seconds)

Acc

eler

atio

n (g

)

Held in Hand Drop with Spin

First Bounce

Bouncing Resting on Floor

Impact

Accelerometer never registers 0g during drop with spin.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0.0 0.5 1.0 1.5 2.0 2.5

Time (seconds)

Acc

eler

atio

n (g

)

Held in Hand Drop with Spin

First Bounce

Bouncing Resting on Floor

Impact

Accelerometer never registers 0g during drop with spin.

Graphic 2. Free fall of a cell phone with spin. Gauging Response Times

Damaging free-fall events consume very little time, but sufficient time is

needed to execute hard drive protection. The accelerometer and associated

microcontroller must recognize that an event is occurring and determine the nature

of the event—free fall or not—and, if appropriate, trigger the hard drive protection

feature while still leaving enough time to park the read/write head in safety. The

challenge is to accomplish this within fractions of a second.

Table 1 presents the time it would take for an object to impact a hard surface

from a range of fall heights.

- 30 -

Height From W hich an Object

Falls (m)

Time Consumed from Fall to Impact (ms)

Maximum Force Experienced (g)

0.01 45.160.02 63.870.1 142.810.2 201.96 2000.3 247.350.4 285.62 4000.5 319.33 5000.6 349.810.7 377.840.8 403.92 10000.9 428.431 451.60 1200

1.1 473.64 15001.2 494.701.3 514.90 20001.4 534.341.5 553.10 28001.6 571.23

Table 1: Fall heights and response times.

Consider how a product might be protected from a 400g-impact threat. To

sustain a shock of 400g, a product need be dropped only from a height of 0.4

meters, and the entire event, from fall to impact, will take approximately 285.62

milliseconds.

The KXM52-1050 accelerometer can determine the onset of free fall in four

milliseconds. If an additional 70 milliseconds is required to park the hard drive head,

there is only a budget of 211 milliseconds during which the microcontroller must

discern free fall and signal the park.

Signaling the Drop Interrupt

A tri-axis accelerometer can reliably detect free fall. Once detected, the

motion of the spinning hard drive must be interrupted, and the drive head must be

signaled to park in safety. This critical function must occur at the appropriate time—

at the point of actual free fall—within fractions of a second and without a misread.

- 31 -

The following considerations are key to determining the correct timing for interrupt

and the choice of low-g or high-g thresholds for triggering the interrupt:

1. Maximum g-forces experienced by the object during typical use.

2. Minimum g-forces experienced by the object during typical use.

3. The length of time an object spends at these levels (dwell time).

4. The maximum allowable g-force on the object.

In other words, we need to determine the maximum and minimum g-force

points along with the length of time spent at these levels in order to determine the

appropriate point to signal a drop interrupt. The time spent above or below a certain

g-force level is the key to differentiating between a typical-use event and a tumbling

drop.

The simple free-fall detection algorithm—an object is in free fall when

acceleration on all three axes equals 0—works only if the object does not spin and/or

the accelerometer is placed at the center of mass. If this placement is not possible,

given the location of the hard drive in a particular product, more complex high-

g/low-g threshold algorithms are required to accurately detect free fall.

The centrifugal force experienced by an object during a tumbling fall is read

by the accelerometer as a high-g event, as depicted earlier in Graphic 2. But,

accelerometer readings can show high-g and low-g events in typical-use activities

such as running and dancing, as depicted in Graphic 3 below.

0

0 .5

1

1 .5

2

2 .5

3

3 .5

0 .0 0 0 . 50 1

Acc

eler

atio

n (g

)

Graphic 3.

Graphic 3. Response Times vs. Height

0

0 .5

1

1 .5

2

2 .5

3

3 .5

4

4 .5

0 . 0 0 0 .5 0 1 .00 1 .50 2 .0 0 2 .5 0 3 .00

D a n c in g

. 0 0 1 .5 0 2 . 00 2 . 5 0 3 .00

T im e (se c o n d s)

R u n n in g

G-force analysis during typical use activities, running and dancing.

- 32 -

Again, the key to differentiating between typical use and a tumbling drop is

the time spent above or below a certain g-force level (dwell time). Graphic 4 below

presents the maximum dwell times for a simple drop, a drop with spin, dancing and

running for a 400g threat.

Table 4. G-Force analysis during typical use activities and tumbling drop.

Max Dwel l Times at g Leve ls

0

50

100

150

200

250

300

350

400

450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2 2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3 3.1

3.2

3.3

3.4

3.5

Force (g)

Tim

e (m

s)

Drop w Spin

Dance

Running

Simple Drop

Max Dwel l Times at g Leve ls

0

50

100

150

200

250

300

350

400

450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2 2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3 3.1

3.2

3.3

3.4

3.5

Force (g)

Tim

e (m

s)

Drop w Spin

Dance

Running

Simple Drop

.

In the abov

for any event lower

for longer than 211

a simple drop (low-

If the objec

operating impact o

threshold would be

dwell time of 211

minimum and max

trigger a park prio

during typical use.

Graph 5 pre

spin, dancing and r

interrupt is triggere

Graphic 4. Maximum dwell times for a 400g threat

e depiction, the “algorithm” is set to trigger a free-fall notification

than 0.7g for longer than 211 ms or any event higher than 1.5g

ms. In this case, the interrupt will be triggered appropriately by

g case) or by a drop with spin (high-g case).

t can be used in an active environment and will withstand an

f 400g, the accelerometer settings would be as follows: The low-g

set at .5g with a minimum dwell time of 175 ms or a maximum

ms, and the high-g threshold would be set at 1.5g with the same

imum dwell times. Set at these levels, the accelerometer will

r to an impact of 400g or greater and will not park the heads

sents the maximum dwell times for a simple drop, a drop with

unning for a 200g threat with a wait time of 127 ms before

d.

- 33 -

0

50

100

150

200

250

300

350

400

450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2 2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3 3.1

3.2

3.3

3.4

3.5

Force (g)

Tim

e (m

s)

Drop w SpinDance

Running

Simple Drop

0

50

100

150

200

250

300

350

400

450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2 2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3 3.1

3.2

3.3

3.4

3.5

Force (g)

Tim

e (m

s)

Drop w SpinDance

Running

Simple Drop

0

50

100

150

200

250

300

350

400

450

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1 1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2 2.1

2.2

2.3

2.4

2.5

2.6

2.7

2.8

2.9

3 3.1

3.2

3.3

3.4

3.5

Force (g)

Tim

e (m

s)

Drop w SpinDance

Running

Simple Drop

Graphic 5. Maximum dwell times for a 200g threat. At these levels and these dwell times, as depicted in Graphic 5 (see red circles),

some typical-use events will cause a false positive.

Conclusions

The convergence of larger and larger capacity hard disk drives in ever

smaller, more portable products is creating a compelling need for drop protection

features. Moreover, the environment for portable platforms is rapidly becoming one

in which these products are operating while in motion, again making drop protection

features an imperative.

Accurate and reliable free-fall detection, with adequate time to park the disk

head prior to impact, is possible with a tri-axis accelerometer. Understanding that

the key to distinguishing between typical-use activities and genuine threats is the

amount of time the product spends above or below certain g-level thresholds, Kionix

has developed tri-axis accelerometers with features specific to hard disk drive drop

detection. These products allow customers to program the part with the high-g and

low-g limits and program the duration of the high-g and low-g periods. Armed with

this information, the accelerometer knows if and when to signal the hard drive to

park.

- 34 -

- 35 -

About the Author:

Dr. Galvin founded Kionix in 1993 to commercialize a novel micromechanical

technology pioneered by researchers at Cornell University. From 1993 to 2000,

Kionix grew from its two founders to over 40 employees and developed products in

inertial sensors, microfluidics, data storage, micro-relays and micro-optics. Late in

2000 Kionix was acquired by Calient Networks of San Jose, CA and renamed Calient

Optical Components. Just prior to the acquisition, a new company was spun out to

then Kionix shareholders to pursue inertial sensor, microfluidics and data storage

markets. This company took back the name Kionix post acquisition. From the

acquisition until June 2002, Dr. Galvin served as president and CEO of Calient Optical

Components and on the boards of both Calient Networks and the new Kionix. In July

2002, he returned full time to Kionix as President and CEO.

Dr. Galvin has a B.S. from the California Institute of Technology in Electrical

Engineering and a Ph.D. in Materials Science from Cornell University (1984). Dr.

Galvin served for over five years as the Deputy Director of the Cornell

Nanofabrication Facility (CNF) in which the Cornell micromechanical research was

conducted. In addition to his doctorate, Dr. Galvin has an M.B.A. degree from Cornell

(1993). Prior to founding Kionix in 1993 he was employed by Cornell University in

the position of Director of Corporate Research Relations, focusing on transferring

technology from the university to industry. Dr. Galvin's graduate research was in the

areas of thermodynamics of silicon under ultrafast melting, ion beam analysis, and

thin film technologies. He has published over 20 technical papers and holds 15

patents. He is a member of several scientific societies and the American Management

Association. Dr. Galvin is a founding member, and former chairman, of the Finger

Lakes Entrepreneurs Forum. He is a member of the Cornell University Council, the

Advisory Council of the Cornell Veterinary College, a Director of the Boyce Thompson

Institute for Plant Research, a Director of the Tompkins County Area Development

Board, and Director of the El Portal de Belén Foundation.

Contact:

Kionix, Inc.

36 Thornwood Drive, Ithaca, New York 14850 USA 607-257-1080

www.kionix.com info@kionix.com

Upcoming IDEMA Events USA Region DISKCON USA 2005, September 19-22: Santa Clara, CA

Charity for Youth Golf Tournament, September 19 Conference and Tradeshow, September 20-21 Keynote Dinner featuring Mike Wingert, Maxtor President & COO, Sept. 20 Global Standards Meetings, Sept. 22

Asia Pacific Region IDEMA Golf Event, Bangkok, Thailand, July 11 IDEMA Breakfast Meeting, Singapore, July18 IDEMA Asia Pacific Members Night, Bangkok, Thailand. Sept. 26 IDEMA Traveling Symposium, China: October 3-5 Japan Region Quarterly Seminar, Tokyo: October 21 - 36 -