Bond Test Fundamentals for Successful Advanced Applications
BY ADRIAN WILSON, MALCOLM COX AND LISA GERBRACHT
In today's technology press, you do not hear much about bond pull and shear testing. However, these test methodologies are more crucial in the industry today than ever. For example, the drive toward 25-µm wire bond pitch demands a tester capable of measuring 5-g bond strengths with a 3-sigma accuracy of 0.01 g. The latest generation of chip-interconnect advances, such as ball grid arrays, ultra-fine pitch and stacked-die wire bonding, all impose their own challenges on bond testing.
Bond Test Systems
Advanced bond test systems available today are high-precision universal systems that can perform pull and shear type tests (Sidebar). Variations to the fundamental tests include wire-pull, bond-shear, bump-shear, ball-pull, stud-pull and die-shear testing. Typical test specifications for testing, as dictated by established industry standards (such as Mil-STD 883d, Mil-STD 750c and ASTM 1269) include pull tests to 10 kg and shear test to 100 kg (Table 1). A universal system means that modules for a specific test are swapped in and out of a common test system mainframe. The mainframe gives the ability to perform all these tests under PC control, providing statistical process control (SPC), networked data management and a convenient user interface. While bond testing originally was a technique primarily used in failure-analysis labs, today the available state-of-the-art tools make bond testing easily applicable on production lines in assembly facilities worldwide.
Table 1. Accuracy and resolution requirements for various wire pull and bond shear tests.
While bond tests seemingly are straightforward, the critical objective in these tests is quantification of test force. Understanding the evolution of the methods of measuring the actual bond-breaking force is critical. Foremost to effectively applying any bond test method is a clear understanding of a tool's resolution and its method for determining accuracy.
The accompanying brief tutorial about pull and shear testing (See Sidebar) provides a basic concept of how pull and shear testing are mechanically different. The nature of these processes dictates that resolution and accuracy of the testing systems is key. While this might seem obvious, in fact the importance of resolution and accuracy in pull and shear testing today often is lost in a casualness of system specifications.
As illustrated in Figure 1, resolution is the smallest increment an instrument can show (Figure 1a). Accuracy is a position within the two concentric circles a distance from the center (Figure 1b). Precision is the tightness of the cluster of accuracy measurements (Figure 1c). It is important to precisely identify the force peak that corresponds to bond failure. Historically, some bond test systems have used analog circuitry to detect a peak in the electrical signal from the transducer. Using this technique, it was difficult for the system to resolve unambiguity between peaks in the signal that caused the actual break and by noise or false peaks caused by electrical noise spikes or mechanical disturbances caused by hook slippage or mechanical vibrations in the local environment. A noise spike is recognized by the analog peak detector as a false peak in the force reading. This is made worse by locating the analog/digital (A/D) conversion external to the load cartridge, inside a PC at the end of a noisy interconnect cable.
A solution is to place the A/D converter inside the test module and use rigorous digital signal processing (DSP) techniques to analyze the transducer signal in real time. DSP techniques implement specially written algorithms to distinguish between a misleading noise signal and the signal signature of the actual break from bond failure. Digital processing needs a sampling rate of more than 100,000 measurements per second, thus enabling the processor to capture the actual force profile. Using a 16-bit A/D converter equates to a theoretical resolution of one part in 65,536. For both conventional and fully digital bond test systems, it may be tempting to specify the theoretical resolution as the system specification. Electrical noise in any system, however, will degrade the usefulness of the theoretical resolution to a somewhat lesser value. For example, one could use a moving average six measurements wide to smooth out any noise, yielding a robust, usable system resolution of 1:10,000, rather than the theoretical resolution of 1:65,000.
Bond Test Accuracy
The accuracy of a bond test system is limited by many small error sources. For example, mechanical sources include transducer linearity, thermal stability and creep. Electrical error sources include amplifier and power supply stability, electronic linearity, and thermal stability and various sources of electrical noise. The best way to measure the accuracy is with a test that best approximates actual machine usage. An accepted method is by a gage repeatability and reproducibility (GR&R) study. GR&R characterizes measuring systems by dividing the equipment's variation in measurement of force by the process variations that result in force variation. The rule of thumb for bond testing is that the equipment variations should be ≤10% of the variation of the process. In general with bond test systems, a GR&R ≤10% can only be achieved with three-sigma accuracy. There are two primary methods for measuring the accuracy of the system in the GR&R equation:
- Static GR&R is performed by repeatedly hanging static weights and determining the repeatability of the results; this test is subject to the care of the person applying the weights, motion of the weights after they are hung and potential environmental vibration, etc. (Figure 2).
- Dynamic GR&R is performed using a test jig that simulates the dynamic bond pull or shear test. This method uses the actual machine test sequence of the bond test itself and most closely simulates the high-frequency force signature of a real bond test (Figure 3).
Figure 2. GR&R.
GR&R testing results in an accurate measure. Repeated tests give a statistical statement of the accuracy of a system. For example, a 15-g wire pull test of a 1.2 mil gold wire with gold ball bond on one side and typical wedge outer lead bond would have a typical process variation of 3 g, meaning that the 3-sigma accuracy of the pull test needs to, by 1.5 g, deliver a GR&R of 10%.
Figure 3. Dynamic GR&R.
The statistical variation of the accuracy of bond testers is also important for advanced applications in packaging. Although often glossed over, accuracy should always be stated at a 3-sigma value. This is directly dictated by today's common use of Cpk (which is derived from 3-sigma values) in semiconductor manufacturing processing specifications.
Although many assembly cleanrooms have excellent environmental temperature control, bond testers often are used in less stringently controlled environments and, therefore, there may be significant departures from the ideal temperature. To compensate, the 3-sigma accuracy of bond testers must be stable over a wide temperature range. Stated accuracy also must be maintained where one test module is used from time to time in several mainframes.
Contemporary universal bond testers are the unsung heroes of package process development. As packaging and technology roadmaps weave their way toward ever more complex interconnects, bond test systems must and will continue to evolve to meet this challenge.
ADRIAN WILSON, director of marketing, MALCOLM COX, president and CEO, and LISA GERBRACHT, head of applications engineering, may be contacted at Royce Instruments Inc., 500 Gateway Drive, Napa, CA 94558; (707) 255-9078; e-mail: firstname.lastname@example.org, email@example.com and firstname.lastname@example.org.
Pull and Shear Bond-test Fundamentals
Wire pull testing (a.k.a. bond pull testing) is a time-zero test for wire bond strength and quality. Simply described, the wire being tested is pulled upward until the wire or the bond to the die breaks. However, there are both destructive (for process setup) and nondestructive pull tests (typically used as an in-line test for high-reliability packaging applications). The system used for pull testing features a mechanism for applying the upward pulling force via a pull hook and a calibrated method for measuring the force, measured in grams-force (equivalent to Newtons), at the point when the wire or bond breaks. Test procedures vary, but the pull hook usually is placed at the highest point along the loop of the wire being testing (wire-to-hook placement repeatability is important) and the pulling force is applied perpendicular to the die surface. The test is quantified by an accurate measurement of the bond strength and a qualification of the failure mode (i.e., ball bond lifting, wire neck/neck break, mid-span wire break or wedge bond lifting). In general, bond lifting is unacceptable and requires further failure analysis.
Bond shear testing assesses the strength of a ball bond, complementing wire pull testing. Bonds often can exhibit high shear strength, but less resistance to wire pull stresses. The module for bond shear testing includes a sample holder, a shearing arm with a chisel-shaped tool at the end and an instrument for measuring the shear strength of the bond. The system measures the force needed to shear a ball off its pad. The shear force reading of a ball bond may be correlated with the ball diameter for proper assessment of its ball shear strength; the shear force reading of a wedge bond must be correlated to the tensile strength of the wire for proper assessment of its wedge shear strength. Bond shear failure modes include bond lifting, bond shearing, cratering and bonding surface lifting (separation of the bonding surface from its underlying substrate).