How microchips work


Microchips – also referred to as integrated circuits – are considered to be among the greatest technological achievements of the last century. Their invention has paved the way for a digital revolution that keeps changing the world to the present day. The range of devices that rely on microchips still increases at an amazing rate and includes (desktop and laptop) computers, servers, tablets, smartphones, smartwatches, flash drives, portable music players, game consoles, cameras, televisions, electronic household appliances, elevators, cars and traffic guidance systems, airplanes, and countless others. It is often integrated circuits operating covertly that make our daily life so incredibly convenient.
They are fascinating pieces of technology that consist of millions or often billions of microscopic transistors. Just as fascinating as transistor counts is the speed that modern integrated circuits operate which can even be in the range of 4 gigahertz which is 4 billion operations per second. In the following chapters, the most interesting secrets of those engineering marvels will be uncovered.

Short History of Integrated Circuits

At around 1940, computers consisted of thousands of vacuum tubes that were individually wired together in a very complex and expensive way. This design not only required a great deal of electricity, but it was also very maintenance-intensive. The ENIAC computer from 1946 had over 17.000 vacuum tubes and suffered a tube failure on average every two days, which was time-consuming to troubleshoot and repair. With the invention of the transistor in 1947 by Bell Labs, the components became significantly smaller, but the transistors were still wired together individually. This reduced power consumption of those computers and their overall size, but not their wiring complexity. It was not before the invention of integrated circuits before computers became way more efficient and easier to operate and maintain. Although an early concept for an integrated circuit goes back to 1949, it was ten years later when Robert Noyce (Fairchild Semiconductor) invented and patented the first monolithic integrated circuit chip based on a silicon substrate. At the same time, Jack Kilby (Texas Instruments) was working on a similar concept. In 1971, the Intel 4004 microprocessor was released. It is considered to be the first commercially produced microprocessor. It accomodated 2.250 transistors in a single chip, and came in a 16 pin Dual-Inline Package (DIP). Since that date, subsequent generations of microprocessors have been released with rapidly increasing transistor counts and operation speeds. An ever increasing demand for integrated circuits by consumers and businesses has allowed semiconductor production to grow to a global industry that made over 400 billion USD in annual sales in the year 2019.

Advantages of Integrated Circuits

The large success of integrated circuits can be attributed to several different factors and advantages.

  • Low size and weight: Integrated circuits may contain millions or billions of transistors in one silicon chip which dramatically reduces the size and weight of electronic devices compared to those built from discrete components.
  • Low power consumption: Integrated circuits require less power than discrete circuit designs.
  • Increased operating speed: Transistors that are located closer together can communicate faster. Therefore, densely packed integrated circuits operate faster than discrete circuits where components are further apart.
  • Low production cost: Although integrated circuits are very complex to manufacture, their mass production has greatly reduced the price of microchips. As the number of transistors in a microchip is typically very high, the cost per transistor is extremely low compared to the production of discrete transistors.
  • High overall reliability: Integrated circuits have a very low failure rate due to the fact that they are produced in a highly controlled and clean environment. They do not rely on solder connections but only have a few micro-welded interconnections that are very durable. Also, they are sealed in a protective and shock-resistant epoxy package.
  • High temperature stability: Integrated circuits can get hot and may need additional cooling, but they still operate at very high performance even at increased temperatures.
  • Low maintenance cost: Damaged or faultly integrated circuits can easily be replaced by a new chip.

Categories of Microchips

Microchips do not all work the same. In fact, there are thousands of different chips for today's large variety of technical devices. Only just a few types of ICs are standardized so they can be bought by consumers to upgrade, for example, their personal computer. However, the majority of ICs produced are highly customized for use in very specialized equipment like industrial machinery, aircraft control systems, medical devices, remote controls, etc. Here are three helpful ways to categorize integrated circuits:

I - Categorization according to the signals processed: Analog ICs vs. Digital ICs vs. Mixed Signal ICs

Analog ICs

Analog integrated circuits (also called linear integrated circuits) are designed to process analog signals. An analog signal can be any voltage, current, or frequency that is used to represent information, and the special thing about analog ICs is that they can perceive and process any value between the lowest signal and the highest signal within their range of operation. For example, an analog integrated circuit that is designed to read analog signals between 0 Volts and 20 volts can read a voltage of 2.6 Volts, or 14.4 volts, and will process these exact values according to its specification. An analog IC very often uses transistors that are optimized for amplifying signals, and therefore analog ICs are very widely used for audio amplification and radio transmission. Other applications include ICs for power supplies, sensors, timers, etc.

Digital ICs

Digital integrated circuits, on the other hand, either use a low signal (0) or a high signal (1) for processing information. For example, a digital processor that uses an operating voltage of 2 volts can only process information by either switching circuits to 0 volts or 2 volts. For that reason, they mainly use transistors optimized for switching signals entirely on or entirely off. Very typical integrated circuits that use digital signals to operate are central processing units (CPUs), graphics processing units (GPUs), image processing units (IPUs), and more.

Mixed-signal ICs

Mixed-signal ICs are a combination of both analog and digital ICs. They include circuitry to process both analog and digital signals, and are often designed to perform conversions from one type of signal to the other. Therefore, two very clear examples of mixed-signal ICs are analog to digital converters (ADCs) and digital to analog converters (DACs). To be able to process both types of signals, they have to use both types of transistors, and often some other semiconductors embedded in the chip. This makes mixed-signal ICs more difficult to design than analog-only or digital-only ICs, because analog and digital components usually have very different power requirements and consumtion characteristics.

II - Categorization according to the structure of integrated circuit

Hybrid Integrated Circuits

A hybrid is a combination or mixture between two elements. In terms of integrated circuits, a hybrid IC is a combination of two rather different approaches to implement electronic circuits: Using discrete components and monolithic structures produced on silicon crystals. The package of a hybrid IC typically consists of a ceramic substrate, and has a layer of metal interconnects applied to the surface. Using production techniques called thin-film or thick-film, resistors are printed onto the substrate, connecting individual metal wires. Various discrete components (single transistors, diodes, capacitors, coils, etc.) are connected to the surface by either soldering or wire-bonding them to the metal interconnects. An isolating solder mask is applied over the wire traces in areas where bond wires could possibly sag and cause short circuits. What makes the structure of an IC chip a true hybrid is that separate circuits that have been fabricated on silicon crystals are glued to the substrate, and wire-bonded to the existing circuitry. Hybrid IC technology became very popular in the 1960s and 1970s, and is still in use today to some extent.

Monolithic Integrated Circuits

The term monolith originates from the Greek words monos, meaning 'single' and lithos, meaning 'stone'. A monolithic IC is a semiconductor where all circuits and all electronic elements of that IC have been fabricated in and on the surface if one single piece (chip) of substrate, e.g. on one piece of silicon wafer. The chip is then usually wire-bonded to the metal interconnects of the package which lead to the connection pins, and is enclosed in a protective epoxy or ceramic package. With this technology, active devices (transistors) as well as passive components (resistors, diodes, capacitors) can all be fabricated on the same piece of silicon crystal. Although the development and fabrication of monolithic ICs is extremely complex, monolithic ICs offer numerous advantages over other IC structures: A smaller size, higher operating speed, lower power consumption, and greater overall reliability. The following information pages will mainly deal with monolithic integrated circuits.

III - Categorization according to the function of use: Logic IC vs. Memory IC vs. Systems-on-Chip

Logic ICs

Logic ICs are semiconductor devices that perform logical operations on digital input signals, and produce a digital output signal. There is a very broad spectrum of logic ICs, ranging from simple logic gate ICs to highly complex special-purpose ICs developed for certain purposes. Here is a list of some very common types of logic ICs:

  • Central Processing Unit (CPU): The device inside a computer that processes all the user input, executes all the instructions and processes that are required to run programs, and produces outputs that can be stored in memory or viewed on a display.
  • Graphics Processing Unit (GPU): A special-purpose unit that is designed to process large amounts of data that can be broken into a swarm of very simple tasks that do the same operation in parallel. As image data usually contains large portions of such data-parallelism, GPUs are mainly used to accelerate the imaging performance of a device. However, GPUs can also be used for simulations, cryptocurrency mining, and other highly parallel tasks.
  • Image Signal Processor (ISP): A device that is specialized in image processing (demosaicing, noise reduction, image sharpening, etc.), and is often found in digital cameras and smartphones.
  • Neural Processing Unit (NPU): Also a very special-purpose IC that is designed for deep learning and machine learning applications.

Memory ICs

Memory ICs store information. They can be specified based on their methods for reading, writing, and storing data. Two broader classes of memory ICs are Random Access Memory (RAM) and Read-Only Memory (ROM):

  • Random Access Memory (RAM) describes any memory chip that can be written to and read from. Common RAM types include Dynamic RAM (DRAM) and Static RAM (SRAM). A special characteristic of RAM is that data stored on RAM is usually volatile, which means that all information stored in RAM is lost when the device is disconnected from power. There are only few exceptions to this rule, such as Ferroelectric RAM (FRAM) which is a special type of non-volatile RAM.
  • Read-Only Memory (ROM) describes chips where stored information can only be read from other components, but no component can write new information onto the chip's memory. ROM is typically non-volatile and is used to store permanent data that is read on a regular basis. Types of ROM include Programmable ROM (PROM), Erasable programmable ROM (EPROM), Electrically erasable programmable ROM (EEPROM), and Flash memory.

System-on-Chip (SoC)

A System-on-Chip, also known as an SoC, is an IC that integrates an entire computer system onto a single platform. Therefore, an SoC often includes power management modules, the CPU, GPU, IPU, memory units, WiFi connectivity module, audio unit, and others, on one single piece of silicon. They are considered to be more powerful and faster than traditional forms of system implementations while being smaller and more power efficient. Therefore, SoCs have become the most popular type of IC for modern smart devices and the internet of things.

From Sand to Silicon - IC Fabrication

To manufacture a modern integrated circuit is one of the most complex and difficult endeavours ever. It can take over 400 steps to process an integrated circuit, as layers of material are deposited and removed over and over again to create desired patterns. In addition, numerous very complex production technologies involving atomic diffusion, metallization, etching, exposure to ions and special light, and others are required to develop structures on a substrate. The diagram below only includes the most important steps.

The raw material for integrated circuits is usually silicon, and it can be produced in different levels of purity:

  • Ferrosilicon: An alloy of iron and silicon with 10-90% silicon content.
  • Metallurgical grade silicon (MG-Si): 98% pure
  • Solar grade silicon (SG-Si) as used for photovoltaics: 99.9999% pure ('six nines pure') That's one impurity atom in 1.000.000 silicon atoms.
  • Electronics grade silicon (EG-Si) as used for silicon wafers: 99.9999999% pure ('nine nines pure') That's one impurity atom in silicon atoms. Only silicon of that purity level can be used in IC production!

Once the manufacturing process has started, even a single dust particle can ruin an entire microchip. For that reason, integrated circuits are fabricated in special clean rooms. These offer a dust-free environment with stable temperature and humidity levels.

Some more facts about IC production:

  • A typical silicon ingot can weigh up to 100 kgs
  • Several thousand chips can be fabricated on a single wafer
  • The size of a die can vary between 1mm² and few cm²
  • Features in today’s most advanced chips can be as small as 4 nm – that’s thousands of times smaller than a grain of sand.

Moore’s Law

Moore’s Law is an observation that was issued in 1965 by Gordon Moore, the co-founder of Fairchild Semiconductor and CEO and co-founder of Intel. He observed that the number of transistors in an integrated circuit doubles roughly every year, and he predicted that this rate of growth would continue for at least another decade. Ten years later, he published an update, revising the doubling period to every two years. The figure shows the validity of Moore’s Law until today. All data points indicate the release dates of Intel microprocessors.