Increasing process reliability in fine-pitch wire bonding: a 2-year study identifies close ties between capillary performance and bonding failures.
In the past, chipmakers have relied on new wire bonding machinery for greater productivity gains. While offering greater speeds and efficiency, new bonders cannot resolve all the issues related to wire bonding failures. More attention must be paid to the type of bonding tools used in a process. Understanding the effects of the tools on the bonding process is important to identify possible factors and solutions for process failures.
A key element of the bonding process, the capillary is the tool that carries the bonding wire during the bonding cycle. After a first ball is bonded to a die, the capillary is responsible for feeding out the wire over the first bond, forming a loop to the second bond target point. The capillary's reliability in accurately forming this loop affects the wire bonding process reliability and costs. While various statistical methods can determine equipment downtime, studies have not been conducted on machine downtimes related directly to capillary performance.
To determine the effects that a capillary has on process reliability, we conducted numerous bonding applications, mapping out all wire bonding processes and those related to the capillary. Results identified failures that occurred during the wire bonding process while the capillary was performing and those that occurred at other times during the process. The capillary-related failures had a direct effect on reducing the mean time between assists (MTBA), the amount of time between bonding failures caused by wire bonding failures. Reduced MTBA results in lower machine uptime and productivity and, as a result, higher production costs.
Looking for a way to improve capillary performance during bonding, we analyzed the interior geometry of a capillary including the hole diameter, IC length and IC angle to determine a better design that would optimize its relationship to the wire bonding process. After two years of study, we developed an advanced capillary design that incorporates a special inner and outer configuration that reduces the number of common wire-bonding failures during bonding. These failures, as termed by K & S, include:
* Premature tail bond break after second bond formation (short tail).
* Unsuccessful free-air ball formation (insufficient tail, spark did not reach wire/EFO-open).
* Bond did not stick or is lifted off bond pad (non-stick on pad/malformation of first bond on pad).
* Bond did not stick or is lifted off lead (non-stick on lead/malformation of second bond on lead).
By reducing these wire bonding process-related assists, the new capillary can increase MBTA, which supports greater process yields. Fewer scrapped devices mean a higher quality process with greater UPH.
LQFP Lab Tests
In one test conducted at a K & S Bonding Tools Application Center, over 8 million bonds and 25,000 devices were used to compare the performance of conventional capillaries with the novel capillary--called NEXXUS--on a 55 [micro]m bondpad pitch (BPP), 166-lead LQFP. The process window used on this LQFP wire bonding process was identical to current mass production processes running at major assembly houses.
The equipment used was a K & S Max[micro]m using a 55 [micro]m BPP. The wire diameter was 20 [micro]m The conventional capillary tested was P/N 418FF-2881-R33. The novel capillary was P/N 48NFF-1000-R33.
The performance of both capillaries were quantified based on two major responses:
* Number of wire bonding process failures (assist), which caused process interruptions (downtime).
* Relative improvement in overall productivity (MTBA) obtained with the novel capillary-based process.
To ensure testing procedures were closely duplicated when evaluating the performance of both capillaries in different applications, the following methodology was maintained.
1. Test conditions such as product line, bonder type, wire type and bonding temperature were recorded prior to test for proper evaluation management.
2. Wire assists data including the number of NSOP, NSOL, EFO Open and SHTLs that occurred during a noted number of Kbonds for each specific capillary on a specific bonder were captured in a logbook.
3. The bonder was set to regular production conditions.
4. All assists counters (capillary and statistics) were reset to zero.
5. Novel and conventional capillaries were installed on prepared bonders and recalibrated.
6. Diagnostic data were recorded for each capillary type, and data were collected for every 200,000 bonds to compute the cumulative assists distribution rate during the capillary life (Table 1).
7. Ball shear, ball diameter, stitch pull and wire pull were recorded according to regular product specifications. Also, we recorded bonding parameter changes such as USG, Time, Force and F.A.B. if changes occured during production runs.
8. We ensured all data were correctly gathered before replacing used capillaries. Steps 2 through 8 were repeated after replacing new capillaries.
Collected data were automatically graphed for conventional and novel capillary results in regards to MTBA improvement, assist rate, total assist count (see results analysis).
LQFP Results Analysis
Wire bonding total assist count. When analyzing each assist type (Figure 1), the novel process produced fewer failures, particularly of the NSOP type, as compared to the conventional process.
Wire bonding assist type and rate. For each capillary design and failure type, Figure 2 presents the following values:
* The average assists count for every 100,000 bonds along the capillary life.
* The overall number of assist types accumulated for every 100,000 bonds.
Both the average and total failure count achieved with the novel capillary show improved process stability. The overall number of failures decreased by 4.8 for every 100,000 bonds.
Wire bonding relative assist improvement. As shown in Figure 3, the novel capillary displayed a relative failure rate improvement over the conventional capillary for each assist type. The lowest failure rate reached 3% for SHTL failure, while EFO-open failure improved by 65%. Overall, the novel capillary enabled an improvement of 29%.
Along with a failure rate improvement for each assist type, the novel capillary prolonged the wire bonding MTBA from 0.27 to 0.38 hrs., which represents 40% more process uptime relative to a conventional process.
The novel capillary achieved superior overall process stability and improved productivity by 40% more uptime (Figure 4).
In another joint test with a K & S customer, dozens of novel capillaries were directly compared to a conventional capillary design-based process in a mass production volume of more than 45 million bonds and 30,000 device units. This time, both conventional and novel capillary types were used in a BGA wire bonding process under identical conditions. The same methodology was implemented and the same equipment used.
The package was an 80 [micro]m BPP, 76-lead BGA. The wire diameter was 25.4 [micro]m. The conventional capillary's P/N was 483FC-3122-R33; the novel capillary's P/N was 48NFC-1002-R33.
Wire bonding total assist count. By assist type, the novel process produced 662 less failures for all assist types as compared to the conventional capillary (Figure 5).
Wire bonding assist type and rate. The novel capillary process produced fewer failures as compared to conventional capillaries (Figure 6). The overall number of failures decreased by 2.6 for every 100,000 bonds.
Wire bonding relative assist improvement. All failure rates improved, with the lowest reaching 13% for EFO-Open, while NSOP failure improved by over 58% (Figure 7). Overall, the novel capillary enabled an improvement of 29%.
Productivity improvement. The novel capillary process' objective is achieved with better overall process stability and improved productivity (Figure 8). A 0.20% yield improvement results in a cost savings of $7,300 for a production run of 1 million packages (Figure 9). Cost-of-ownership computations were based on:
* Wire bonding production yield.
* Device unit cost.
* Number of wires per device.
* K-bond per capillary.
When used in fine-pitch applications, the novel capillary generates real cost savings. In addition, its greater machine utilization reduces the number of bonders needed. And through productivity gains, it also reduces wire bonding costs of goods and SG & A costs.
Yair Alcobi is bonding tools director of marketing at Kulicke & Soffa Industries (kns.com); email@example.com. He has a bachelor's in mechanical engineering from Technion, Israeli Institute of Technology, and an executive MBA from Haifa University, Israel.
*Out of 4M bonds for each capillary type [Cumulative Assists count] Novell Capillary Conventional capillary SHTL 159 164 EFO-Open 6 17 NSOP 276 451 NSOL 57 69 Overall 498 701 FIGURE 1: Wire bonding total assist count, LQFP. Note: Table made from bar graph. Table 1. Sample Diagnostic Data At K-Bonds NSOP NSOL SHTL EPO OPEN (units) (count) (count) (count) (count) 0 200 400 600 End of cap life At K-Bonds TOTAL ASSISTS MTBAp UPH net (units) (total count) (Hr) (devices) 0 200 400 600 End of cap life [Assist / 100 Kbond] Novel Capillary Conventional capillary SHTL 3.80 3.90 EFO-Open 0.14 0.40 NSOP 6.59 10.74 NSOL 1.36 1.64 Overall 11.90 16.69 FIGURE 2: Wire bonding assist type and rate, LQFP. Note: Table made from bar graph. Improvement [%] SHTL 3% EFO-Open 65% NSOP 39% NSOL 17% Overall 29% FIGURE 3: Wire bonding relative assist improvement, LQFP. Note: Table made from bar graph. MTBA [Hr] Conventional capillary 0.27 Novel Capillary 0.38 Improvement [%] Novel Capillary Improvement [%] 40% FIGURE 4: Productivity improved, LQFP. Note: Table made from bar graph. [Cumulative Assists Count] Conventional Novel Capillary SHTL 88* 683 EFO-Open 242 202 NSOP 443 178 NSOL 472 317 Overall 2,042 1,380 * [UNREADABLE IN ORIGINAL SOURCE.] FIGURE 5: Wire bonding total assist count, BGA. Note: Table made from bar graph. Assist 100Kbond Conventional Novel Capillary SHTL 3.84 3.0* EFO-Open 1.05 0.91 NSOP 1.92 0.81 NSOL 2.05 1.44 Overall 8.85 6.25 * [UNREADABLE IN ORIGINAL SOURCE.] FIGURE 6: Wire bonding assist type and rate, BGA. Note: Table made from bar graph. Improvement [%] SHTL 19% EFO-Open 13% NSOP 58% NSOL 30% Overall 29% FIGURE 7: Wire bonding relative assist improvement, BGA. Note: Table made from bar graph. MTBA Conventional 0.23 Novel Capillary 0.33 Improvement Novel Capillary Improvement 42% FIGURE 8: Productivity improvement, BGA. Note: Table made from bar graph. Device Type BGA No. of I/O 300 wires Bond Pad Pitch 60 [micro]m Wire 23 [micro]m Average Capillary Life Span 850 Kbonds Average Package Cost 3.65 $USD Regular Novel Design Capillary Average No. of assists in 35 15 No. of assists 24 hrs. Actual UPH 53.6 54.4 Units per hour Total units bonded 1286 1306 Units per day per machine Productivity increase by 19 Units per day per machine Value of productivity $69.35 Per day per machine increase W/B overall yield 99.20% 99.40% Yield improvement in % 0.20% Yield improvement in ppm 2,000 Value of yield $7,300 For production of 1 improvement milion packages Novel Capillary Value 1 GOOD Device for every 500 units produced EACH NOVEL CAPILLARY SAVES $10.34 FIGURE 9: Cost-saving exercise for a 60 [micro]m application.