SPC (Statistical Process Control)

Notes

Control limits describe what the process does; specification limits describe what the customer needs, and confusing the two is the classic SPC error.

Overview

Statistical process control (SPC) is the practice of charting a process's measurements over time against control limits worked out from the process's own normal variation, so that unusual changes are spotted and acted on before they produce bad material. In a fab, SPC runs on thousands of measurements, from film thickness and line width to tool parameters, and is the first line of defence against excursions.

How it works

Every process varies a little from run to run. SPC separates this common cause variation from special causes, such as a worn part, a contaminated chemical or a drifting sensor. A control chart plots each measurement or subgroup statistic, such as a mean and range, with a centre line and upper and lower control limits, usually three standard deviations from the mean, calculated from data taken while the process was stable.

Points outside the limits, or patterns such as a run of points on one side of the centre line or a steady trend, signal that something has changed. The response is set in advance in an out-of-control action plan: hold the lot, inspect the tool, check the measurement system, and release or scrap material once the cause is known.

Control limits are not specification limits. A process can be in control and still not capable of meeting the specification, which is what capability indices such as Cpk measure. The approach dates from Walter Shewhart's work at Bell Telephone Laboratories in the 1920s, and automotive suppliers follow AIAG and VDA guidance on SPC.

Examples

• A control chart of gate oxide thickness that flags a slow upward trend before any wafer goes out of specification.
• An etch tool held automatically after a chamber pressure reading falls outside its control limits.
• Charting wafer sort yield per lot to spot a sudden drop linked to one tool.
• Calculating Cpk for a critical dimension to show a customer that the process comfortably meets its specification.

Why it matters

With hundreds of steps per wafer and weeks of cycle time, a drift caught late can affect many lots. SPC engineers, process engineers, process control engineers and quality engineers run and respond to the charts, and automotive customers expect SPC on key characteristics as part of control plans.

In the chip lifecycle SPC is set up as a process moves from development into production, tightened during ramp, and runs continuously in volume manufacturing at the fab, assembly and test.

Related terms

• Cpk
• Ppk
• Control plan
• Six Sigma
• Excursion
• APC

Where it appears on the Hub

Used on 43 pages of the Hub. In each group, the pages that mention it most come first.