In this section, consideration is given to the variables which are involved in the resolution-of-forces calculation. These are the position of the hook, E, the angle ẞ (the angle formed between the direction of the pulling force and the normal to the line joining the bond terminals), and the loop height, h. In studying each variable, an extensive series of experiments was carried out in order to compare the prediction of the resolution of forces with the measured pull strengths.

An experiment was performed to demonstrate the effect of the position of the pulling hook on the measured pull strength of a single-level ultrasonic bond pair. The results of the experiment are shown in figure 7 in which the measured pull strength, F, of the bond pair is plotted against the position of the hook along the span of wire. The plotted points are the mean pull strength and the error bars represent the 95 percent confidence intervals for the mean.

These experiments indicate that when the wire is pulled on the side nearest the first bond, it is this bond that breaks at the heel due to tensile failure. Conversely, the second bond breaks by tensile failure when the force is applied nearest to it. A transition point, where the probability for either bond to break is nearly the same, occurs at the maximum value of measured pull strength. In the present case this point is found to be displaced more toward the second bond, as might be expected since the bonding machine used in these tests yields a stronger second bond than first.

It should be noted that limited tests on aluminum wire doped with 1% magnesium rather than silicon indicated, as expected, that there is no difference with respect to the dependence of the pull strength on the geometrical variables. The resolution of forces calculation relates the forces to geometrical variables only, while wire metallurgy determines the tensile strength of the wire [6].

4.1.2 TWO-LEVEL BONDS

The effect of varying the position of the pulling hook on pull strength was studied on two-level bond pairs where the first bond was made either to the low pad or to the high pad. Specimen preparation procedures restricted the study to a single value of H, the height of the terminal above the die. The results, plotted in figure 8 as measured pull strength normalized to its value at midpoint as a function of hook position, are in agreement with calculations based upon tensile failure except for positions between midpoint and the low pad. It should be noted that in this region the pulling hook was observed to slip along the wire loop toward the midpoint position; thus the measured pull strength lies above the predicted value [5].

Figure 7.

Measured pull strength of single-level bond pairs for different hook positions. (The first bond is at 0 and the second bond is at 1 mm).

Figure 8. Measured pull strength of two-level bond pairs as a function of the position of the pulling hook. (The values are normalized to the mean pull strength at mid span, 0.5 mm. Solid points are for the first bond made to the high pad; open points are for the first bond made to the low pad. In each case the second bond is at 0 mm and the first bond is at 1.0 mm. The data points represent the mean of 10 bonds all of which ruptured by tensile failure at the heel of the first bond. Error bars indicate one sample standard deviation above and below the mean. The solid curves were calculated by resolution-of-forces.)

While the discussion of the resolution-of-forces has been cast in terms of the pull angle it is often more convenient to represent the results of pull test experiments (particularly for two-level bonds) in terms of a different angle. This angle, denoted by 8, is defined as that formed between the direction of the pulling force and the normal to the line joining the bond terminals (positive in the direction toward the second bond). The angle B is related to the angle by the equation 8 = B B+ where the positive sign obtains if the first bond is the higher (in which case ß, the angle between the substrate and the line joining the bond terminals, is negative), and the negative sign obtains if the first bond is the lower (in which case is positive). The various angles are illustrated in figure 9.

4.2.2 SINGLE-LEVEL BONDS

O

=

Single-level bond loops were pulled at the midpoint of the loop at angles between 0 and 45 deg. in both directions. It should be noted that B O for single-level systems and hence 8 = $. The results are shown in figure 10. The mean values of the measured pull strength are denoted by the plotted points and the error bars represent the 95 percent confidence intervals for the mean. The results show higher pull strengths for 8 negative which is to be expected since the second bond is stronger than the first. When the larger component of the force is applied to the first bond, the measured pull strength is lower and when the larger component of the force is applied to the second bond, the measured pull strength is higher [6].

4.2.3 TWO-LEVEL BONDS

The results of the single-level experiments for the larger values of ẞ yielded some insight into the pull test for two-level bond pairs. For a two-level bond system, there are two ways the bond loop may typically be positioned for pulling as shown in figure 11. First, the substrate or package may be held level, which places each of the bonds at different heights, and the force applied in a direction normal to the substrate.

Alternative

ly, the substrate may be tipped to bring both bonds to the same vertical height; the configuration produced is more nearly similar to a single-level bond case, and the force is applied in a direction normal to an imaginary line joining the two bonds. The configuration in the level substrate case is similar to that of a single-level bond system where the pulling is done at some angle B It is apparent that tipping or not in the two-level bond system is essentially equivalent to varying the angle B. Although this comparison of single-level and two-level bond systems is in terms of bond loop geometry only, one would expect that the general results shown in figure 10 would be applicable to the twolevel system. That is, if the angle of pull is inclined toward the second bond (applying the larger component of force to the first bond) the measured pull strength is less than if the angle of pull is inclined toward the first bond (applying the larger component of

Figure 9. Geometric variables for the pull test for bonds made on two-level substrates for pulling normal to the substrate. (For this case, the angle between the normal to the substrate and the pulling direction is zero. The distance d is the total bond-to-bond spacing in the plane of the substrate.)

Figure 10. Measured pull strength of single-level bond pairs as a function of the angle of pull in the plane of the bond loop. (The inset shows the relationship of the angle of pull to the normal to the line joining the bonds. Note that the angle is positive when the pull force is inclined toward the second bond and negative when the pull force is inclined toward the first bond.)

Figure 12. Measured pull strength of bond pairs as a function of the angle of pull in the plane of the bond loop. (The values are normalized to the mean pull strength at 0 deg. Solid points are for two-level bond pairs made with the first bond to the high pad (circles) and with the first bond made to the low pad (triangles); open circles are previously reported data for single-level bond pairs. Except for the solid circles, which represent the mean of 30 bonds, the data points represent the mean of 10 bonds, all of which ruptured at the heel of the first bond. Error bars have been omitted to reduce clutter; typically the sample standard deviation is between 0.05 and 0.1.)