What does OCAP mean in a chip fab?

OCAP stands for "Out-of-Control Action Plan": when something on the line looks abnormal, such as a measurement outside the control limits or a tool that alarms and stops, the people on the line follow this procedure, written in advance: what to stop first, which lots to hold, what to check, and who to tell. The same four letters mean other things in other fields; this article is about how factories, chip fabs included, use them.

What do the four letters stand for?

Taken apart:

• Out of control: in a factory this doesn't mean a tool running wild. It means "not like usual" in a statistical sense: a number lands outside the range it normally falls in, or keeps drifting to one side.
• Action: the things to do.
• Plan: written down ahead of time.

Put together, it's a plan for what to do the moment you see a "not like usual" signal. An engineering statistics handbook describes it as a document that sets out the action to take once an out-of-control situation is detected, sometimes drawn as a flowchart that walks people through it step by step.

In the fab, people usually just say OCAP.

It's a bit like an earthquake drill at school: when the alarm sounds, nobody holds a meeting; everyone does the moves they practiced. An OCAP is the script a factory writes in advance for each kind of alarm.

What does an OCAP say?

Here's an example. A tool runs a monitor wafer every day: not customer product, just a wafer used to check how the tool is doing. Today's measurement shows far more particles than usual, past the control limit. The OCAP for that tool might look something like this (details differ from fab to fab; this is only an illustration):

1. Stop the tool, and don't load any more lots into it.
2. Put the lots it ran in that window on hold: they stay where they are and don't go on to the next step.
3. Check the number itself: was the data entered right, is the metrology tool off? Measure again if needed.
4. If the number really is out, notify the engineer for that tool, and say clearly what you saw and which steps you've already done.
5. Write down every step: what was done, and when.

Notice the order: stop and hold first, then check. The number might be a measuring mistake, but if the tool keeps running while you check, another lot may come out bad. Stopping the damage before anyone knows the cause is the same idea as the containment step in an incident report; the cause can be found a little later.

Textbooks split this kind of procedure into three parts: the signal that triggers it, the possible causes to check, one box at a time, and what to do once the cause is found. The order of the checks matters: the most likely cause, or the easiest one to check, often goes first. So it's less like a manual and more like an "if this, then that" map.

Why write it in advance instead of working it out on the spot?

Because trouble doesn't wait for a time when people are thinking clearly.

A chip fab runs around the clock (why: "Why do chip fabs run 24 hours a day?"). At night the line is run by shift crews, and engineers mostly take turns being on-call from home. When a tool turns red at three in the morning, the people on the floor are tense, and the engineer the phone just woke up isn't fully awake yet. Occupational safety guidance points out that the brain and body slow down at night and in the early morning, and that being overtired makes it harder to concentrate, so mistakes become more likely.

Writing it down in advance helps in three ways:

• No relying on memory: follow the steps on the page one box at a time, and none gets skipped. Pilots facing an emergency work through a checklist item by item for the same reason: when people are stressed and tired, memory is the weakest link, and a written list doesn't forget.
• No waiting: the people on the line can stop the tool and hold the lots right away, without waiting for an engineer to call back at every step. One university study did exactly this: it turned the causes of past abnormalities into an OCAP so operators could deal with them on the spot instead of waiting for an engineer.
• Everyone does the same thing: whichever shift or person meets the problem, the first step is the same. That also makes it possible to compare how it was handled this time with the next.

What if the OCAP doesn't cover it?

No OCAP covers everything. When it doesn't, that's the moment to call: phone the on-call engineer, say what you saw and what's been done, and let them decide the next step.

Afterwards, the new situation often gets added to the OCAP, so next time there's a ready answer. An OCAP isn't written once and left alone; it grows with experience.

How does an OCAP relate to SPC and SOPs?

These three abbreviations often show up together. Each answers a different question:

Abbreviation	The question it answers
SPC	Is the process behaving abnormally?
OCAP	It is. What do we do now?
SOP	How is each step done normally?

Start with SPC, Statistical Process Control: measurements are plotted point by point on a control chart to see whether the process is drifting. A point outside the control limits, or several points in a row drifting to the same side, is a signal. SPC does the noticing; what happens next is the OCAP's job. In one line: SPC tells you something is wrong, the OCAP tells you what to do next. The handbook puts it the same way: when a process is out of control, the engineer looks for the cause by following the OCAP tied to that control chart. SPC itself has its own article: What is an SPC control chart?

Being outside the control limits (OOC) doesn't always mean being outside the product spec (OOS); tap OOC and OOS for the difference. Both need action, and the OCAP sets the order of that action.

Apart from a control chart, the tool itself can set off an OCAP: an alarm, a red tower light, a tool that has stopped. For what red and yellow mean, see "What do the red, yellow and green lights on a tool mean?"

Then there's the SOP, the Standard Operating Procedure: it covers normal work, how each step is done so the result is the same whoever does it. The OCAP covers the abnormal times. Both are written in advance for everyone to follow; one is for every day, the other for the few minutes when something goes wrong. For how technicians work with them day to day, read "What's the difference between a fab technician and an engineer?"

In fab24hr

Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab" and swipe to Manufacturing. Which of the four shifts you get depends on the time you open the site. On the late-night shift, at 02:47, Chamber A of the etch tool ETCH-03 turns red with a hot lot wafer stuck inside. Hao, on the night shift, says a reset and another try usually gets it through; the OCAP says: tool alarm, notify the engineer first. Follow the OCAP and call the on-call engineer, or try one reset first: it's your call. Once the call goes through, Lin asks you to mark Chamber A down, leave the stuck wafer where it is, do a lot split for the first 6 and send them to wet clean before their Q-time runs out, and put the rest of the lot on hold.

The same events look different from another role. As the quality engineer, at ten on the morning shift you take the OCAP to the etch area and check the night against it step by step: three items don't match, and one of them is something the OCAP never covered, what to do when a wafer is stuck in the chamber. As the etch process engineer (swipe to the process engineer and tap "Etch" under the name), after the 16:30 review meeting on the afternoon shift, you find someone to add that case to the night shift's procedure.

The company, people, tools and procedures in the game are fictional; what an OCAP is and why it's written in advance come from public, general sources. For other jargon, see Chip fab jargon in plain English.