Operators of ultrasonic wedge bonders desire that bonders using the same process inputs achieve identical process outputs. In practice, it has been found that process outputs can differ between bonders operating with the same inputs. A common practice to bring the performance of bonders into closer agreement is to make small adjustments to the input parameters. These adjustments are looked upon unfavorably because they are typically determined using non-standardized, non-automated procedures, and further, the process inputs for each bonder then need to be tracked separately. Therefore, there is a strong desire to normalize the performance of one bonder to another using an automated procedure. This paper presents a method to normalize the performance of a number of bonders based upon bond deformation measurements taken automatically by the bonder. Several case studies are presented that demonstrate the effectiveness of this normalization procedure in improving the similarity of process outcomes across a group of bonders. The results show that bonders after Ultrasonic Normalization make bonds with more similar end deformation, shear strength and bond width than the bonders before Ultrasonic Normalization.

A modeling approach is developed to better describe the relation between input electrical power and the physical reaction of the bonding system during ultrasonic bonding. The major distinction between this analysis and previously published works is to attempt to eliminate empirically driven correlations between the input power and the kinetics of the bonding process. Two models, a piezoelectric model and an ultrasonic bonding model, are combined in order to reach this goal. The piezoelectric model is used to calculate the desired forcing, amplitude, and frequency that is created by the piezoelectric transducer during the actual ultrasonic bonding process. For this process, a lumped parameter model, taken from literature, is used, that converts input current and voltage to velocity and position of the bonding tooltip, respectively. This model is then combined with an updated model of the relative amplitude between the bonding material and substrate as the ultrasonic bond is being formed. Our model differs from existing friction power models by utilizing the Archard Equation to account for the removal of the natural oxide film. The integrated model provides a relationship between the bond growth and the driving power. The analysis enables comparison between the transverse force on the bond tool and amplitude of the bond tool’s motion for different electrical input powers.