What do chip packages look like, and what do names like BGA and QFN mean?

What a chip package looks like comes down to two things: where the pins come out, and how they are soldered to the circuit board. Going by those two things, common packages fall into roughly five kinds: ones that plug into holes in the circuit board, ones with thin leads sticking out of two or four sides, ones with no leads and only metal pads underneath, ones with a whole grid of solder balls underneath, and ones about the same size as the chip. Names like DIP, QFN and BGA describe these shapes.

For why a chip has to be packaged at all, start with "What is chip packaging?".

How are package types sorted?

A packaged chip has two layers of connections. The inner layer connects the chip to the package. The outer layer connects the package to the circuit board, through metal leads, metal pads or solder balls. This article is about the outer layer: what a package looks like from the circuit board's side.

Plugged into holes in the board: DIP

A small black rectangle with a row of leads along each long side, bent downward, a bit like a tiny centipede. The leads go into holes drilled in the circuit board and are soldered on the other side of the board. They are about two and a half millimeters apart.

Early circuit boards mostly had their parts mounted this way: pushed through holes, then soldered. The leads are thick and far apart, so they're easy to solder by hand, and the chip can also sit in a socket, which makes it easy to swap.

New products rarely use it now: the holes have to be drilled first, the parts take up room, and machines are slower at mounting them. One place you'll often see it is in school lab classes, plugged into a breadboard that needs no solder and can be used again.

Thin leads out of two or four sides: SOP and QFP

These don't go through the board. They sit flat on solder pads on the board's surface and are soldered there, which is called surface mount. The leads come out of the sides, bend down, then flatten out, like a seagull's wings. If only two sides have leads, it's an SOP; if all four do, it's a QFP.

With no holes to drill, parts can be smaller and packed closer, machines place them quickly, and the whole board is heated at once to solder them ("How are chips soldered onto circuit boards?"). Most electronic parts today are surface mount. Using all four sides fits about twice as many leads: on a QFP the leads are a millimeter apart or less, the tightest under half a millimeter, and a big one has a few hundred.

Because the leads are on the outside, you can check the solder with a magnifier. Chips with a small to medium number of pins often use this kind.

No leads, just metal pads underneath: QFN

A small, flat square with no leads sticking out of its sides. Turn it over, and along the four edges underneath is a row of small metal pads, soldered straight onto the circuit board. In the middle there's often a large exposed patch of metal, soldered to the board as well, so the chip's heat can flow down through it. A version with pads on only two sides is often called DFN.

With the leads tucked underneath, the whole part is small and thin, and signals have a shorter path. Like the QFP, it's built on a metal lead frame, and it's cheap to make. In small electronic products, chips with few pads use it a lot.

The downside is that the solder joints are hidden underneath: it's hard to solder by hand, and hard to see whether the joints are good.

A grid of solder balls underneath: BGA

Turn it over, and the entire bottom is covered with tiny solder balls, like the squares of a chessboard. The balls are about a millimeter apart, the dense ones just a few tenths of a millimeter, and one package can have hundreds to more than a thousand connections.

Not just the edges but the whole bottom carries connections, so it can connect a lot; the paths are short, and heat passes easily into the circuit board. During soldering the balls melt, and surface tension often pulls a slightly misplaced package back to the center. Chips with lots of connections, such as processors and memory chips, mostly use it.

The price is that every joint is hidden underneath, out of sight, so it's checked with machines such as X-ray. And solder balls can't bend a little the way thin leads can: when the circuit board and the package expand by different amounts as they heat up, all the strain lands on the balls.

About the size of the chip: chip-scale packages

Go smaller still, and the package is only a little bigger than the chip. This is called a chip-scale package (CSP). A common definition is an area no more than about 1.2 times the chip's. Underneath there are usually solder balls too.

One step further, the wafer isn't cut first: the solder balls are built right on the whole wafer, and it's cut apart only at the end. This is wafer-level packaging, and the package is the same size as the chip.

It's the smallest and thinnest, but all the balls have to fit under the chip itself. The chip and the circuit board expand very differently with heat, and the bigger the chip, the harder the outer balls get pulled. So it's mostly used for small chips with few pads.

What do the letters in the names mean?

Name	Stands for	In plain words
DIP	Dual In-line Package	Two (dual) rows of leads, each in a straight line (in-line)
SOP	Small Outline Package	A package with a small outline
QFP	Quad Flat Package	Leads on all four (quad) sides, and flat
QFN	Quad Flat No-lead	Four sides, flat, no leads sticking out
BGA	Ball Grid Array	Solder balls laid out in a grid, an array
CSP	Chip Scale Package	About the scale of the chip

The final P is just "package". An extra letter in front usually marks a variant: T, for example, means thin. The same shape often goes by several names.

Why do more pins mean a grid of balls underneath?

Let's work it out with made-up numbers. Take a square package 2 centimeters on a side:

• Leads around four sides, one every 0.5 millimeter: 40 per side, 160 in all.
• Solder balls across the whole bottom, one every 1 millimeter: 20 rows of 20, which is 400.

The balls are twice as far apart as the leads, yet there are two and a half times as many: the edges only grow with the side length, while the bottom grows with the area. It's the same reason flip chip fits more connections than wire bonding.

And the leads can't keep getting closer together. The thinner and tighter they are, the easier it is for solder to join two neighbors, and that's a short circuit. So chips with few pads that need to be cheap do fine with leads on four sides or a QFN; once there are many connections, it's BGA.

Does the outside shape have anything to do with how the inside connects?

Yes, but they're not locked together. A common pairing: DIP, SOP, QFP and QFN are built on a metal lead frame, with the chip glued in the middle and connected to the leads or pads around it by fine wires (wire bonding); a BGA uses a small board with the chip on top, connected by either wire bonding or flip chip. For the inside, see "What is wire bonding?" and "What is flip chip?".

Flip chip plus BGA means two layers of balls: tiny bumps under the chip, and the BGA's much bigger solder balls under the board.

What does it look like when several chips are packaged together?

A chip packed into a small module with resistors and capacitors is called a SiP, and from the outside it looks like an ordinary chip ("What is SiP?"). Several chips side by side or stacked is the advanced packaging in the news (see The news says "advanced packaging". What is it?). However many chips are inside, the layer that connects to the circuit board is often still a grid of solder balls underneath.

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 walks you along the assembly line: Glowmere is packaged as a small SiP module. The chip is glued onto a small board and wire bonded, then goes through molding, is cut into single modules and goes on to final test. The game doesn't say whether the bottom of the Glowmere module has solder balls or metal pads.

Take the wire bonder apart in the codex's Machines list, and there's a copper lead frame: a chip glued in the middle, with a row of leads around it. That's the frame used for packages with leads on four sides or QFNs; the game doesn't say which kind it ends up as.

Tap "Clock in", choose "A day in the fab", then on the "Your role today" card tap "Test-and-assembly house" and swipe to the customer liaison. You get a shift by the time on your device. On the late-night shift, the customer's production line calls to say the orientation on the shipping drawing doesn't match their machine's settings, and you check the "Pin 1" mark in each pocket of the tape and reel. On real packages, a small dot or notch usually marks where pin 1 is.

Names like BGA, QFN and DIP never appear in the game. The test-and-assembly house, the chip and the people in the game are fictional; the package shapes follow public, general sources, and the size and spacing used to count pins are made-up numbers. For other jargon, see Chip fab jargon in plain English.