Reading time ( words)
Traditional test strategies can no longer meet the demands of newer, more complex board design; looking at the test/inspection process from a new perspective may help address these issues.
Today, increasingly complex printed circuit board assemblies (PCBA) are being manufactured because of market demands for sophisticated electronic products. More components on a PCBA typically mean higher cost per board, significantly more defect opportunities, lower yields and greater difficulty in diagnosing faults. These boards have pushed the traditional test strategies beyond their limits to effectively detect faults. Smaller components are more difficult to place correctly. Visual access to solder joints is drastically limited. The traditional assumption of 100 percent electrical access for in-circuit test (ICT) is no longer valid. Understanding how to address these issues effectively in the face of time-to-market pressures is extremely important. To address the trends of higher complexity, it makes sense to look at the test/inspection problem from a new perspective.
Defining PCBA Complexity
Everyone believes their PCBAs are complex, but clearly some boards are significantly more complex than others. To bring objectivity to the phrase and to help select the most appropriate test strategies, Agilent Technologies proposes a manufacturing complexity index (Ci).
PCBA complexity increases because:
- As more components are added, more solder joints need to be soldered correctly.
- The manufacturing environment becomes higher mix/lower volume. A high-mix environment does not allow the luxury of repetition for reducing defects through finely tuned manufacturing process control.
- More manufacturing steps are added to create double-sided SMT boards instead of just single-sided boards.
The higher the PCBA complexity, the more likely it is that defects will be introduced and the more difficult it is to achieve high yields without test and inspection. A guideline for identifying high-complexity boards is found in Figure 1.
The complexity index (Ci) can be expressed as:
Ci = ((#C + #J)/100) x D x M
Ci = Complexity index
#C = Number of components
#J = Number of joints
D = Double-sided: For double-sided boards
D = 1, and for single-sided boards D = 0.5
M = High mix: M = 1 for high mix and M = 0.5 for low mix.
A board earning a score of 75 or higher can be called a high-complexity PCBA. A score of 25 to 74 describes a medium-complexity board. A score below 25 indicates that the board is low-complexity. The test strategy described next is most cost-effective and justifiable for high-complexity boards, and for some medium-complexity boards.
Test Economics and Board Complexity
Consider the boards of different complexities in Table 1.
Apply a formula for yield of:
Yield = [1 - (DPMO/1,000,000)]N
DPMO is defects per million opportunities.
N = Number of defect opportunities on the specific PCBA.
The high-complexity board has 20,000 defect opportunities (2,500 components + 17,500 solder joints). At a manufacturing process with a respectable defect level of 200 DPMO, the high-complexity board without any test and repair would have only a 2 percent yield out of SMT.
To emphasize the economic impact of testing, assume that 40,000 high-complexity boards are manufactured per year at a manufacturing cost of $4,000 each. Apply the simplest "go/no-go" test, i.e., the boards that work are shipped, and boards that fail are not repaired but simply scrapped. The annual cost of scrapped boards is 40,000 boards produced at a 98 percent scrap rate at a cost of $4,000 per board, or $156.8 million. It becomes clear that investing in effective test strategy adds significant value, particularly for the high-complexity board.
The Deficiencies of Traditional Test Strategies
It is common knowledge that it costs much less to locate, diagnose and repair defects earlier in the manufacturing process than it does later in the process. Therefore, it is important to have high fault coverage as early as possible. With fine-pitch components and double-sided SMT boards, the prevalent type of manufacturing fault is solder joint defects.
The most common test/inspection strategy used for SMT manufacturing today is a sequence of manual visual inspection (MVI), followed by ICT and then some form of functional test (Figure 2). While each test method has significant fault coverage, the strategies overlap and miss fault coverage. The grids in Figure 2 represent fault coverage, and the uncovered space within the box represents missed fault coverage. Overlapping or redundant fault coverage potentially increases total test cost, while missed fault coverage means lower quality of shipped products and more field failures.
Significant reduction in test overlap can be achieved if high fault coverage of shorts, opens and missing components can be achieved before electrical test is done. On high-complexity boards with fine-pitch components and many hidden solder joints, MVI and even automated optical inspection (AOI) simply do not have adequate visual access to achieve high enough fault coverage. The test method with the highest possible fault coverage of solder defects is automated X-ray inspection (AXI).
ICT for high-complexity boards has shortcomings of its own. High fault coverage is usually achieved after a long, costly programming effort. In addition, turnaround time for building the required fixture can be significantly longer when the node count is more than 3,000. Both of these characteristics negatively impact the manufacturer`s time-to-market.
For high-complexity boards, ICT fixtures are not only expensive, but the probability of fixture contact problems increases with the number of test probes needed to hit shrinking test targets. This can result in numerous retests and difficulties in differentiating between real faults and fixture-induced faults. Several manufacturers have tried traditional ICT on high-complexity boards with less than optimal results.
A new test strategy that has a high fault coverage of solder defects as early as possible, provides solutions to limited electrical and visual access, uses simpler and more reliable fixtures, and has higher yields into functional test is needed.
New Paradigm for Intelligent Testing
The new recommended test strategy for high-complexity PCBAs has multiple steps. The first is to replace visual inspection with AXI. After X-ray inspection, the boards still go through ICT and functional test - with a new twist. The new test strategy is shown in Figure 3. Higher fault coverage and lower redundancy are indicated. Less redundant test is possible for two reasons:
1. AXI has an extremely high fault coverage of solder joint defects, such as opens, shorts, insufficient solder and other marginal solder joints; AXI also identifies missing components.
2. Because AXI has such a high fault coverage on solder defects, ICT can be simplified to focus on component placement, functionality and orientation.
The AXI/ICT/functional test strategy is already used today by leading-edge manufacturers. These users have commented about the complementary nature of AXI and ICT:
"The biggest single impact is yield at ICT; we saw this rise from an average of about 75 up to 95 percent ... It was literally a step change."1
"Neither the AXI or ICT system alone could have achieved the net effect of the complementary test strategies. Incorporating both systems has increased manufacturing defect coverage on products to 97 percent, lowered field failures, warranty and repair costs, and allowed greater design flexibility. Products move to market faster, costs are lower and quality is higher. A complementary approach is key."2
Recently, a Hewlett-Packard study found higher yields into functional test, fewer field returns and lower overall costs if an AXI and ICT approach is used to find manufacturing faults.3 This study focused on six different board types that are manufactured by a contract manufacturer (CM). The CM provided MVI, AXI and ICT test while the HP division provided functional test. For one board type, no AXI or ICT were performed - only minimal MVI - because of lower complexity and anticipated lifetime volume. For the other five board types, ICT was used. The gathered data were from normal production and field failures over a significant amount of time. The 6,928 boards in the study were of the high-mix, low-volume category.
In Table 2, each test strategy`s effectiveness at detecting solder-related defects, non-solder-related defects and field failures is shown. The defects per million component opportunities (DPMOc) reported defects detected at functional test. As different test strategies were added, the DPMOc was reduced significantly when AXI replaced MVI. (For this study, no data was available regarding the effectiveness of AXI alone, but experience indicates that AXI is extremely effective in finding solder-related defects.) However, the DPMOc stayed basically the same when both AXI and ICT were used to detect non-solder-related defects. This is not surprising because AXI is not well suited to detect these defects. For this sample of boards tested with AXI, ICT and functional test, the field failure rate decreased dramatically. Figure 4 is an overview of the different test steps and their effectiveness, while Figure 5 is a graphical illustration of the different strategies` effectiveness.
AXI was shown to be effective at finding solder-related defects. ICT was effective at finding non-solder-related manufacturing defects, so yields are much higher into functional test, and field returns are significantly reduced, with ICT cutting field failures by up to half. Adding AXI to the strategy reduced field failures close to an order of magnitude because AXI is so effective at finding marginal solder joints - those that are electrically correct during test but are likely to fail during thermocycling or mechanical stress. For high-complexity boards with many defect opportunities, the results are significant.
Keeping Tests and Fixturing Simple
If all shorts, opens and missing components have been removed, ICT is much simpler. The remaining manufacturing faults are typically wrong value component, wrong component, misoriented component, damaged component, etc. Consider the circuit in Figure 6, an octal buffer with series resistors on all inputs. The series resistors are in one resistor pack. The traditional ICT strategy for such a circuit is to test for shorts, test all resistors to make sure they have the right value and that there are no opens on the resistor pack pins. Finally, all pins on the octal buffer are tested to verify right component and that all pins are soldered correctly. Testing the circuit in this way requires 25 test probes.
With the new strategy of AXI before ICT, in-circuit testing can be limited to right component, functionality and orientation. Only one resistor in the resistor pack and one of the eight buffers needs to be tested; the number of test probes required is reduced to four.
It should be noted that some component failures and silicon failures could go undetected from the AXI and ICT test steps. While overall impact will be higher yields into functional test, a few faults that could have been detected at ICT under the traditional strategy will now be detected at functional test under this strategy.
In actual tests on several boards, these techniques have reduced node count on high-complexity boards by 28 to 76 percent. Early indications show that total test development time for this new test strategy is shorter than the conventional test strategy development and that debug is also shorter and easier.
Of course, the node reduction technique does not need to be applied to all components. For instance, a Flash-RAM or a critical component can be 100 percent probed as usual. The node reduction technique offers more flexibility and a design for test tool when electrical access is reduced and when lowering overall test costs is a priority.
1 David Evans, "X-ray Vision Improves Yield," TEST, November/December 1996.
2 Steve Bonham and Wayne Singbeil, "No Single Test Strategy," Circuits Assembly, November 1997.
3 Agilent Technologies, "A New Test Strategy for Complex Printed Circuit Board Assemblies - an Update," by Stig Oresjo, August 1999 (internal HP study).
STIG ORESJO is a senior test strategy consultant with the Manufacturing Test Div. of Agilent Technologies Co., a subsidiary of Hewlett-Packard Co., P.O. Box 208, Loveland, CO 80539-9951.
Figure 1. How to identify high-complexity PCBs.
Figure 2. The most common SMT test/inspection strategy. The grids represent fault coverage.
Figure 3. Combing AXI and ICT results in higher fault coverage.
Figure 4. Different test steps and their effectiveness.
Figure 5. AXI and ICT are more effective than MVI or ICT alone.
Figure 6. Using AXI before ICT reduces the number of test probes needed.