A mother brakes and stops her car as she drops her children off for dance lessons. At the time, she doesn’t notice anything out of the ordinary, but when she takes her car in for her regular maintenance appointment, the mechanic performs a diagnostic check and discovers that the car’s main brake system has failed. failure due to a faulty brake controller without anyone noticing. Fortunately, the car was able to successfully shut down due to vehicle system redundancies, and dealer diagnostic testing confirms that since that first chip failure, another has not occurred. The braking systems behave normally.
Following this, the dealer sends the brake failure information to the manufacturer, where an analyst notes that in the last 60 days, and across the country, six other brake failures traced back to the same controller system have been reported. for the same brand. and model. In each of these situations, the back-up system successfully brought each car to a complete stop. And, as in the case of the mother who dropped her children off at dance class, the analyst reviews the sample reports for those six other failures and determines that each is isolated and non-recurring.
So what happened, and what may be the common factor shared by these otherwise isolated events?
To find those answers, the manufacturer starts running a Failure Modes and Effects Analysis (FMEA) on the brake systems for the particular make and model, and they do it, in part, using analysis with genealogy and traceability capabilities. Based on the results, the manufacturer will determine if a general recall is necessary.
After a thorough analysis, the manufacturer identifies that the defective chips came from the same supplier. Additionally, the chips come from a one-time shipment that the company had received from the system parts supplier. Once the parts supplier is notified, they perform their own analysis and determine that several components of the six brake systems were manufactured at approximately the same time at the same overseas semiconductor plant and that the chips from all six looked good. during the first inspection. In fact, the performance and electrical results of the faulty chips were within specifications and matched the expected distributions.
The factory is approached to determine if there are any unusual signals surrounding the materials used in the construction of the faulty brake systems. Beyond the normal 100% pass checks, outgoing visual quality assurance, and electrical test structure (WAT) wafer acceptance testing, no other online data is directly recorded on this material . In this case, this is where the analysis stops due to a lack of direct data on the construction of the faulty chips. But that doesn’t fix the problem.
Beyond the electrical performance characteristics of the parts in question, semiconductor factories generally do not capture sufficient or representative online metrology data for all the parts they manufacture in their facilities.
In today’s manufacturing environments, performing part-level analyzes on a process-by-process basis is impossible given the most common measurement sampling strategies, strategies that have been configured at process control and maintenance purposes (Figure 1). These monitoring and control strategies are woefully inadequate when applied to analyzing products for parts-per-million traceability. Even with good data extrapolation and expansion packages, error bars on all representative data created by these approaches would create insufficient results to draw conclusions. To better understand why data extrapolation doesn’t work, consider the multiple sources of error that naturally occur in manufacturing. Extrapolated programs must account for measurement tool variation (repeatability and reproducibility of instrument) and film variation across wafer, batch, batch-to-batch, and tool-to-tool . Accounting for these errors from a practical perspective means that any extrapolated value could have error bars consuming more than 50% of the allowed process variation window.
Fig. 1: A highly reliable system requires knowing the origin of each component. An easily accessible genealogy is a prerequisite.
Although the initial investigative event described above is fictional, the described approach to isolating the problem is a likely scenario, as is the lack of data needed to perform full diagnostics. In the future, manufacturers, especially those supplying parts for the automotive and medical industries, will have choices. Collectively, they may decide to continue to design solutions with inherent redundancies and create ever more restrictive Guard Band and Part Average Test (PAT) systems to weed out questionable parts and mitigate failures when they occur. in the field. Or they can increase the frequency with which they perform metrology in their manufacturing operations to ensure that they have enough data for each unit produced in their factories. In this situation, the days when sampling 13 points on a wafer, two wafers in a batch, and one batch in 20 would be sufficient would be over.
Manufacturers today need to create metrological sampling plans for device analysis, making this data available and traceable to downstream component manufacturers and customer-facing system vendors. In addition, linking these more comprehensive metrology coverage plans to the results of Defect Detection and Classification (FDC) software available from process tools, as well as maintenance records and source chemicals and gases process (GAC), will ultimately provide diagnostic analysts with a comprehensive and actionable analysis. FMEA.
Ultimately, it will be the end customer who comes up with the best solution.
Mike McIntyre is Director of Software Product Management at Onto Innovation.
#Metrology #sampling #plans #essential #device #analysis #traceability