Low-cost method of making leadless packages

The demand for quad flat no-lead (QFN) packaging has dramatically increased during the past three years. It has been replacing older packages like SOIC and TSSOP mainly because of the added thermal enhancements from the exposed pad, as well as the elimination of coplanarity issues. QFN features a relatively bigger pad size, allowing more die integration. As the requirement for more I/Os in a smaller footprint increases, current QFN structure will have some limitations. Currently, there is a solution for two rows of I/O but any higher than that, the available solution becomes more expensive. This article describes an alternative methodology of making a low-cost QFN with multiple rows.

Leadframe-based packaging is a low-cost IC solution. However, a conventional leadframe has limitation in the ever increasing complexity of products. On the other hand, a substrate-based solution offers a lot of flexibility through routing and multi-layer metal construction but is more expensive. Multiple-row QFN has brought the package to another step of satisfying the increase in I/O. The inner leads could be produced by using a narrow tie bar attached to the connecting bar that is holding the outer leads (Figure 1, Type A). However, the inner leads are challenging to make and the narrow tie bar lacks rigidity, making the wire bonding process more difficult. This results in lower productivity and yield. Dual-row QFN can also be produced by having the inner leads connected to the die pad (Figure 1, Type B). This is more robust, but it requires a two-pass sawing process, whereby the first pass uses partial cut to isolate the inner lead from the die pad and the second pass is the final cut to singulate the units. Due to this two-pass process, the saw capacity is reduced by 50 percent and is likely to produce rejects.With the current challenges on the existing dual-row QFN, an alternative multi-row leadless packaging – called etched leadless packages (ELP) – was developed to offer not only dual-row but also higher I/O low-cost packaging with the advantage of better wirebond stability with patronizing standard manufacturing processes such as the single pass sawing process.

MULTI-ROW QFN VERSUS ELP

Leadframe design and bondability: In a standard multi-row QFN, majority of the leads and features are created by etching through the top and bottom of a copper stock, which is about 0.20mm thick. For Type A multi-row QFN, the inner leads are attached to a half-etched connecting bar. This makes the structure weak and prone to bent leads. A common problem for a typical QFN with tape is the stability during the second bond formation. The use of Type A QFN makes this even more difficult due to longer lead lengths and thinner connecting bars. Dual-row Type B QFN provides better stability of the leads during wirebond since the inner leads are shorter and connected to a more rigid structure. However, this solution requires a dual-pass package saw singulation process that results not only in low yield and productivity, but also in design limitations such as a smaller die pad and a smaller number of leads to allow path for a saw blade during inner cutting. Both Type A and Type B only allow a maximum of two rows of leads. On the other hand, ELP is a multi-row leadless package that utilizes a partially etched leadframe. The top portion is half-etched, which defines the desired features such as leads, power, ground rings, and die pad. The bottom portion of the leadframe is solid. Due to the rigidity of the leadframe, multiple heavy wire bonding such as 2mil Au wire, aluminum wedge/ribbon bonding, and copper wire bonding are viable.

Post mold cleaning: Typical QFN comes with a tape underneath to prevent mold flash. Depending on the tape used, this will require an additional cleaning process after mold to remove the tape residue and mold flash. The extra cleaning and the tape required add to the cost of the overall package by as much as 15 percent. ELP has an advantage in this area because it does not require a tape.

Package saw singulation: The QFN requires metal cutting between units. There are several issues associated with this. First is low productivity. Saw speed is typical at the range of 35-60mm/sec, which is very slow compared to a standard punch singulation process. Second is high blade consumption due to metal cutting. Third is unavoidable cosmetic concerns such as copper burrs and smear. ELP undergoes back-etching process after mold. At this stage, the copper in between units is fully etched out, leaving only 100 percent mold compound left for sawing. This results in an improved saw speed up to 120mm/sec, longer blade life, and absolutely burr-free units.

ELP PROCESS FLOW

Generally, ELP has similar processes and applicable controls as an existing QFN package except for the etching process and immersion plating process needed to isolate and plate the leads respectively. For QFN, the isolation is done by either sawing or punching.

In backside etching, the features are pre-defined by a pre-plated mask. The process of masking can be done during leadframe assembly or after the mold assembly process. Since masking is an established process in leadframe assembly, it is preferred to perform the masking during leadframe preparation for process simplification and better control such as placement accuracy. The mask is either NiPdAu or Ag and can be the same material plating used on the leadframe bonding side. Ag masking is used because it employs straightforward stripping before final plating. For NiPdAu, the stripping process is complex and costly due to the multiple stripping processes needed (i.e., separately stripping Ni, Pd and Au). For this reason, Ag masking is selected, as the process is simple and takes only one stripping procedure. To isolate the leads, rings and pads, the leadframe material needs to be back-etched. This is done after the mold process. After etching the backside to define the features, it is optional to leave the Ag mask as a final finish. However, the Ag mask will have overhang after etching. Since the mask is very thin, it will tend to deform at any direction, which is cosmetically unacceptable. Stripping off the Ag mask and re-plating it with the desired finish is suggested.

There are two industry standards available for copper etching: acid-based and alkaline-based. So far, there is no known process that uses Ag as a mask. In selecting the etching process, it is important to understand the chemical’s effect on the Ag mask. Ferric chloride and cupric chloride are the most commonly used acid-based copper etchants, and ammonium chloride is used as an alkaline-based copper etchant. Based on the study, both chemistries were able to etch the copper, but acid-based etchants easily attacked the Ag mask. After etching, the Ag mask turned a dull color and was difficult to strip off.

Ammonium chloride slowly etches the Ag when exposed to it for more than 1min. The etch time can speed up by increasing the temperature to about 55ºC to make the chemical more aggressive. However, it shortens the life of the bath due to the high loss of ammonia that results in inconsistent etching. Increasing the pH will also increase the etching rate of the chemical. This eliminates the Ag attack and can operate at a lower etching temperature which stabilizes the chemical. Another observation is that ammonium chloride does not affect the Ag mask and thereby enables the Ag stripping process.

After backside etching, surface finishing is applied on the isolated circuitry to prepare the material for board mounting. While the leads and pads are already isolated, conventional electrolytic process is not applicable and electroless/immersion process is the suitable process. There are two designs in the market: horizontal and vertical immersion plating processes. The difference between the two is that for horizontal process, the material is fed through a conveyor, while for vertical, racks or trays are used to carry the materials and dip them from station to station. Immersion or electroless process is a longer process compared to electrolytic. To achieve the plating thickness required for electroless/immersion plating, the materials must be submerged in the chemical for a long period of time. This is where the decision must be made to either go vertical or horizontal. If there is a space constraint, the vertical process is recommended since a horizontal process could be as long as 40 meters. If there is no space limit and the capacity requirement is high, the horizontal process is the best choice. Both processes should have similar output in terms of quality.

There are many immersion processes that can be used for ELP: electroless Ni and immersion Au (ENIG), immersion tin (ISn), organic solderability preservatives (OSP) and immersion Ag (IAg). Each has its pros and cons.

ENIG is a very stable plating solution. However, the cost is high and the process control is challenging since there are two metal components involved: nickel and gold. Also, waste treatment must be considered due to the cyanide component of Au. OSP is the cheapest solution but there is concern of visual in-process control since it is colorless. There is also a concern for electrical testing since it is non-conductive. Immersion Ag is a cheaper solution than ENIG. It features low-temperature with plating thickness only above 0.127µm, and the plating time is very quick. However, IAg is sensitive to the environment and is easily tarnished. Good material handling and environment control must be employed. Immersion Sn is comparable in cost to immersion Ag. The thickness of ISn is also relatively thin but doesn’t easily tarnish compared to IAg. In IPC -4554 section 3.2.1, the immersion Sn thickness shall be 1µm minimum at 4 sigma. To achieve ISn thickness, the plating process must be extended. Typically, it requires about 30min dwell at 70ºC in the main plating chemistry. This will define the length of the plating system.

The plating thickness of ISn being only over 1µm may affect solderability due to quick intermetallic growth. It is then suggested to have a six-month shelf life to ensure solderability. There is also a question on whisker growth common to Sn plated products. ELP is considered as leadless package and is exempted from the whisker growth category based on JEDEC JESD201, as the leads will be fully covered by solder during board mounting. However, the chemistry selected has a whisker-limiting agent. Studies on the kinetics of whisker growth state that whisker is produced by compressive stress coming from the buildup of Cu-Sn IMC during storage. This stress is relieved by diffusion of Sn atoms resulting in whisker growth. This whisker can occur along the Sn grain boundaries or near lattice defects.

Whisker can be controlled by thermal excursion or what is commonly called baking. This process increases the oscillation of atoms within a crystal and at the same time heals the lattice defects. However, since ISn is very thin, thermal excursion or baking is not ideal since the temperature will speed up the IMC growth that affects solderability. With this as limitation, change in the chemistry must be made. The chemical identified for ELP has a proprietary additive that changes the Sn crystal structure to control diffusion. The additive blocks the pathway along the grain boundaries and heals the lattice defect, thus preventing whisker growth. With solderability still a question due to the thickness, the chemistry process requires pre-plating of a dense Sn layer prior to final plating. This was done at a low temperature of about 25ºC. The purpose of this pre-plating is to minimize IMC growth, ensuring better solderability after long storage and multiple reflow processes.

PACKAGE CONSTRUCTION

The etching process makes ELP construction different, as shown in Figure 2. Since the leadframe is half-etched at the beginning, a conventional locking feature is not applicable, such as “lip” on lead tips and pad edges. A different approach can be applied, such as slots on pad and irregular lead shapes and profiles (Figure 3). Also, the external leads will be tapered as a result of the etching process. The lead will have higher standoff to about 2-4mils (50-100µm). This feature will have stronger soldering to the PCB and at the same time will provide helpful space for flux cleaning when required.

Since the lead has a narrow tip compared to conventional QFN and has more leads, test socket manufacturers need to design good and reliable contacts. Several suppliers have reviewed the design requirement and suggested that it can use conventional socket design. So far, two sockets were tested using 228L 12x12 ELP with triple-row, 0.5mm pitch. The test insertion is only one pass with 100 percent yield.

Second-level reliability studies were performed to define the limitations of the new lead design. These were done at an independent lab in Hong Kong. The tests included package pull test, mechanical drop test, mechanical bending test and temperature cycling test. The package pull test was done in comparison with 10x10 72L QFN package. To have one-on-one comparison, only the outer leads of ELP were soldered to the PCB and both die pads were unsoldered. The resulting pull forces were very close to each other, with ELP showing two failure modes (lead and pad). Lead failure mode is observed when the pull force is <150N. Therefore, ELP lead pull strength is comparable with QFN.

In the mechanical drop test, a high-speed oscilloscope was used to monitor the daisy chain resistance in real time. The oscilloscope can only measure voltage, not resistance. Therefore, an extra bridge box setup is needed. The voltage across the 100Ω resistor was measured. If the package failed during the drop test, the resistance of the daisy chain would increase and the voltage across the resistor would drop to zero. The units passed the mechanical drop test since no failures were observed on all samples after dropping each PCB up to 30 times.

In the mechanical bending test, the crosshead speed was 2.54mm/s. A strain gauge was installed in the middle of the PCB to monitor the load, displacement and strain reading in real time. The samples passed the mechanical bending test since no failure was observed in all the test boards.

In the temperature cycle test, each unit in the board was monitored in real time inside the chamber. The ELP samples tested has reached >1,300 cycles at -45ºC/+125ºC without any failure and was still an ongoing test. Based on board-level reliability design studies, an 8x8 QFN package was able to reach 1,126~1,356 cycles in -45ºC /+125ºC conditions.RELIABILITY

Internal features such as pad and lead designs promote interlocking with the mold compound and contribute to package integrity during reliability tests. The ELP internal feature design is different from the QFN internal feature design, but it offers the same interlocking strength. As the construction of ELP limits routing of bond post, it is expected to have long wire lengths, especially for triple-row applications. The die sizes are typically small with fine pitch bond pads requiring small bonding wire diameter. This impacts the wirebond and molding capability.

In wirebond process, the wire length is increased by 30 percent to 50 percent and requires multi-row bonding. Standard or square loop type for QFN is practically not adequate for ELP, as the latter package requires good vertical clearance between different loop tiers and extra kinks to support the longer wire lengths. It is therefore recommended to use a loop profile capable for >3 kinks for better loop stability.

In mold process, it is necessary to determine wire sweep performance after mold. Parameter optimization was conducted to get the best mold response to eliminate risks of wire short and wire sagging. Also, a new mold compound type was considered to ensure robustness in the molding process. The property of this new mold compound is such that it puts less pressure on the wires during the mold transfer process. It has lower narrow gap flow pressure that is most suitable for longer wires with multiple wire loop tiers.

CONCLUSIONS

ELP is a solution to bring leadframe-based packaging closer to substrate-based packaging. It offers flexibility that a standard leadframe package cannot support such as heavy wire bonding, split pad, SIP, etc. Its design improves yield by eliminating known issues at assembly such as second bond stability, mold flash and tape residue, all of which has a direct impact to packaging cost. Though the ELP package-locking feature is different from QFN conventional locking design, the package- and board-level reliability test results are comparable.

ELP introduces new processes that need to be considered, such as back-etching and immersion plating. Back-etching uses alkaline-based chemistry and plating uses immersion type. These processes are not new but were used in different industries. It is shown that with fine-tuning, these processes will be able to support ELP requirements.

With the limitation of routing and with the higher I/O design, the wire length will be longer and bring challenges to wirebond and molding capability. It is needed to optimize the looping parameter and mold parameter to ensure good wire clearance. Using a low-pressure mold compound also helps improve the manufacturability of ELP. The simplicity of the process and the advantages at saw singulation and mold processes makes ELP the best available alternative solution for low-cost multiple-row leadless packaging.

Click here for Illustrations:

Figure 1, Figure 2, Figure 3