Solutions for Today's Most Advanced Rework Challenges

Central to each of these challenges is the need to provide rework and repair solutions that:

• Deliver heat where it’s needed and not where it’s not. It does no good to rework one component, only to disturb several others during the process.
• Demand high accuracy - as area-array features fall below 4 mil, sub-10 µ placement, accuracy will become a commonplace requirement.
• Provide system to system reproducibility - Why? Surface mount processes are being stretched to their limit and errors will multiply. Rework volumes will increase, multiple units will be necessary - all running identical profiles and expecting identical results. And most importantly, with expectations of reproducing the same results on different machines spread out over the globe.

This paper focuses on several of today’s more challenging rework processes, including small passives, single ball reballing, stacked die, package-on-package, QFN paste print, and the need for small-volume dispense.

Fig. 1: 01005 Thermodic Nozzle

An 01005 device (Fig. 1) is much smaller than a fine soldering needle tip. As a result, the simplest form of manual rework (without machine assist) is an unrealistic production process. To provide some idea of the size of an 01005 component, a single unit weighs less than 0.1 mg. and is essentially the same weight as a small grain of sand. Failures that occur during automated small passive placement/reflow are generally due to inaccurate positioning, faulty solder paste printing, and the influences of vibration and shock. These result in common failures such as rotated, tombstoned, billboarded, broken or missing components.

While the price of such passive devices is commensurate with their size, they typically are functionally integrated into vastly more expensive modules and circuits.

Their replacement requires:

• Thermodic nozzles that primarily use heat conduction through the tip to pick up, reposition and replace rotated and other non -“flipped” errors
• A vacuum nozzle to remove tombstoned and billboarded faults, and the residual solder from the pads.
• Integrated paste dispense to deposit fresh solder paste on pads as small as 0.250 mm. (0.010") in diameter.

By integrating these processes into one platform, the complete cycle can be performed on an array of components - sequentially removing, pasting fresh solder paste and replacing the component, or carrying out the same step on each module before moving to the next process step.

Fig. 2: Solder ball on pickup tool

Single ball reballing is a technique to replace a defective solder sphere in an area-array type component (BGA/CSP/flip chip), in which the sphere connection is defective, and if replaced, restores to good health an otherwise scrap component.

Due to the array characteristics of these devices, visual identification clearly shows the location of the defective ball. The defect is removed by melting the solder and removing it via a vacuum nozzle. After fluxing a replacement sphere by dipping into a flux tray with the appropriate flux depth, the sphere is aligned to the solder pad and reflowed. It is key to control the gas flow to ensure that neighboring balls are not disturbed. Things get trickier as the sphere diameter approaches 100 µm. Anti-static pick-up tools, together with a “puff” of air, are needed to release the ball from the tool tip. At such dimensions, both accuracy and optical resolution play a more important role, and material handling problems become just as important as the rework and repair itself.

Because this is a serial process, it is unlikely to appeal to large area-array packages with hundreds of I/Os. But, as the ball count and diameter decrease, where replacement packages may not be available or prohibitively cost effective, the process offers a simple and immediate solution to an “annoying” problem. Furthermore, this technique is simply an add-on to an otherwise established process, adding significant incremental value at small cost outlay.

Fig. 3: Stacked die rework nozzle

Flip chip, stacked die rework provides the ability to rework stacked packages that are being adopted in more applications than just the memory module arena. The process can separate die from one another, or the complete package from the substrate/interposer. The problem here is the ability to control the airflow rate, direction and temperature gradients to such an extent that the “stack” separates in the required location.

The process requires special tooling to directly apply hot air or nitrogen to the bond that requires rework. Using a high magnification camera at a low incidence angle to observe the position of the tooling relative to the die-stack becomes advantageous. The ability to remove the uppermost package or one from lower down depends on the design of the nozzle and, in particular, the location of the air channels with respect to the solder interface between the die. Removing the residual solder from a stacked package allows the package to be reused or to be studied for failure analysis identification.

For some applications, vacuum and hot-gas flow are still not sufficient to overcome the bond between devices. Under such circumstances it may be necessary to use a “gripper” device built into the nozzle.

Fig 4: Clamping nozzle with actuated tweezers in small clearance between PoP and small passives

PoP rework requires a nozzle with a clamping solution, as well as an effective combination of selective bottom and top heating, in order to avoid PCB and package warping.

Requirements for successful PoP rework include:

• no mechanical lateral movement
• sufficient force to lift the package from the liquid solder
• no contact with or misalignment of neighbor components, especially in densely populated PCBs
• a single nozzle for de-soldering and soldering
• precisely-controlled bottom and top heat that ensures minimal stress to component
• lead-free repair according to the JEDEC guideline (DoE)

Fig. 6: Solder paste allied in-situ to QFN component

Increasingly, micro lead frame (MLF) components, such as QFNs, are incorporated into area-sensitive products - with contact pads directly attached to the bodies via lead frame technology. Because leadless components do not provide any solder coating, new solder paste must be applied during the rework process. When integrated into a rework system, a newly developed direct component printing module provides the one process step needed to create an “all in one“ solution for reworking QFN and MLF components. The component to be pasted is placed in the component carrier, snapped into the module, and flipped 180° . A stencil-handling tool is loaded into the reflow arm of the system and locked into place. This tool is used to pick up the stencil, align it, and drop it into place over the part to be printed. It can be used for any component and varies with the overall size of the stencil frame. Stencil openings and component pads are viewed through the split-vision optics and aligned. The stencil is lowered toward the component surface, vacuum clamped to the module, and freed from the pick-up tool. Solder paste is spread with a spatula in typical fashion.

Once the paste is printed, the stencil surface is realigned to the pick-up tool, vacuum applied, and then lifted from the component surface prior to being removed from the tool. The reflow nozzle for the specific MLF is inserted, and the module is flipped back 180° so that the backside of the die faces the pick-up tool. The component is picked from the fixture and the freshly pasted surface faces the MLF pads. Using split-vision optics, the pads are aligned to the substrate and the component is placed prior to reflow.

The direct component printing module is integral to the process. Split-vision optics are used not only to align the stencil to the LAN array (to ensure accurate paste placement on the component), but also to align the component to the substrate. This ensures that once it is picked from the module, the process consisting of align, place, and reflow can continue automatically.

It is not difficult to predict the direction that rework system manufacturers must follow - simply track dimension size, device proximity and vertical integration within a system architecture that minimizes temperature variations from machine-to-machine. Advantageously, it should possess an open architecture capable of integrating the various process steps necessary to ensure that a single platform can perform the complete rework cycle. And, one that does so in a financial environment that still considers rework a “non-value-added” process.

Then what? As assembly processes leading to the need for rework are automated, so to will certain segments of the rework market itself seek an automated solution. Following test and AOI, the rework/repair station will be required to accept databases defining and detailing the failure location and device type, and be expected to handle product arrays of “like” faults. Removal, surface conditioning, paste/flux dispense and device replacement will be automatically sequenced.

Logically, this applies more to OEM situations whose defective product is identified and corrected prior to distribution. Such situations will result in the separation of the types of rework solutions - those aimed at “board-by-board” repair in which each repair may differ from the job before, a new board or a different component, and those that address the volume production need. Fulfilling both needs will be the domain of a few progressive manufacturers who recognize the need to develop processes in tandem with advances in packaging design and assembly integration. Satisfying only the former applications will be the domain of many standard, array-package-oriented systems where technology demands are significantly more lenient on the rework system design.

As the 40^th anniversary of Moore’s Law passes, and the future continues to develop ever smaller, designers must satisfy the ever-growing need for speed, bandwidth, functionally and whatever else Madison Avenue and its counterparts introduce in to the marketplace. As a result, rework and repair is one technology with a predictably busy future, for to err is human, even for machines.

Author: Dr. Chris Underhill, General Manager
FINETECH – Tempe, Arizona
480-893-1630
www.finetechusa.com