Yield improvement is one good way to reduce costs.

A recent case study illustrates the impact of defects on cost. For this particular company, the DPMO value was calculated to be 3,500 parts per million (ppm) at a production rate of 1,000 boards a day and 4,200 defects per shift. Time to rework these defects was 3.5 hrs./day. The cost per year for this assembly line was calculated to be $30,000 (not including the cost of scrap boards). This example clearly shows the importance of recording DPMO values. Moreover, it would be possible to optimize line utilization, reduce cycle time, prioritize the deployment of constrained resources, and reduce cost of assembly, test, rework, scrap, and warranty. Manufacturers will be able to benchmark their DPMO rates against others in the industry.

This matter is of such importance, the 2002 iNEMI Roadmap specified cost reduction in the cost per I/O of each electronics assembly over time. A group of companies affiliated with iNEMI then started the Defect per Million Opportunities (DPMO) Project, which benchmarked assembly quality by calculating DPMO for a number of different products.

To calculate DPMO, the group followed two standards: IPC-9261 and IPC-7912. (Refer to these standards for proper definitions of terminology used here.) These standards define DPMO as the number of defects divided by number of defect opportunities, multiplied by 106. Two indices are used to combine the DPMO rates. Index1 includes all the component, placement, and termination DPMO rates and Index2 adds assembly DPMO rates to Index1.

DPMO Index1 = [dc+dp+dt]/ [oc+op+ot] x 106

The group defined 22 defects, divided into four categories:

• The component defect applies to components that do not match the component specifications. Only one defect can be counted per component. Examples: missing lead, bent lead, component damage, bare board defect, etc.
  
  
• The termination defect applies to any joint terminations that do not match the requirements established by the inspection standards. Only one defect can be counted per termination. Examples: bridging, open, solder balling, insufficient solder, etc.
  
  
• The placement defect applies to a presence or positioning defect of any component on the printed circuit board during the manufacturing process. Only one defect can be counted per placement. Examples: component misorientation, component missing, component misalignment, billboard, etc.
  
  
• The assembly defect category pertains to defects that affect the entire assembly and printing process. Only one assembly defect can be counted per assembly. Examples: insufficient paste, paste bridging, paste scooping, etc.

The opportunities are also categorized as assembly, placement, termination and component opportunities and defined by IPC-9061 and IPC-7912. Once the defects and opportunities are registered, and depending on other data recorded, the DPMO index can be calculated per package type, process operation type, etc.

The calculation of DPMO also can be used to monitor a manufacturing process and define if a process is statically controlled or not. In SPC, control charts allow users to monitor processes and eliminate sources of variations that can lead to out-of-control processes. When using DPMO values, the center line, upper and lower control limits can be calculated:

Center line = dpmo* = average of the dpmo values



This chart has a constant center line, but control limits vary as a function of sample size.

DPMO values were calculated from a total of more than 335,000 boards. A comprehensive database was built to collect the information. The architecture of the database can be found at inemi.org/cms/projects/ba/dpmo.html. DPMO indexes from the different companies were calculated. The data showed DPMO values ranging from seven to 150.

In addition, DPMO per package type and manufacturing process also were evaluated (Figure 1). From this, it can be observed that press-fit assemblies show the best DPMO values, while hand soldering is characterized by the worst DPMO.


Au. note: I would like to acknowledge the critical participation of the following companies: Agilent, Celestica, Delphi Automotive, Georgia Tech, Hewlett Packard, Mack Technologies, Motorola, Nortel, Plexus, Sanmina-SCI, Shipley, Solectron (now Flextronics) and Teradyne.

Ursula Marquez de Tino is process and research engineer at Vitronics Soltec, based in the Unovis SMT Lab (vitronics-soltec.com); umarquez@vsww.com.