Search

Since starting production with Copper wire bonding for almost a decade, there are still many findings in order to achieve the most stringent quality requirements. Furthermore, some older machine platforms were also converted to Copper wire bonding in order to increase production capacity. In the cases of running extremely sensitive devices, there were many occurrences of lifted bond and pad peel defects at the same time. From the initial assessments, even those bond pad structure which was improved until reasonable robust level could still fail for pad peeling issue. On top of that, some pad peeling cases could not be simulated, rejects could not be duplicated. In order to understand the pad peeling issue, comprehensive study was conducted in order to eliminate this issue. The first set of evaluations was performed on bonding parameters, and results met most of the expectations. The results showed parameters of bonding time and bonding force did not provide much significant effect as long as they were operating within reasonable levels. As for the bonding power, its effect exhibited weird scenario. When power increased, ball shear strength was also increased accordingly. However, pad peeling defect was not able to be correlated linearly to ball shear strength, other than higher power made it becoming worse. In order to understand this issue, another new approach was initiated, equipment characterization along with process optimization. With new optimum process window, equipment characterization was performed in depth. From all the characterized machine parts, it was found that capillary installation was the most suspicious root cause. With the Copper kit installed at the firing location, normal installation jig was no longer useful. The capillary setup became inconsistent, and capillary could be slanting in any direction. The most common setup would be either capillary slanting towards operator or away from operator. One of this slant directions might interact with the transducer movement and vibration, resulted severe pad peeling rejects. In order to resolve this root cause, a breakthrough capillary installation jig was invented and deployed in production. This new jig enabled stress-free and self-center capillary setup. The proper installed capillary would allow desired ultrasonic energy transfer to the Copper ball, without any mechanical stress point. The overall improvement was combining optimum process window and consistent capillary setup. With deployment in production, pad peel issue has reduced to zero.

With the Copper wire bonding already occupying majority of production lines, there are many significant factors continue to surface out. The Copper wire bonding has also becomes very susceptible to production variability of small leaded packages such as Thin Small Outline Package (TSOP). The TSOP utilizes thin and flimsy leadframes, causing a lot challenges at 2nd bond (stitch). The stitch bonding shows high occurrences of pre-mature failures or wire breaks, while struggling to maintain good stitch bonding strength. Furthermore, the inconsistent stitch bonding can also cause inconsistent tail formation, subsequently affecting the formation of Free Air Ball (FAB) and 1st bond quality. With many DOEs (Design of Experiments), several improvements were executed to address the stitch forming issues. Some of these major actions include optimizing several parameters such as the 2nd bond force, wire looping speed, wire loop mode change and wire sequencing. However, even with the implementation of these improvement actions, the Copper wire bonding performance was only improved by 50%. More comprehensive mapping was extended into possible impact of leadframe location and design. Surprisingly, die attach process was found to be the confounding with the leadframe design. It looks like wire stitch formation at ground lead is not consistent, mostly affected by the die attach epoxy bleed and outgas. In order to obtain more robust Copper wire bonding performance, the die attach process was optimized accordingly. Without losing any die attach quality, optimized process provided much larger operating margin for Copper wire bonding. All the combined improvements managed to eliminate wire break occurrence, while maintaining the good bonding strength of 1st and 2nd bonds. The lessons learnt were very important and already embedded to all new Copper wire bonding projects in order to realize 1st pass success into production mode.

Over years, the crack die defect was proven cannot be screened effectively through Final Test, therefore it has high risk reaching the end applications. This issue can affect many devices, including some miniature devices such as Small Outline Transistors, SOTXXX Eutectic. In order to eliminate the die crack defect, entire assembly process required optimization; including Saw, Die Attach, Wire Bond, Mold, Plating, Trim/Form and Final Test. Each process step has its own risks and therefore full analysis was conducted on all the possible opportunities. At wafer sawing process, several sawing methods were evaluated, including single cut, step cut, 3 channels cutting, cutting directions and selection of various blade types. With the optimum sawing process, die chipping was minimized to lowest level (only few microns), that resulting to minimum die cracking risk. With that performance, further risk investigation was carried on to understand the stress amount inside the package. There were 2 significant stress factors, die location on flag and trim/form impact to the leads. In order to minimize the effect of the stress, die was attached to the location with the least amount of stress. As for trim/form, this involved new design leadframe fabricated and experimented in production. The leadframe design optimization was completed after compliance to the reasonable stress level, as calculated in the trim/form stress analysis. When new leadframe was subjected to actual trim/form process, actual performance was verified by reliability testing. With the optimum assembly configurations, further safety action was depended to the Final Test capability. There were 2 additional test parameters added to increase effectiveness of screening potential die crack rejects. With the total compilation of all the optimum assembly and test configurations, actual performance monitoring showed total elimination of die crack occurrence. This project becomes good benchmark of any die crack reduction or elimination project.

The TDMA constructed protocols have been planned (i.e., proposed) by various researchers. These protocols are generally two time periods or three time periods based, and have shown significant reduction in Bit Error Rates (BER), because of low diversity order. Consequently, the Improved Amplify-and-Forward (IAF) protocol has been proposed. The IAF protocol uses three time periods, to transmit data from transmitter to receiver. Moreover, the transmitter is stimulated during the second as well as third time periods, to broadcast to receiver, and the diversity order is improved. It is shown from results that IAF demonstrated significant reduction in BER and rise in BER-Gain, over the previous proposed two time periods as well as three time periods protocols. The IAF protocol demonstrated substantial improvement in Global-Percentage-Error-Decrease (GPED) and Local-Percentage-Error (LPED) values, both at severe as well as at low fading conditions, over to previous proposed TDMA constructed protocols (i.e., two time periods as well as three time periods based).

The epoxy dispensing process is one of the most commonly used die attach methods in the semiconductors assembly process. Some major packages such as QFN package can only utilize epoxy die attach method due to the temperature constraint of the leadframe tape (withstand less than 300DC). However, the epoxy dispensing die attach method has encountered several major quality issues even after running production for long time. These issues could not be eliminated due to lack of possible effective permanent corrective actions. Such quality issues including application/field failures due to intermittent epoxy-shorting, epoxy overflow and insufficient epoxy. Some of the common root causes were nozzle clogged or deformed, pressure effect inside syringe changed with different content level, variations of epoxy material mechanical properties (viscosity, modulus,…), … and etc. In order to eliminate all these quality issues, the conventional epoxy dispensing method requires some evolutional changes. From the comprehensive study on various die attach methods, and one of the best possible options is Wafer Back Coating (WBC). The WBC method requires epoxy paste printing on the wafer backside and then sawn together with the wafer. At the die attach machine, the singulated die (with epoxy coated underneath) would be bonded directly to the carrier leadframe. Therefore, the WBC method enabled the lowest dependency level of epoxy handling in the die attach process. Most of the epoxy dispensing major issues that dependent to human setup and judgement can be eliminated. Unfortunately, the high conductive WBC does not exist yet in the semiconductors industry, mainly because such material was considered impossible to be developed in production mode. The WBC material available is only for non-conductive and medium conductive applications. The main challenge of developing high conductive WBC material is due to its large weight content must be filled with Silver (Ag) to achieve high conductivity. Such high Ag filler content can cause severe quality issues in multiple assembly process steps. In the printing process, Ag filler may protrude out causing uneven surface. This caused poor adhesion to the mylar surface that is holding the wafer. During the high speed wafer sawing process, those poorly adhered die will fly-off, chip or even crack. Meanwhile, the Ag filler can also affect the die attach adhesion quality as it is blocking the function of resin to adhere to the leadframe and wafer surface. The adhesion strength can be very bad until die lifted in the wire bonding process. As for reliability performance, crack line was found propagating from leadframe surface, through WBC, then cracked at wafer backside surface. The failures occurred in both adhesive and cohesive strength. A joint efforts project was initiated with a major epoxy supplier in performing comprehensive WBC development study. In the initial stage, there were a lot of quality issues occurrences, sometimes ended up with zero assembly yield. Furthermore, some slight changes in formulation can cause strong confounding effect, resulted in getting unpredictable bad results. In the later stage, proper evaluations were designed and executed by following the Six Sigma DMAIC methodology. Many engineering samples were created in the screening stage, deriving each confounding effect to details independent effect. Some new interim tests were also created in order to evaluate the initial mechanical strength and electrical performance of any new WBC epoxy samples. Some of the breakthrough actions included optimization of catalyst types (accelerate cross linking), Ag filler types, Ag filler size mixture, Master Batch robustness, molecule chain length, hardener types (promote cross linking reaction), etc. Eventually, some of the new WBC engineering samples started to pass most of the assembly requirements. However, new issues occurred when the electrical performance failed inconsistently on some extremely sensitive devices. Further deep dive study showed that the WBC inconsistent performance was related to the variations found in both epoxy manufacturing process and assembly printing process. All the previously optimized process window needs to be further tighten. Some new controls were also introduced in both processes, then WBC material performance becoming more repeatable in production scale. This new die attach method with WBC is the ultimate solution to eliminate all the epoxy dispensing quality issues, from assembly process to reliability performance. Furthermore, the WBC method also provides other positive contributions such as throughput improvement (epoxy dispensing is the normal bottleneck process step), consistent electrical performance, competitive cost and utilization of maximum die size (zero clearance required between die edge and flag) for better functionality. In the near future, the new WBC die attach method will replace the entire epoxy dispensing process. This change will enable the opportunity to gain the most distinguished competitive advantages to run the most robust assembly processes in comparing to other semiconductors companies.

Derivatives of eigenvalues and eigenvectors of parameter-dependent matrix eigenproblems play a key role in the optimum design of structures in engineering, and in the solution of inverse problems, such as the problem of model updating, which arises, for example, when information on the normal modes of vibration of a structure is used to detect structural damage. Both these applications often involve quadratic eigenvalue problems. Most existing methods for the computation of derivatives of quadratic eigenvalue problems are based on the assumption that repeated eigenvalues have well separated first order derivatives. In this paper we propose new algorithms for computing derivatives of eigenvalues and eigenvectors for quadratic eigenvalue problems under much more general conditions than existing methods, whose effectiveness for repeated or tightly clustered eigenvalues are confirmed by some numerical examples.