Justifying inspection in a test strategy
Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 11/1/2005
When examining the role of inspection in manufacturing test, you must look at both its costs and its benefits. Managers are notoriously unwilling to authorize the purchase of capital equipment without some assurance that the expenditure can be justified.
As bed-of-nails access continues to decline in many products, manufacturers must increasingly rely on functional test, hot mockup, and creative use of self-tests to verify that a circuit works. Developing comprehensive test programs for such situations is challenging, and the resulting fault reports may flag a board as bad but often cannot pinpoint the exact failure location closer than a particular section or logic block. Diagnostic techniques such as guided probing must zero in on the actual fault.
Government demands to eliminate lead from components and solder complicate the situation further. Higher melting points and other inherent characteristics of lead-free solder will ensure higher defect rates, at least at the beginning.
Also, because lead-free solder is more brittle than its tin/lead counterpart, companies may limit physical probing on faulty boards for fear of creating problems rather than correcting them. In that case, it is very likely that more boards will be scrapped rather than repaired.
Inspection to the rescueAn effective inspection step can dramatically reduce the cost and quality penalties introduced by these manufacturing trends. Because inspection "looks" at the board to determine whether it was built correctly, it can more easily identify the exact culprits for the faults that it finds.
Automated optical inspection can spot bent and broken leads, solder shorts, and other faults on visible nodes. Techniques like x-ray inspection will see missing or misplaced balls under ball-grid arrays, bad solder joints, and other problems hidden from more conventional techniques. Adding inspection to a test strategy will ensure that more boards can be repaired successfully and fewer end up as scrap.
The impact of fix vs. scrap is far from trivial. Most economic analyses will show that if there is more than a small amount of scrap, its cost overwhelms all other costs.
Manufacturers often measure process quality in defects per million opportunities (DPMO). Stig Oresjo, senior test strategy consultant with Agilent Technologies, presents the number of defect opportunities as the sum of the number of components and the number of solder joints. You can calculate expected yields with this equation: where:
DPMO = defects per million opportunities
N = number of defect opportunities
Running the numbersTo justify investing in an inspection system, you need to calculate the number of faulty boards you produce and the expected savings from repairing them rather than scrapping them. Consider a board with a manufacturing cost of $400 that contains 550 components and 4450 solder joints—a total of 5000 defect opportunities—built with a process that produces a DPMO of 75. Using the equation above, yield from assembly (before any test or repair) would be 68.7%.
Assume that without an inspection step, you can identify and successfully repair half of the defective boards. In that case, the effective yield after test and repair is 84.35%. To ship 160,000 good boards per year, you would have to build 189,686 boards. At $400 per board, the scrap costs the operation almost $12 million (Table 1).
Suppose you add an inspection step to diagnose the failures more effectively so you can repair more boards. Assume that fixing the boards costs 25% of their value. If inspection allows you to repair even half of the previously scrapped boards, the cost picture would look like the "Fix 50% of boards" column in Table 2.
The inspection and repair steps reduce the number of board starts required to ship 160,000 boards to 173,583 and saves more than $5 million. The need for fewer board starts also increases the effective capacity of the production line and can postpone the need for labor overtime or new production facilities, representing huge potential additional savings.
Repairing a larger proportion of the boards improves this picture further. If you can repair 90% of the previously scrapped boards, as in the third column of Table 2, effective yield increases to 98.4%.
The number of board starts drops to 162,544. The total cost for board scrap and repair now totals only $3,307,000, so you save more than $8.5 million in a single year. Numbers like these go a long way toward convincing company controllers that inspection systems will pay for themselves in short order.
Increasing the benefitsFor many boards, especially expensive ones, the average repair cost is much less than 25%. Repairing a board worth $2000 would rarely exceed a few hundred dollars, reducing the total repair cost and making inspection an even more attractive option.
The calculations in Table 2 hinge on meeting a 75 DPMO specification. If the number of defects were even double that, first-pass yield drops to 47.2%. You would need to build almost 340,000 boards to ship 160,000, and the other costs would skyrocket as well.
Of course, adding an inspection step is not always cost-effective. Repairing very simple, inexpensive boards could actually exceed the cost of scrapping them.
Similarly, repairing boards that are conformally coated for high-reliability applications requires recoating the boards after repair, so the cost of fixing them may also exceed their value. In these cases, scrapping the boards remains the most economical alternative.
Getting there from hereThe object of any test strategy is to ensure that you ship good products to customers. With the advent of lead-free solder, which has a higher melting-point and is more brittle, you also need to be reasonably sure that merely shipping the products will not make them fail.
Incorporating inspection can alleviate some of those concerns. But convincing company managers to spend what can be a significant sum on capital equipment requires numbers to show how the purchase will ultimately benefit the company's bottom line.
Fortunately, generating convincing numbers is not difficult. The only task remaining is selling those numbers to the managers who must sign the checks.
| Number of components | 550 |
| Number of solder joints | 4450 |
| Defect opportunities | 5000 |
| Nodes | 500 to 1500 |
| Yield from assembly (before test or repair) | 68.7% |
| Repaired boards, as a result of test only | 50% of failures, or 15.65% of board starts |
| Yield from test and repair (no inspection) | 84.35% |
| Number of boards planned to ship per year | 160,000 |
| Board starts, including scrap, needed to reach planned shipment | 189,686 |
| Cost per board | $400 |
| Value of scrapped boards, 15.65% of board starts | $400 x 29,686 boards = $11,874,400 |
| Fix 50% of boards | Fix 90% of boards | |
| Number of boards planned to ship per year | 160,000 | 160,000 |
| Yield from assembly (before test or repair) | 68.7% of board starts | 68.7% of board starts |
| Repaired boards, as a result of test only | 50% of failures, or 15.65% of board starts | 50% of failures, or 15.65% of board starts |
| Yield after test and repair (no inspection) | 84.35% of board starts (leaving 15.65% scrap) | 84.35% of board starts (leaving 15.65% scrap) |
| Repaired boards as a result of inspection | 50% of scrap, or 7.825% of board starts | 90% of scrap, or 14.085% of board starts |
| Yield after test/inspection and repair | 92.175% of board starts | 98.435% of board starts |
| Board starts needed to reach planned shipment of 160,000 | 173,583 | 162,544 |
| Number of scrapped boards per year | 13,583 | 2,544 |
| Cost per board | $400 | $400 |
| Value of scrapped boards | $5,433,200 | $1,017,600 |
| Cost of repairing extra boards as a result of inspection ($100 each) | $1,358,300 ($100 x 13,583 boards, or 7.825% of board starts) | $2,289,400 ($100 x 22,894 boards, or 14.085% of board starts) |
| Value of scrapped plus extra repair | $6,791,500 | $3,307,000 |
| Savings over no inspection ("Value of scrapped boards" from Table 1) | $5,082,900 | $8,567,400 |
