US3641660A

US3641660A - The method of ball bonding with an automatic semiconductor bonding machine

Info

Publication number: US3641660A
Authority: US
Inventors: Anthony L Adams; Gerald A Yearsley; Marion I Simmons; Billy P Yager
Original assignee: Texas Instruments Inc
Current assignee: Texas Instruments Inc
Priority date: 1969-06-30
Filing date: 1969-06-30
Publication date: 1972-02-15
Anticipated expiration: 1989-02-15
Also published as: DE2032302A1; FR2048054B1; DE2066210A1; KR780000596B1; GB1323331A; NL172806C; JPS565062B1; DE2032302C2; FR2048054A1; NL7009641A; NL172806B

Abstract

An automatic machine for ball bonding a lead wire to an expanded contact of a semiconductor device and stitch bonding the wire to a lead of the semiconductor package is described. The machine automatically aligns an X-Y table in predetermined relationship to the semiconductor device by means of an electro-optical system. A bonding needle mounted on the table is then moved in a predetermined manner to press the balled end of the wire against the expanded contact. Wire is then positively payed out as the bonding needle is moved to the lead where a stitch bond is performed. After the stitch bond, the needle is raised to pull out a length of wire, then the wire is clamped to break the wire near the stitch bond. The length of the wire protruding from the needle is then passed through a flame to form a new ball preparatory to a new cycle. The wire is payed out to the bonding needle from a spool supported by an air bearing. After the ball is formed, reverse torque is applied to the spool to pull the ball up against the end of the bonding needle. After the ball bond, forward torque is applied to pay out the wire, and the wire is positively fed through the needle by an airstream. After the stitch bond is completed and the ball reformed by the flame, reverse torque is again applied to the spool to draw the ball up against the end of the needle preparatory to the next bonding cycle.

Description

Unitedv States Patent Adams et al.

[54] THE METHOD OF BALL BONDING WITH AN AUTOMATIC SEMICONDUCTOR BONDING MACHINE 72 Inventors: Anthony L. Adams, Dallas; Gerald A. Yearsley, Garland; Marion I. Simmons, Richardson; Billy P. Yager, Lake Dallas,

all of Tex.

[73] Assignee: Texas Instruments Incorporated, Dallas,

Tex.

[22] Filed: June 30, 1969 21 Appl. No.: 837,485

[52] U.S. Cl ..29/47l.l, 29/4975, 29/589, 29/628, 228/3 [5]] Int. Cl ..B23k 31/02 [58] Field of Search ..29/470.l, 471.1, 470.3, 497.5, 29/589, 628, 470; 228/3, 4, 5, 15, 16

[56] References Cited UNITED STATES PATENTS 3 ,006,067 10/ 1 961 Anderson ..29/497.5 3,006,068 10/1961 Anderson ..29/497.5 3,087,239 4/ 1963 Clagett ..29/497.5 3 ,235 ,945 2/1966 Hall ..29/470.l 3,250,452 5/1966 Angelucci.. ..228/3 3 ,289,452 12/1966 Kollner ..29/470.l 3 ,313,464 4/1967 Avedissian. ..29/470.l 3,357,090 12/ l 967 Tiffany ..29/497.5 3,397,451 8/1968 Avedissian et al. ..228/3 3,430,835 3/1969 Grable et al. ..29/59] 10/1969 Holzl ..228/ 3 I Primary Examiner-John F. Campbell Assistant Examiner-Donald P. Rooney Attorney-James 0. Dixon, Andrew M. Hassell, Harold Levine, Mel Sharp, John E. Vandigriff, Henry T. Olsen and Michael A. Sileo, Jr.

[5 7] ABSTRACT An automatic machine for ball bonding a lead wire to an expanded contact of a semiconductor device and stitch bonding the wire to a lead of the semiconductor package is described. The machine automatically aligns an X-Y table in prcdetermined relationship to the semiconductor device by means of an electro-optical system. A bonding needle mounted on the table is then moved in a predetermined manner to press the balled end of the wire against the expanded contact. Wire is then positively payed out as the bonding needle is moved to the lead where a stitch bond is performed. After the stitch bond, the needle is raised to pull out a length of wire, then the wire is clamped to break the wire near the stitch bond. The length of the wire protruding from the needle is then passed through a flame to form a new ball preparatory to a new cycle.

The wire is payed out to the bonding needle from a spool supported by an air bearing. After the ball is formed, reverse torque is applied to the spool to pull the ball up against the end of the bonding needle. After the ball bond, forward torque is applied to pay out the wire, and the wire is positively fed through the needle by an airstream. After the stitch bond is completed and the ball reformed by the flame, reverse torque is again applied to the spool to draw the ball up against the end of the needle preparatory to the next bonding cycle.

12 Claims, 22 Drawing Figures PATENTEBFEB 15 m2 3.6M 0660 sneer 01 or 13 1 1 I I I I I FI 2b 28 I r r I I I I I I I z L I x r [4 I I r 1 1 I I 1 1 1 V FIG; 2C

I f I l I 1 1 I I 1 1 I l4 A I 1 I I 1 l INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLEY MAR/ON SIMMONS BILLY P YAG ER ATTORNEY ArEmmrw 15 m2 SHEET INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLEY MAR/ON I. SIMMONS BILLY F. YAGER-v ATTORNEY PATENTEDFEB 15 I872 SHEET UBUF 13 INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLE Y MARION l. SIMMONS BILLY P. YAGEE.

V/QM- /ATTORNEY PATENTEIJFEB 1 5 I972 I saw uuur 12.

u VQN PMENTEBFEB 15 I972 SHEET OSUF 13 FIG. 6

INVENTORS ANTHONY 1.. ADAMS GERALD A. YEARSLEY MAR/ON I. SIMMONS BILL; P. YAGEB, f ATTORNEY PATENTEBFEB15 I972 3.641 .660

SHEET 0? or 13 INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLEY MARION SIMMONS BILLY P. YAGER FIG. Q WQ ATTORNEY PATENTEUFEB 15 1912 3.641.660

SHEET 08 0F 13 INVENTORS ANTHONY 1.. ADAMS GERALD A. YEARSLEY MAR/ON I. SIMMONS BILLY P. YAGERO ATTORNEY PMEENTEDFEB 15 I972 MEET USOF 13 INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLEY MARION l. SIMMONS BIL Y P YAGE ATTORNEY PATENTEDFEB 1 5 I972 3,641,660

SHEET llflF 13 GRECTION STEPPING PLKSES FIG 14a PATENTEDFEB15 1972 SHEET IBM 13 ENABLE I 406) Y L ENABLE JI\ FIG /7 INVENTORS ANTHONY L. ADAMS GERALD A. YEARSLEY MARION l. SIMMONS BILLY P YAGE}? ORNEY perature of about 320 THE METHOD OF BALL BONDING WITH AN AUTOMATIC SEMICONDUCTOR BONDING MACHINE This invention relates generally to method and apparatus for manufacturing semiconductor devices, and more particularly, relates to a bonding methodadapted for automation and fully automatic apparatus for carrying outthe method.

In the typical process for manufacturing a semiconductor device, such as a transistor, a large number of devices are formed on a semiconductor slice nominally about 2 inches in diameter. Expanded metal contacts-are then vapor deposited on the top surface of the individual semiconductor devices before the slice is divided into discrete semiconductor chips which may be on the order of 0.020 inch square. The substrate typically forms the collector, a first diffused region forms the base, and a second diffused region forms the emitter, with separate expanded metal contacts for the base and for the collector regions.

The semiconductor chip is then alloyed to a metal surface of a header, which forms acollector lead of sufficient size to be soldered or otherwise connected into an electronic system. The base and emitter expanded contacts are then typically electrically connected to isolated leads of the header by extremely small gold wires approximately 0.00] inch in diameter. The machine used for this purpose is commonly referred 1 to as a ball bonder. Thetypical ball bonder utilizes a hypodermic-sized tubular needle to press aball formed on the "end of the wire against the respective expanded contact. The

contact, wire and bonding needle are heated to a bonding tem- C. so that the pressure of the needle produces a bond. The needle is manipulated with the aid of a microscope and micromanipulator. The wire passes upwardly through the 'needle and between a pair of resilient pressure pads to a supply spool. The pressure pads tend to tension the wire as it is pulled. After the ball is bonded to theexpanded contact, the needle is raised which pulls the wire from the needle. The needle is then moved over and down until the edge of the lead is pressed against and bonded to the lead. This is called the stitch bond. As the needle is again raised, the wire is pulled from the needleby the stitch bond. The exposed length I of wire is then severed by a flame, which simultaneously forms a new ball on the end of the wire. The length of wire left between the stitch bond and the pointsevered by the flame is referred to as the pigtail," which is subsequently removed by another worker using a pair of tweezers. The new ball is then pulled against the end of the needle by the drag of the pressure pads as the needle is lowered preparatory to making the next ball bond.

There are a number of typical failure modes in a manual bonding system of the typedescribed. If the bonding needle is misaligned at the time the ball is bonded to the expanded contact, either a short circuit or an open circuit can occur. After an attempted ball bond, the bond may fail as the needle is raised because the expanded metal contact separates from the semiconductor chip, because the ball separates from the expanded contact, or because the wire separates at the neck formed between the ball and the wire. These failures are enhanced by the fact that the manner in which the needle is raised is under the control of the operator and may not be smooth and at the proper speed. Any jerky movement will tend to accentuate failure. If the wire separates from the ball at any time, tensioning pads will pull the wire from the needle and disrupt the bonding operation until the wire can be reset and threaded through the needle and a ball reformed. Threading the very fine wire is a tedious and time consuming operation. Such a failure in a high-speed automatic system is much more serious.

The present invention is concerned with a completely automatic system for ball bonding a gold wire to an expanded contact of a semiconductor device, or other surface, then stitch bonding the wire to a second surface. The system includes an electro-optic means for precisely positioning the needle at a predetermined position relative to the semiconductor device, then a means for moving the bonding needle precisely in a predetermined manner to make the bonds. This substantially eliminates faulty bonds due to misalignment. The needle is moved in a manner to minimize wire breakage. The system eliminates the necessity to manually pull pigtails. The system has a sequence of operation which insures con'tinuousoperation even in the event of bond failure or wire separation during a cycle. The bonding machine of the present invention is=ca'pable of making a complete cycle in approximately-one second.

The novel features believed characteristic of this invention are setforth in the appended claims. The inventionitself, however, as well as other objects and advantages'thereof, may best be understood byreference to the following detaileddescription of illustrative embodiments, when read in conjunction with the accompanying drawings, wherein:

FIG. I is a plan view of a typical workpiece for the method and apparatus of the present invention;

FIGS. 2a-2d are simplified side elevational views of the workpiece of FIG. 1 which illustrate the method of the present invention;

FIG. 3 is a side elevational view, partially broken away, of a bonding machine in accordance with the present invention;

FIG. 4 is a front elevational view, partially broken away, of the bonding machine of FIG. 3;

FIG. 5 is a sectional view taken substantially on line 5--5 of FIG. 3;

FIG. 6 is an enlarged side view of a portion of FIG. 3, in cluding a portion of the structure broken away in FIG. 3;

FIG. 7'is a top view of that portion of the apparatus shown in FIG. 6;

FIG. 8 is a sectional view taken substantially on lines 7a-7a of FIG. 7;

FIG. 9 is a 'front view of the apparatus shown in FIGS. 6 and FIG. 10 is a side view, partially in section, of the bonding head of the apparatus of FIG. 3;

FIG. 11 is a top view of the bonding head of F IG. 10;

FIG. 12 is a front view of the bonding head of FIGS. 10 and l FIG. 13 is a schematic circuit diagram of the control system for operating the bonding machine;

FIG. 14'is a schematic circuit diagram of a portion of the circuit shown in FIG. 13;

FIG. 14a is a ruth table for the counter of the circuit of FIG. 14;

FIG. 15 is a schematic circuit diagram of a portion of the circuit illustrated in FIG. 13;

FIG. 16 is a schematic drawing of the control disk illustrated in FIG. 15;

FIG. 17 is a schematic circuit diagram of another part of the circuit of FIG. 13; and

FIG. 18 is a timing diagram which illustrates the operation of the bonding machine of FIG. 3.

BONDING METHOD Referring now to FIG. 1, a transistor header 10 includes a transistor chip 12 which is alloyed to a flattened head on a metal pin or lead 14. The lead 14 together with similar leads l6 and 18 are held in a glass header 20.

Lead wires

24a and 24b extend from the expanded base and emitter contacts to the leads l6 and 18, respectively. The transistor 12 is typically gold and are 0.001 inch in diameter. The present invention is concerned with a method and apparatus for bonding the

lead wires

24a and 24b between the respective expanded contacts on the transistor 12 and the

leads

16 and 18. It is to be understood, however, that the method and apparatus of the present invention can be used to bond wires to any surfaces of any semiconductor device. FIGS. 2a-2d illustrate the method of the present invention for bonding the lead wire 24a to the transistor 12 and lead 18. The semiconductor chip 12 is first positioned at a work station, then the bonding mechanism centered on a predetermined axis 26 of the chip by an electrooptical servosystem hereafter described. The bonding needle 28 is positioned well out of the field of view of the optical device during positioning of the bonding mechanism, the position being illustrated by dotted outline 28a. The wire 24 extends through the hypodermic needle to a supply spool (not illustrated in FIGS. 2a-2d). The wire 24 is maintained under tension by reverse torque on the spool, and ball 25 formed by passing the wire through a flame prevents the wire from being withdrawn from the needle.

After positioning of the eIectro-optical system relative to the chip, the needle 28 is moved to a predetermined position with respect to the axis 26. The predetermined position is selected to align the needle over the appropriate expanded contact of the transistor 12 and the needle is then lowered to press the ball 25 against the expanded contact as shown in solid outline in FIG. 2a. Both the ball 25 and the expanded contacts are heated to bonding temperature so that the pressure of the needle 28 on the ball 25 bonds the wire 24 to the expanded contact.

The needle 28 is then raised and translated so that the point is moved along dotted path 32 (FIG. 2b) and lowered at a second predetermined position relative to the axis 26 selected so as to press the edge of the wire 24 against the upper surface of the lead 18. The torque on the supply spool is reversed as the needle is moved along path 32, and the wire is positively payed out through the needle 28 by means of a jet of air to eliminate tension on the wire and bond, and thus permit the needle to be moved at a high rate of speed without breaking the bond or the wire. The pressure of the needle on the lead 18 forms a stitch bond 33.

Next, the needle 28 is raised to pay out a length of the wire 24c as shown in FIG. 2c, then the wire 24 is clamped relative to the needle 28. As the needle 28 continues to rise, the wire is broken at the point where weakened by the stitch bond 33. The needle 28 then proceeds along path 34 so that the length of wire 24c passes through a flame 36 which forms a new ball 27 as illustrated in FIG. 2d. After passing the flame 36, the wire is unclamped so that reverse torque on the wire spool draws the ball 27 up against the tip of the needle 28, and the needle proceeds on to the initial position represented at 28a in FIG. 2a.

There are important advantages to the method just described and illustrated in FIGS. 2a-2d. As a result of reversing the torque on the spool supplying the wire and the positive pay out of the wire from the needle by an air jet, the tension on the bond between the wire and the expanded contact of the transistor device 12 is materially reduced or eliminated. This reduces the likelihood of failure either because ofthe expanded contact separating from the chip, the ball separating from the expanded contact, or the wire separating from the ball. Even more importantly, in the event of such a failure, the wire 24 is not pulled out of the needle 28 i by reverse torque on the spool as has heretofore been the case in connection with manual bonding machines. Instead, the positive feeding of the wire 24 insures that a length of wire is protruding from the end of the needle so that a stitch bond can be made whether or not a ball bond is achieved. The forward torque on the spool is maintained until such time as the wire is clamped in the step of FIG. 2c so that a premature failure of the wire 24 after the stitch bond or a failure to accomplish a stitch bond will not result in the wire being withdrawn from the needle. Once the wire is clamped, the clamp is maintained until the wire is passed through the flame 36 and a new ball 27 will be formed preparatory to the next cycle. If a successful ball bond or stitch bond has not been effected, the flame 36 will eliminate the excess wire and thus, even through there is a failure to make either a ball bond or a stitch bond, the bonder is conditioned to continue the next cycle. The bonding cycle can be completed in approximately one second, thus making some 60 bonds per minute a reality. The importance of maintaining operability of the bonder even though a bond step should fail is readily evident, since the loss of the wire 24 from the needle would require that the bonder be shut down until the wire could be rethreaded and a new ball formed. Such a period would result in a substantial loss ofproduction.

MECHANICAL APPARATUS The bonding apparatus in accordance with this invention for carrying out the method illustrated in FIGS. 2a-2d is shown in FIGS. 3-18, and is indicated generally by the reference numeral 50. The bonding apparatus 50 is comprised generally of a U-frame support 52 which can best be seen in FIGS. 3 and 4. A first table member, indicated generally by the reference numeral 54 in FIG. 3, is mounted on low friction crossed-roller slides 56 and 58 for movement in what is hereafter termed the X" coordinate direction. The first table member is driven in the X coordinate direction by means ofa ball screw (not illustrated) which in turn is driven by an X" stepping motor 60. Motor 60 is mounted on the frame 52. A second table member 62 is mounted on member 54 by low friction crossed-roller slides 64 and 66 (see FIG. 4) for movement relative to the first table member in what will hereafter be referred to as the Y" coordinate direction. The second table member 62 is driven by a second ball screw (not illustrated), which in turn is driven by a second stepping motor 68 mounted on the first table member 54. Limit switches 71 and 73 (illustrated only in FIG. 13) are mounted in the housing 70, and limit switches 75 and 77 (see FIG. 13) are mounted in housing 72 and are engaged by cams (not illustrated) mounted on the respective ball screw shafts to limit movement of the first table member 54 in the X direction along the slides 56 and 58 and movement of the second table member 62 in the Y direction along the slides 64 and 66 in a manner which will hereafter be described. Thus, the second table member can be moved to any X-, Y-coordinate position relative to support 52, and thus constitutes what is commonly referred to as an X-Y table.

An electro-optical pattern recognition system 74, which is of the type described in copending U.S. application entitled Alignment System, Ser. No. 564,917, filed on July 12, 1966, and assigned to the assignee of the present invention, is mounted on the X-Y table member 62 by a cantilevered arm 76. The operation of such an eIectro-optical system is fully described in the referenced application. In general, the system 74 compares the pattern within the optical field of view, i.e., the semiconductor chip 12, with an identical reference pattern and produces a positive or negative X-error signal and a positive or negative Y-error signal which indicates the direction the eye must be moved along the X- and Y-axes to align the reference pattern with the scanned pattern.

The X- and Y-error signals are used to control operation of the X- and Y-stepping

motors

60 and 68 and thus automatically align the electro-optical eye 74 along the predetermined optical axis 26 (see FIG. 2a) of the semiconductor chip 12 mounted on the header 10 (see FIG. 3) by the circuitry illustrated in FIG. 13 which is hereafter described in detail. The header 10 is supported by a chuck 40 carried on an indexing chain which positions the header generally at the work station so that the semiconductor device will be positioned within the field of view of the optical aligning system.

The X-Y table member 62 also supports a bonding mechanism indicated generally by the reference numeral 80, which will now be described. A platform 82 is slidably mounted on the base plate 84 of the X-Y table member 62 by crossed roller slides 86 and 88 (see FIG. 3) for movement in what is hereafter termed the H direction. The platform 82 is driven in the H direction by an H" stepping motor 92 which is mounted on a depending leg 94 of the table 62 and which drives a shaft 98. The shaft 98 is journaled in the legs 100 and 102 of the table 62. A double acting disk-shaped cam 96 is mounted on shaft 98 and engages

cam followers

104 and 106 which are mounted on the plate 82 by

brackets

108 and 110. Cam 96 is sometimes hereafter referred to as the H cam.

A vertical plate 112 is slidably mounted on the upright leg 114 of an L-shaped member 116 by vertically disposed crossed roller slides 118 and 120, as can best be seen in FIG. 5. The vertical plate 112 is raised and lowered by a stepping motor 122 which drives a shaft 126 which carries a cam 124, which is sometimes hereafter referred to as the V cam. A

cam follower 128 rides on the periphery of the cam 124 and is journaled on the end of an arm 130. The arm 130 is pivotally connected to the vertical plate 112 by a pin 132. The height of the cam follower 128 relative to the plate 112 may be adjusted by screw 134 which is threaded through the downwardly extending extension 136 of the arm 130 against the edge of the plate 112. This permits vertical adjustment of the needle 28 over the chuck 40.

A disk 138 (see also FIGS. and 16) is mounted on the cam 124 by a tubular member 140. The edge of the disk 138 passes through a photoelectric sensing assembly 416 (see also FIG. 13). Apertures at selected points around the periphery are photoelectrically sensed to indicate the position of the V" cam 124 and these signals used to control the V" cam 124 and these signals used to control the Vcam stepping motor 122. A similar disk and photoelectric sensing mechanism is provided to sense the position of the H cam 96 and to control the H" cam stepping motor 92, but has been omitted from the apparatus drawings for simplicity, and is shown only in FIG. 13 at 457.

The bonding needle 28 is mounted on the end of an arm 142 which in turn is pivotally connected to a yoke 144 fixed to the lower end of the vertical plate 112 by pivot pin 146. The downward limit of travel of the arm 142, and thus ofthe bonding needle 28, is limited by an adjustable screw 148 on the horizontal leg of an L-shaped bracket 150 which is fixed to the yoke 144. A solenoid 152 is also mounted on the L-shaped bracket 150 and when energized holds the arm 142 down against the stop 148.

As previously mentioned,'the X-Y table member 62 is automatically aligned with an optical axis on the transistor device 12 by the electro-optic eye 74 and servo system including X- and Y-

motors

60 and 68 which is hereafter described. Also, the slide 112 from which the bonder mechanism 80 depends constitutes, together with platform 82, an H-V table which effects translation of the needle-28 through the various positions depicted in FIGS. 2a-2d. The positions of the fH-V table relative to the X-Y table and hence the position on the transistor at which the bonding needle 28 is first lowered onto the semiconductor chip 12 (to make the ball bond) relative to the optical axis is set in the X- and Y-coordinate directions by

micrometers

154 and 156, respectively, which can best be seen in H0. 5.

The housing of the micrometer 154 is connected to a bracket 158 which is mounted on the horizontal plate 82. The rod 160 of the micrometer is connected to a key 164 which slides in a dove-tailed keyway in horizontal plate 82 (see FIG. 3). The plate 162 is held in the selected position by bolts 166 which extend through elongated openings in the plate 162 and are threaded into the plate 82. Thus, by loosening screws 166 and adjusting the micrometer 154, the position of the

cam followers

104 and 106 relative to the plate 82 can be adjusted, and thus the X" coordinate position of the bonding needle 28 predetermined for any position ofthe H cam 96.

Similarly, the housing of micrometer 156 is mounted by bracket 168 on plate 82, and the rod 170 is connected to bracket 172 on the horizontal leg of the L-shaped member 116. A dove-tailed key 174 is mounted on the bottom of the L- shaped bracket 116 and rides in a keyway 176 in the plate 82. After adjustment in the Y" direction, the bracket 116 can be secured in place by screws 178 which pass through elongated openings in the L-shaped bracket 116 and are threaded into the horizontal plate 82.

A wire supply indicated generally by the reference numeral 180 is mounted on the L-shaped bracket 116 for movement with the platform 82 and is shown in the detailed drawings of F 16$. 6, 7, and 8. The mechanism 180 includes a replaceable wire spool 182 which is mounted on a carrier spool 184. The carrier spool 184 is journaled on a tubular shaft 186 and is held in place by the annular shoulder of a threaded cap 188. Air under pressure is supplied to the interior passageway 190 of the shaft by means of a fluid passageway 192 formed in the arm 194 which supports the shaft 186. The arm 194 is mounted on the Lshaped bracket 116 so as to move with the bonding head assembly. Air is supplied to the passageway 192 by a flexible hose 196 and coupling 198. As can best be seen in FIG. 8, a plurality of orifices 200 extend from the interior passageway of the shaft 186 to the annular space between the shaft 186 and the spool 184. A sufficient volume of air is pumped through the orifices 200 so that the spool 184 is continuously supported only by a layer of air, thus providing an extremely low-friction bearing. Additionally, it will be noted that the orifices 200 are offset slightly from the radial position so that the air circulates around the circumference of the annulus between the shaft 186 and spool 184, thus imparting a reverse torque to the spool 184 which normally tends to wind the wire back up onto the spool. A nozzle 202 periodically receives air under pressure from a flexible hose 204 and directs a jet of air onto a surface 1840 of the spool 184. This jet of air is of sufficient magnitude to overcome the reverse torque produced by the air through orifices 200, and produces a net forward torque on the spool 184 tending to positively feed wire from the spool. Air to nozzle 202 for reversing the spool 184 is controlled by a solenoid-operated valve (not illustrated).

in normal operation, the wire supply spool 184 rotates at a relatively slow rate. If, however, the wire should break for any reason, the air through the jets 200 will soon causethe spool 184 to rotate at a high rate in the reverse direction. Alternate dark and

light segments

205 and 206 on the rim of the spool 184 permit this catastrophic failure to be sensed so that operation will be terminated. Separate fiber

optical bundles

208 and 210 are combined and directed toward the portion of the rim having the light and

dark areas

205 and 206. Light is directed through the bundle 208 onto the rim and the reflected light returned through bundle 210 to a photodetector. The frequency of pulsations and the intensity of the light reflected from the light and dark segments is then electronically detected and a wire failure signal produced when the frequency exceeds a predetermined value. The wire is fed from the spool 182 down to the bonding needle 28 by means of the wire feed assembly shown in detail in FIGS. 9, l0 and 11.

The wire feed assembly includes a means for positively forcing the wire through the bonding needle 28 by air pressure, which is indicated generally by the reference numeral 220, and clamp means for gripping the wire, indicated generally by the reference numeral 222. The clamp means is comprised of a fixed jaw 224 and a movable jaw 226. The fixed jaw 224 is fixed to a bracket 228 by a screw 230. The bracket 228 is in turn connected to the arm 142 by the screw 232. The movable jaw 226 is mounted on the end of a rod 234 which is reciprocated by a solenoid 236. The rod 234 passes through a bushing 238 disposed within a sleeve 240. The sleeve 240 is held in a bore in the arm 142 by a setscrew 242. A compression spring 244 is disposed between the end of the sleeve 240 and a keeper 246 and biases the jaw 266 open whenever the solenoid 236 is not energized.

The positive wire feeding means 220 is also supported by the bracket 228 and is comprised of an inner tubular member 250 which passes through an aperture in the bracket 228and is threaded into an outer tubular member 252. The member 250 has a small diameter tube 250a which extends only partially through a tube 252a extending downwardly from the outer tubular member. Air under pressure is introduced to the cavity 254, and thus to the annulus between the

tubes

250a and 252a, through a fitting 255. The source of air is controlled by a solenoidoperated valve (not shown).

The needle 28 is mounted on an electrical heating assembly 256 which in turn is suspended from the tubular support 258 which passes through the bracket 228 and is connected to the cap 260 on the sleeve 240.

A flame assembly for forming a new ball on the end of the wire after each stitch bond is indicated generally by the reference numeral 261 in FIG. 5. This assembly includes an arm 262 which contains the necessary plumbing to provide combustible gases to a nozzle 264. The arm 262 is supported by the X-Y table member 62 and moves with the table member. The nozzle 264 is pivoted between the operative position shown in solid outline, and an inoperative position shown in dotted outline 264a by means of a solenoid 464 shown only in FIG. 13. When in the inoperative position 264a, the nozzle directs the gases onto an electric igniter 266 which is continuously energized during operation. The nozzle 264 is moved from the inoperative position 264a to the operative position 264 when the solenoid is energized. The position of the igniter 266 is also illustrated in FIGS. 4 and 11.

ELECTRICAL SYSTEM The apparatus heretofore described is operated automatically by the circuitry illustrated in the schematic block diagram of FIG. 13. A first clock 300 produces pulses at the rate of 4,000 per second. One output goes to a divide-by-two counter 302 and the other to a divide-by-l counter 322. The output from divide-by-two counter 30 goes to a master counter 304. The master counter is inoperative until set in operation by a reset pulse on input 305, then proceeds from a count of one to a count of 2,000. The outputs from the master counter 304 are fed in parallel to count decoders 306-318 which produce logic outputs as is hereafter described in connection with FIG. 18.

The logic outputs from decoder 306-308 are used to operate the positioning servo system for the X-Y table 62. The pulses from the divide-by-two counter 302 are gated out through gate 320 when the Servo Enable decoder 306 produces a logic output. When the Low Speed decoder 307 produces a logic output, gate 320 is disabled and pulses from a divide-by-IO counter 322 are gated out from gate 324. The outputs from

gates

320 and 324 are applied to X- and Y-servoamplifiers 326 and 328.

As described in the above-referenced copending application, the scanning eye 74 produces an X-axis error signal on channel 330 which is applied to the X-servoamplifier, and a Y- error signal on channel 332 which is applied to the Y-servoamplifier 328.

The servoamplifier 326 produces a logic signal on channel 334 indicating the direction in which the X-motor must rotate in order to align the reference image in the scanning eye 74, and thus the XY table 62 on which it is fixed, with the semiconductor device 12. This logic signal is routed through the normally closed contacts of

limit switches

71 and 73 and through one pole of a double-pole, three-position manual log switch 336a to the X motor drive circuit 338.

When a logic 1" level appears on output channel 334, the X-motor 60 drives the X-Y table toward limit switch 71 so that upon reaching the limit switch, the logic input to circuit 338 is reversed to a logic 0." The logic 0" level at the input of X-motor driver 338 then starts the table in the reverse direction towards limit switch 73. If the X-Y table engages limit switch 73, the logic signal is changed to a logic "1 level, again reversing the direction of the X-drive motor 60. The pulses from either

gates

320 or 324 are conditioned by the X-servoamplifier 326 and output on channel 340 and through the other pole 336b to the X-motor driver circuit 338.

The driver circuit 338 is illustrated in FIG. 14 and is essentially a four step reversible counter comprised ofJ and K flipflops FF and FF having logic outputs T,, C,, T and C The logic input for determining the direction in which the X-motor is driven is applied to input 344. The stepping pulses are applied to input 346. NAND-gates 348356 perform the necessary logic to cause flip-flops FF and FF to complement in a sequence as successive pulses applied at input 346 to cause the outputs to assume the logic states indicated in the truth table of FIG. 14a. When the input 344 is at a logic I, the counter steps in the forward direction, and when the input 344 is at a logic 0, the counter steps in the reverse direction. The true output T drives transistor 358 which in

turn switches transistors

359 and 360 to either connect output 362 to the positive voltage supply or to ground, thus in essence inverting the logic signal. Similarly, output C, controls transistor 364 which in turn switches transistors 36S and 366 to either connect output 368 to the positive voltage supply or to ground. Logic output T controls transistor 370, which in

turn switches transistors

371 and 372 to control current to output 374. Logic output C controls transistor 376 which in

turn switches transistors

377 and 378 to control current to output 380. Line 382 is a common return from the motor 60. The

outputs

362, 368, 374 and 380 are connected through the resistors shown in FIG. 13 to drive the X-motor 60, the common return 382 being shown in the center. As the four

outputs

362, 368, 374 and 380 are stepped as described, the motor rotates a predetermined amount.

The X-Y table 62 can be manually actuated to move in either the positive or negative direction along the X-axis by throwing the switch 336 to either the upper or lower contact, respectively. When the switch is thrown upwardly, pole 336a is connected to the positive voltage indicating a logic I" level, causing the counter of the motor driver circuit to step in the forward direction each time that a low-speed pulse from the divide-by-IO counter 322 is applied to input 346 through the lower pole 336b. Similarly, when the switch 336 is thrown into the downward position, logic input 344 is connected to ground, which is a logic 0," and the X-motor driver complements in the reverse direction as the pulses from the divide-bylO counter 322 are still input through the lower pole 336b.

The Y-stepping motor 68 is operated the same as the X- stepping motor 60 by the servoamplifier 328, reversing limit switches 75 and 77, three-position, double-pole switch 384, and a Y-motor driver 386, all of which are identical to the corresponding components described in connection with the X- stepping motor.

The V-cam stepping motor 122 and H-cam stepping motor 92 are driven by

circuits

388 and 390 which are identical to the driver circuit 338 illustrated in FIG. 14. The

driver circuits

388 and 390 are controlled by

circuits

392 and 394, respectively, which are illustrated in FIG. 17 and which now will be described.

Pulses to the motor control 392 are derived from a second clock 396 which is operated at 2.4 kHz. and a divide-by-five counter 398 (see FIG. 13). These pulses are received on line 400 (see FIG. 17) and these are applied to inputs of a pair of NAND-

gates

402 and 404. The outputs of

gates

402 and 404 are OR wired to a common output 452. The logic signal on output 406 from V-cam enable I decoder 315 is applied to an input of gate 408. The output of gate 408 is connected to inputs of

gates

402 and 410. The other input to gate 410 is connected through a resistor 412 to a positive voltage supply, and to output line 414 of the V-cam position sensor 416, which is shown in detail in FIG. 15 and will presently be described.

The logic signal on output 418 from V-cam enable ll decoder 316 is applied to an input of gate 420. The output of gate 420 is connected to inputs of

gates

404 and 422. The other input of gate 422 is connected to output 424 from the V- cam position sensor 416, and by resistor 426 to the positive voltage supply.

The V-cam position sensor 416 includes a light source 430, the photodisk 138 and first and

second photodiodes

436 and 438. Light from the source 430 passes through apertures 432 disposed at selected circumferential positions on the same radius from the center of disk 138 as shown in FIG. 16 onto photodiode 36. Light from source 430 passes through a single aperture 434 disposed on a different radius to illuminate the second photodiode 438 and is designated as the starting point of a cycle. An aperture 432 is also provided at the starting point so that the V cam motor can be manually cycled by the enable pulse from a single one-shot circuit.

Photodiode 436 forms a voltage divider with resistor 440 which controls the base of the transistor 442 which is connected as an emitter-follower stage to output 414. A resistor 444 limits current through the transistor 442, Similarly, photodiode 438 forms a voltage divider with resistor 446 and controls the base of the transistor 448 which is connected in current through transistor 448.

Whenever the photodiode 436 is illuminated by light passing through one of the apertures 432; the output 414 approaches ground potential, which is a logic level. Whenever the photodiode 436 is not illuminated, however, thebase of transistor 442 and hence the output 414 becomes sufficiently positive to represent a logic 1 level. The diode 438 operates in the same manner to produce a logic 0" level on output 424 when the diode is illuminated by light passing through aperture 434, and a logic 1" level when the photodiode is dark.

Button 454 provides a means for manually actuating the V" cam motor. When button 454 is closed, one-shot circuit 455 produces an enable pulse on line 406 of the same character as the pulse from V-cam enable I decoder 315, and the pulse produces the same results which will now be described.

Clock pulses are always present on clock input line 400. When photodisk 138 is positioned so that light illuminates the first photodiode 436, line 414 is a logic 0" and the output from gate 410 is therefore a logic I Enable I line 406 is normally at a logic l so that both inputs to gate 408 are a logic l," and the output is a logic 0, thus disabling gate 402 by holding its output at a logic l level. This results in a steady state logic l at the output 452, so that no pulses are applied to the pulse input of the V-cam motor driver circuit 392, which input would correspond to input 346 in FIG. 14. When V-cam enable I decoder 315 detects the appropriate count on the master counter 304, line 406 momentarily goes to a logic "0 level, thus producing a logic 1" level at the output of gate 408 which enables gate 402. Then as the clock pulse line 400 swings from logic 0 from logic 1 to logic 0." This causes the driver circuit 388 to step the V cam motor and move the V-cam until the photodisk 138 blocks light to photodiode 436. This causes output 414 to go to a logic 1" level, which coupled with the logic l fed back from the output of gate 408 causes the output of gate 410 to go to a logic 0 level, which latches the output of gate 408 at a logic l level. As a result, the V-cam motor continues to step even after the enable I pulse line 406 returns to a logic 1" level, until such time as the photodiode 436 is illuminated through the next aperture 432. The line 414 again goes to a logic 0 level, output of gate 410 .goes to a logic 1,

Gates

404, 420 and 422 function in combination with photodiode 438 in the same manner in response to a pulse from V-cam enable II decoder 316. The purpose of the

dual apertures

432 and 434 is to insure that the V-cam always returns to the reference position defined by aperture 434 preparatory to the start of a cycle. If only a single set of apertures 432 were employed and the motor should continue to run past one of the apertures 432 for any reason, the V-cam to logic l, output 452 swings and output of gate 408 goes to a logic 0 to hold .the output ofgate 402 at a steady state logic 1" level.

would be out of step for all succeeding cycles until manually reset. However, by producing the enable ll pulse to move the photodisk 138 from its position at the last aperture 432 in the cycle, the V-cam motor will continue to operate until stopped by registry of the aperture 43 between the light source 420 and photodiode 438.

The H-cam motor 92 is driven by the same circuit components as the V-cam motor 122, except that-the stepping pulses are derived from the clock 396 by a divide-by-four counter 456. The H-cam position sensor 457 is identical to the V-cam position sensor 416 and the one-shot circuit 458 is identical to the one-shot circuit 455.

Air through the jet 202 to produce forward torque on the wire spool is controlled by a solenoid 460 which is energized by the wire spool forward decoder 309. Air to the cavity 254 of the wire feed mechanism is controlled by solenoid 462 which is energized by the wire feed decoder 310. The wire clamp solenoid 236 is energized by an output from the wire clamp decoder 312. The flame nozzle 264 is moved from the inactive position to the active position by solenoid 464, which is energized by the flame active decoder 313. The holddown solenoid 152 is energized by the holddown decoder 314. Provision is made to manually selectively energize each of the five solenoids as represented by pushbuttons 466-470 and the bank of diodes 472. The diodes 474 connected in parallel with the various solenoids protect the circuit from inductive pulses.

OPERATION OF APPARATUS In order to set the bonding machine 10 up for operation, a reference corresponding to the image of the semiconductor chip 12 is inserted in the photoelectric scanning mechanism 74. The X- and Y-coordinate positions of the expanded contact where the ball bond is to be made relative to the optical axis of the image is then set by adjusting micrometers l54'and 156. The wire 24 from the spool 182 is then threaded through the clamp 222, tube 250a and the needle 28 and a ball formed. The apparatus may be manually operated for this and other purposes by manual switches 466470, and the switches 454 of the V-cam position sensor circuit 416 and H cam position sensing circuit 457. In addition, the X-Y table 62 can be manually jogged in either direction on either axis by

switches

336 and 384.

The automatic sequence of operation can best be understood by referring to FIG. 18. As the chuck 40 moves the device 10 into position, a reset pulse is mechanically generated on line 305 to reset the master counter 304 which immediately begins to count. The servo enable decoder 305 produces a logic signal on the count of 002 which gates pulses from the divide-by-two counter 302 to the X- and Y-servoamplifiers 326 and 328. The X- and Y-stepping

motors

60 and 68 are then driven in a direction to align the reference pattern in the scanning eye 74 with the semiconductor chip 12 as a result of the X- and Y-error signals fed back on

lines

330 and 332. The indexing cycle is completed on the count of 025 at which time the eye zoom decoder 308 produces a logic signal which causes the scanning eye 74 to effect an optical zoom in on the chip 12. On the count of 125, the low-speed decoder 307 disables gate 320 and enables gate 324 so that the low-speed pulses from the divide-by-l0 counter 322 are applied to the X- and Y-servoamplifiers 326 and 328. The logic signal for both the servo enable decoder 306 and the low speed decoder 307 terminate on the count of 225. The eye zoom logic signal terminates on the count of 250.

During the alignment by the X- and Y-stepping

motors

60 and 68, the needle is positioned out of the optical field of view represented by the dashed lines 480 as best seen in FIG. 12, and as represented at 28a in FIG. 2a. Before the termination of the period during which the eye is zoomed in on the chip, the holddown solenoid 152 is energized at the count of to stabilize the pivoted arm 142. The H-cam enable I decoder 317 produces a pulse at the count of 206 to commence moving the needle 28 horizontally to the bail bond position. The enable pulse lasts for only a sufficient period of time for the aperture in the photodisk of the position sensor 457 to move so that the photodiode is no longer illuminated. The l-l-cam motor continues to rotate until another aperture is aligned with the photodiode to produce a pulse to the H-cam motor control 394 to terminate operation of the motor. This occurs approximately at the count of 365 and the bonding needle is then positioned over the expanded contact on chip 12 to which the ball bond is to be made.

The holddown solenoid 152 is 345.

Next the V-cam enable l decoder 315 produces an enable pulse at the count of 375 which sets the V-cam steppingmotor 122 in operation. As the V-cam rotates, the vertical plate 112 is lowered so that the ball on the end of the wire held by the bonding needle 28 is pressed against the expanded contact. The pivoted arm 142 insures that only the correct amount of pressure is applied as the plate 112 travels below the level at which the ball engages the expanded contact. The V-cam is deenergized at the count of stopped at approximately the count of 497 when an aperture 432 in the photodisk 138 comes into register between the light source 430 and the photodiode 436. This movement of the needle is illustrated in FIG. 2a.

While the V-cam is lowering the needle, wire spool forward decoder 309 energizes solenoid 460 on the count of 425 to produce a jet of air from nozzle 202, which produces a net forward torque on the wire spool 182. The flame active decoder 313 energizes solenoid 462 at the count of 500 so that the flame nozzle is pivoted into the active position illustrated by the flame 36 in FIG. 2d.

After a short dwell period to insure that the ball bond is firm, the V-cam enable I decoder 315 produces a second enable pulse which again sets the V-cam motor in operation, and the V-cam motor 122 remains in continuous operation until approximately the count of 1,268, when the second aperture 432 registers between the light source 430 and the photodiode 436. H-cam enable 1 decoder 317 also produces a second enable pulse on the count of 650 which sets the H-cam motor 92 in operation. By the count of 673, the V-cam has raised the needle to its maximum height for travel and the needle then dwells in this position from approximately the count of 673 to the count of 815, although the V-cam continues to rotate. At approximately the count of 790, wire feed decoder 310 energizes solenoid 462 so that a pulse of air is introduced to the wire feed mechanism 220 to positively feed wire from the needle 28. The V-cam starts lowering the needle at the count of 815 and the H-cam motor is stopped at the count of 857. By the count of 912, the V-cam has moved the needle to the fulldown position against lead 18 and dwells until the count of 1,062 to produce the stitch bond. From the count of 912 to the count of 1,000, holddown decoder 314 energizes solenoid 152 so that the additional force produced by the solenoid increases the pressure of the needle on the edge of the wire to make the stitch bond. This portion of the travel of the needle is illustrated in FIG. 2b.

By the count of 1,062, the V-cam is raising the needle to pay out wire and on the count of 1,200, the wire clamp decoder 312 energizes wire clamp solenoid 236. As the V-cam continues to raise the needle, the wire is broken at the point where it has been weakened by the stitch bond. This movement is illustrated in FIG. 2c.

As soon as the wire has been broken, the H-cam enable 1] decoder 318 produces an enable pulse at the count of 1,225 to start the H-cam motor 92 and move the needle back to the original position out of the field of view of the scanning eye 74. Shortly thereafter the V-cam motor is stopped at 1,268 by feed back from V-cam position sensor 416. The wire clamp remains energized until the count of 1,475, by which time the wire has passed the flame 36 and a new ball has been formed. The travel of the needle between the counts of 1,225 and 1,525 is illustrated in FIG. 2d.

As soon as the wire clamp solenoid 236 is deenergized, V- cam enable ll decoder 316 produces an enable pulse which again sets the V-cam motor 122 in operation and moves the needle downwardly to the intermediate dwell position preparatory to the start of the next cycle. As the needle moves downwardly, the newly formed ball on the end of the wire is pulled upwardly against the tip of the needle by the reverse torque on the spool caused by the air continually passing through jets 200 to the air bearing. As soon as the needle is lowered below the level of flame 36, the flame active decoder 313 deenergizes solenoid 464 at the count of 1,525 to pivot the flame nozzle back to the igniter 266.

At the count of 1,525, H-cam position sensor 457 produces a pulse from the second photodiode to stop the H-cam motor 92, and at the count of 1,653 the V-cam position sensor 415 produces a pulse from the second photodiode to terminate operation of the V-cam motor.

The chuck 40 begins to index at the count of 1,225 and the movement of the chuck generates a new reset enable pulse on input 305 at approximately the count of 1,800 and the cycle is repeated. Thus, it will be noted that the complete cycle takes less than one second.

Although preferred embodiments of the invention have been described in detail, it is to be understood that various changes, substitutions, and alterations can be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

What is claimed is:

1. A method for bonding a wire to interconnect first and second metallized surfaces, at least one ofsuch surfaces being on a semiconductor device, comprising the steps of:

a. passing the end of the wire from a supply spool through a tubular needle;

b. passing the wire across a flame to form a ball on the end of the wire having a diameter greater than the inside diameter of the tubular needle;

c. pulling the wire back through the tubular needle until the ball touches the end of the needle;

(1. pressing the ball against the first surface by means of the needle to bond the ball to the first surface;

e. moving the needle from the first surface to the second surface;

f. pressing one side of the wire against the second surface to bond the wire to the second surface;

g. raising the needle from the second surface;

h. gripping and pulling the wire with sufficient force to break the wire at the bond on the second surface;

i. passing the wire across the flame to form a new ball; an

j. pulling the wire back through the tubular needle until the new ball touches the end ofthe needle.

2. The method of claim 1 wherein the wire is pulled back through the tubular needle by applying a reverse torque to the spool.

3. The method of claim 1 including the step of applying a forward torque to the spool while moving the needle from the first surface to the second surface to reduce the tension in the wire as the needle is moved.

4. The method of claim 2 wherein the wire is positively fed from the needle by a gaseous stream while the needle is being moved from the first surface to the second surface.

5. The method of claim 1 wherein the distance between the first and second surfaces is a fixed distance.

6. A method of interconnecting first and second metallized surfaces, at least one of which is on a semiconductor device, with a jumper wire stored on a spool and passed through a tubular bonding needle, comprising the steps of;v

a. forming a ball on the end ofthe wire;

b. pulling the wire back through the tubular bonding needle until the ball touches the end of the bonding needle;

0. heating the wire and the first and second surfaces to a bonding temperature;

d. pressing the ball against the first surface by means of the needle to bond the ball to the first surface;

e. moving the needle to the second surface;

f. pressing the wire against the second surface by means of the bonding needle to bond the wire to the second surface;

g. raising the needle from the second surface; and

h. clamping and pulling the wire to break the wire adjacent the bond to the second surface.

7. The method of claim 6 including the step of passing the wire protruding from the needle after the second bond across a flame to form a new ball on the end of the wire.

8. The method of claim 6 wherein the wire is clamped relative to the needle and the needle further raised above the second surface to break the wire adjacent to the bond on the second surface.

9. The method of claim 6 wherein:

a. the needle is raised from the second surface a predetermined distance to pull a predetermined length of wire from the needle; then b. the wire is held relative to the needle and the needle raised to break the wire adjacent to the bond and leave said predetermined length of wire protruding from the needle.

Claims

1. A method for bonding a wire to interconnect first and second metallized surfaces, at least one of such surfaces being on a semiconductor device, comprising the steps of: a. passing the end of the wire from a supply spool through a tubular needle; b. passing the wire across a flame to form a ball on the end of the wire having a diameter greater than the inside diameter of the tubular needle; c. pulling the wire back through the tubular needle until the ball touches the end of the needle; d. pressing the ball against the first surface by means of the needle to bond the ball to the first surface; e. moving the needle from the first surface to the second surface; f. pressing one side of the wire against the second surface to bond the wire to the second surface; g. raising the needle from the second surface; h. gripping and pulling the wire with sufficient force to break the wire at the bond on the second surface; i. passing the wire across the flame to form a new ball; an j. pulling the wire back through the tubular needle until the new ball touches the end of the needle.

6. A method of interconnecting first and second metallized surfaces, at least one of which is on a semiconductor device, with a jumper wire stored on a spool and passed through a tubular bonding needle, comprising the steps of; a. forming a ball on the end of the wire; b. pulling the wire back through the tubular bonding needle until the ball touches the end of the bonding needle; c. heating the wire and the first and second surfaces to a bonding temperature; d. pressing the ball against the first surface by means of the needle to bond the ball to the first surface; e. moving the needle to the second surface; f. pressing the wire against the second surface by means of the bonding needle to bond the wire to the second surface; g. raising the needle from the second surface; and h. clamping and pulling the wire to break the wire adjacent the bond to the second surface.

9. The method of claim 6 wherein: a. the needle is raised from the second surface a predetermined distance to pull a predetermined length of wire from the needle; then b. the wire is held relative to the needle and the needle raised to break the wire adjacent to the bond and leave said predetermined length of wire protruding from the needle.

10. The method of claim 9 wherein the predetermined length of wire is then passed through a flame to form a new ball and the wire is pulled back through the tubular bonding needle until the new ball touches the edge of the bonding needle.

11. In a method for thermocompression bonding a small diameter jumper wire between two surfaces, the steps of: a. bonding the wire to one of said two surfaces by means of a bonding needle; b. positively feeding the wire from the bonding needle while the needle is being moved from said one surface to the other of said two surfaces; c. bonding the wire to said other surface by means of said bonding needle; wherein d. the wire is positively fed from the needle while the needle is being moved from said one surface to said other surface by passing the wire and a gaseous stream through a small diameter passageway in said needle such that the gaseous stream of air exerts a frictional force on the wire tending to feed the wire from said needle.

12. The method of supplying small diameter wire to a thermocompression bonding needle for bonding a jumper wire between first and second surfaces which comprises: a. storing the wire on spool; b. normally applying reverse torque on the spool to maintain tension on the wire; c. bonding the wire to said first surface by means of a needle; d. applying forward torque to the spool to overcome said reverse torque and reduce the tension on the wire as the needle is moved between the first and second surfaces; e. bonding the wire to said second surface by means of said needle; wherein f. the forward and reverse torques are applied to the spool by the friction of a gaseous stream acting against the spool.