e-broidery: design and fabrication of textile-based computing

by E.R. PostM. OrthPN. Gershenfeld. R. Russo

Highly durable, flexible, and even washablemultilayer electronic circuitry can be constructedon textile substrates, using conductive yarns andsuitably packaged components. In this paperwe describe the development of e-broidery(electronic embroidery, i.e., the patterning ofconductive textiles by numerically controlledsewing or weaving processes) as a meansof creating computationally active textiles.We compare textiles to existing flexiblecircuit substrates with regard to durability,conformability, and wearability. We also reporton: some unique applications enabled by ourwork; the construction of sensors and userinterface elements in textiles; and a completeprocess for creating flexible multilayer circuits onfabric substrates. This process maintains closecompatibility with existing electronic componentsand design tools, while optimizing designtechniques and component packages for use intextiles.

Everyone wears clothing. It conveys a sense ofthe wearer’s identity, provides protection from

the environment, and supplies a convenient way tocarry all the paraphernalia of daily life. Of course,clothing is made from textiles, which are themselvesamong the first composite materials engineered byhumans. Textiles have mechanical, aesthetic, and ma-terial advantages that make them ubiquitous in bothsociety and industry. The woven structure of textilesand spun fibers makes them durable, washable, andconformal, while their composite nature affords tre-mendous variety in their texture, for both visual andtactile senses.

Sadly, not everyone wears a computer, althoughthere is presently a great deal of interest in “wear-

able computing.”1 Wearable computing may be seenas the result of a design philosophy that integratesembedded computation and sensing into everydaylife to give users continuous access to the capabil-ities of personal computing.

Ideally, computers would be as convenient, durable,and comfortable as clothing, but most wearable com-puters still take an awkward form that is dictated bythe materials and processes traditionally used in elec-tronic fabrication. The design principle of packag-ing electronics in hard plastic boxes (no matter howsmall) is pervasive, and alternatives are difficult toimagine. As a result, most wearable computingequipment is not truly wearable except in the sensethat it fits into a pocket or straps onto the body. Whatis needed is a way to integrate technology directlyinto textiles and clothing. Furthermore, textile-basedcomputing is not limited to applications in wearablecomputing; in fact, it is broadly applicable to ubiq-uitous computing, allowing the integration of inter-active elements into furniture and decor in general.

In this paper we describe our efforts to meet this needin three main parts. First, we present examples ofapplications that can be realized only through theuse of textile-based circuitry and sensing, includinga piecework (or quilted) switch matrix, an electricaldress, a musical jacket incorporating an e-broidered

!Copyright 2000 by International Business Machines Corpora-tion. Copying in printed form for private use is permitted with-out payment of royalty provided that (1) each reproduction is donewithout alteration and (2) the Journal reference and IBM copy-right notice are included on the first page. The title and abstract,but no other portions, of this paper may be copied or distributedroyalty free without further permission by computer-based andother information-service systems. Permission to republish anyother portion of this paper must be obtained from the Editor.

POST ET AL. 0018-8670/00/$5.00 © 2000 IBM IBM SYSTEMS JOURNAL, VOL 39, NOS 3&4, 2000840

E-broidery:Design and fabricationof textile-basedcomputing

keypad and fabric buses, a musical ball with e-broi-dered pressure sensors, and a flexible, multilayer“No-Soap Radio” built using textile-ready compo-nent packages. Next, we discuss the materials usedin realizing these applications, both to report on ourexperience with them and to illustrate the range ofpresently available electronic textiles. Finally, we de-scribe a process for making flexible multilayer cir-cuits in fabric.2 This is a detailed description of stepsinvolved in fabricating the No-Soap Radio.

Throughout this work we hope to demonstrate thattextile-based circuitry compares favorably with con-ventional printed circuitry when applied to user in-terfaces, and furthermore that its mechanical prop-erties present several advantages over conventionalflexible circuit substrates.

Overview of applications

Our research into textile circuits was driven not onlyby an interest in wearable computing, but also by adesire to create new sensor materials for physicalcomputer interfaces. For example, commerciallyavailable electronic musical instruments, with theirbuttons, knobs, and sliders, are excellent examplesof the physically limiting sensors available when cre-ating physical interfaces for the computer. Textilesare typically soft, conformal, durable, and highly plas-tic. Turning them into computers and sensors wasa compelling way to create better wearable comput-ers and better physical computing interfaces. Theseapplications reflect many of our motivations.

Row and column fabric keyboard. The row and col-umn fabric keyboard (Figure 1) is a fabric switch ma-trix sewn from conducting and nonconducting fab-ric. The keyboard consists of two layers of highlyconductive metallic organza with a resistance of ap-proximately 10 !/m (ohm/meter) and nonconduct-ing rows separated by an insulating layer of nylonnetting (also known as tulle). When pressed at theright point, the two conducting layers make contactthrough spaces in the nylon netting and current flowsfrom a row electrode to a column electrode. Com-mercial gripper snaps are used to connect wires froma microcontroller to the organza. The keyboard canbe repeatedly rolled up, crushed, or washed withoutaffecting its electrical properties.

The microcontroller used to read out the keyboardwas programmed to generate standard MIDI (Mu-sical Instrument Digital Interface) control informa-tion, allowing players to easily trigger different notes

on an attached synthesizer. This keyboard did not mea-sure pressure information (which is important forexpressive control in music). Because there was nononlinear element present at the switching intersec-tions, false or “phantom” contacts were observed bythe microcontroller when multiple keys were pressed.Finally, the piecework nature of the keyboard re-quired more manual labor than was hoped for. Forall its drawbacks, however, users found it very sat-isfying to make music by crushing and crumpling thiskeyboard.

Firefly Dress. The Firefly Dress (Figure 2) is a cre-ative application that embellishes the wearer’s mo-tion with an ever-changing display of light. Its firstpart is a skirt, handmade from two layers of conduct-ing organza (one supplying power and the otherground) separated by a layer of nylon netting. Light-emitting diodes (LEDs) with fuzzy conductive Vel-cro** ends for electrical contacts are placed through-out the netting. When both ends of an LED brushagainst the power and ground planes, the circuit iscomplete and the LED lights.

The bodice (with a conductive front panel) and thenecklace form a second dynamic element. The neck-lace is a simple analog computer, powered when anyof its conducting tassels brush against a plane of con-

Figure 1 Row and column fabric keyboard

IBM SYSTEMS JOURNAL, VOL 39, NOS 3&4, 2000 POST ET AL. 841

ducting organza sewn to the front of the bodice. Eachtassel has its own resistor network and provides adifferent color bias to the red, green, and blue LEDson the face of the necklace.

The dress demonstrated the visual, tactile, and me-chanical potential of sewing circuitry into clothing.The difficulty of adding ordinarily packaged LEDs tothe skirt pointed out the need for more appropriatecomponent packages. The time-consuming piece-work involved in creating the dress also motivatedour search for a better means of creating textile cir-cuitry.

Musical Jacket with embroidered keypad. The Mu-sical Jacket (Figure 3) turns an ordinary denim jacketinto a wearable musical instrument, allowing the

wearer to play notes, chords, rhythms, and accom-paniment using any instrument available in the Gen-eral MIDI scheme. It integrates directly into thejacket an embroidered fabric keypad, a MIDIsequencer/synthesizer, amplifying speakers, a fabricbus sewn from conductive organza, and batteries topower the above subsystems.

The embroidered keypad is flexible, durable, andhighly responsive to touch. It was mass-produced us-ing ordinary embroidery techniques and a mildly con-ductive stainless steel and polyester composite thread(Bekaert BK50/2, described later), and demonstratedfor the first time that a textile keypad could be madeby e-broidery on a single layer of fabric in one step.This process allowed us to precisely specify the cir-cuit layout and stitch pattern in a computer-aideddesign (CAD) environment, from which any numberof articles can be sewn under machine control.

The MIDI jacket keypad used a PIC** microcontrollerfrom Microchip Technology Inc., to perform capac-itive measurements of connections to sewn elec-trodes on a denim substrate. The measurement wasimplemented almost entirely in software as an ex-ercise in developing electric field sensors with a min-imum of hardware, but this minimalism also bene-fits the fabric circuit designer by reducing the numberof components (and hence interconnections) re-quired in a circuit. The electrode array was sewn inthe pattern of a telephone keypad, with traces lead-ing from the symbols to an array of connection padsintended to mate with a conventional circuit board.

The resistivity of the electrodes and their associatedconnecting traces is reduced by the precise place-ment of individual stitches. Each electrode is sewnfrom a single continuous thread that crosses itselfmany times, leading to a multiplicity of intersectionsthat create parallel resistances and increase the over-all conductivity of the sewn circuit element. By us-ing this “e-broidery stitch” we obtain resistances perunit trace length that are dramatically lower (by morethan an order of magnitude) and more consistentthan those obtained with conventional embroiderystitches (e.g., satin or tatami stitches) that are cho-sen for their appearance or texture.

The embroidered keypad3 uses a capacitive sensingtechnique4 because of the high impedance of thesewn electrodes. The measurement algorithm is sim-ilar to that used in the Museum of Modern Art(MoMA) table described elsewhere in this issue,5 but

Figure 2 Firefly Dress

POST ET AL. IBM SYSTEMS JOURNAL, VOL 39, NOS 3&4, 2000842

here the measured capacitance is compared to athreshold to distinguish contact events. This mea-surement is implemented using a single bidirectionaldigital input/output (I/O) pin per electrode and aleakage resistor (which can be sewn in highly resis-tive yarn). Key contact events are output as a serialdata stream by the microcontroller. The circuit boardmakes contact with the electrodes at the circular padsonly at the bottom of the electrode pattern. A viewof the component side of the circuit board has beensuperimposed to show its extent and its connectionsto the fabric. A flexible circuit board can be substi-tuted for the rigid one used in this implementation.

Fifty denim jackets were embroidered with this pat-tern and outfitted with sewn fabric buses that carrypower, ground, serial data and analog audio signalsacross the backs of the jackets. These jackets alsoprovided an AC ground for the keypad, to ensure con-sistent capacitance measurements. In all 50 jackets,

both the fabric keypad and bus worked reliably andheld up to repeated mechanical stresses better thanthe more traditional wires and connectors also used.The interface between the sewn keypad and the elec-tronics led into our work to do away with the circuitboard entirely.

Musical Ball. The Musical Ball (Figure 4) is a plushstuffed toy that allows users to play music by apply-ing pressure in any of eight regions as they squeezethe ball. There are ten electrodes made by e-broi-dery on its surface, eight of which capacitively sensepressure and two of which provide a ground refer-ence for the capacitive sensing. The eight channelsof pressure data are sent as serial data to a host run-ning an application that uses these data to generatemusic through either external MIDI devices or an in-ternal sound card. This application would have beenimpossible to realize using conventional pressuresensors. By using sensors made by e-broidery, we

Figure 3 On the left is the Musical Jacket, comprising a fabric keypad on one side, a MIDI synthesizer “boat” on the otherside, speakers behind speaker grills in the pockets, and fabric buses visible inside the jacket. On the right is acomposite image of both sides of the circuit board attached to the back of the fabric keypad.


were able to create a light, hand-held, flexible, mul-tichannel, continuous control instrument in a singlemanufacturing step. We were also able to directlytailor the sensor’s size, shape, and appearance.

The impedance in the electrodes in the ball is dra-matically reduced by using highly conductive threadin the machine bobbin during e-broidery. In contrastto the electrodes in the musical jacket, whose con-ductivity is improved and controlled through stitchpattern, the conductivity of the electrodes in the ballis limited by the conductivity of the bobbin thread.

Electronic Tablecloth. The Electronic Tablecloth(Figure 5) is a sensor surface that allows users to in-teract with a computer and with each other in thecontext of a social function. Attendees identifiedthemselves to the system by setting their assigned“coaster tags” on the swirling pattern, above whichare embroidered the words “Place your coasterhere.” The coaster is in turn a capacitively coupledradio frequency identification device (RFID) tag, andis read by the tag reader in the tablecloth when theuser sets it on the swirling pattern and touches thefabric electrode on top of the coaster. The computerthen enters a dialog with the user through the ap-propriate fluorescent display and keypad.

We combined five keypads and five capacitive tagreaders all made by e-broidery to create a tablecloth

(and centerpiece) that turns an ordinary table intointeractive furniture. The electronics for this appli-cation are housed in the centerpiece. We encoun-tered difficulty using our capacitive keys near theelectrostatic tag reader electrodes, which in normaloperation present a 50-volt rms (root-mean-square)sine wave at 125 kHz to power up and read out tags.By applying a shielding layer of metallic organza un-der and around the swirling electrodes, we were ableto terminate most of the electric “field lines” beforethey reached the keypad electrodes.

No-Soap Radio. This washable, crushable keypadmade by e-broidery (Figure 6) is known as the No-Soap Radio, because it was designed to be dry-clean-able and to transmit data to a nearby AM/FM radio(via near-field electrostatic coupling). When the usertouches a symbol on the keypad, the microcontrollercapacitively senses the contact and then transmits aparticular audible tone (depending on whichsymbol was touched) by FM-modulating a 455 kHzsquare-wave carrier in software and presenting theresulting signal on all of the keys connected to thecircuit (Pad0 through Pad3). The user then acts astransmitting antenna for the No-Soap Radio. Twopins are reserved for intrasensor communications toallow the construction of larger keypads by network-ing several instances of the basic four-key sensor.Nodes can communicate with each other and a hostprocessor over a serial bus on these two reserved pins.

The only external components used in this circuitare two 2M! resistors, which allow the charging timemeasurement to be done in software. These are usedin lieu of resistors that could be included directly inthe integrated circuit. The resistors are soldered ontothe PIC12C509 leadframe after steel threads are mi-crospotwelded onto the leads. The resistors are insurface-mount 0402 packages, with dimensions 0.040inch " 0.020 inch " 0.010 inch. The fabrication ofthis and other, similar devices is discussed in detailbelow.

Textile circuit fabrication techniques

Our first step in manufacturing textile circuitry wasto identify textiles and yarns suitable for use in fab-ric circuitry, and then to find a way to design andfabricate such circuitry. As piecework is time con-suming and imprecise, our primary method of cir-cuit patterning has been e-broidery, i.e., numericallycontrolled embroidery using conductive thread. Weuse e-broidery to stitch patterns that define circuittraces, component connection pads, or sensing sur-

Figure 4 Musical Balls


faces. By using commercial embroidery processes,we have built on prior art in the textile industry thatpermits precise control of the design, layout, andstitch pattern of the circuitry through CAD processes.

Yarns and threads usable in high-speed embroideryprocesses must make a trade-off between electricaland mechanical properties. These yarns must be con-ductive, but also strong and flexible enough to besewn at high speeds without breaks in either thethread itself or in the electrical continuity. Theseyarns must also maintain their electrical propertieswhen washed (we consider dry cleaning to be the tar-get washing process, and exposure to water, alcohol,and sweat as an everyday occurrence).

These processes also allow control and integrationof yarns with different electrical properties, for in-stance, different resistances. This level of stitch con-trol and yarn variety brings up the possibility of re-placing discrete components such as capacitors,resistors, and inductors with specific combinationsof thread and stitch pattern. Such components might

lead to reduced manufacturing cost and higher re-liability from improved mechanical properties.

Figure 6 The No-Soap Radio. An AM broadcast receiver(right) is used to listen to the transmitted signal.


First, in order to work with available machine sew-able threads we have chosen not to require low-im-pedance component interconnections, as is done inconventional printed circuit design. Instead, wechoose circuit elements and design principles withlow power consumption and high input impedances.

Second, our next step was to design packages op-timized for use in fabric circuitry. In this report, wefocus on the design of the “Plastic Threaded ChipCarrier” (PTCC), a package that resembles the Plas-tic Leadless Chip Carrier (PLCC). The PLCC is a well-known surface-mount integrated circuit (IC) pack-age style, designed to be soldered onto printedcopper circuit traces on a rigid or semi-rigid circuitboard. Conversely, the PTCC is designed to be stitchedor woven into a fabric circuit, and for this purposehas long, flexible conducting leads. Our primary goalwas to demonstrate a durable interconnection andattachment scheme for electronic components usedin circuits made by e-broidery.

The third step was to develop multilayer circuitry.Even if we can sew a circuit, its complexity will belimited if it is restricted to lie in a plane. By makingmultilayer e-broidery, we hope to demonstrate thattextile-based circuitry compares favorably to printedcircuit boards and conventional flexible substratesin terms of design complexity and reliability, partic-ularly in user interface applications.

The first step in integrating electronic technology intotextiles and clothing might be to simply remove thebulky and stiff housing of most electronic gear. Wemight also make the circuit board smaller and re-movable by redesigning it with components in smallerpackages, and using conductive fasteners, like snaps,to connect the boards to the fabric. (This way theboards can be removed when a garment goes intothe wash.) Unfortunately, circuit boards neither looknor feel good when they become part of clothing,nor does this approach scale well because of the manylayers of packaging and connective elements thatcome between the silicon and the textiles.

To overcome these limitations, we have chosen tocombine electronic components and conductive tex-tile circuitry as directly as possible. We replace stan-dard printed circuitry on rigid or flexible substrateswith circuitry made by e-broidery and componentpackages optimized for use on fabric substrates.These packages can be washed without harm to ei-ther their internal electrical properties or their con-nections to the substrate and other components.

Finally, we would like the interface between com-ponents and circuitry to withstand the flexing andstretching that clothing and textiles are prone toexperience.

One goal of our work is to create electronic garmentsand textiles that exhibit integrity, that is, garmentsthat look, feel, wash, and wear as well as ordinaryclothing. In this context, system integration becomesthe art of partitioning a digital system across severalitems of clothing. For example, most people haveonly a few pairs of shoes, a few more pairs of pants,several shirts, and a few outer garments. It makessense to put the most expensive parts of a system inthe shoes and outer garments, and the least expen-sive parts in the shirts, pants, and undergarments.

In this section, we discuss the properties of severaltextiles we have used. In the next section on com-ponent integration and packaging, we detail a strat-egy for making electronic circuitry disappear intoclothing and textiles in several steps. First, we re-place standard printed circuit boards with e-broidery(embroidered textile circuitry) designed with tradi-tional CAD tools for circuit layout and rendered onnumerically controlled embroidery machines usingconducting and insulating threads. Second, we re-design electronic component packaging for use incircuitry made by e-broidery. Third, we combine lay-ers of circuitry to form multilayer circuitry of arbi-trary complexity.

Existing flexible circuit substrates. Many methodsalready exist to fabricate circuits on flexible sub-strates. Most of these rely on the metallization of aflexible polymer substrate that can withstand the hightemperatures of conventional soldering processes.Kapton** (polyimide) film, for example, is one ofthe most commonly used flexible substrates, typicallyfinding applications in cameras (where circuitry mustfit intimately into available space in a small pack-age), printers and portable computers (where a num-ber of connections must cross a hinge or other ro-tating joint), and nonplanar antennas (where theantenna elements must have accurate dimensions yetbe shaped to fit on conical forms).

When circuitry is designed for such a substrate, caremust be taken to ensure that in the final applicationmechanical stress will not be applied to the compo-nent-circuit solder joints. Because of this, the param-eters of motion of the flexible substrate must be wellunderstood at the time of design. That is, the mo-tion of conventional flexible substrates must be con-


strained to ensure that they continue to work overtheir expected lifetime. In general, conventional flex-ible substrates are flexible only in regions where com-ponents are not attached.

Our dissatisfaction with conventional flexible sub-strates stems from the fact that they cannot be crum-pled nondestructively in the way that cloth can be.To be clear, by “crumpling” we mean simultaneouslyfolding the substrate along multiple nonparallel axes.

Flexible multichip modules. An interesting excep-tion to our ruling against conventional flexible cir-cuitry might be the flexible multichip module (MCM),an example of which is shown in Figure 7. This mod-ule is the result of collaboration with workers at theCharles Stark Draper Laboratory at MIT to producesmall, semi-flexible circuit elements that may be em-bedded in textile products to form a sensor readoutnetwork. We make an exception in this case becauseof the module’s small size.

The major part of this readout element is a PIC16F84microcontroller IC, visible underneath the titanium-copper interconnections. In addition to the PIC16F84there are four 3M! chip resistors and a 16 pF (pi-cofarad) chip capacitor. The schematic for the MCMis similar to the circuit used in the fabric keypad inits simplicity and choice of microcontroller. In thiscase, the host attaches to the RB7 and RB6 pins,which form a full-duplex (RS232) serial data link op-erating at 9600 baud. The connections are formedby drilling out the centers of the pads and insertinga small pin or rivet to pressure-fit each pad to a mat-ing circuit trace made by e-broidery. Power for thecircuit is supplied across the Vdd and Vmm pins, andelectrodes are connected to the Trode[0–3] pins tobe capacitively read out just as was done in the orig-inal fabric keypad.

The entire flexible MCM measures 0.600 inch " 0.300inch, is 0.016 inch thick, and has a permissible bend-ing radius of curvature of about 50 centimeters (cm).In a fabrication process similar to that developed bythe General Electric MCM-E/F group,6 the constit-uent components were thinned to 0.006 inch by me-chanical grinding and lapping of the rear face of eachcomponent. The components were then placed ona rigid substrate, a laser-cut 0.006-inch Kapton maskwas applied to planarize the top surface, and theTi/Cu (titanium/copper) interconnect was depositedthrough a stencil mask. This process extends readilyto multilayer circuitry by the addition of thin (0.001-inch) insulating layers. Finally, a protective Kapton

cover was glued to the top of the circuit module. Themodule was then removed from the rigid substrate,and a Kapton cover glued to its bottom, resulting ina completely encapsulated element.

Previous work in conductive textiles. The state ofthe art in textile circuitry seems to be largely unpub-lished and proprietary. While a great deal of workhas been done in the field of electrically active tex-tiles, we have not been able to find (at the time ofthis writing) any published prior art relating to ei-ther the layout and sewing of active circuitry on tex-tiles with conductive yarns, or to the integration andattachment of off-the-shelf or specially designed elec-tronic components onto textile circuitry. However,there is related work which could enable the devel-opment of conductive textiles with active circuitry.A survey of the U.S. patent literature reveals muchongoing work focusing on the development of syn-thetic conductors compatible with existing textileproduction processes. For example,7,8 there are con-ductive polymers that remain fairly conductive (#1!/cm) after being spun into threads having theweight, strength, and appearance of ordinary nylonthread.

In other potentially enabling developments, for ex-ample, workers at Motorola have patented IC pack-ages with hook-and-loop fasteners9 as a means ofattaching components to conventional flexible sub-strates.10 Such packages—if successful—might doaway with the complex art of surface-mount solder-


ing and its associated environmental burdens. In an-other patent, workers at IBM proposed a zipperedconnector to simplify the assembly of wide backplaneinterconnections.11

Other researchers are embedding photonic conduc-tors (optical fibers) into clothing to build architec-tures for sensing and data distribution. The sensateliner12 is one such undergarment, in which a wovenfiber optic matrix is used to sense disruptions of theweave (by measuring optical continuity) and to per-form spectrographic analysis of the materials presentat intrusions. Such analysis can be used to determinethe type and severity of wounds suffered by thewearer of the liner. While this work is promising be-cause of the variety of fiber optic sensors that canbe made, drawbacks arise at the interfaces of the op-tical fibers to the conventional electronics that arerequired for signal origination and processing.

Conductive yarns, fibers, and textile substrates. Thetextile circuitry in this paper includes sewn or stitchedtextile circuit elements that function as conductivetraces, sensor elements, electrodes, and discrete com-ponents (e.g., resistors and capacitors). Many tex-tiles and yarns have properties that suggest their usein such fashion. There are also a number of man-ufacturing processes that could produce these fab-ric circuit elements.

An ideal yarn or textile for fabric circuits would havecompletely adjustable electrical properties and would

maintain those properties while being sewn, flexed,and worn. In this section, we identify and describethe strengths and limitations of some of the mate-rials that have been useful in our work. In outliningour practical knowledge of a variety of materials andprocesses for creating fabric circuits, we can at besthope to inspire the creation of new textile forms op-timized for use in circuitry.

Metallic silk organza. In its primary form, this is afinely woven silk fabric with a thin gold helix wrappedaround each thread that runs along the weft of theweave (Figure 8). Believed to have originated in In-dia, this sort of fabric has been produced for cen-turies and appeared in Western fashion as early asthe mid-18th century.13 The warp of this fabric con-sists of parallel silk threads. Through this warp, theweft is woven with a silk thread that has beenwrapped in a metal foil helix. This metallic threadis prepared just like cloth-core telephone wire, andis highly conductive (#0.1 !/cm). Because the con-ductors in this fabric only run in one direction, thefabric is anisotropically conductive. The spacing be-tween these fibers also permits them to be individ-ually addressed, so a strip of this fabric can functionlike a ribbon cable. If a section of organza is sub-jected to shear, the cells formed by the weave alsoshear to form parallelograms, keeping the conduc-tive fibers parallel and separated at all times. Thesilk fiber core has a high tensile strength and canwithstand high temperatures.

Circuits fabricated with components attached to or-ganza only need to be protected from folding con-tact with themselves, which can be accomplished bycoating, supporting, or backing the fabric with an in-sulating layer which can also be cloth. However, ev-idence shows that coating can disturb the conduc-tivity of the fabric. Circuits formed in this fashionalso have many degrees of flexibility (i.e., they canbe wadded up), as compared to the single degree offlexibility that conventional substrates can provide.In the microcontroller circuit shown in Figure 9, aPIC16C84 microcontroller and its supporting compo-nents are soldered directly onto a patch of metallicorganza. This circuit uses bidirectional I/O pins onthe PIC to control LEDs and to sense touch along thelength of the fabric, and uses audible feedbackthrough a piezoelectric speaker to reinforce the senseof interaction. All of the components are soldereddirectly onto the surface of the metallic organzaweave except for the microcontroller, which sits ina socket soldered to the organza.

Figure 8 Micrograph of woven metallic organza. The twohorizontal threads are wrapped with a metalfoil.


While the copper, gold, or silver foil wrap the silkthread, the foil’s fragility makes it impossible to ma-chine sew without damaging the wrap and breakingthe electrical continuity. Consequently, this mate-rial was used to create labor-intensive piecework cir-cuits, or as the substrate for component attachment.In piecework, many drawbacks of the material ap-peared. Because the material is anisotropically con-ductive, it was necessary to tie the conductors to-gether either with a perpendicular conductive bandor by folding the material over itself. All this was timeconsuming and labor-intensive. Finally, the standardcopper and silver coatings on the material will cor-rode over time unless they are properly passivated.

Conductive yarns. The conductive stainless steel yarnsoutlined in this subsection were initially manufac-tured to make filters used in processing fine pow-ders. Stainless steel presents some immediate advan-tages. It is inert and therefore not sensitive to washingor sweat. These stainless steel yarns vary in compo-sition, from 100 percent continuous conductive steelfibers to feltings or composites of polyester and withshort steel fibers interspersed throughout. Varyingthe ratio of the two constituent fibers leads to dif-ferences in resistivity. The conductivity of these yarnsis also ultimately limited to the conductivity of the

stainless steel fibers themselves. Their conductivityis limited by their manufacturing process, by coldworking, and their fine diameter, which makes themsewable.

Bekaert Fibre Technologies14 credits a 1936 U.S.Patent15 with the original description of the processwhereby a bundle of fine metal fibers may be drawncontinuously and simultaneously from source met-als. Bundle-drawn fibers can be produced in manydifferent morphologies by translating the polymer-based methods of the synthetic textile industry intotheir metallurgical equivalents. Continuous and bro-ken bundles, cut fibers, spun yarns, threads, etc., areall produced in this way.

The fibers themselves are available in diameters from100 !m (micron) to 12 !m, and as small as 2 !m forvery metallurgically clean (free of large inclusions)alloys. Alloys commonly used in these processes in-clude stainless 316L steel, stainless 302 steel,Inconel** 601, NiCr (nickel-chromium) and FeCr(iron-chromium) alloys, and titanium.14 These 100percent metal fiber yarns are used in knitting, weav-ing, needlepunching, and braiding processes, withsewing and embroidery being notable omissions fromthe list.

Figure 9 Circuit fabricated on metallic organza


Bekaert Fibre Technologies14 cites the primary ap-plications for these materials as being filter media,antistatic textiles, heat-resistant textiles, burners, andconductive plastics. Their product range includes 100percent metal fiber products as well as blends ofmetal fibers combined with natural or human-madefibers. Steel threads were chosen for their strength,resistance to corrosion, biological inertness, andready availability in textile form at low cost. The ma-jor drawback of using steel and steel-composite

threads is the difficulty involved in attaching themto existing electronic components.

Composite yarns of short steel fibers and polyesterwere chosen for machine embroidery because an all-steel continuous fiber thread (used to form compo-nent leads), cannot be sewn by machine. In a con-tinuously drawn steel fiber bundle, individual fibersare meters (rather than centimeters) long. Stressesare no longer local to a small region of sewing, andtensions rapidly mount without resolution, resultingin bunching of the thread as it feeds through the sew-ing machine. Some of the bunched threads will even-tually twist into an obstruction that will not passthrough the sewing needle’s eye, and the entire pro-cess will grind to a halt. In contrast, threads spunfrom short (staple) fibers are able to accommodatethis variation in tension by stretching. Following isa list of fibers we have studied:

" BK 50/2 (#50 !/cm)—This composite thread,while being highly resistive, performs the best un-der machine sewing and passes through ordinarysewing needles better than any other conductivethread we have used. Both the magnitude and con-sistency of this thread’s conductivity can be con-trolled by careful stitch placement, which deter-mines the intrathread conductivity as a functionof thread crossings as well as interthread conduc-tivity as a function of thread tension. Use of thisthread in both the needle and the bobbin dramat-ically increases the conductivity of stitched traces.

The thread comprises a spun felt of short (#1-inch) steel and polyester fibers (see Figure 10).Because the steel fibers are short and extend fromthe body of the thread, short circuits can occur be-tween sewn traces. This can be corrected after thetraces have been sewn by brushing the fabric witha sponge-covered magnet. These same extendingsteel fibers provide an excellent contact surface.This is important for pads that make contact withcircuit boards (as in the Musical Jacket), padsmeant to contact skin for sensing, and contact be-tween separate threads and traces (particularly inthe case of vias in multilayer circuitry made by e-broidery).

" Bekintex 100 percent stainless steel continuous fi-bers (#1 !/cm)—These threads comprise bundlesof continuous cold drawn stainless steel fibers (seeFigure 11). The fibre is highly conductive, but can-not be machine sewn because of its low elasticity.The single-ply version of this thread will not hold

Figure 10 Photomicrograph of BK 50/2. The dark, unrulyfibers are stainless steel, in a polyester mesh.

Figure 11 Photomicrograph of Bekintex single-plycontinuous-drawn stainless steel fiber bundle


a knot, but the two-ply version (BK 12/2 "275/175S) will, because the plies are tightly twisted.These threads are suitable for microspotweldingto component leadframes and are used to repack-age components with threaded connections. Al-though they are unsewable, these threads can becouched (attached to a substrate by a coveringstitch); this is the technique used to attach com-ponent leads to e-broidered circuits.

" Bekintex 15/2 100 percent stainless steel spun fi-bers (#1 !/cm)—An alternative to the continu-ously drawn steel fiber bundle is an all-steel yarnspun from short steel fibers. This thread is ma-chine-sewable and performs well in the bobbin ofcommercial embroidery machines. Its main draw-backs are due to its “untidiness”; although it con-tains a far larger number of short stainless steelfibers than its BK 50/2 counterpart, these fibers arenot held in place by a polyester mesh. As a result,many more steel fibers free themselves from thethread body, causing short circuits between tracesand irritation of the wearer’s skin.

" VN 140 nyl/35 " 3 (#10 !/cm)—This thread isa nylon core wrapped with three continuous cross-ing stainless steel fibers. It is relatively sewablethrough needles and highly sewable in the bobbinof commercial embroidery machines. Electrodessewn with this thread are limited by the conduc-tivity of the steel fibers. Because it has no loosefibers, it creates fewer short circuits between traces.Electrodes sewn with this thread are highly dura-ble and conductive. Because it has no loose fibers,it also presents less electrical contact area, mak-ing electrical connection difficult.

" Aracon** metal clad aramid fibers (MCAF) yarns(#0.001 ohm/cm)—Recently the DuPont companyannounced their Aracon MCAF yarns.16 Theseyarns comprise anywhere from 24 to 200 conduct-ing fibers, each approximately 15 !m wide. Thearamid core of these fibers is usually Kevlar**,while the cladding metal is silver (Ag), nickel (Ni),copper (Cu), gold (Au), or tin (Sn). Ag and Ni clad-ding are preferred for the aerospace applicationsthat DuPont has targeted as their initial market,where the benefit of a 40–60 percent mass reduc-tion outweighs the added cost of what is presentlya very expensive composite fiber.

Aracon exhibits excellent mechanical and electri-cal stability over repeated flexures, radiation ex-posure, and changes in temperature. Because the

aramid core can withstand high temperatures andhas a thermal coefficient of expansion similar tothat of the cladding metals, Aracon can be sol-dered like normal wire. Because of these prop-erties, as Aracon comes to be mass-produced andits availability increases, it is likely to emerge asa material of choice for unexposed fabric circuitry.

Composites made by e-broidery. One technique weare developing is the stitching of composite traces.Suppose that a Bekintex 15/2 stitch is covered by aBK 50/2 overstitch. Segments of the trace that coverthe steel bundle will exhibit much lower resistanceper unit length than segments of BK 50/2 alone. Asdiscussed above, the conductivity of the all-steel yarnis about 100 times greater than that of the compos-ite yarn, which implies that low-precision resistor net-works can be constructed simply by combining runsof stitches in the two materials. It also leads to themore general notion of a composite trace or sensorcomprising materials combined by stitching.

This technique can also be used to build coaxial struc-tures. One particularly useful example is that of cov-ering a conductive trace with an insulating stitch toprotect the former from coming into accidental con-tact with other conductors.

Component integration and package designin textile circuitry

As a consequence of the design principles mentionedabove (i.e., distributed sensing and computation, thininterconnect, small packages with fully integrated cir-cuitry), components destined for placement in fab-ric circuitry do not demand high connection densi-ties. Most component packages are optimized forjoining to printed circuit boards, so we instead op-timize them to interface with textile circuitry. Ulti-mately, we hope this work will lead to an actual pack-aging standard used by the integrated circuit industry.

Fabrication techniques. There are several ways toembed electronic circuitry in fabric, depending onthe choice of substrate. Some possibilities exploredin the course of this work include:

" Soldering surface-mount components directly ontometallic organza

" Bonding components to a substrate using conduc-tive adhesives

" Stapling components into a conductive stitched cir-cuit (i.e., pressure-forming their leads to grip cir-cuit pads)


" Joining a component’s “threadframe” directly toa stitched fabric circuit (where components areformed with a single conductive thread per pin)

The last possibility is explored in some detail in thefollowing section. We have ruled out the others forvarious reasons indicated below:

" Solders used with electronic components are softalloys of lead (Pb), tin (Sn), and sometimes silver(Ag). Such compounds are not suitable for use inapplications in which they could potentially be inconstant contact with a user’s body, because oftheir toxicity. Worse yet, while it is possible to sol-der components onto metallic organza and achieve(very) good electrical contact, the mechanicalproperties of the joint are unsuited to the flexureto which it will be subjected as an item of apparel.

" Conductive adhesives are better suited to this ap-plication than solder, because it is possible to en-vision adhesives that are nontoxic, highly conduc-tive, highly durable, and moderately flexible to actas a “mechanical impedance match” between aflexible fabric substrate and a rigidly packagedcomponent. They remain an open possibility andshould be the subject of further study.

" Stapled components are an interesting compro-

mise where a component lead grips a sewn con-ductive trace by being pressed into shape aroundit. When the substrate flexes, the trace is free tomove within the clasp of the formed lead, forminga self-wiping contact at every junction between fab-ric circuitry and component pins. However, the di-mensional rigidity of the component is a poormatch for freedom of motion enjoyed by a fabricsubstrate, and the mismatch is likely to stretch openpins that have been formed into clasps and to ac-celerate wear and tear of the fabric substrate.

Gripper snaps make excellent connectors betweenfabric and electronics. When a two-piece grippersnap is placed on fabric, the first piece has severalmetallic “teeth” that pierce the substrate and anyconductors, making a wiping contact that scrapesthrough surface contaminants (e.g., oxides) to makea metal-on-metal bond. They are already commonlyused to connect circuit packages to fabric, but allowonly a low connection density.

The case for round packages. Consider electroniccomponents with threads leading out from the pack-age rather than a rigid leadframe. These threads maybe attached to the substrate by covering stitches thatalso form electrical connections to circuitry made by

Figure 12 Square and radial PTCC packages

STEEL THREADS

STEEL/NOMEX COMPOSITE THREADS


e-broidery. Alternatively, the threads emanatingfrom a component may be punched or woventhrough the substrate on their way to connect toother components. In either case, these threads con-strain the components to their location on the sub-strate. Because we would like to balance as evenlyas possible the forces conveyed by these threads, wenow consider round component packages. Figure 12illustrates components in square and round packagesstitched onto a fabric substrate, whereas Figure 13shows an actual stitched prototype componentcouched onto a fabric substrate.

For a given lead spacing (s) at the periphery of apackage and a given package diameter (d) or sidelength (l ) the maximum number of threads comingout of each type of package is found to be

nsquare " 4! ls # 1"

nround "$ds

For a given thread spacing around the periphery ofthe package, it is clear that for a large number ofpins the square package will have a smaller diam-eter than the round package. If the number of pinsis less than 16 (or to be exact, 16 (4 $ $) $ 4), thenthe round package makes better use of the availablespace.

The finest practical thread spacing achieved in ourwork has been 0.050 inch (1.27 millimeter), whichfor 16 leads implies a package diameter of 0.254 inch(6.48 millimeter). It is practical to expect that a dieof about half this width will fit into the package, whichgives an upper bound of 10 mm2 of single- or multi-chip real estate to work with in a typical sewable but-ton package.

The plastic threaded chip carrier. Now we look tothe practical aspects of building packages like thosethat have been described so far. To manufacturethreaded packages, we would start with a bare die,add a “threadframe,” and connect the near ends ofthe threadframe to the bondout pads of the die.

Figure 14 shows the possible construction of such acomponent package. Some details have been omit-ted for clarity, such as the metal base that helps tokeep the die substrate at near-uniform temperatureand electrical potential across its area.

Connections directly to the die are made in the usualway; fine gold wires are thermocompression-bondedon one side to the die’s gold-plated “bondout pads”and on the other to the stubs of a conventional cop-per leadframe.

Figure 13 Prototype PTCC package couched onto fabricsubstrate

Figure 14 Details of round PTCC construction

THREADFRAME(STEEL)

LEADFRAME(CU)

BONDOUT WIRES(AU)

BONDOUT PADS(AU-PLATED)

INTEGRATED CIRCUIT


In most packages, the leadframe continues out ofthe plastic package and is ultimately soldered or cold-welded to external circuitry. Instead of having a solidexternal leadframe, the plastic threaded chip carrier(PTCC) has flexible, corrosion-resistant threads in-tended to connect to external circuitry.

In this case, the threads that leave the package arebundles of approximately 100 continuous steel fibers,each about 5 !m in diameter. These fiber bundlesare microspotwelded to the leadframe stubs, and theentire structure is hermetically sealed in a plastic car-rier.

Prototyping the PTCC. To test the proposition of athreaded chip carrier, prototypes were built to eval-uate their electrical, mechanical, and sartorial per-formance.

Since part of the PTCC’s assembly process is commonto most IC packages, we can test the design princi-ples behind the PTCC without building one com-pletely from scratch. Leaded surface mount (SMT)IC packages already incorporate the die, the lead-frame, and bonding wires in a sealed plastic carrier,and their electrical characteristics are well under-stood.

Prototypes have been built by starting with SMT pack-ages, adding a few components directly to the leads,then microspotwelding steel fiber bundles to theleads. Finally, the threads are arranged in a radialpattern, and the entire assembly is encapsulated inan epoxy resin.

Packages made this way are larger than they needto be, but all of the new design principles involvedare represented and open to refinement.

Spot welding. Elihu Thompson pioneered electricalor resistance welding in the late 19th century,17 tak-ing advantage of the fact that an electrical currentpassing through any resistive material evolves heat.

For constant-current welding the heat evolved perunit area %Q takes the form

%Q "Iw

2 RA %t

where Iw is the weld current, R is the weld junctionresistance, A is the weld junction area, and %t is theweld time.

The welding apparatus is arranged so that most ofthe weld power is dissipated at the faying interface,which is the junction of the materials to be welded.Inspection of the heat evolution equation reveals thatfor a given weld current and time, the weld heat in-creases directly in proportion to the junction resis-tance and inversely in proportion to the junctionarea. This suggests that the resistivity of the weld pathshould be concentrated at the weld junction, and thatthe junction area should be small. Resistive heatingis only part of the story, however. Although a largearea of contact is desirable between the electrodesand the workpiece to reduce heating at those points,a small electrode-workpiece junction area is pre-ferred to reduce heat diffusion from the faying in-terface. When two uniform, smooth pieces arewelded together, the heating will be greatest alongthe center axis of the applied current, and a weldwill form between the pieces along this axis. Theheating is greatest along the center of this axis, be-cause at that point the weld is most surrounded byheated material. Different types of welds will arisefrom different levels of heating. The “coldest” weldsform when the grain boundaries of the workpiecesshift to form large interfaces. As heating increases,grain boundaries shift and dovetail, forming a com-plex interface. Still greater heating results in a break-down of grain boundaries and a variable alloying ofthe materials making up the junction. The leadframeused in this case was copper (Cu) plated with a tin-lead (SnPb) alloy. Diffusion alloying takes place atthe Cu-SnPb interface, and the result after severalhours is a Cu-CuSn-SnPb interface. This chemistryis common to the process of soldering with SnPb al-loys on Cu or CuSn surfaces.

It is worth noting that while the process describedhere was developed using Cu plated with SnPb, ourgoal is to produce component packages that exposeonly biocompatible materials (e.g., the stainless steelthreads). Furthermore, the process is compatiblewith uncoated Cu leads, making SnPb altogether un-necessary.

Bondout metallurgy. Weld schedules are often dif-ficult to optimize, and depend crucially on many fac-tors including electrode geometry, contact area, andelectrode-workpiece metallurgy. For instance, wemight wish to attach steel threads directly to thebondout pads of an IC die without using an inter-mediate Cu/CuSnPb leadframe. This is impracticalfor many reasons.


Steel threads are too strong to be thermocompres-sion bonded to metallized pads on a crystalline sil-icon substrate without causing the substrate tocrack.18 More important, however, is the metallurgyof bondout pads on silicon chips.

Most ICs use Al (aluminum) metallization for on-chip interconnections, as well as for bondout pads.Nearly all pad-to-leadframe wirebonds take place us-ing Au (gold) or Al wire (actually 99 percent Al &1 percent Si). Au is used more frequently becauseit requires less preforming during the wirebondingprocedure and because it forms less of a surface ox-ide layer than Al wire does. Bondout pads are alsousually plated with Au (over the Al) to promote goodwire-to-pad bonding.

The Au-Al interface does, however, represent a com-promise between the need to achieve a strong wire-bond and the need to develop a process insensitiveto contamination. The most common reliability prob-lem in wirebonding results from the formation of in-termetallic compounds at the Au-Al interface, knownby their color as “purple plague” (AuAl2) or “whiteplague” (Au5Al2). These compounds are very brit-tle compared to the metallic Au and Al that surroundthem. Because of this brittleness, wire flexing (as a

result of vibration or thermal cycling) more easilyinduces metal fatigue and stress cracks.18 These andother intermetallic compounds and alloys deleteri-ous to the strength of the Au-Al interface are knownto form more readily in the presence of all of themetals that Bekaert produces in textile form. It istherefore essential to ensure the purity of the goldused in plating bondout pads, as well as the purityof gold wire used to bond to the gold-plated pad.

For the present work, weld schedules have been de-veloped to maximize the pull strength of the thread-leadframe junction. To prevent interference thatwould be caused by fraying of the end of the fiberbundle, it is first welded by one or two weld oper-ations. The tab is then cut in the middle and the ex-cess length is discarded. To weld this tab to the lead-frame, the tinned copper leadframe (Cu/SnPb) isplaced on the lower electrode, the steel fiber bundletab is placed on top of the lead, and the upper Cuelectrode is brought down to weld the tab to the leadat an applied force of 16 pounds.

Other weld schedules, geometries, and electrode ma-terials have been tried with varying degrees of suc-cess. To minimize the number of tool changes nec-essary during weld process development, an

Figure 15 Two views of microspotwelding rig


electrode geometry has been chosen to simulta-neously provide a good mix of thermal, electrical,and mechanical properties, and is shown in Figure15.

The upper electrode is formed from 18 -inch diam-

eter copper rod stock, with the end ground down toa 1

8 -inch " 116 -inch rectangular cross section. The

lower electrode is formed of 18 -inch copper rod with

a 1-inch " 34 -inch " 1

16 -inch copper “anvil” solderedon top. The anvil provides a thermal sink that drawsexcess heat away from the weld, decreasing the ten-dency of the electrodes to stick to the weld and pro-moting even heating (and therefore, uniformity) ofthe weld joint. A representative weld joint of a stain-less steel thread to Cu/CuSnPb lead is shown in Fig-ure 16.

Weld evaluation. The mean shear strength of a weldformed by the above procedure is 36.8 N (standard de-viation ' 4.3 N, n ' 9), determined by pull testing.Weldswithapull strength less than10Nwerediscarded(3 out of 12 samples tested) as being the result of poorprocess control or weld contamination. During thesetests, most failures occurred at the point where theleadframe entered the plastic package, not at thethread-leadframe weld itself. That is, the weld jointproved to be stronger than the leadframe itself.

A pull strength of 36.8 N (or equivalently, 8.25pounds) is more than sufficient for the applicationdescribed above, e.g., that of welding steel threadsto leadframe stubs prior to encapsulation in a her-metically sealed plastic package. It also suffices forthe purpose of prototyping such packages, wheresteel yarn is welded to the leads of an existing pack-age and the resulting construction is encapsulatedin epoxy. In both of these applications, the pullstrength of the original weld is augmented by the ad-hesion of the thread to the package encapsulant.

Multilayer textile circuitry. The leap from single lay-ers of e-broidery to multilayer circuitry is motivatednot only by the desire to develop nontrivial inter-connects and circuitry, but also by the need to con-trol the outward appearance of fabric interfaces. Fab-ric-based user interfaces should at large display onlydevices and symbols relevant to user interaction, andnot the underlying complexity of their implementa-tion. In adopting the modular approach to circuitfabrication, we have so far developed circuit designtechniques, component packages, and a method of“printing” an interconnect in fabric. All of this leads

to the final step of stacking fabric layers with an in-sulating layer and adding an interlayer stitched in-terconnect. The way this is done with conventionalcircuit boards is to print each layer in copper on oneside of a thin substrate, then to laminate the stratatogether, drilling holes as is appropriate, and to platewith copper to interconnect traces on different strata.The interlayer connections are commonly referredto as vias, and the metallurgy and chemistry of thisprocess is messy and expensive.

To do the same thing in fabric, however, we have aneasier time. Each layer of the interconnect is madeby e-broidery on its own fabric substrate. Then in-sulating fabric layers are placed between the layerswith circuitry, and vias are stitched between layersusing more conductive thread.

Suppose we want to build an interface that has onlybuttons on its top surface, such as the No-Soap Ra-dio. How can this be done?

First, we make a plane by e-broidery with nothingbut the interface objects sewn in. The next layer issimply a thin, sturdy insulating layer, followed by an-other layer with circuitry. This second layer of cir-cuitry includes pads that line up with the interfacesymbols on the top layer. The next layer is again aninsulator, and the last layer includes the componentsthat have had their leads couched onto the fabricsubstrate connecting to traces, some of which includepads to connect with other layers. Finally, the layersare assembled and stitched together at points wheresignals must cross between layers.

To enhance the final appearance, the topmost layeris the last one to be quilted onto the remainder. Thisavoids sewing vias through the top layer for any con-

Figure 16 Top view of a stainless steel thread welded toCu/CuSnPb lead


nections other than those required by the top layer.The principle is illustrated in Figure 17.

Capacitive contact and proximity sensing. Briefly,the increase in a sewn electrode’s capacitance dueto finger contact is measured in terms of the timetaken to charge the electrode/finger capacitance fromzero to the switching threshold voltage of a CMOS(complementary metal-oxide semiconductor) logicbuffer (1/2 Vcc or supply voltage).

One form of the instrumentation for this measure-ment is illustrated in Figure 18. The main advantagesof this technique are that it is a time-domain mea-surement, it provides a dynamic range of 104, andit can be implemented with little more than standardCMOS logic. Not only can direct contact be measured,but the dynamic range also permits noncontact sens-ing to occur over a distance roughly twice as far awayfrom an electrode as the electrode is wide.

There are also a number of ways to refine the dy-namic range of the measurement in software. For

example, the measurement granularity of the charg-ing interval is ultimately limited by the length of thesoftware loop which performs the time count. Nor-mally one might discharge the capacitors, then turnon the charging signal and count the number of it-

Figure 17 Exploded view of No-Soap Radio showing stitched vias

STITCHED VIAS

TOP ELECTRODE LAYER

INSULATING LAYER

BOTTOM CIRCUIT LAYER

Figure 18 Schematic of No-Soap Radio

2M 2M

Key0

Key1

Key3

Key2

TO OTHER SENSOR NODES

TO POWER SUPPLY

VCC

PAD0

PAD1

RXD

GND

PAD3

PAD2

TXD

PIC12C509/SO8


erations of a fast loop that elapse before the capac-itor voltage exceeds the input high threshold.

Instead of letting the capacitor charge continuously,however, the software timing the charging curve cantransfer small pulses of current on each iteration,meaning that each tick of the counter correspondsto a smaller portion of the charging curve.

For the MIDI jacket keypad, such careful measure-ments were not used. Instead, a simpler procedurewas employed, with each electrode permanently con-nected to analog ground through a 1-M! resistor,to allow the measurement of the discharge curve ofan electrode. For later systems, however (includingthe flexible MCM and the washable keypad), the tech-nique described above was employed.

Future work

What remains to be done? The present work col-lects our observations and experience over threeyears of work with fabric circuitry in a desire to pointthe way toward scalable processes for the manufac-ture of washable computing systems. It is expectedthat the primary uses of such systems will be as in-terfaces to more commonplace devices, such as pag-ers, phones, and wearable computers.

Regarding washability, e-broidery will not currentlyfunction when wet, but works fine when dried again.Whether the wetting agent is water, isopropyl or ethylalcohol, or one of the common dry-cleaning fluids,once the circuit is dry it functions as specified. Al-though rigorous testing remains to be done, the earlyindications are very promising. Even human sweat(a mildly corrosive saline condensate) does not ap-pear to harm e-broidery over several cycles of ex-posure, while it does in fact corrode conventionaltinned-copper printed circuitry.

The threadframe prototypes have been surprisinglyrobust, and further work should be done to charac-terize their integrity over many stress cycles. Sincethe stainless threads used in this work are composedof many long, thin fibers, we would like to know howlikely an individual fiber is to break when the entirethread containing it is subjected to stress.

What if you want to change your clothes? This is anopen research question that might best be addressedby ongoing efforts into building environments thatsupport mobile agents in embedded applications.One vision of the washable computer of the near fu-

ture is a jacket with many small computing elementsdistributed throughout, all interconnected by astitched network. Each node would provide sensingin its physical locale, some storage for a distributed,redundant database, and a computing element ca-pable of executing small mobile applications. Ajacket or vest could easily host a hundred such smallnodes (an added weight of about 50 grams or 2 ounc-es), and be powered all day by a battery slipped intoa pocket. When you hang up your jacket, it joins thecloset network and synchronizes itself with other ar-ticles of clothing and the network at large. The sce-nario is easy to construct given an embeddable sub-strate for mobile code.

Clothing is also interesting because it can augmentintrabody signaling19 and power transduction20 tech-niques. The capacitive links to the human body thatthese depend on are limited by the total area of ca-pacitive contact available at each link. By transmit-ting power harvested in the shoes through the bodyto a large receiving electrode in the jacket, efficien-cies should be realized that are sufficient to replacethe need for batteries.

Much work remains to be done in replacing Beki-nox and Bekintex with other composite textiles de-signed to withstand corrosion, to allow easy weldingand sewing, and to have higher conductivity. We alsoforesee a way to replace batteries, by collecting thetriboelectric currents that cause charge migration and“static cling,” but this will require careful attentionto the topology and chemistry of weaves intendedto generate power.

Many interesting proposals to make wearable infor-mation infrastructures suggest the application ofhighly redundant optical fiber networks in clothing.We do not see e-broidery as a replacement for op-tical fibers, but rather as a complement. E-broideryexcels at certain things when compared to opticalfiber, but fiber also has its fine points. Many success-ful systems are likely to stem from a clever combi-nation of the two.

Apart from e-broidery, electrospinning21 may also bea viable way to place electronic textiles on a substrate.It may turn out to be a good way to bond compo-nents to circuitry made by e-broidery, for example,by allowing one to deposit polymer meshes of dif-ferent properties onto an existing cloth substrate. Inelectrospinning, the force of electrostatic repulsionis used to generate a fine, splaying jet of polymer(from a droplet of melt or solution) that forms a mesh


at an electrically grounded collection surface. Thismay be better than spot-welding or applying conduc-tive adhesives because electrospinning is guided byelectric fields, so the spinning process will presum-ably extend previously deposited conducting pathsand ensure their continuity.

Electrospinning is compatible with the needs ofwashable computing because it deposits a permeablemesh, rather than a solid, impermeable layer (aswould silk-screening or printing). Materials spun intomesh form can have a better strength-to-weight ra-tio than in the solid form, as well as greater flexi-bility and breathability. By electrospinning with con-ductors, semiconductors, and insulators, one mighteffectively print and encapsulate multilayer electronicstructures. Wearable displays could also be formedby combining electronic ink with an electronic tex-tile substrate.22 Large sensor surfaces are worth ex-ploring. Textiles range in scale from small pieces ofneedlepoint (10$4 m2) to tapestries and carpets (102

m2), with a corresponding range of feature sizes. Thisindicates that textile processes may be well suited tothe production of large, flexible sensing surfaces thatconform to any underlying shape.

Finally, a large interesting area of work to follow willbe in using these techniques to construct meaning-ful and compelling user interfaces and applications.Some of the more obvious applications include cloth-ing that can use electric field measurements to de-termine its fit and shape, keyboards invisibly andcomfortably stitched into sleeves, cuffs, and pock-ets, and hybrid structures that incorporate electronicink into clothing to create highly interactive apparel.

Acknowledgments

The authors are grateful for the support of the MITMedia Lab and its sponsors in pursuit of what ap-peared to be an impossible endeavor (i.e., the mat-ing of textiles and electronics). Neil Gershenfeld andTod Machover were the first to believe we might beable to accomplish these goals. Emily Cooper, JoshSmith, and Josh Strickon contributed to our initialefforts to realize fabric interfaces. Derek Lockwoodhelped design the Firefly Dress. Gili Weinberg hasinfluenced the development of the Musical Ball. BenVigoda, Pamela Mukerji, and Chris Metcalfe all con-tributed to the Electronic Tablecloth, while AndyLippman provided financial and moral support. RickTumminelli, John LeBlanc, and Paul Rosenstrachof MIT’s Draper Labs were responsible for the ini-tiative to produce the flexible MCM fabric keypad

readout. Many stimulating conversations with ThadStarner, Leonard Foner, and Bradley Rhodes shapedthis work in its early form. And, of course, theremight be no such thing as e-broidery were it not forthe inspiration provided by Michael Hawley’s Post-Stitch, an embroidery system based on the Post-Script** graphic programming language.

**Trademark or registered trademark of Velcro Industries B.V.,Microchip Technology Inc., E. I. du Pont de Nemours and Com-pany, Inco Alloys International, Inc., Textron Inc., or Adobe Sys-tems, Inc.

Cited references

1. T. Starner, S. Mann, B. Rhodes, J. Levine, J. Healey,D. Kirsch, R. W. Picard, A. Pentland, “Augmented RealityThrough Wearable Computing,” Presence—Teleoperators andVirtual Environments 6, No. 4, 386–398, MIT Press, Cam-bridge, MA (1997).

2. E. R. Post, E-broidery: An Infrastructure for Washable Com-puting, M.S. thesis, MIT Media Laboratory, Cambridge, MA(1998).

3. E. R. Post and M. Orth, “Smart Fabric, or Washable Com-puting,” Proceedings of First International Symposium on Wear-able Computers, Cambridge, MA, IEEE Computer Society,Los Alamitos, CA (1997), pp. 167–168. See http://www.media.mit.edu/#rehmi/fabric/.

4. L. K. Baxter, Capacitive Sensors: Design and Applications,IEEE Press (1997).

5. O. Omojola, E. R. Post, M. D. Hancher, Y. Maguire,R. Pappu, B. Schoner, P. R. Russo, R. Fletcher, and N. Ger-shenfeld, “An Installation of Interactive Furniture,” IBM Sys-tems Journal 39, Nos. 3&4, 861–879 (2000, this issue).

6. P. E. Garrou and I. Turlik, Multichip Module TechnologyHandbook, McGraw-Hill Book Co., Inc., New York (1998).

7. R. V. Gregory, W. C. Kimbrell, and H. H. Kuhn, “Conduc-tive Textiles,” Synthetic Metals, 28, No. 1, 2, C823–C835 (Jan-uary 1989).

8. R. V. Gregory, M. W. Gilpatrick, and L. W. Adams, Jr., Fab-ric Having Non-Uniform Electrical Conductivity, U.S. PatentNo. 5,316,830 (1992).

9. G. F. Urbish et al., Solderless Electrical Interconnection In-cluding Metallized Hook and Loop Fasteners, U.S. Patent No.5,774,341 (1998).

10. L. L. Bernardoni and T. J. Swirbel, Low Profile MechanicalInterconnect System Having Metallized Loop and Hook Area,U.S. Patent No. 5,457,610 (1995).

11. H. Ohno, F. Narui, and S. Hayashi, “Zipper-Type ElectricalConnector,” U.S. Patent No. 5,499,927 (1996).

12. E. J. Lind, R. Eisler, G. Burghart, S. Jayaraman, S. Park,R. Rajamanickam, and T. McKee, “A Sensate Liner for Per-sonnel Monitoring Applications,” Proceedings of First Inter-national Symposium on Wearable Computers, Cambridge, MA,IEEE Computer Society, Los Alamitos, CA (1997), pp. 98–105.

13. D.Harvey,TheFourSeasons(1997),http://www.costumeinstitute.org/four.htm.

14. Bekaert Fibre Technologies, Find Out More About MetalFibres (1999), see http://www.bekaert.com/bft/.

15. S. J. Everett, Bundle Drawing to Produce Metallic Fibers, U.S.Patent No. 2,050,298 (1936).

16. E. W. Tokarsky, M. G. Dunham, J. E. Hunt, D. E. Santoleri,and D. B. Allen, “Metal Clad Aramid Fibers for Aerospace


Wire and Cable,” Third NASA Workshop on Wiring for SpaceApplications, NASA Lewis Research Center (1995), pp. 197–210.

17. T. W. Eagar, Resistance Welding: A Fast, Inexpensive, and De-ceptively Simple Process, Technical Report, MIT Departmentof Materials Science and Technology (1992).

18. N. G. Koopman, T. C. Reiley, and P. A. Totta, “Chip-to-Pack-age Interconnections,” Microelectronics Packaging Handbook,R. R. Tummala, E. J. Rymaszewski, and A. G. Klopfenstein,Editors, Van Nostrand Reinhold, New York (1989), Chap-ter 6.

19. T. G. Zimmerman, Personal Area Networks (PAN): Near-FieldIntra-Body Communication, master’s thesis, MIT Media Lab-oratory, Cambridge, MA (1995).

20. E. R. Post, M. S. Reynolds, J. Paradiso, and N. Gershenfeld,“Intrabody Buses for Power and Data,” Proceedings of FirstInternational Symposium on Wearable Computers, Cambridge,MA, IEEE Computer Society, Los Alamitos, CA ( 1997), pp.52–55.

21. D. H. Reneker and I. Chun, “Nanometre Diameter Fibresof Polymer, Produced by Electro-Spinning,” Nanotechnology7, No. 3, 216–233 (1996).

22. J. Jacobson, personal communication regarding sewable co-axial displays using electronic ink, May 1997.

Accepted for publication May 9, 2000.

E. Rehmi Post MIT Media Laboratory, 20 Ames Street, Cambridge,Massachusetts 02139-4307 (electronic mail: [email protected]).Mr. Post is presently a doctoral candidate in the Physics and Me-dia Group of the MIT Media Laboratory. He received an M.Sc.in media arts and sciences from MIT in 1998 and a B.Sc. in phys-ics from the University of Massachusetts, Amherst.

Maggie Orth MIT Media Laboratory, 20 Ames Street, Cambridge,Massachusetts 02139-4307 (electronic mail: [email protected]).Ms. Orth is a Ph.D. candidate. She earned her B.A. in paintingand design from the Rhode Island School of Design and anM.S.V.S. from the Center for Visual Studies at MIT. Ms. Orthhas specialized in designing innovative interfaces to encouragecreativity and expression, especially—but not exclusively—in amusical context. She is currently working on a series of musicaltoys to improve how children are introduced to music and is thedesign director of the Toy Symphony project.

Peter R. Russo MIT Media Laboratory, 20 Ames Street, Cam-bridge, Massachusetts 02139-4307 (electronic mail: [email protected]). Mr. Russo is currently a sophomore at MIT studyingphysics and electrical engineering. He is an undergraduate re-searcher at the MIT Media Laboratory, where he works on con-ductive fabric touch sensors, and near-field electric and magneticfield sensors.

Neil Gershenfeld MIT Media Laboratory, 20 Ames Street,Cambridge, Massachusetts 02139-4307. Dr. Gershenfeld leads thePhysics and Media Group at the MIT Media Laboratory, and di-rects the Things That Think research consortium. His laboratoryinvestigates the relationship between the content of informationand its physical representation, from developing molecular quan-tum computers, to smart furniture, to virtuosic musical instru-ments. Author of the books When Things Start to Think, The Na-ture of Mathematical Modeling, and The Physics of InformationTechnology, Dr. Gershenfeld has a B.A. degree in physics withHigh Honors from Swarthmore College, a Ph.D. from Cornell

University, was a Junior Fellow of the Harvard University So-ciety of Fellows, and a member of the research staff at AT&TBell Laboratories.


e-broidery: design and fabrication of textile-based computing

Documents