Wire Bonding Process Control Targeting 100% Yield
01/01/2008
BY HERBERT STÜRMANN, Hesse & Knipps GmbH
Current methods of monitoring wire bond quality for both heavy and fine wire applications fall short of enabling the goal of 100% yield. A process integrated quality control (PIQC) system that was developed to monitor the most applicable and significant measures for judging bond quality addresses these limitations.
Several different methods of monitoring bond quality have been developed and implemented throughout the 20-year history of wire bonding. However, the few methods currently used in actual production - including non-destructive pull test and shear test methods, optical inspection methods and monitoring of transducer current and wire deformation - fall short of enabling the goal of 100% yield.
A mechanical, non-destructive pull test can be integrated into heavy wire bonders for applications with 5 mil wire size or above. These tests raise questions about the right choice of force to apply during the pulling process. Uncertainty regarding the amount of force required to judge bond quality adequately, along with the risk of weakening a bond by the pull test process itself are concerns. Wire bonding equipment users must also consider that integrated pull testing is only able to detect selected failure modes. In addition, the time necessary for this test is added to the total cycle time, reducing throughput. Understandably, this mechanical test is often only applied for statistical purposes where the loss of throughput can reach 20 to 30% - and not frequently used during regular production.
 Figure 1. Wire-bonding head with integrated pull and shear test. |
In addition to the mechanical pull test, a shear test integrated into the heavy wire bonding head exists (Figure 2). This bondhead technology leads to significant improvement in detecting possible failure modes. For example, a wire bonded onto a contaminated surface might hold just enough strength so that it is not detected by the pull method, however may easily pop off when undergoing a shear test. Due to the stiffness in the loop absorbing the pull force, any wire that is connected just slightly is likely to withstand a pull test but can be detected as a failure with a shear test.
In contrast, shear testing measures the quality of the inter-metallic connection. During the shear-test process, the original looping geometry remains untouched and no force is used on the wire. Therefore, the risk of weakening and reducing the quality of the bond by the quality test itself is reduced. The impact on process time with shear testing falls into the same magnitude as with integrated pull testing.
Integrated shear and pull tests aid in improving bond quality control, however neither can detect all possible failures. In addition to the time for testing and the potential to weaken the bonds by the tests themselves, these methods are inherently uncertain because it is difficult to determine the force and time adequate to test bonds without damaging them.
Deformation measurements, although a large contributor to improving quality, are only partially suitable for quality monitoring or active control systems.
Another significant contributor is seen in the transducer current. Deviations in the progression of ultrasonic impedance can indicate when welding has not occured or the wire has been lost. However, if the wire is partially or poorly connected, this method does not detect a failure.
The PIQC Method
The limitations of these methods can be eliminated with PIQC. Once implemented, existing mechanical non-destructive tests may become unnecessary, or can be used on selected conspicuous bonds indicated by a low quality index from the PIQC test. Such selective non-destructive tests are more efficient than 100% pull or shear tests, and can confirm failures detected as a result of 100% PIQC testing (Figure 2).
 Figure 2. The PIQC concept |
The goal of PIQC is to monitor the most applicable and significant measures for judging bond quality and can include mechanical vibration, transducer current, wire deformation, resonance frequency, and scrub behavior during the welding process. Acquiring data in real-time during the bond process avoids impact on machine throughput, allowing throughput-neutral, 100% quality control. This quality control system relies only on actual data, without any statistical assumptions.
 Figure 3. PIQC oscillation sensor |
To reach PIQC functionality, an additional sensor was added to the transducer to provide relevant feedback data for calculating a bond quality value (Figure 3). This assures real-time feedback of the conditions at the wedge tool tip - providing feedback on its movement. The real-time feedback from these sensor signals is gained through an additional interface in a proprietary ultrasonic generator which uses recently developed FPGA circuitry.
PIQC’s multi-dimensional control system includes decision-relevant oscillations at the wedge tool tip, the transducer current, the resonance frequency, friction, and wire deformation to complete the quality assessment.
All acquired real-time feedback signals are statistically analyzed in the PIQC box based on a mathematical decision model. The control algorithms are implemented in very high speed integrated circuit hardware description language (VHDL). The PIQC box allows derivation of extensive quality statements.
PIQC Quality Indices
Signal feedback and processing allows detailed analysis of welding, and translation into an optimized reference process. The PIQC box calculates a quality index for each bond based on actual feedback and reference data, which can be displayed graphically at any time.
After the learning phase, the PIQC box can directly recognize deviations in real-time, which can then be classified and interpreted by the user. It is possible to link signal deviations directly to a specific failure mode, enabling process specialists to react faster on problems with incoming material, contamination, or insufficient clamping. Without PIQC, these problems are detected only after destructive statistical testing that is not in real-time. With PIQC, the real-time signal feedback for friction and wedge tip mechanical oscillation is mathematically transformed, and along with additional information from the ultrasonic generator, provides indicators for surface contamination.
Conclusion
Under laboratory conditions, 6000 bonds were analyzed on contaminated and non-contaminated bond surfaces. The PIQC box was able to detect the pre-defined area with “good” and “bad” bonds (Figure 4). Data collected through extensive testing reveals process variations due to changes in surface conditions, and also detects variations in clamping conditions of both the substrate and the wedge tool.
 Figure 4. PIQC sample results |
PIQC is a multi-dimensional control system for wire bond quality. All physically relevant measures are evaluated in real-time with an individual quality index with which to compare ongoing data.
This method defines possibilities for ultrasonic bonding quality control. In the future, mechanical pull and shear testing will become unnecessary as their associated yield losses disappear while PIQC brings manufacturers closer to a 100% yield.
* developed and patented by Hesse & Knipps GmbH.
HERBERT STÜRMANN, sales and marketing manager, may be contacted at Hesse & Knipps GmbH; Vattmannstrasse 6 D-33100 Paderborn, Germany; +49/0 5251 1560-0; E-mail: [email protected]