Chip innovation and education at DTU Compute

Most Danish companies involved in chip design are located in the Copenhagen area. For the past 50 years, DTU has educated most of the engineers working in chip design. Now, DTU Compute is further strengthening its digital chip design activities.

DTU Chip Day 2024. Credit: Søren Bjørn-Hansen
DTU Chip Day 2024

At DTU Compute, we welcome the increasing focus on chip technology. This requires engineers specialised in the field. For decades, DTU has worked on chip design in various ways.

The majority of Danish companies involved in chip design are located in the Copenhagen area. For the past 50 years, DTU has educated most of engineers working in chip design. At the same time, DTU is the university in Denmark that offers the most courses in chip design and chip technology and has the most students in chip-relevant fields. As the number of transistors on a chip has grown, chip design has become much more than tape-out and prototype manufacturing (link to page). Digital chip design is a special discipline. It also emphasizes the annual DTU Chip Day, where up to 150 students meet the chip industry in Denmark. Now, DTU Compute is further strengthening its activities in chip design.

The pandemic highlighted Europe's deep dependence on mass-produced chips from factories in Southeast Asia and the United States, when production was put on hold and shipping became difficult. That delayed the production of almost every digital electronic device, because these microchips, made from semiconductors (a material that has electrical properties) like silicon, are essential for creating integrated circuits that power devices such as smartphones, computers, and even everyday appliances.

With the EU's Chips Act, the European Union aims to support the development and production of microchips within Europe. The goal is to bring chip production back to European soil.

In this theme, we delve into the current research in chip design at DTU Compute. We look into the role of education to see, how we are educating future engineers and researchers in close collaboration with the Danish chip industry to meet not only the Danish but also the global challenges in chip design. Find also facts about microchips.

Research in chip design at DTU Compute

At DTU Compute, our researchers in the Embedded Systems Engineering section work with chip design in broader term.

Networks on chip (NOC)

Since their introduction in 2007, Networks-on-Chip (NoCs) have become a cornerstone of modern chip design, now integrated into many systems. Acting as on-chip packet-switched networks, NoCs ensure efficient data exchange between cores while optimizing critical factors like area, speed, and power consumption.

At DTU Compute, we conduct research on time-predictable and asynchronous NoCs like Argo, Argo 2, and S4NOC. These solutions offer guaranteed service and best-effort platforms, as well as real-time systems, which are critical in applications like industrial automation, robotics, and signal processing, where precise timing and reliability are essential for seamless operation. With multicore platforms growing ever more complex, NoCs research plays a vital role in many application areas and remains at the forefront of research and development for efficient and high-performance systems.

Targeted systems: Homogeneous multicore platforms, heterogeneous multicore platforms (combining CPUs, GPUs, and accelerators), embedded systems, energy-efficient systems, high-performance computing.
Application areas: Signal processing, industrial control systems, automotive healthcare.

Lead researchers

Jens Sparsø Professor emeritus

Luca Pezzarossa Associate Professor

Martin Schoeberl Professor

Time Predictable computer architectures and networks

Time-predictable computer architectures are specially designed to ensure systems run consistently and reliably, even in situations where every millisecond counts. These architectures make it possible to calculate the longest time a task could take, known as worst-case execution time (WCET), providing the certainty needed for safety-critical applications. In addition to processing, time-predictable networking, such as real-time Ethernet, ensures that data is transferred on time, every time, meeting strict timing requirements. These technologies are essential in fields like industrial automation, aviation, and telecommunications, where precise timing and predictable behavior are critical for safety, efficiency, and reliability.

At DTU Compute, researchers are working on developing hardware and software solutions for time-predictable systems. An example consists of the T-CREST multicore platform and the Patmos processor ecosystem, which form the foundation of this research topic. Both T-CREST and Patmos are used as target platforms for many research projects as well as for educational purposes.

Targeted systems: Real-time systems, Safety-critical systems, Compiler for time predictability.

Application areas: Industrial automation, Avionics, Telecommunication.

 

Lead researchers

Martin Schoeberl Professor

Jens Sparsø Professor emeritus

Luca Pezzarossa Associate Professor

AI in resource-constrained embedded devices

Artificial intelligence is transforming technology, but its high energy demands create challenges for integrating AI into compact devices like hearing aids, sensors, and mobile phones. Research is focusing on making AI solutions more efficient, with innovations such as circuits inspired by biological neurons and synapses, specialized accelerators designed for specific tasks, and streamlined algorithms that eliminate unnecessary computations.

At DTU Compute, we research the use of AI with limited resources (design and architecture of the chip area, energy consumption, etc.). One example is a collaboration about AI chips with hearing aid manufacturer Demant, where AI was used to enhance speech recognition and reduce background noise. Another example is the work conducted in developing techniques that produce AI models optimized for devices with limited resources, such as microcontrollers used in smart sensors and wearable technology.

Targeted systems: Real time systems, Embedded systems, AI accelerators, ASIC and FPGA implementations, Microcontrollers, Sensors and actuators.

Application areas: Hearing aids, Headsets, Video-conferencing, Predictive maintenance, Building automation.

Lead researchers

Xenofon Fafoutis Professor

Jens Sparsø Professor emeritus

Luca Pezzarossa Associate Professor

Additional research: Microfluidic Platforms

Microfluidics refers to the technology that manipulates tiny droplets of fluids on a chip, guided by precisely controlled sequences of electrodes. These droplets can be moved, mixed, split, or analyzed within a compact device, reducing the need for bulky laboratory equipment. This innovation has vast applications, including faster disease diagnosis, personalized medicine, and high-throughput biochemical analysis. Inspired by advancements in traditional chip design, microfluidic platforms and Electronic Design Automation (EDA) tools for such platforms may represent a significant advancement in the landscape of biomedical technology.

At DTU Compute, our Bioware Systems microfluidics platforms and EDAs exemplify the research efforts in this cutting-edge technology. By bridging the fields of electronic chip design and biotechnology, microfluidic platforms and EDA tools can offer solutions for real-time diagnostics and biochemical research.

Targeted systems: Lab-on-a-chip, Point-of-care devices, High-throughput screening systems, Automated biochemical processing systems.

Application areas: Medical diagnostics, Personalized medicine, Biochemical research, Synthetic biology.

Lead researchers

Paul Pop Professor

Jan Madsen Head of department, Professor

Luca Pezzarossa Associate Professor

Additional research: Design of efficient asynchronous circuits

Asynchronous circuits, which operate without a global clock, offer unique advantages such as low power consumption and high speed, making them ideal for applications requiring energy efficiency and/or rapid processing.

At DTU Compute, researchers are working on the design of efficient asynchronous circuits. Such circuits are particularly valuable in energy-constrained environments, such as IoT devices, wearable technology, and medical implants, where battery life is critical. They might be also used in high-speed computing systems and signal-processing applications, where increasing the processing speed can significantly enhance overall performance. New researcher hired from April 2025.

 

Targeted systems: Multicore platforms, Hardware accelerators.

 

Application areas: Packet-switched interconnection networks, Audio/Hearing aids, Medical, IoT devices.

 

 

Lead researchers

Jens Sparsø Professor emeritus

Additional research: Energy Efficient Numerical Processors  

Energy Efficient Numerical Processors are specialized processors designed to perform numerical calculations with high efficiency, consuming less power compared to traditional processors. These processors are particularly important in fields such as high-performance computing, digital signal processing, and machine learning, where large amounts of data need to be processed quickly and efficiently. The goal is to achieve maximum computational performance while minimizing energy consumption, which is crucial for both environmental sustainability and operational cost savings.

At DTU Compute, researchers have been working extensively in this field, focused on Floating-Point (FP) units, application-specific processors, and accelerators for several application domains, including machine learning, signal processing, and edge computing.

Target systems: ASIC and FPGA platforms.

Application areas: High-performance computing, Energy-efficient computing, Digital Signal Processing, Machine Learning.    

Lead researchers

Alberto Nannarelli Emeritus

Chip design at DTU in brief 

Chip technology and digital chip design have a long history at DTU.

Back in the 1960s-70s, the focus was on technology and the underlying physics, and the activities were anchored in the semiconductor laboratory that at one point became part of the electronics department. As the number of transistors on a chip continued to grow, the task of designing chips started to play a more prominent role, and the semiconductor laboratory established a design centre. Focusing on how to support the design process.

Later, in 1985, the computer science department established research and teaching with a focus on large digital chips based on principles pioneered Carver Mead and Lynn Conway (link to Mead and Conway story). In the fall of 1990 DTU established the Microelectronics Center (now DTU Nanolab) with clean rooms etc., a significant ramp-up of the activities in the semiconductor laboratory. It was decided to focus the effort on MEMS (micro-electro-mechanical systems), and the design centre was left behind and subsequently merged into the Department of Computer Science. At that point all chip design activities at DTU were based in the computer science department, now DTU Compute.

Later, around 1990, the electronics department started research and teaching in design of analog chips. Briefly, this means that as of today, three DTU departments are involved in chip technology and chip design: fabrication technology and MEMS are covered by DTU Nanolab, the design of digital systems is covered by DTU Compute, and the design of analog circuits is covered by DTU Electro.

DTU Chip Day 2023. Credit: Hanne Kokkegård

Education

For the past 50 years, DTU has educated most of the engineers working in chip design. At the same time, DTU is the university in Denmark that offers the most courses in digital chip design and chip technology and has the most students in chip-relevant fields. In these years, DTU Compute is further strengthening its activities in education, specifically in the field of digital chip design.

BSc Computer Engineering

Since 2023, students have been able to choose the new bachelor's programme BSc Computer Engineering, which is primarily academically anchored at DTU Compute but has been developed in close dialogue with DTU Electro.

The programme is specifically designed to meet the industry's demand. In the study programme, students build a deep understanding of computer engineering through both digital theory and hands-on programming. They learn about the structure of digital computer systems and how to analyse and design software systems and digital circuits, while also understanding the fundamentalprinciples behind everything from analog circuits to the latest chip technologies.

Computer Engineering qualifies for several MSc programmes at DTU as well as other universities and gives students many opportunities for the future. Whether they want to design chips, work with embedded systems, advise on information technology, or develop computer systems for use in hearing aids, sound processing devices, industrial equipment, or something else entirely.

Read more about BSc Computer Engineering

Open Source Chip Design for All

Alongside the new bachelor's programme and Edu4Chip (look for text about that), starting in February 2025, all students can follow a new course on open-source chip design. In this course, they will have the opportunity to produce a chip through the Tiny Tapeout programme, an educational initiative designed to make it easier and more affordable for individuals to create and manufacture their own custom chips.

Furthermore, selected (larger) projects can also be produced during the yearly tape-out within Edu4Chip.

DTU Chip Day 2023. Credit: Hanne Kokkegaard

Edu4Chip - Joint education for Master’s in advanced chip design   

 

Edu4Chip is an initiative aiming to strengthen Europe's chip design capabilities through harmonised university programs. Specifically, Edu4Chip provides a cutting-edge Master's degree in advanced chip in Europe and offers students end-to-end chip design experiences, from conceptualisation to manufacturing and testing.  In addition, Edu4Chip offers lifelong learning opportunities for IT professionals.

A consortium of five top European universities is facilitating Edu4Chip: Technical University of Denmark | DTU Compute, Technical University of Munich (lead), KTH Royal Institute of Technology in Stockholm, Tampere University in Finland, and Institut Mines-Télécom in France in corporation with one research institute, three SMEs, and, informally, by several additional companies. The initiative is funded by the European Union for four years and started on 1 October 2023.

One of the mist important objectives of Edu4Chip is the joint development of new advanced circuit design master-level course programs at the university level and the improvement of existing ones, which can provide students with the theoretical and practical skills to become chip designers entering the European labour market and help in this way to close the skills gap.

In the Edu4Chip programme, the students can tailor their learning experience by choosing from a wide range of specialised courses at each partner university. Student exchange opportunities, together with the aligned programmes, make it easy to take advantage of each university’s strengths, ensuring students to get the most out of their education.

Edu4Chip will also offer three summer schools. In 2025, the first Edu4Chip summer school will be held at the Technical University of Denmark | DTU Compute from Monday 18 to Friday 22 of August 2025. The summer school offers an introduction to the chip design process from specification to testing including the use of open-source tool chains.

The summer school is designed for Bachelor and Master level students with a background in electrical engineering, computer science and engineering, or physics. The summer school places particular emphasis on practical learning activities. Participants will acquire foundational knowledge of chip design workflows, develop skills in tools and methodologies, and explore cutting-edge developments in the field.

Overall, the summer school follows the Edu4Chip project objectives: to advance education in chip design and to prepare students for the European semiconductor industry. Additionally, it also provides an ideal opportunity to meet and engage with students from all five partner universities while promoting chip design and motivating students to pursue further studies and careers in this field. 

The programme includes keynote speeches, expert-led lectures, hands-on laboratory sessions, and social activities. 

Learn more about Edu4Chips here

Danish Chips Competence Centre

DTU has a long tradition for design of microchips and is a part of the Danish Chips Competence Centre (DKCCC), headquartered at the Technical University of Denmark. The centre is designed to support the entire ecosystem of chip production, from research and development to prototyping and small-scale manufacturing in Denmark and Europe.

Supported by both the EU and national funding through the European Chips Act (through The Chips Joint Undertaking (Chips JU), the Chips Competence Centre provides Danish and European SMEs and startups with streamlined access to state-of-the-art facilities, research expertise, and specialised services for semiconductor and quantum chip design and fabrication.

It leverages facilities like the National Center for Nanofabrication and Characterization at DTU, which is one of the largest university-owned cleanroom facilities in Europe. This centre plays a crucial role in transitioning research into industrial applications, ensuring that Denmark can contribute significantly to the European chip agenda. Learn more in this theme

Starting in January 2025, the Chips Competence Centre will begin its four years mission to drive innovation and strengthen commercial chip development.

The Chips Competence Centre's consortium includes Technical University of Denmark (DTU Nanolab), Aarhus University, the University of Copenhagen (Niels Bohr Institute), Danish Fundamental Metrology (DFM). The private business and employers' organisation Danish Industry representing approximately 20,000 companies in Denmark will serve as a key hub. It will connect Danish companies to both Danish and European networks and resources, helping to foster growth, collaboration, and market opportunities in the chip sector.

In November, the Danish government and the parties in the parliament Folketinget have agreed on the allocation of DKK 5.5 billion for research and innovation in relation to “Forskningsreserven” of which DKK 140 million in 2025 will be dedicated to supporting Danish participation in the Chip JU, the European research and development initiative within semiconductors, microchips, and nanochips.

How to design microchips

 

Chip technology (design and production) involves many types of employees who look at different aspects. Even if, for example, chip design is not listed on researchers’ CV, their work may still contribute to the development of chips through circuit calculation, circuit optimisation, etc.

The exponential growth in the number of transistors that can be manufactured on a chip means that chips are evolving from being components to housing complete systems. Therefore, it is necessary for chip design to be handled by system designers rather than semiconductor engineers; design must be separated from fabrication, and Chip Design and Chip Technology have developed into two very different disciplines. Many of the solutions needed for chip design are derived from computer science. This is evident in concepts such as Computer Aided Design (CAD) tools and silicon compilation.

Chip design workflow:
Today chip design is all this, showing that a lot of people are involved in the chip design  workflows. At DTU Compute our researchers covers all aspects of the design process for microchips. 

Digital electronics:
This field focuses on circuits that use digital signals, which are represented by binary numbers (0s and 1s). It includes designing logic gates, adders, multiplexers, and other digital components.

Analogue electronics:
Designing circuits that handle continuous signals. Examples include amplifiers, oscillators, and filters. Analog IC design focuses on signal fidelity and power efficiency.

Computer architecture:
The design and organisation of a computer's core components, including the CPU, memory, and input/output systems. It involves creating efficient data pathways and processing units.

Hardware accelerators (for example for AI):
Specialised circuits designed to perform specific tasks more efficiently than general-purpose CPUs. Examples include GPUs for graphics processing and TPUs for AI computations.

Timing organisations and asynchronous circuits:
Managing the timing of signals within a chip to ensure proper synchronization. Asynchronous circuits operate without a global clock, which can reduce power consumption and improve performance in certain applications.

Techniques for low power:
Aim to reduce the power consumption of chips. Methods include clock gating, power gating, and voltage scaling to minimize both dynamic and static power.

DTU Chip Day 2023. Credit: Hanne Kokkegård

How to design microchips

 

Networks on chip:
A communication subsystem on a chip that connects different IP cores using a network-like structure. It improves scalability and power efficiency compared to traditional bus architectures.  


Layout of circuits:
Arranging the geometric shapes that represent the various components of an integrated circuit (IC) on a chip. The layout must meet performance, size, density, and manufacturability criteria.

Verification:
Ensures that the chip design meets all specifications and functions correctly. It involves simulation, formal verification, and other techniques to identify and fix design errors.

Tape out and fabrication of chips:
The final stage of the design process where the design data is sent to the fabrication facility to create the photomask for the IC. Fabrication involves the actual manufacturing of the chip using processes like photolithography.

Software executing on dedicated heterogeneous multi-core platforms:
Software designed to run on systems with multiple types of processing cores (e.g., CPUs, GPUs). These platforms can handle diverse workloads more efficiently.

Application domain: Signal processing/AI/Communication:
Involves designing chips for specific applications like signal processing (e.g., audio and video processing), artificial intelligence (e.g., neural networks), and communication (e.g., wireless transceivers).

The ecosystem for chip production

The ecosystem for chip production today is a complex and global network. It begins with signing a Non-Disclosure Agreement (NDA) with a fabrication plant, commonly known as a “fab.” These fabs do not design chips themselves but manufacture them based on the designs provided by their clients. After extensive simulations, the chip design is sent to the fab, which then produces the chips. The production process typically takes about three months, after which the chips are tested to ensure they function correctly.

This global ecosystem means that many of these processes are not localised. For instance, Denmark does not have its own fabs, although there are some in Europe. Often, it is more cost-effective to use fabs in Asia. This production process has been in place for around 50 years, and while it has become slower and more expensive over time, the speed of processing has increased, and power consumption has decreased.

The exponential development in chip technology has been largely driven by consumer electronics. However, as technology has advanced, it has become increasingly expensive to produce smaller and more powerful chips. This has led to the introduction of various chip acts by the USA and EU (The European Chips Act) to address these challenges.

Short about a chip: A microelectronics chip, a microchip, or integrated circuit (IC) is a small piece of mono-crystalline silicon. On its surface, a circuit composed of transistors, wires, and other components is fabricated as a single monolithic structure. This allows the chip to perform various electronic functions efficiently.

How to produce microchips?

The area required for a single transistor has been reduced many thousands of times over the past 35 years. We can now make chips with enormous computing power containing many billions of transistors. The structures have become so fine that we are down to atomic dimensions.

Therefore, building a chip factory has become complicated. We need machines that can control structures that are a few hundred atoms in size, and we need to add materials with a thickness of just a few atomic layers. It also makes demands on how small vibrations are tolerated and on the temperature fluctuations to which the machine is exposed. The transistors on a modern microchip are about 10 nm (one nanometre is one billionth of a metre).

When laying layers on top of each other, the individual layers must be placed very precisely. The slightest vibration makes this impossible. And the temperature must be stable. If the temperature increases by just one degree Celsius, a chip of 1 centimetre will become about 50 nanometres larger, and the next layer will not fit correctly. And a single speck of dust will destroy the chip. It also means that there is a significant waste of chips that do not work.

 

So, the manufacture of microchips places high demands on a particle-free environment. The National Center for Nanofabrication and Characterization at DTU is one of the largest university-owned cleanroom facilities in Europe.

Source: https://www.dtu.dk/english/news/topics/chip-technology and DTU Compute.

 

 

 

What is a microchip?

A microchip is a collection of electronic circuits on a small flat piece of silicon. These electrical circuits contain transistors. Transistors are small electrical contacts to turn power on and off. The transistors are controlled by other transistors. When many transistors are connected to each other and can switch on and off quickly enough, a computer chip or microprocessor can be built.

To make such a transistor, small patterns are created on a silicon wafer with addition and removal of materials to create a multilayer structure. Electrical connections to the different parts of the structure are then created using a thin strip of metal.

Silicon is used due to its electrical properties, abundant availability, and low price. It is extracted from ordinary sand consisting of silicon dioxide. By melting the sand, removing the oxygen, and cleaning thoroughly, pure silicon is achieved.

Silicon can act as a semiconductor by changing the electrical properties using small amounts of other elements (typically boron or phosphorus). In some cases, silicon will then behave like an electrical conductor and in other cases like an insulator. These options can be controlled electrically. Electric current can thus be switched on and off with (another) electric current.

Microchips for - for example - mobile phones or computers are following a trend in which they are becoming increasingly small and more and more powerful.

Making the transistors smaller makes it possible to produce a very large quantity of them without using more materials, thus reducing the price per transistor. The more transistors that are located in the same area, the greater the computing power per microchip and the higher the optimization of power consumption.

But the development also means that the machines used in the manufacture of microchips and nanochips become significantly more complicated. Extreme precision is required when using billions of transistors per square centimetre.

In 1965, Intel co-founder Gordon Moore predicted a doubling (exponential growth – Moore’s law) of transistors on an integrated circuit, it will double each 1,5-2 years with minimal rise in cost. This has been a guiding principle for the semiconductor industry for close to 60 years. Even though researchers say that we can no longer improve them, new ways of constructing the circuits on the material are found, making chips even more efficient, e.g. by raising the electrical circuits with transistors into vertical on the flat chip.  

Source: https://www.dtu.dk/english/news/topics/chip-technology and DTU Compute.

Stories 

chip production

Denmark needs a national long-term strategy for chip design

In Denmark, there is an ongoing discussion about how the country can contribute to the future of chip production in Europe. At the Digital Tech Summit, universities and industry discussed the issue and received some advice from Meta: Go for the white space instead of the red ocean. The discussion highlighted the need for a vision to better steer the direction of the work.

quantum technology

DTU to develop quantum technology in NATO centre

As part of a strengthening of the technology development in NATO, a new quantum centre will be established in Denmark. DTU will be responsible for the development and fabrication of products.
Professor Emeritus Jens Sparsø. Credit: Hanne Kokkegård

Chip design

A journey back to the early days of chip design with Professor Emeritus Jens Sparsø

From component catalogs to complex systems: Through more than 40 years, Jens Sparsø has been involved in the development of chip design at DTU. In his first chip, the number of transistors was about 2,000. Today, a chip has several billion transistors, just as predicted by Moore's law, the number of transistors doubles every 1.5-2 years.

Chip Technology

DTU's cleanroom provides reassurance for chip manufacturers

The startup SBT Instruments rents DTU's cleanroom in order to produce measuring chips for its customers and avoid supply problems and fluctuating quality.

Chip Technology

High-tech chip can improve diagnosis

DTU spin-out company Octlight has developed a laser chip for 3D measurement of the eye background, which can improve the diagnosis of and surgery for eye diseases.