What is tape-out, and what happens after a chip design is handed off?

Tape-out (also written tapeout) is the moment a finished, checked chip design is handed off to be made into photomasks. A photomask is the "negative" a circuit is printed from; once the masks exist, a fab can build the circuit layer by layer on a wafer. Changing the design after that means making new photomasks: expensive, and months of extra waiting. So for the people who design a chip, tape-out is the biggest day of the whole project.

Why is it called tape-out?

The usual explanation is that the final design files were once stored on paper tape, and later on reels of magnetic tape, and handed over; the name stuck. Others say that even earlier, the enlarged artwork for a photomask was laid out by hand with strips of black tape on clear sheets. Both stories are told. The files have long since been sent by computer, but the name never changed.

Where does tape-out fit in a chip's life?

A chip roughly goes like this: design → tape-out → masks → wafer start and months in the fab → packaging and testing → shipping. Tape-out is the handoff line between design and manufacturing. Before it, the chip exists only on a computer; after it, it starts to become a real thing.

For how a design gets to tape-out step by step, see "How are chips designed?". For how a chip design company works with a fab, see "How does a chip design company work with a foundry?".

What happens before and after tape-out, step by step?

Step 1: Final checks and sign-off

• What: Special chip design software ("What is EDA?") goes over the whole design again: does any shape break the fab's rules, is the drawn layout the same circuit as the original, do signals arrive in time? The function has to be verified too ("What is chip verification?"). When everything passes, the people responsible approve the release. This gate is called sign-off.
• Why: After tape-out, the design is fixed. It can't simply be changed any more.
• Otherwise: a missed bug shows up only when the chips come back months later, and the only fix is new masks.

Step 2: Hand off the design files

• What: The shapes of every layer go out as files. The full set of every layer's shapes, stacked together, is called the layout, and each layer later needs at least one photomask.
• Why: The masks are made from these files, and every pattern later printed on the wafers comes from them.
• Otherwise: if a layer is left out or the wrong version goes out, the masks won't match the design that was checked.

Step 3: Change the patterns so they print right

• What: For the finest layers, the patterns are first changed on a computer. This is called OPC (optical proximity correction).
• Why: The lines on a chip are narrower than the wavelength of the light used to print them. Printed as drawn, corners come out rounded and line ends come out short.
• Otherwise: what prints won't match the design, and things that should connect won't. See "What is OPC?".

Step 4: Make the masks

• What: A mask shop writes the masks one by one from the corrected patterns, then inspects and repairs them. Each layer needs at least one, and a full set often runs to dozens.
• Why: The pattern on a wafer is printed by a scanner, which shines light through the mask and shrinks the image.
• Otherwise: one defect left on a mask prints on every wafer. See "What is a photomask?".

Step 5: Wafer start

• What: Once the masks arrive, the fab puts the first wafers on the line. This is the wafer start. A new chip's first lots are a pilot run.
• Why: Tape-out hands off a design; only at wafer start do real wafers begin. You have to make it before you know whether the design and the fab's process get along.
• Otherwise: if you go straight to mass production and the design has a bug, every lot is wasted. For how tape-out differs from a wafer start, see "What is a wafer start?".

Step 6: Wait months for the chips

• What: The wafers move station by station through hundreds of steps. After that, they're usually cut up and packaged before the first chips reach the designers.
• Why: A chip is built layer by layer, and at every station the wafers wait in line for a tool. The common industry figure is about three months on average just in the fab ("How long does a wafer take to make?").
• If you try to rush it: even with priority at every station, the wafers only spend less time in line; none of the hundreds of steps can be skipped. So the day of tape-out more or less decides when the first chip arrives.

Step 7: Test, and tape out again if something's wrong

• What: The chips are hooked up to test equipment and boards, and checked item by item against the spec ("How are chips tested?").
• Why: However much you simulate on a computer, it's never quite the same as the real chip.
• Otherwise: problems turn up only in the hands of the people using it. When testing finds a bug, the design is changed and taped out again, a new version often called a respin. Many chips, it's commonly said, go through several.

Why is tape-out such a big deal?

Because once a mask is made, its pattern can't be changed; the only way to change it is a new mask. Each redo costs two things at once:

• Money: a full set runs to dozens of masks, each drawn extremely fine with not a single defect allowed, and they're expensive ("Why are chips so expensive?").
• Time: once the new masks are ready, the wafers start over from the first station, and that's months again.

So before tape-out, design teams would rather check one more time, and tape-out day is often one a team celebrates. And because masks are expensive, university research, or designs that need only a few chips, often share one mask set with other designs: they're taped out together, made on the same wafers, and split the cost. This is called a multi-project wafer (MPW).

Does a small bug mean starting over?

Not always. At the bottom of a chip are the transistors; above them are layer after layer of wiring ("How is the wiring inside a chip made?"). The layers that make the transistors have the finest patterns, and their masks cost the most.

So many designs scatter some spare little circuits that aren't used yet across the chip before tape-out. If a small bug turns up afterwards, only the wiring above the transistors is changed, hooking a spare circuit in. The bottom masks are used as they are, and only the wiring-layer masks that changed are remade, which costs less than a whole new set. But only so much can be fixed this way; if the change is too big, the whole set has to be redone.

In fab24hr

There's no chip-designer role in the game; the main story, The life of a chip, starts after tape-out. The first time you play it, you see the prologue, "Tape-out": half a year earlier, "Glowmere", the first chip a small hearing-aid company has designed itself, is finished and checked, and goes to the fab to become a set of photomasks. In the codex, on The life of a chip page, the chip passport starts with Tape-out; and when you finish the finale, the timeline on the last card starts with Tape-out too.

Chapter 1, "How to Split 12", is the day before wafer start. You're the R&D integration engineer on the R&D line, and you decide how to split the first 12 wafers. Chapter 3, "The shape of a failure", is a month or two after wafer start, and the next mask set is finalized next week: you decide whether to leave two spots inside the chip for your own test keys, though the chip's area belongs to the customer.

The companies, chip and people in the game are fictional; what tape-out is and the steps around it follow public, general sources, and every company does it differently. For other jargon, see Chip fab jargon in plain English.