What is an SOP? Standard operating procedures in a chip fab

SOP stands for standard operating procedure: a document that writes down each step of a task, so that different people on different shifts do it and get the same result. A public guide to writing SOPs defines one as a set of written instructions that document a routine or repetitive activity an organization carries out. In a chip fab, how a tool is started, how it is taken apart for maintenance and how a gas cylinder is changed each have their own SOP.

It's a bit like the instructions that come with flat-pack furniture: step one, take these two boards; step two, put in these screws. Follow them and anyone builds the same thing. The difference is that in a fab, one wrong step can cost an entire lot of wafers.

What do the three letters in SOP mean?

• Standard: everyone follows the same version.
• Operating: the everyday work being done.
• Procedure: a way of doing it, one step after another.

Chip packages have an SOP too: short for small outline package, a package with thin leads along two sides. That has nothing to do with the documents in this article (see "What do chip packages look like?").

How is an SOP different from a recipe or an OCAP?

A recipe is the set of numbers a tool runs by: temperature, gases, seconds (see "What is a recipe in semiconductor manufacturing?"). An SOP is written for people: which actions to take, in what order, and what to check. An OCAP is written for people too, but only covers what to do, in what order, once something is abnormal (see "What does OCAP mean in a chip fab?"). So people load and check lots by the SOP, the tool runs by the recipe, and when an alarm goes off, people follow the OCAP.

What do SOPs cover in a chip fab?

Every fab divides them differently, but these are common:

• Running a tool: starting it up, what to check before loading a lot, which numbers to look at after a run, how to shut it down. Some university cleanrooms publish their rules online, and they work this way too: each tool has its own SOP, and users must be trained on that tool before they may run it alone.
• Maintenance: during PM, what to take apart, what to replace, in what order to put it back, and what to check afterwards (see "What is PM (preventive maintenance)?" and "A tool has been fixed. Why does it still need a qual run?"). Workplace safety guidance points out that common maintenance errors happen when things go back together and start up again: parts put back wrong, a step left out, a safety feature left disconnected.
• Rare jobs that must still be done right: an emergency shutdown, or starting back up after one. Safety rules for highly hazardous chemical processes require operating procedures to cover these situations too.
• Chemicals and gases: changing a specialty gas cylinder, refilling a chemical, lockout/tagout before a repair. Each step, what to wear, and whether it takes two people are all in the SOP (see "Is working in a chip fab dangerous?").

Where does an SOP come from, and how does it change? Step by step

The names, forms and sign-offs differ from fab to fab. A common order looks like this:

1. Write it: the people who actually do the job write it with an engineer who knows the tool, one action per step, in the order it's done. That guide also asks for people who know the work to write it, short and unambiguous. If the writer doesn't know the floor, the SOP won't match the real work, and people fall back on their own habits.
2. Review it: someone else who knows the job reads it and approves it before it goes out. The writer is the person least likely to see the step they left out. Without a review, a wrong step simply becomes "the standard."
3. Train and check understanding: an experienced person teaches the job and watches the trainee do it; only after they can do it, with a training record kept, do they work alone. Reading is not the same as being able to do it, so the same kind of safety rules require the employer to make sure each worker understood the training and to record how that was checked. Skip this, and someone's first practice run is on real product.
4. Follow it and keep a record: the SOP is at hand while working, each step is ticked off as it's done, and numbers are written down on the spot. Memory is where steps get lost; Rules for making medicines likewise require written procedures to be followed and recorded at the time, with any deviation written down and justified. Without a record, the next shift can't tell how far the work got (see "What gets handed over at a chip fab's shift change?").
5. Speak up when something's wrong, and revise it through a formal change: anyone who finds a step that's wrong, missing or impossible raises it, and the SOP is revised through the set process: the new version is reviewed and approved too, followed by retraining. If the change touches a process condition, it also goes through change control (see "What is change control in a chip fab?"). The next person follows the new way only once the step on paper has changed. If everyone quietly does it their own way instead, the paper says one thing and the floor does another.
6. Pull the old version: as soon as the new version is issued, the old one is removed from the floor and the system, or clearly marked obsolete. The international quality management standard also requires documents to be available where they're used and changes to be controlled. With two versions around, someone will always pick up the old one.

Even when nobody raises anything, SOPs are reviewed from time to time to make sure they still match how the work is really done.

Why does "following the SOP" still go wrong?

Having an SOP doesn't mean nothing goes wrong. Four common reasons:

• A step is unclear or missing: after maintenance a valve has to go back to auto, but the SOP has no line for it. Every step was followed, and the valve is still on manual.
• An old version: the SOP changed, but the copy in hand is the old one. Public workplace safety guidance lists common reasons procedures aren't followed: they are wrong, out of date, hard to use, or not at hand when needed.
• A skipped step: in a hurry, or because the step seems pointless ("it's brand new, it shouldn't leak"). Research on human error finds that rules that can't be met and time pressure are common reasons people skip rules; after a while, nobody even notices they're skipping.
• A situation the SOP never imagined: an SOP describes the normal case. When something isn't covered, stop and call someone instead of guessing (see "What's the difference between a fab technician and an engineer?").

How does a good SOP make the right way easy?

• Short, clear, in the floor's own words: one action per step, written and checked with the people who use it.
• Where the work happens: posted next to the tool or on the cabinet door, not back in the office.
• The reason, where it matters: for example, what happens if a limit is exceeded and how to correct it, which those safety rules require operating procedures to include. Research on procedures also points out that people who don't understand the reason for a step sometimes decide it's pointless and skip it.
• Two people for the critical steps: one reads, one does. Flight-deck checklists often work the same way: one pilot reads the item, the other checks and answers.
• Let the system stop it, not only a reminder: making the wrong step impossible is called mistake-proofing (see "What does MO mean in a chip fab?").

When people keep skipping the same step, ask why first. Often it's the SOP that needs changing, not the person.

In fab24hr

Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab", swipe to the equipment engineer and tap "Etch" under the name. On the morning shift, "Chamber open", you swap the ESC (electrostatic chuck) in Chamber A of ETCH-03. Taking it apart and putting it back are each a step list, from lockout/tagout and waiting for the chamber to cool, to installing the new one, pumping down and a leak check. Each step shows how many minutes it takes, and Yun from Manufacturing keeps calling to ask, "When will it be up?" Some steps carry a "Shortcut" note, such as the leak check: "It's all brand new. It shouldn't leak." If you do the step, Kai says, "The leak check caught one. Brand-new parts are the ones that leak most." If you first tap "R&D line" on the card and play the vendor field service engineer, and on the evening shift, "Quarter end", ask the customer for one more day to test again, your service report says, "After an ESC swap, leak-check the fittings." Then this step list gains that line, and the leak check can no longer be skipped.

Swipe to the Facilities engineer and tap "Gases & chemicals" under the name. On the late-night shift, "Night watch", the acid refill valve for WET-02 was left on manual after maintenance, and the PM record has no line saying "switch back to auto when done." At 03:30 you can add that line, or leave it for the day shift. On the morning shift, "Cylinder change", Hui-Ju says, "The procedure is posted on the cabinet door. I read a step, you do a step." As the etch process engineer, at the 16:30 review meeting on the afternoon shift, the department manager asks, "The SOP clearly says to verify before running. So why did it still go wrong?" The conclusion: "So what needs to change is that step, not the person."

The companies, people, tool numbers, steps and minutes in the game are fictional. What an SOP is and how it's written and revised come from public, general sources; the actual formats and who signs differ from company to company. For other jargon, see Chip fab jargon in plain English.