values of q are somewhat larger than previously reported for the conventionally mounted tools but maintain the same trends (NBS Spec. Publ. 400-8, table 7). Because the tool as mounted in the inverted position has a uniform cross section, the present values of q are expected to be more accurate than those previously reported. (J. H. Albers)

9.2. Beam-Lead Bonding Force Measurement

The bonding force is a critical parameter in beam-lead bonding. Changes in bonding force can markedly affect the strength and integrity of the beam lead bonds. In order to determine the bonding force with most conventionally designed beam-lead bonding machines such as the one pictured in figure 22, a calibration curve must be prepared. In this procedure, the measured force at the bonding head is plotted against the reading of a dial setting (A). It should be noted that the dial setting is not a direct reading of the bonding force as the mechanism used is not that of a force gauge.

In attempting to calibrate the force mechanism while setting up a beam-lead bonder, several problems were encountered with the bonding force assembly as supplied by the manufacturer. First, a non-linear relation was found between the measured bonding force and the dial setting. Second, the bonding force could not be adjusted in the range below about 500 gf (4.9 N) because of mechanical inertia of the spring (B) in the force adjusting assembly. Within this range only a single force of about 130 gf (1.3 N), which corresponds to the sum of the weight of the bonding tripod assembly and the tension of the tripod spring, could be applied. Third, for the range where the force could be calibrated it was found that differences in substrate heights change the force calibration so that an individual calibration must be performed for each substrate height. Fourth, any change in the tension of the

spring in the force adjusting assembly is not reflected in the dial setting and, hence, any such changes and the subsequent bond force changes would go unnoticed. Suc problems appear to be common to most wobbletool beam-lead bonding machines.

In light of the above, a modified force assembly was developed to give a linear bonding force read out and hence eliminate these problems. This assembly is shown in figure 23. When the bonding tool (K) is lowered :: contact with the beam-lead device (L), the uppermost rod (F) of the tripod assembly (" presses against the lever arm (D). As the distance between this point of contact and the pivot point (E) is the same as that of the between the pivot point and the compensating spring (C), the force due to the ter sion of the compensating spring (which is

H

Q

Schematic diagram.

directly read from the force gage) is added to the weight of the tripod assembly (H) and the tension of the tripod spring (J) to give the total bonding force. These latter contributions to the force may be measured by disconnecting the compensating spring (C) from the lever arm (D) and then lifting the tripod assembly (H) from below at the tip of the bonding tool (K) with a force gauge until the rod (F) makes contact with the lever arm (D). The force gauge reading at this point is the weight of the tripod assembly plus the tension of the tripod spring. The compensating spring is then reinstalled on the lever arm.

The contribution to the force arising from the compensating spring may be varied by loosening the force gage holder (A) from the rod assembly and then raising the force gage mounting to increase the force or lowering the force gage mounting to decrease the force. The range of the gage may be extended by a factor of 2 or 4 by mounting the compensating spring at position 2 or 4 10cated a distance of 1/2 or 1/4 the length of the lever arm of the force gage (B). To obtain an accurate reading on the force gage, its lever arm (B) must be adjusted so that it is perpendicular to the rod (F). This is accomplished by pivoting the force gage holder (A).

a force of 10 mN (1 gf) or more was applied to each of the beams in the peel direction by pushing the die downward with sufficient force. The block diagram of the system used for this purpose is shown in figure 24.

Two test series have been carried out thus far. The first was a feasibility test in which 10 to 20 beam-leaded devices were stressed by pushing, probing, and pulling. Individual beams were peeled back and acoustic emission noise counts were taken. It was determined that the acoustic emission apparatus had sufficient sensitivity to detect the breaking or peeling of a single bonded beam and that the probing system did not itself produce any measurable acoustic emission.

The second test series was conducted on 72, 14-beam devices which, except for 10 wellbonded control devices, had one or two of their beams bonded to areas rendered intentionally defective. These defective or contaminated areas included thinned substrate metallization, oxidized surface layers of chromium, graphite on the surface, or a fluorocarbon grease-stop on the gold bonding pads.

The output of the sensing equipment was read as noise counts. The background was typically only three to five counts. The wellbonded control devices either gave no signal above background or a burst of 10 to 20 counts at the point of complete beam collapse. For these particular devices this typically occurred when a force of about 450 to 550 mN (45 to 55 gf) was applied to the device. The majority of those devices with one or more beams bonded onto defective substrates produced a series of two to five acoustic bursts in the 20 to 40 count range with some bursts as high as 100 counts. For the most weakly bonded beams, such noise counts appeared at an applied force of about 100 mN (10 gf), the minimum force available from the test apparatus. Surprisingly the devices bonded over the fluorocarbon greasestop gave little or no noise counts with applied force. Later examination revealed

The experiments were performed with the help of Dr. M. Linzer, of the Inorganic Materials Division at NBS, where acoustic emission studies have been carried out on both brittle and ductile materials for many years.

10.1. Helium Mass Spectrometer Method for Leak Detection

Anomalous results were obtained from the second stage of the interlaboratory evaluation of the helium mass spectrometer method [42] for testing fine capillary leaks in large volume (~ 1.5 cm3) containers being conducted in cooperation with ASTM Committee F-1 on Electronics (NBS Spec. Publ. 400-4, p. 67). This stage of the experiment was intended primarily to check the Howl and Mann theory [43] for relating measured to actual leak rate. Following stage one, which consisted of the direct measurement of the 50 borosilicate glass capillary leaks in open tubulated capsules, the tubulations were sealed off at NBS and tested by the participants by the back pressurization technique [42]. Pressurization of the capsules was at 75 psia (5.2 × 105 Pa) for 20 h in helium. Measured leak rates on the resultant "packages" were found to be orders of magnitude greater than anticipated and of approximately identical value independent of true leak rate as measured in stage one.

All 50 leaks were returned to NBS for diagnosis of the problem. For the diagnosis, 20 of the capsules were pressurized and measured as prescribed. Again, uniformly high

and approximately identical threshold values of 10-6 atm.cm3/s (10-7 Pa⚫m3/s) were measured. Measurements, however, were then continued on the leaks over a four-day period. Leak rates declined, but the smallest indicated leak rate at the end of this period was still of the order of 10-8 atm cm3/s (10-9 Pa⚫m3/s). Measurements were then made on 10 capsules which had been pressurized and measured one month previously. Leak rates were about 5 x 10-10 atm cm3/s (5 × 10-11 Pa m3/s).

Such values as found are obviously outside the prediction from the Howl and Mann theory and must be due to surface effects. To isolate the surface effects, ten sealed-off capsules were constructed with approximately the same dimensions as the test leak capsules. These were then pressurized and measured in the same way as the test leaks. The measured leak rates were of the same order of magnitude as those of the test leaks and also were tightly grouped in value. The resultant data as measured over a period of time are shown graphically in figure 25. The measured leak rates were plotted on a logarithmic scale against time on both a linear scale and a logarithmic scale in order to discriminate between the gas emission mechanisms. Surface desorption