A client came to us with concern about poor right-first-time product test results from their contract manufacturer. They had reason to be concerned: 10% yield is not good! Clearly, the manufacturer was having trouble. Our task was to identify the root causes and recommend how to fix the problems.
The design company initially asserted that the SMT process was not good enough. However, we were to disprove this.
The process flow to the first test rig was SMT, firmware load and then test on a pogo-pin test rig. All components were standard, off the shelf, and from reputable suppliers. Quality controls were well implemented covering purchasing, receipt, handling and storage. In addition, our initial assessment of the SMT process indicated that the SMT process was also well under control.
However, firmware load files, whilst being correctly “pulled” from the design company, did not have record of having been validated. Therefore, we prompted investigation of the firmware itself.
Specialists identified that the FPGA manufacturer support centre held knowledge of issues and work-arounds with regard to timing and other issues involved in firmware load, and these strongly influenced loading success.
Accordingly, the firmware was improved and, after further validation of firmware, a much-improved manufacturing yield was obtained.
This issue was clearly one of poor design team processes.
We also found poor or omitted design transfer to the manufacturer, and of poor or non-existent validation of test fixtures, including of the pogo-pin tester noted above. Gauge R&R was therefore initiated.
At a higher level, it became apparent that the design company did not understand or have in place procedures to ensure comprehensive and validated transfer of knowledge to its contract manufacturer. Without clear guidance, consistent manufacture cannot be expected.
And without consistent manufacture, reliability is lost. Not surprisingly, I always include consideration of manufacturing capability when working with clients.