In this video we're going to look at a brief history of the semiconductor packaging industry. Packaging refers to the integrated circuits, carrier, and enclosure. It protects the silicon die inside from physical damage while also allowing it to be connected to other devices. The industry has long been overlooked by the sexier work of wafer fabrication, but it's important and we should know about it. Let's go.
But first let me talk about the agenometry Patreon. Early access members see new videos and selected references for them before the release of the public. They help support the videos and I appreciate every pledge. Thanks and on with the show.
ICs are precious little snowflakes. They need protection from the real world, which is full of damaging influences. Particles can get in and interfere with operations. Moisture in the air can cause their metals to corrode. High temperatures generated during operations can cause the IC to degrade or even fail outright. Vibrations or jolts can damage the chip's capabilities and so on. The packaging is for protecting from all of these threats.
Let's talk about what's inside. First we have your die. The die attaches to a metal support called a lead frame, oftentimes through the use of clips, adhesive, or straps. The lead frame has wires attached to it, which connect it to the outside world. These are called bond wires. And then finally there is a plastic or ceramic or metal enclosure surrounding the die, the lead frame, and the bond wires. This whole setup is considered by the industry as only the first level of semiconductor packaging.
The second level refers to the circuit board, and the third and final level refers to the system's final enclosure, like the desktop PC's chassis. The process of encasing the fabricated die into its level 1 package is referred to as the back end. We can further break down back end processes to two major steps, assembly and test.
Processes differ based on the technology, but here is the vanilla workflow. In assembly, we start by cutting the dies out of the completed wafer and inspecting them. After this we have to put the dies onto the packages lead frame called die attach or die bonding or die mounting. Then we attach the very small bond wires to the dies, which allow the chip to communicate with the outside world.
This step is called wire bonding and was pioneered by Bell Labs in the 1950s. After that we put the die into its ceramic or plastic package to protect it from the outside world. At the start this process had to be done manually, making it one of the first things to be outsourced abroad. Since then machines have taken over, and a swath of new bonding techniques have been introduced, discussing them is beyond the scope of this video, but you should know they exist.
Throughout its history, packaging schools have stayed rather consistent, helped the IC achieve its full functionality and don't get in its way, be as small as possible, and be as cost-effective as possible. In the beginning, only military or aerospace companies used semiconductors. Price was less of a concern than total reliability. There go most semiconductor manufacturers, packaged their dies, in hermetically sealed cans made out of ceramic or steel. This prevented any contaminants from messing with the chip, but there were serious drawbacks.
Ceramics and metal cost a lot. They're also heavy, which meant that the circuit boards they were attached to had to support all that weight. This made them heavy too. Faire child semiconductor, when Faire child started producing their revolutionary planar ICs in the late 1950s, they packed them in the same T-O5 and flat pack packages they used for transistors, just with a few more leads. We will talk about flat packs a little later in this video, but first let us talk about the T-O5.
The T-O stands for transistor outline. A T-O5 kind of looks like a squid, basically a metal can and metal lid. Faire child's first micrologic chips were accordingly etched in a round shape. But the T-O5's awkward round shape made it difficult to attach onto and arrange on a board, adding to the labor cost. Worse yet, T-O5's limited how much data the chip can take in and push out. They maxed out at just 8 to 10 pins, Faire child's new logic circuits needed far more than that.
T-O5 packaging diluted the ICs advantages and worsened its disadvantages. With this situation, however, are ICs ever going to get to price parity with germanium devices and vacuum tubes. In response, Faire child developed two new packaging techniques. The first was plastic encapsulation. This sounds fancy, by the way, they did it back then was to put the IC die on a ceramic bead and cover the whole thing with a plastic blob. This method had some issues.
Plastic can be permeable to certain chemicals leading to concern surrounding contamination. And plastic can shrink or grow depending on outside conditions like heat, inflicting mechanical stress on the chips inside. However, plopping plastic onto a chip required very little skill, this meant that Faire child circuit outsources assembly work to cheap manual labor in Hong Kong or South Korea. One of the first such electronics companies to do so.
Faire child's second idea would forever change semiconductor packaging. Faire child's digital systems laboratory head, Rex Rice, created a new lead configuration for an IC. The leads were spaced about 100,000 of an inch apart. The industry term for this measurement is lead pitch and came out in a single line. You would attach the package to the circuit board using what is called through hole mounting.
Here, you insert the packages leads through literal holes drilled into the boards. The holes have large round solder contact areas surrounding them. After insertion, you solder the leads on the backside to secure them. The method for doing this was called wave soldering. Tests found that this new inline packaging system fit all the criteria. It can accommodate more IO, some 12-16 pins at the start. It simplified the circuit board layout, and finally it took less time and skill to assemble. The inline arrangement was far simpler than the T05 cans circle shaped leads.
Before deploying the product, Faire child first consulted with customers and end users, who did not like the single line arrangement. However, they would be okay with two lines. So Faire child revised the prototype to two rows of inline pins, dual inline packaging or DIP. Introduced in 1965, the simplicity of dual inline packaging cut the cost of assembly by a factor of 4. The industry widely adopted it throughout the 1970s and 80s. During those 20 years, DIPs, which includes ceramic and plastic versions, held some 80-90% market share when measured by value.
However, as the industry moved into the 1970s and the era of very large scale integration or VLSI, demands for IO density kept growing. Some VLSI chips would require as many as 300 leads. More leads forced the DIPs to grow increasingly large. DIPs can have 64 or even more leads, but doing so gave them in practically large board footprints. If trends continued the way they were, the DIP packages would end up being far larger than the dies themselves. This invalidates the various miniaturization gains in semiconductor manufacturing.
In response to this trend, a series of new packages capable of accommodating far more leads emerged, the packaging industry's second big revolution. This generation of packages was defined not by their shape, but rather how they were attached to the circuit board.
Surface mounting technology describes a style of mounting packages onto the board using flat patches of solder already on the board. Surface mounting had several advantages over through-hole mounting. First, because through-hole mounting required you to solder the backside of the boards to secure the DIPs, you can only use one side of the board. With surface mounting, you can use both. This alone increases the theoretical number of packages on a single board by 35 to 60%.
Second, it allowed us to move the leads closer together, or in other words, for a smaller lead pitch. This is because you no longer have the solder contact areas around the holes. And third, you can use cheaper boards, drilling a hole into a board costs money, and there can be up to a thousand holes on a standard board. Ray estimates found that a surface mount board costs 13 cents per square inch, while an insertion hole mounting board costs 15 cents. Furthermore, the boards aren't peppered with holes anymore, which means they don't need so many layers to maintain their structural integrity, another cost saving measure.
Previously, all that surface mounting had to be done by hand, which, considering the smaller lead pitches, required a great deal of skill and training. This was the primary reason why surface mounting didn't take off at first. New automatic machines finally closed the loop on this. Manufacturers now use convection heating, hot gas, or even infrared heat rays to solder the packages onto the boards, basically ovens.
The origins of modern surface mounting technology are murky, and there isn't one significant inventor or eureka moment. I mean the concept of attaching stuff to a board is not exactly groundbreaking. The earliest credible mention is a British patent filed in 1960. It describes resistors, coils, and such attached to a printed board using an adhesive. The connectors were connected using solder.
Early adopters in the 1960s included the US military, which used flat packs for their missile guidance computers. The flat pack was a stackable rectangular glass and ceramic package that was surface mounted onto a board. The flat pack was invented in 1962 by Young Tao of Texas Instruments, so actually a few years older than the venerable DIP. I couldn't find much more about inventor Young Tao or his life, which is unfortunate.
In the late 1960s, a Swiss watch industry adopted surface mounting as a way to reduce some number of electronics in their watches. They popularized what is called the small outline integrated circuit, sometimes also known as the Swiss outline. The small outline looks like the DIP and is made from the same molded plastic materials. However, it is far smaller, a third is high and half as long. This was because as 28 gold wing leads are spaced far closer together, possible again, because we are directly mounting these onto the board. They are called gold wing leads because their shape is somewhat reminiscent of seagulls wings. Seagulls are birds that steal your hot dogs at the beach.
In the 1970s, the Japanese electronics industry began adopting surface mounting technology for their car radios, TV tuners and TV cameras. It is likely they got the idea from Europe, Germany or Britain perhaps. By the 1978 and 1980 period, the semiconductor packaging industry had started moving towards surface mounting, with numerous published articles discussing its use for increasing packing density.
In the early 1980s, the Japanese revived the old US military flat pack to create an updated surface mount compatible version called the Quad Flat Package or QFP. With the Quad Flat, we have the leads projecting down and away from all four sides of the square package. Depending on their size, the Quad Flat can offer anywhere from 20 to 240 leads. Pitches shrank to as small as 1 or 0.65 mm. You can go guess how many inches that is.
The Quad Flat quickly evolved. Push along by consumer demands for smaller and smaller consumer electronics, the Japanese industries introduced yet smaller packages like the shrink quad flat package, the very small quad flat package and the thin quad flat package. On the other end of the spectrum, new structures outpaced the Quad Flat on the lead count front. People realized that having the leads come out of the sides meant that you couldn't use the real estate underneath the package itself.
So for high lead count devices, the packaging industry revived an old IBM invention, the pin grid array or the PGA. First introduced in 1971, the square shaped pin grid array can handle hundreds of leads on its underside. IBM invented the pin grid array for high IOUs in the computer industry. Those packages were made from ceramic, which not only made them more expensive, but also their circuit boards, since those boards had to get thicker to handle the added weight. Plastic variants were later introduced.
Pin grid arrays were invented before the surface mount revolution. So a surface mount compatible cousin emerged in the 1980s, the ball grid array. With these solder balls are used rather than pins to connect the chip to the circuit board. One big disadvantage of the ball grid was that you could not easily visually inspect the solder connections underneath the package. X-ray and electrical tests are often used instead with badly soldered packages removed, re-balled and re-applied. solder ball technology appear in another packaging innovation that popularized at about the same time as the BGA.
Flip chip. This is where the silicon die is flipped, so that it faces downwards. To build the interconnects, we ditch wire bonding entirely and connect using bumps and balls on the chip's pads. Flip chip technology is still used today and offers several benefits. First, there is more contact with a heat sink, which dissipates more heat. Second, electrical signals have a shorter distance with flip chip interconnects than bond wires.
Starting in the 1990s, consumers started buying smaller electronics like mobile phones. Smaller devices mean smaller components, chips, and packages. Space became a premium in a chip package. The industry measures this using what is called packaging efficiency, the ratio of the area occupied by the active device, the die. In response to this, the packaging industry developed the chip scale package, which entered the market in the mid 1990s.
Chip scale packaging is an evolution of the aforementioned ball grid array, and refers to any package where the bear die occupies 80% or more of the total package area. Taiwanese OSAT vendors like ASC Group rose up on the back of such ultra compact packages. One prominent implementation of this concept is called wafer level packaging. This is where we hook up all the die interconnects before we cut those dies out of the wafer. It's not only very cost effective, but gives you very small packages.
As I mentioned earlier, the advent of surface mounting technology created a new generation of packages and techniques. Has Moore's law slows down, packaging technologies have seen yet another surge of investment, creating what we can call advanced packaging solutions. These new packaging technologies now have a much more direct role to play in the systems overall performance.
Semi-analysis has a fantastic multi-part breakdown on advanced packaging that I think drops the mic on the subject. Perhaps the most well known of these are chiplets. This is a variant of what are called multi-chip modules or hybrid circuits or system in packages. I did a video about it previously. Most prominently AMD used them to create a disruptively great product.
One special area of the multi-chip worlds involves 2.5D and 3D integration. These are especially cool. 2.5D integration is where we put multiple die side by side on top of an interconnect substrate called a silicon interposer. This silicon interposer doesn't have any logic, but is just made up of many embedded interconnects. There are a few consumer products using 2.5D integration today. For instance, AMD's Radeon R9 Fury GPU first introduced in 2015. As a name implies, 2.5D is an intermediate step towards 3D packaging, stacking and connecting multiple dies.
We are now able to use a whole new dimension to achieve packaging efficiencies of over 100%. Connecting these dies may require something more heavy than your standard wire bonds, bringing forth new concepts like the Thru Silicon Vias or TSVs. I am working on a future video on 3D integration and die stacking, so keep an eye out for that, if I ever finish it.
The semiconductor packaging industry is truly a chaotic one. There are so many different technology trees growing concurrently with one another, each branch developing and splitting for a certain significant need. New technologies are built on top of those addressing particular niche applications, and then suddenly before you know it, you have no idea why this new thing, which you saw fail a long time ago, is now such a big deal all over again.
In the semiconductor packaging world, ideas are created and then vanish off the mainline, only to return to the forefront when their time is right. Love it. Expect more deep dives into semiconductor packaging in the near future, as we familiarize ourselves with this new world.