The reason that I chose to discuss wire bond pull testing after examining shear testing is that, in concurrence with the development process of a wire bond, we perform pull testing only after shear testing has been performed. The destructive pull test is explicitly what it sounds like. A wire bond loop is pulled until it fails in one of three ways: bond lift, heel break, or mid-span break. The pull test measures the force required to break the wire from tensile force, and the remnants of the wire bond are used to determine the failure mode. Typically used to develop the wire bond process, destructive pull testing is also used as a statistical process control in production bonding. The wire bond pull test is not a replacement or substitute for the shear test, but rather a completely different test with a different purpose and intent. Just like the shear test, the pull test has very specific characteristics that it measures. If we assume that shear testing was used to develop a robust weld at the bond foot, a weld that is stronger than the wire itself, then the pull test should never result in a wire bond lift, as this occurrence would be considered a rejectable condition and cause for immediate action. We don’t just focus on the force required for the wire to fail or break, we are also interested in the failure mode. For a robust and well-defined process, that failure mode would be a heel break or a mid-span break. Unlike shear testing, pull testing is covered in MIL-STD-883. This standard, however, covers wire bond pull testing at only a most basic level.
The destructive wire bond pull test is performed by a complex instrument. The tool accessories and settings are selected based on the theoretical breaking pull force range of the wire loop. Once selected and calibrated, the process involves placing a hook under the apex of the bond loop and exerting an upward force on the bond loop until it fails. The loop geometry is very important to deciphering the data from the pull test. The tensile force within the bond wire is a function of the pull force and the angle each wire forms with the substrate. In fact the resultant tensile force is a trigonometric function of the pull force and the angle formed by each wire. So if we assume that the weld formed at the foot of the wire is stronger than the wire itself, as it should be based on rigorous shear testing and development, then in theory, the resultant pull force can vary from zero to two times the tensile strength of the wire based on wire geometry alone. Just as the shear test is intended to properly develop a robust weld between the bond wire and the bond pad, the pull test is intended to develop the optimum wire formation and loop geometry.
The destructive wire bond pull test is intended to measure the force required to break the bonded wire using tensile force. It is also possible to perform a non-destructive pull test, but without the results from the wire break, the usefulness of the data collected is drastically decreased. The non-destructive pull test should be used only after a defined and rigorously tested wire bond process has been developed. In destructive pull testing for wire bond development and process control, we record the actual force required to break the bond many times over, but as you may begin to notice it is important to know how that data can be assessed. This developed measurement method specifically relates to the tensile strength of the wire being bonded. We typically explicate our destructive pull data with the use of non-destructive pull force. By placing a small non-destructive pull force on a wire loop to straighten it, we can then measure the geometry, calculate the angles, and with the known tensile strength, calculate the theoretical pull force required for wire breaks.
In addition to actual and theoretical breaking force, we are also interested in the failure mode. Figure 2 indicates how we at SMART Microsystems generally categorize the different failure modes. A wire bond lift is considered a rejectable condition, but with a properly formed wire heel and a roughly 30 degree angle at the wire-substrate interface, we expect the failure mode to be primarily heel breaks with some mid-span breaks. The reason that heel breaks are most common is simply that the heel is the part of the wire that gets deformed the most, and the wire will always break at the weakest point. There are five distinct modes of failure : i) 1st Bond lift, ii) 1st Bond Heal Break, iii) Mid Span break, iv) 2nd Bond heal & v) 2nd Bond lift. Each one of the failure modes has a group of potential causes associated with them. A bond lift on either bond foot is considered an immediate cause for concern and is a rejectable condition, but nonetheless provides valuable information. A mid-span break near to either the first or second bond is considered to be the most desirable outcome, but can be difficult to achieve with many loop geometries. Heel breaks are the most common failure mode by far, but this could easily indicate an incorrectly formed wire bond foot or improper loop geometry, so although common, this type of failure should never be overlooked.
MIL-STD-883 includes a section on destructive wire bond pull testing that includes the minimum pull strength of different wire diameters. At SMART Microsystems we use MIL-STD-883 as a general guide only. Our upper and lower pull strength limits are developed internally, and are derived from the wire manufacturer’s tensile strength for the specific wire being used and the wire geometry. Using the wire bond pull test as a process control can be very useful, and each of the failure modes can be a leading process indicator. A pull test breaking strength that is too low or too high can indicate a wide variety of root causes, including change in loop geometry, or even the wrong wire installed in the bonder. A low pull test breaking strength with poor heel formation could indicate a change in settings or a fouled tool, among many other possible root causes. Pull strength should never be considered a single sided limit, it should have defined upper and lower limits.
Wire bond destructive pull testing has a valued and indispensable place in the world of wire bonding. It can provide real actionable data to improve, develop, and regulate the wire bond process. Simply stated, pull testing speaks to the quality of the formation of the wire bond loop. The data that it does not speak to is the quality of the weld between the wire and the substrate, which is derived from shear testing. If I could only choose one test for process development it would be the shear test, and if I had only one test to choose for process control it would be the wire bond pull test.
Ball Bonding V. Wedge Bonding
For many years, wire bonding has been the most robust and commonly used method for chip-to-die interconnection of lead-frame, integrated circuit (IC) packages, RF microwave packages and optoelectronic packages.
A question often gets asked concerning what kind of application we use: ball bond or wedge bond? Why would a process engineer choose a wedge bonder over a ball bonder or vice versa? This has been the question by most process engineers because, generally speaking, electrical characteristics of the package are affected by the method of wire bonding. However, there are cases where certain packages have physical constraints such as temperature limitation (low heat or no heat applications), use of aluminum wire instead of gold, use of ribbon instead of wire and fine pitch application. This is where the proper selection of wire bond process comes into play.
Typically, “ball bonding” applications are associated with thermo compression (T/C) and thermo sonic (T/S) joining methods. T/C utilizes pressure and temperature from about 150 degrees C (for most common applications) to create an inter metallic bond. T/S, on the other hand, adds ultrasonic energy from the previous process. With both methods, however, a “free air ball” is being created by a spark from an “electronic flame off” or EFO underneath the capillary before bonding takes place. This free air ball then gets deformed when the capillary touches the surface of the bond pad and applies force and ultrasonic with a given amount of time to deform the ball. Thus the inter-diffusion between the wire and the bond pad metallization occurs, which makes the inter metallic bond.
In general, “ball bonding” offers faster speeds of about 5 -12+ wires per second. Types of wire used for this application include gold, palladium-coated and copper wires. Typical packages and applications for this process include BGA, QFP, SOP, MCM hybrids and wafer level bumping. The ball bonding process is suited for fine pitch applications down to 40 microns or less.
The wedge bonding process, on the other hand, utilizes ultrasonic energy and pressure to create a bond between the wire and the bond pad. When gold wire is used, wedge bonding uses temperatures up to 150 degrees C, and similar to ball bonding this technique is called thermo sonic (T/S) bonding. The most predominant process for wedge bonding is a low temperature process or ambient temperature bonding, where aluminum wire is used to make the interconnection of the die and package. This “welding process” (either hot or cold) deforms the wire into a flat elongated shape of a wedge, however depending on the type of wedge tool used, the shape can sometimes have a cross groove (typically for 1 mil gold wire). Unlike ball bonding, the first bond for a wedge bond does not have a ball, which is why this wire bonding process is called wedge-wedge bonding.
The absence of the ball on the first bond gives wedge bonding an advantage for much finer pitch applications of 40 microns or less. Aluminum wire is the most predominant wire used for this process, followed by gold wire. Typical packages and applications include high-power applications, RF microwave, optoelectronic packaging, BGA, QFP, SOP, MCM hybrids and temperature-sensitive applications. Wedge bonding speeds typically fall within 3-6 wires per second, comparatively slower than a ball bonder.