Method of testing wire bonds for defects using a magnetic field

METHOD OF TESTING WIRE BONDS FOR
DEFECTS USING A MAGNETIC FIELD

BACKGROUND OF THE INVENTION 5

This invention relates to integrated circuit packages of the type in which multiple wires are bonded to an integrated circuit chip; and more particularly, it relates to methods of testing the wire bonds for defects in such packages. 10

Typically, the above type of integrated circuit package includes a ceramic body having a cavity to which the back surface of the integrated circuit chip is attached. Various circuitry such as bipolar or CMOS or NMOS logic circuitry is integrated into the front sur- 15 face of the chip. In order to transmit electrical signals to and from the circuitry on the chip, a plurality of discrete wires are bonded between bonding pads of metal signal lines in the body of the package and the front surface of the chip. 20

One way in which these bonds are formed is by a thermocompression method. In this method, a bonding tool presses the wire against the bonding pad at an elevated temperature which is below the melting temperature of the wire. This causes the wire to laterally 25 spread in the bonding area and there make intimate contact with the bonding pad which results in the bond.

Wire bonds are also formed by an ultrasonic bonding method. In this method, a bonding tool presses the wire against the bonding pad and simultaneously vibrates the 30 wire at ultrasonic frequencies. These vibrations disrupt and disperse any oxides and other films between the wire and bonding pad and thereby bring the wire and bonding pad into intimate contact which forms the bond. 35

To determine the quality with which these bonds are fabricated, the bonds between every wire and pad on the chip, and between every wire and pad on the body of the package, must be tested. To that end, mil standard 833c specifies a nondestructive bond pull test, the pur- 40 pose of which is to reveal nonacceptable wire bonds. In this test, an apparatus having a hookshaped member is utilized to hook onto and pull each bonded wire with a predetermined force which the wire should be able to withstand. For example, bonded aluminum wire having 45 a diameter of 0.00125 inches should withstand a pull force of 2.5 grams.

Such a test procedure works well for SSI and MSI chips of the prior art. However, modern VLSI chips have a very large number of bonding wires, (e.g., over 50 one hundred wires). Consequently, pulling on each of the bonding wires one at a time with a hook is very time-consuming. Further, as the number of bonding wires increases, the space between the bonding wires is forced to decrease. Bonding wires on modern VLSI 55 chips can be spaced by only 0.006 inches. As a result, placing the hook of the testing apparatus around a single bonding wire without damaging or shorting some of the adjacent wires is very difficult and tedious.

Accordingly, a primary object of the invention is to 60 provide an improved method for testing the wire bonds of an integrated circuit package which overcomes the above described problems.

BRIEF SUMMARY OF THE INVENTION 65

According to the present invention, the above object and others are achieved by a method which includes the steps of: placing an integrated circuit package in a mag

2

netic field; generating a current through the bonding wires of the package while the package is in the field to thereupon induce a magnetic force; and, monitoring the current through the bonding wires to determine if a bond to any wire breaks under the magnetic force thereby causing the current to stop.

BRIEF DESCRIPTION OF THE FIGURES

Various features and advantages of the invention are described herein in conjunction with the accompanying drawings wherein:

FIG. 1 illustrates the claimed method by which wire bonds are tested;

FIG. 2 is a top view of the integrated circuit package of FIG. 1 illustrating how the bonding wires deflect when their bonds are being tested;

FIG. 3 shows a mathematical analysis of how the stress, B field, and current interact during the wire bond test of FIGS. 1 and 2;

FIGS. 4A-4C illustrate the steps by which current can be generated in the bonding wires of a chip which contains bipolar logic circuitry;

FIG. 5 illustrates the steps by which current can be generated in the bonding wires of a chip which contains field effect logic circuitry; and

FIG. 6 shows equations which describe how stress on the bonding wires vary as a function of the frequency of the current through the wires.

DETAILED DESCRIPTION OF THE
INVENTION

Referring now to FIG. 1, the details of the claimed method of testing wire bonds for defects in an integrated circuit package will be described. In FIG. 1, reference numeral 10 indicates the integrated circuit package whose wire bonds are being tested. This package includes a ceramic body 11, an integrated circuit chip 12, and a plurality of discrete wires 13 which are bonded at their ends between bonding pads of metal signal lines in the body 11 and chip 12. Those bonds 14 are what the claimed invention tests.

To perform the test, electrical signals are applied to the package 10 in a fashion which causes a current "i" to flow through the wires 13. This is achieved by placing the input/output terminals 15 of package 10 into a socket 20 and by applying appropriate voltages to a connector 21 of the socket.

While the current i is flowing through the conductors, a magnetic field B is directed through the package. In FIG. 1, this magnetic field B is produced by a permanent magnet 30 having a north pole 31, a south pole 32, and an air gap 33 in which the package 10 and socket 20 are placed. Preferably, magnet 30 has a retentivity of at least 1.0 Weber's per meter squared such as that of Tungsten steel or Alnico V and VI.

In operation, the magnetic field B and current i interact to produce a force on the conductors 13 which is perpendicular to the plane of FIG. 1. As a result, the wires 13 are deflected sideways as illustrated by the dashed lines 13' in FIG. 2. Preferably, the magnitude of the current i is generated to change in a cyclic fashion to cause the wire to move back and forth.

By moving the wire 13 back and forth for several cycles, the magnitude of the magnetic force which is required to test the bond 14 is substantially reduced over that which is required by a nondestructive pull test. This is because the force which a wire bond can withstand decreases with an increase in the number of times which that force is applied to the bond.

To move wire 13 back and forth, it is not necessary that the current i be made to flow in both directions through the wire. All that is needed is that the magni- 5 tude of the current i be made to vary in a cyclic fashion from a value at or near zero to a predetermined maximum. Then, as the current i decreases from its maximum value, the mechanical stresses in the wire will move it back to its rest position. 10

Turning now to FIG. 3, suitable values for the magnetic field B and current i in the above wire bond test will be mathematically estimated. In this mathematical analysis, the wire 13 is treated as a beam which is fixed (bonded) at its ends between chip 12 and package 11. A 15 static force is uniformly applied to the beam, and F is the force per unit length. This is shown in the force diagram of FIG. 3; and equation 1 of FIG. 3 gives the stress which is present at the ends of the wire in that diagram. 20

In equation 1, S is the stress at the ends of the wire; M is the bending moment at the ends of the wire which is caused by the distributed load F; r is the radius of the wire, and I is the moment of inertia of the wire. Equation 2 restates quantities M and I in terms of the radius 25 of the wire r, the length of the wire 1, and the force per unit length F on the wire.

Substituting equation 2 into equation 1 yields equation 3 which expresses the stress in the ends of wire 13 . in terms of the parameters F, 1, and r. This equation can 30 be further modified by using the relation: magnetic force on a wire equals the length of the wire times the cross product of the wire's current i and the wire's surrounding B field. Thus, equation 3 can be rewritten as equation 4. 35

For annealed aluminum bonding wire, the fatigue strength S is 5,000 PSI at 500 million cycles. This means that the wire can be flexed back and forth 500 million times from a stress of — 5,000 to 5,000 PSI before breaking. This is stated as equation 5. Thus, a suitable nonde- 40 structive fatigue test for the wire would be to cycle the - stress in the wire from 0 PSI to 1,000 PSI about a million times. This is stated as equation 6.

Typical values for the length 1 and radius r of a bonding wire respectively are 0.100 inches and 0.00125 45 inches. Converting these values for stress S, length 1, and radius r into RMKS units and the result into equation 4 yields equation 7. Suitable values for B and i can then be picked which meet the constraint of equation 7. For example, as stated in equation 8, the maximum 50 value of the current i can be set equal to 40 milliamps, and the magnetic field B can be set equal to 1.0 Weber's per square meter.

Consider now in further detail the steps by which the current i is generated in the bonding wires 13 in the case 55 where chip 12 is a bipolar logic circuit. This case is depicted in FIG. 4A in which some of the bonding wires 14a carry input logic signals while other bonding wires 136 carry output logic signals. Also in FIG. 4A, reference numeral 40 indicates the bipolar logic on the 60 chip; and reference numerals 41,42, and 43 respectively indicate an inverter and two bipolar transistors which are intercoupled in a push-pull fashion to drive the output conductor 136.

One way to generate the current i in the conductors 65 13a of the FIG. 4A circuit is to apply high logic signals to all of those input conductors. Thereafter, current can be generated in the output conductor 136 by applying a

set of logic signals to the input conductors 13a such that the push-pull transistor 42 turns on.

However, as an alternative, the current i can also be generated in the output conductor 136 independent of the signals which are applied to the input conductors 13a. How this is achieved is illustrated in FIG. 4B. In that figure, the physical makeup of the bipolar transistors 42 and 43 are shown in cross section; and C(N) indicates an N type collector region, B(P) indicates a P type base region, and E(N) indicates an N type emitter region.

Under normal operating conditions, the supply voltages Vcc and Vee are such that the collector-substrate junctions are reverse biased. Typically, Vcc is 5 volts and Vee is 0 volts. Under these conditions, no current flows through the collector-substrate junctions.

However, by choosing the supply voltages differently, the collector-substrate junctions of transistors 42 and 43 can be forward biased; and this in turn will cause current to flow through the output conductor 136 independent of the input signals. For example, the substrate voltage Vee can be biased at +5 volts and the collector supply voltage Vcc biased at 0 volts.

FIG. 1 illustrates the current loop for one of the output conductors 136 as an example. This loop consists of a conductor 22 in socket 20, the right-hand terminal 15 of package 10 which connects via a conductor 16 to the back surface of chip 12, chip 12 itself, the left bond wire 13, the left-hand terminal IS of package 10, and a conductor 23 of socket 20. Generating the current i through the back of the chip is very attractive when a large number of chips having different logic sections 40 are to be tested, or when a single chip which generates many different output signals is to be tested because the input signals don't have to be tailored to the particular chip.

Usually the input of a bipolar logic circuit is coupled through a protection diode 44 to the substrate as illustrated in FIG. 4C. For those circuits, when the substrate is biased at a higher voltage than the voltage on the input conductor 13a, current flows through the back surface of the chip, then through the input protection diode 44 and then through the input conductor 13a. An attractive feature of this is that the magnitude of the current which can flow through the protection diode 44 is usually bigger than the normal input current to a logic circuit, so consequently the Bi constraint of equation 7 is more easily met.

In the case where chip 12 contain CMOS or NMOS field effect type logic, a similar method may be used to generate the current i in the conductors 13. Such a circuit would be depicted in the same fashion as FIG. 4A but with block 40 reading "field effect logic" and transistors 42 and 43 being field effect transistors. A cross-sectional view of such output field effect transistors 42' and 43' is shown in FIG. 5.

Under normal operating conditions, the FIG. 5 circuit is biased such that supply voltage Voois +5 volts and substrate bias voltage Vss is 0 volts. However, by changing the substrate bias voltage Vss to +5 volts and the supply voltage Vdd to 0 volts, current will flow through output conductor 136 independent of the signals on the input conductors 13a. At the same time, current will also flow through the input conductors since the input to field effect gates is almost always connected to input protection diodes just like those of FIG. 4C.

7

stantially less than a pre-established force which each bond must be able to withstand for one pull.

10. A method according to claim 8 wherein said conductors which are bonded to said chip surface and said body portion of said package have a natural frequency; 5 and wherein said method includes the step of cyclicly varying the magnitudes of said currents in said conductors near said natural frequency.

11. A method according to claim 8 wherein said circuits include: a plurality of logic gates which are inter- 10 connected on said chip to generate multiple output signals, bipolar transistors with bases which receive said output signals, and collectors which are bonded to said conductors; and wherein said method further includes the step of generating said currents through said con- 15 ductors by applying a voltage to the substrate of said

8

chip which forces said currents to flow serially through said substrate, said collectors, and said conductors independent of the salts of said output signals.

12. A method according to claim 8 wherein said circuits include: a plurality of logic gates which are interconnected on said chip to generate multiple output signals, field effect transistors with gates which receive said output signals, and drains which are bonded to said conductors; and wherein said method further includes the step of generating said currents through said conductors by .applying a voltage to the substrate of said chip which forces said currents to flow serially through said substrate, said drains, and said conductors independent of the state of said output signals.

*****

20

25

30

35

40

45

50

55

60

65