Defect-detection strategies: today's operations require a high-speed 3-D approach.

Ed.: This article is abridged. For the complete version, visit circuit-sassembly.com.

A system throughput increase has a direct impact on conversion cost by reducing the number of systems required to meet manufacturing volumes. Typically, throughput increases have come incrementally by increased efficiency in panel handling, image acquisition or image analysis. Increases in panel handling efficiency include faster movement of the panel within the system, panel vibration reduction and panel loader use. Image acquisition speed increases can include elements such as reduction in image overlap, parallel image capture, faster detector readout and improved x-ray to light conversion. Image analysis speed is typically addressed by using efficient analysis algorithms and by leveraging Moore's Law. As an alternative to increasing the basic speed of the system, some solutions rely on using a combinatorial approach of 2-D and 3-D x-ray inspection. 2-D x-ray inspection has been traditionally much faster than 3-D. This is achieved by using a large field of view (FOV). For single-sided or simple double-sided boards where the designer has followed strict design-for-x-ray guidelines, 2-D is an effective solution for detecting soldering defects.

A combinatorial strategy is to inspect most of the board with 2-D and then augment defect coverage with 3-D on a small portion of the board as necessary because of joint shading. This strategy breaks down quickly as the complexity of the board increases. In a recent evaluation of different board types, the percentage of overlapping solder joints caused by double-sided boards ranged from approximately 8 to 40%. The 8% was on a typical, low complexity automotive double-sided board. These boards are usually designed to be inspected by 2-D x-ray. This shows how difficult it is to design a double-sided board without overlapping solder joints. The combinatorial strategy can currently provide high throughput and high defect coverage on this type of product. In typical medium-to-high complexity communications products, 25 to 35% of solder joints are overlapping. Using 2-D alone on these products has an immediate and significant coverage loss. Furthermore, these overlapping joints are often spread throughout the board. This requires a significant amount of (much slower) 3-D inspection for coverage.

The general manufacturing trend of increased joint density is in conflict with the overall long-term success of a combinatorial strategy. To meet manufacturing needs then, high-speed 3-D inspection is needed. This approach would provide the necessary throughput to reduce the capital investment while still maintaining the high defect coverage provided by 3-D inspection.

To Sample or Not

Some manufacturers use either board- or device-based sampling techniques to increase automated x-ray inspection (AXI) throughput. The implication of any sampling strategy is that a user is trading off throughput against defect coverage. Several studies (2) have indicated x-ray inspection is the single best step for capturing solder-related defects. In many cases, defects detected cannot be found at any other test step. Therefore, the very nature of sampling in a continuous flow environment means that defects that would have been caught at x-ray are passed to the next test step. This is true of board- or device-based sampling. Thus, sampling can be an effective test strategy when the end product has three characteristics:

* Very high first-pass yield into ICT or functional test.

* High defect coverage at ICT or functional test.

* Low cost/risk associated with field failures.

Typically, repair cycles at either of these steps take longer than x-ray. At functional test, which typically has poor repair diagnostic resolution, repair time can be 10 times greater than repair time at x-ray. The second condition is necessary to reduce the number of escapes that will make their way to the end-user.

Manufacturing defect phenomena have traditionally been categorized as either systematic or random. Systematic defects can be those caused by situations like a clogged stencil at paste deposition or an incorrect reel at the pick-and-place system. These types of defect causes affect multiple boards in sequence. Typically, the earlier in the process that this type of defect cause is detected, the lower the overall repair cost. AOI-based solutions for solder paste or pre-reflow inspection have been developed specifically for finding these structural types of defects quickly and efficiently. As the name suggests, random defects have no readily identifiable cause. The frequency of this class of defect, product cost and overall warranty cost of an escape drive the manufacturing defect containment solution. X-ray has been shown to be an excellent solution for capturing the broadest range of defects. The level of sampling at x-ray directly correlates to the number of random defects that escape to the next test steps. The following formula (3) approximates the turn-on rate based on the number of defect opportunities (n) and the defect rate of the process (DPMO).

Yield = [1-(DPMO/[10.sup.6])][.sup.n]

This formula is a reliable predictor of yield for totally random defects. Figure 3 shows the estimated yield of a board with 20,000 solder joints; an excellent solder joint defect rate of 50 DPMO will have a turn-on rate of about 37%. Without a solution to effectively detect these defects, the result will be a large number of boards failing functional test or escaping to the end-customer with the potential for field or warranty failures.

James Benson is AXI product marketing engineer at Agilent Technologies (agilent.com); [e-mail].