What is equipment uptime, and how is a tool's up and down time counted?

Equipment uptime is how much of a tool's time goes to production. It's also called availability. A tool that can run product is up; one that can't is down. Availability comes from recording which state the tool is in for every stretch of time and turning that into a percentage. Fabs calculate it a little differently, as we'll see below.

Fab tools are very expensive and lose value every day whether they run or not, so fabs watch this number daily (see "Why do chip fabs run 24 hours a day?").

What states can a tool's time be in?

A public international standard sorts every minute of an observation period into six states that never overlap:

State	Meaning
Productive	Running wafers that will ship, as scheduled
Standby	The tool is fine and able to run, but there are no lots right now
Engineering	The tool is fine, but it's being used for experiments or to try new conditions
Scheduled downtime	A stop planned in advance, such as maintenance
Unscheduled downtime	A stop nobody planned: a breakdown
Non-scheduled time	Time the tool was never meant to be used, such as shutdowns, holidays, or a new tool still being installed

Think of a taxi's day: carrying a passenger is productive, waiting empty at the taxi stand is standby, a visit to the garage for servicing is scheduled downtime, a breakdown on the road is unscheduled downtime, and the driver's day off is non-scheduled time. Maintenance is PM; for why tools stop on a schedule, see "What is PM (preventive maintenance)?"

When does a breakdown stop counting as downtime? In fab24hr, only once the tool can really run product: it may be repaired, but until the qual run after the repair passes, no lots can go in, so that time counts as downtime too ("A tool has been fixed. Why does it still need a qual run?").

Why split it so finely? Because "not producing" has different causes, and each needs a different fix: a breakdown goes to Equipment, badly placed maintenance means changing the maintenance plan, and no lots means talking to whoever schedules them. Log it all as "stopped", and you can't tell where to start.

How is availability calculated, and why do fabs get different numbers?

Here is an example with easy numbers (invented, not from any fab): in one 24-hour day, a tool spends 18 hours productive, 2 hours on standby, 1 hour on engineering, 2 hours in maintenance and 1 hour broken.

• Counting only the time it runs lots: 18 ÷ 24 = 75%.
• Counting the time the tool is healthy: productive + standby + engineering = 21 hours, 21 ÷ 24 ≈ 88%.

Same tool, same day, two numbers. The second is often called uptime; availability and utilization are common words too, but what they cover varies from place to place. The denominator can differ too: some use all the time, others first subtract the time the tool was never scheduled to be used. So before comparing two tools or two fabs, ask: "How was it calculated?"

For breakdowns alone, two more numbers are common:

• MTBF (mean time between failures, how often it breaks): the hours the tool worked normally over a period, divided by the number of failures. Only unplanned failures count as "breaking once"; scheduled maintenance doesn't.
• MTTR (mean time to repair, how long each fix takes on average): the total hours of breakdown downtime, divided by the number of failures. Some count only the hands-on repair; others count from the moment the failure is found until the tool can run product.

The longer the MTBF and the shorter the MTTR, the less time breakdowns eat.

How do you raise availability? Step by step

Step 1: Record the state of every stretch of time

What: whenever the tool changes from one state to another, record the time and the reason; many state changes are reported by the tool itself to the fab's computer system ("What is MES?"), while what a stopped tool is waiting for often has to be filled in by a person. Why: you can only see where the time went if there's a record. If you skip it: all you know is "availability fell", not whether the tool broke, was in maintenance or had no lots; record the wrong state, and the number looks better while the problem hides.

Step 2: Sort the time and start with the biggest piece

What: total the hours in each state over a month and sort them from largest to smallest; then split the breakdown time by cause, such as which part or which kind of alarm ("What are tool alarms and warnings?"). Why: people and time are limited, so the biggest piece comes first; and the busiest step on the whole line, the bottleneck, comes before the others, because one hour lost there is one hour lost for the whole line ("What does production control do in a chip fab?"). If you skip it: effort goes into small things, and the whole line's output doesn't change.

Step 3: Breakdowns: make them rarer (longer MTBF)

What: find out why it breaks and fix the cause: replace parts that fail easily, change what the maintenance covers, and watch for warning signs, such as a yellow light that keeps coming back or a number slowly drifting in FDC. Why: a fault that isn't removed keeps coming back. If you skip it: you just press reset to get the tool running, and next time it may stop in the middle of the night again ("What is root cause analysis?").

Step 4: Breakdowns: fix them faster (shorter MTTR)

What: break one stop into its pieces (noticing it, calling someone, waiting for them to arrive, finding the cause, waiting for parts, the repair, the qual run) and see which takes longest; keep common parts in stock, write the procedure as an SOP, and have someone on-call at night. Why: part of the downtime is spent waiting, not repairing. If you skip it: the longer it's stopped, the more lots collect behind it, and some lots may run past their Q-time.

A textbook example: two tools have the same availability, but one stops briefly and often, the other rarely but for a long time. The lots behind the second one wait much longer on average, because one long stop makes them all collect at once ("How long does a wafer take to make?").

Step 5: Maintenance: pick the right time, do just enough

What: the maintenance date can usually move a little, so choose a time when there are few lots and other tools at the same step can run; have parts and tools ready beforehand, and the maintenance itself gets shorter. Why: maintenance is downtime too, and badly timed, it blocks the lots behind it just the same. If you skip it: maintain too often and maintenance eats the time; too seldom and the tool breaks more ("What is PM (preventive maintenance)?").

Step 6: Waiting for lots: look at it with the people who schedule them

What: when the tool is fine but has no lots, the cause often isn't the tool: the step before is slow, or lots were sent to another tool. This is something to look at with dispatch in Manufacturing and with production control ("What does the manufacturing department do in a chip fab?"). Why: Equipment can't repair standby time away. If you skip it: when a bottleneck tool sits waiting, the whole line simply makes less; for a tool that isn't the bottleneck, though, some spare time is normal.

Is higher availability always better?

Not always. The number can be made to look better by cutting corners, and the cost comes back later:

• Skipping maintenance: fewer stops this month, then more breakdowns and longer stops.
• Skipping the qual run: the tool goes back to production sooner, but if it isn't really fixed, a whole lot of a customer's wafers comes out wrong and needs rework, or even scrap.
• Loading every tool to the full: the closer a tool gets to never being free, the longer lots wait.

The industry also watches OEE (overall equipment effectiveness): on top of "is it running", it multiplies in "is it running fast enough" and "is what it makes good".

In fab24hr

Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Production fab", swipe to the process engineer and tap "Etch" under the name.

• Morning shift, "Firefighting": before the morning meeting, you ask Kai from Equipment to push Chamber B's 10:00 maintenance to tomorrow: "Chamber A is down. If both chambers stop, the whole tool has no capacity." At the 08:30 morning meeting, two cells on the production board are red: Moves vs. plan at 82%, and "ETCH-03 availability: Chamber A down, since 02:47". The chief asks you: "What time will it be back up?" Once Chamber A is back up, that cell changes to "Chamber A back up, Down 9.8 h today".
• Tap the etch tool ETCH-03: its card has the table "The six states of tool time", and today's red light counts as "Unscheduled downtime".
• Open your phone: the row of notes at the top uses the same six states (UP, IDLE, ENG, PM, DOWN, OFF), and you can pick one for yourself.

The companies, people, tool numbers and figures on the board in the game are fictional; what availability is and how tool time is divided come from public, general sources, and the actual state names and calculations differ from fab to fab. For other jargon, see Chip fab jargon in plain English.