What is wire bonding, and how does a chip connect to the outside?

Wire bonding is the step where metal wire finer than a hair connects the pads on a chip, one wire at a time, to the outside: to a small board inside the package, or to a metal frame called a lead frame. It is one step of packaging, after die attach and before molding (every step of packaging is in "What is chip packaging, and why does a finished chip still need a package?").

Why does a chip need wires to connect?

A chip talks to the outside world through small metal patches on its surface, called pads. Power goes in and signals go in and out through them. A pad is often only a few tens of micrometers across (a micrometer is a thousandth of a millimeter), about as wide as a human hair, or smaller.

The solder joints on a circuit board can't be made that small. So the chip is first stuck onto a board or lead frame, and a fine wire runs from each pad to it: that is wire bonding. The board or lead frame then turns those connections into much bigger metal pins or solder balls on the outside.

The wire itself is commonly only twenty to thirty micrometers thick, a fraction of a hair.

How is one wire bonded?

At the very front of a wire bonder is a thin tube called the capillary, with the wire coming out through its middle. Bonding one wire goes roughly like this:

1. Make a ball: the machine fires a high-voltage spark at the wire tip and melts a little of it. Surface tension pulls the melted metal into a round ball, a bit wider than the wire.
2. First bond, on the chip: a heated stage underneath has already warmed the chip to over 100 °C. The capillary presses the ball onto the pad while ultrasound makes the capillary vibrate in tiny, fast strokes. With heat, pressure and ultrasound together, the metal of the ball and the pad weld to each other without melting into liquid. The ball is squashed flat, like the head of a nail.
3. Draw an arc: the capillary lifts and follows a set path to the board, paying out wire as it goes, so the wire forms an arc in the air, called the loop.
4. Second bond, on the board: the capillary presses the wire straight onto a pad on the board or lead frame. No ball this time: the wire is flattened into a small, thin spot.
5. Break off and get ready: a small clamp closes on the wire, the capillary lifts, and the wire snaps at the thinnest, flattened point. A short tail of wire is left under the capillary, and the next ball is made from it.

It's a lot like sewing: the first stitch goes into the chip, the second into the board. An automatic wire bonder can make several wires a second.

If the loop is too low, the wire can touch the edge of the chip and short. Too high, and later, during molding, the flowing plastic can push it sideways into the wire next to it.

What's the difference between ball bonding and wedge bonding?

What you just read is ball bonding: the first bond is a ball, and most chips are bonded this way. Because the ball is round, the wire can leave it in any direction, so it's fast.

Wedge bonding makes no ball. A tool with a flat, wedge-shaped tip presses the wire flat onto the pad, so both ends are flat. The wire feeds in at an angle from the back of the tool, so the second bond can only land straight ahead along the wire; to change direction, the tool or the part has to turn, which is slower. The upside is a smaller bond, so very closely spaced pads still fit. Wedge bonding often uses aluminum wire and can bond at room temperature with only ultrasound and pressure. High-current power devices even use aluminum wire hundreds of micrometers thick.

Gold wire or copper wire?

Gold was the traditional choice. It doesn't oxidize, so it forms a clean ball in ordinary air, and it's fairly soft and easy to bond. The drawback is price.

So many products now use copper wire: much cheaper, and a better conductor than gold. But copper is harder, and it oxidizes when the ball is formed. So the ball is made in a protective gas (often nitrogen with a little hydrogen), and the settings have to be held tighter: a bit more force and energy than for gold, but never so much that it damages the chip under the pad.

How do you know a bond is strong enough?

The trouble is that a wire that isn't bonded firmly usually still conducts right after bonding, so an electrical test may not catch it. After some use, as the temperature goes up and down, the wire, the chip and the board expand by different amounts and pull on one another, and the bond slowly works loose. So bonds also get a physical check, with real force:

• Pull test: a small hook catches the wire near the middle of the loop and pulls upward until it breaks, recording how much force it took and where it broke. A break in the middle of the wire means the wire itself reached its limit, which is normal. If the whole ball lifts off the pad, it wasn't bonded firmly enough. A pulled wire can't be used, so a few parts are usually sampled at regular intervals.
• Ball shear test: a small blade, set just above the pad, pushes the flattened ball off from the side, measuring how much force that takes. The pull test looks at the whole wire, the shear test directly at how well the ball sticks to the pad, and the two are often used together.

Public military test methods and industry standards set out how to do both. After molding, the parts are also X-rayed to check whether any wires were pushed out of place.

Why does a bond fail to stick?

There are two common reasons. One is a pad surface that isn't clean: a trace of something you can't see is enough to keep the ball from sticking. That's why many assembly lines clean the surface with plasma before wire bonding. Left too long after cleaning, though, the surface picks up contamination from the air again, so there is often a time limit between cleaning and bonding, the same idea as Q-time in a wafer fab (What is Q-time?).

The other is the machine's settings: the force, the strength and length of the ultrasound, the temperature. Too weak and the bond won't hold; too strong and it damages the pad. The capillary also gets replaced after a while, and a new one means tuning the settings again.

Why don't all chips use wire bonding?

Wire-bond pads usually sit along the four edges of the chip. The more connections a chip needs, the more crowded those edges get, and every wire has some length, so signals take a detour on the way out.

So chips with a great many connections, or that run especially fast, often use flip chip instead: small bumps are built on the pads, the chip is turned upside down, and it's soldered straight onto the board. The whole face of the chip can hold connections, and the path is much shorter. The advanced packaging in the news, when it places several chips side by side on a thin slice called an interposer, also turns them upside down and solders them on with tiny bumps (The news says "advanced packaging". What is it?).

Wire bonding's strengths are that it's cheap, mature, and more forgiving when a design changes. So for chips that don't need many connections or top speed, wire bonding is still very common today.

In fab24hr

In chapter 11 of the main story, The life of a chip, "The Ball Lifted", you are the assembly process engineer at a test-and-assembly house. Shu-Hui, the line leader, walks you from dicing past SMT and die attach to the row of wire bonders. Guo, the wire-bond technician, brings out six close-up photos from today's pull-test sample check on WB-03 and asks you to tap the wrong ones: a wire that broke in the middle is normal, a ball that lifted off is not. The capillary was replaced three days ago, and it could also be the pad surface of a new product's first batch. Stop the machine, change the capillary and retune it, and re-sample every tray bonded that day, or keep it running with more frequent sampling: the decision is yours.

Tap "Clock in", choose "A day in the fab", tap "Test-and-assembly house" on the "Your role today" card and swipe to the assembly process engineer. You get one of four shifts by the time on your device: late at night, boards fresh from plasma cleaning are waiting for the wire bonder, and the ones that waited too long need cleaning again; in the morning, chips coming off the dicing saw have small nicks along their edges, and the blade may need changing; the afternoon is the chapter 11 shift; in the evening, the plastic in molding may push fine wires out of place, and you find them in the X-rays of eight mold cavities.

The test-and-assembly house, the "Glowmere" chip, the people, the machine numbers and figures like "bond within so many hours of cleaning" are fictional settings; how wire bonding is done and checked follows public, general sources. For other jargon, see Chip fab jargon in plain English.