SBO project on Resource-Efficient Deep Learning

semiconductor industry

While deep learning is quickly gaining traction in server platforms, we struggle to deploy this machine learning method on embedded devices. Researches reach out to you to join the advisory committee of their project and tackle this challenges.

by KU Leuven

Deep learning revolutionized many disciplines of signal processing, going from advanced image processing of AR/VR, over natural language processing to visuomotor robot control policies. While the fundamental concepts were already articulated in the 1940s, the major breakthroughs were only realized in the past decade. Besides algorithmic innovations, the recency of the breakthrough can be attributed to the fact that only now the necessary computational hardware acceleration and large corpora of labelled data have become available. Also smartphones and embedded devices are increasingly equipped with dedicated DSPs for embedded AI, often named “tensor processing units”.

Despite the success of deep learning, executing these deep learning workloads on embedded resource constrained devices still poses severe challenges. These devices suffer at the same time from extremely tight energy consumption specs, low memory footprints, and a lack of large (labeled) data sets. Connecting these embedded systems with the cloud however also brings many drawbacks such as the need for constant and high-bandwidth connectivity with the cloud, increased latency for execution, and privacy and security concerns of processing sensitive data (e.g. health-related data). As a result innovation is required towards efficient on-chip learning from small, ill-labeled data sets. These workloads can moreover be tackled by either devices individually, or by collaborative learning devices organizing themselves in a distributed and possibly hierarchical setting.

Research teams from Ghent University, the University of Antwerp and KU Leuven are now teaming up in an FWO SBO project proposal to tackle exactly these challenges. The project will look at novel ways to tackle the resource bottlenecks in deep learning on embedded devices. We will design efficient training routines and appropriate hardware platforms for continuous on-device learning with few labelled data. We will look at ways to optimize the distribution of intelligence in connected multi-device environments across resource-rich and resource-constrained devices.

The researchers now reach out to the members of DSP valley to join the advisory committee of the SBO project.

Committee members will be regularly updated on the research progress, can propose research directions (e.g. via specific use cases or needs), and will be invited to workshops and hands-on tutorials, providing not only employee training but also unique opportunities for networking with other Flemish and Brussels companies. For a company to join the committee, the FWO mandates a limited financial contribution that depends on the size of your company. For an SME, the yearly fee is € 250, for large companies the fee is € 1000 per year.

For more information or indicate interest, please contact, preferably before April 15th.

Ultra-Low Power Transceiver for Bluetooth 5 and Beyond with Lowest Supply Voltage Ever Reported

At the 2018 International Solid-State Circuits Conference in San Francisco (US), imec, together with Renesas Electronics Corporation, announced a low-voltage (0.8V) ultra-low-power Bluetooth 5 transceiver for IoT applications. The low supply voltage enables a longer battery life.

by imec

In the Internet of Things (IoT), numerous tiny sensor nodes are embedded in our environment. Reducing the power consumption and enabling operation at low supply voltage of IoT radio transceivers is key to enable reliable multi-year battery life of the billions of IoT devices in our environment. Furthermore, new applications are empowered by reducing power consumption and supply voltage, such as leave-behind sensors or wearables.

Only 0.8V

Imec, the world-leading research and innovation hub in nano-electronics and digital technology, together with Renesas, a premier supplier of advanced semiconductor solutions, developed a fully-integrated, highly energy efficient BT5 transceiver which showcases the best RX figure-of-merit (FoM) and the lowest supply voltage among the state-of-the-art (0.8V). A low supply voltage of 0.8V is beneficial because it extends battery life by up to 50%, reduces the complexity of the power management unit, and allows the use of a wider range of energy sources such as energy harvesters.

A low supply voltage of 0.8V is extends battery life by up to 50%.

The transceiver consists of a novel phase-tracking RX, a digital TX based on an all-digital PLL (ADPLL) and a PHY-layer digital baseband. A hybrid loop filter enhances interference tolerance, in conjunction with precise frequency control enabled by the ADPLL. The receiver fronted achieves a sensitivity of -95dBm, while consuming only 2.3mW (non-duty-cycled) at 0.8V. The transceiver chip is implemented in 40nm CMOS, with a core area of only 0.8mm2, including on-chip matching.

Further leap for wireless IoTmarket

“This world-class BT5/BLE transceiver for IoT applications is the result of a successful collaboration with Renesas,” commented Kathleen Philips, Program Director at imec/Holst Centre for IoT. “We are constantly looking for ways to improve battery lifetime of IoT sensors and radios and a low supply voltage is yet another key element in the roadmap towards successful IoT radio solutions.”

“Renesas has been producing high-performance devices leveraging low-power wireless communication technology, which will play a key role for expanding the upcoming wireless IoT business moving forward,” said Shin Saito, Senior Director, Industrial Analog & Power Business Division, at Renesas. “Renesas believes that by jointly developing with Imec, we can achieve a more innovative system architecture, enabling a further leap for ultra-low power wireless technology and the wireless IoT market.”

Rapid Product Development with ANSYS Discovery

To make simulating a product’s physical performance as easy as using Google Search is the vision of every engineer and every product. The ANSYS Discovery family is a big step forward towards that dream.


The ANSYS Discovery family includes ANSYS Discovery SpaceClaim, ANSYS Discovery AIM, as well as Discover Live. These three, bundled together, can turn your design team into a powerhouse. Some examples.

Rossignol, winter sports equipment

With the 2018 skiing season in full force, what better example is there than Rossignol, the maker of top of the line winter sports equipment, used by many of the athletes in the competition.

Rossignol has a team of engineers working with skiers, designers and others to create new products. Of course, because of the ski season, time to market is critical for the company. With Discovery the Rossignol engineers assessed changes in real-time and discussed them with the other team members to rapidly explore designs. What was previously a longer and more individual exercise became collaborative not just among engineers, but with every team member benefiting from simulation insight. The engineers literally captured feedback from the skiers as they came off the slopes and Discovery gave them a joint language to explore product performance.

To learn more about their story, watch this video

Innovation is now involving everyone in the company – Nicolas Puget, Lead Research and Innovation Manager, Rossignol

Wibotic, intelligent, wireless battery charging solutions

In a different part of the world and with a very different product, Wibotic also discovered the value of real-time simulation for design exploration. Wibotic designs and manufactures intelligent, wireless battery charging solutions for drones, robots and underwater autonomous vehicles.

For the company, dissipating the significant amount of heat produced during charging of the batteries can be a challenging design problem. They began using ANSYS Discovery to simulate more design iterations in a short amount of time resulting in decreased prototyping costs and less time to market. They too encountered increased collaboration because they could use Discovery as an interactive tool to illustrate the engineering improvements based on simulation results. You can watch the full story here

I was able to design on the fly with ANSYS Discovery. In a matter of 10 minutes I could be analyzing a completely different model and test thermal characteristics and airflow. The time we’ve saved in using Discovery can now be spent on other tasks related to product development, like usability, aesthetic design, and design for manufacturability – Chasen Smith, Lead Mechanical Engineer, Wibotic

Every engineer and product benefit

In both cases design engineers were able to immediately pick up and use simulation – thanks to the radical ease of use of Discovery — and interactively explore product options due to the groundbreaking speed. Their products are not turbines nor rockets, that traditionally benefit from simulation, but skis and battery chargers.

Perhaps more interesting is the way the product development process changes radically with intuitive and real-time simulation. Innovation thrives when making mistakes is merely a matter of seconds and fixed by an undo button. When it does not imply breaking costly prototypes or redoing complex multi-day simulations. And that is exactly what ANSYS Discovery stands for.


You too can benefit from interactive design exploration in minutes.

You are welcome to try ANSYS Discovery Live right now for free. When downloading Discovery you will automatically be taken to an online forum with lots of material to help you get started, ask questions, and share your experiences.

How to manage test equipment in R&D?

Since productivity is the new watchword in R&D, the testing phase of a design needs to be done correctly and with all requisite equipment available when needed. Microlease, a new member of DSP Valley, can help you with that.

By Microlease

Today, the semiconductor sector influences more industry sectors than any other aspect of the electronics industry. It permeates just about every electronic device on the market today and is heavily influenced by the need to deliver products to market rapidly, thereby capturing early market share.

Hence productivity is the new watchword in R&D. A recent report by McKinsey and Company also showed there is a very strong connection between small, highly focussed, R&D teams based on a single site and high levels of productivity.

While often the focus of productivity is on the design phase, the design is not market-ready until testing and verification is complete, so this project phase must also be completed efficiently.

More diverse test requirements

Productivity requires the testing phase to be correctly and fully resourced in terms of equipment, with all requisite equipment being available when needed. However in smaller R&D teams there is scant spare resource to manage the planning and sourcing of test equipment without impacting productivity. Likewise, multi-site and larger teams face similar issues where the costs of inefficient test management are higher.

As companies seek greater versatility to enhance their competitiveness, so test requirements become more diverse. For example, the test requirements for a microcontroller are vastly different to those for a sensor device, yet many fabless semiconductor companies could well be required to develop technologies as diverse as this to meet customer demand.

Productivity requires the testing phase to be correctly and fully resourced.

Microlease test equipment

This versatility impacts test equipment requirements significantly. Even though modern simulation tools are highly sophisticated, every project requires detailed hardware testing to verify the simulation and assess parametric spread due to the production process. Teams of all sizes increase productivity with the required equipment being available when and where it is needed and budget holders benefit from greater visibility of available equipment, ensuring duplicate purchases are avoided.

Accuracy needed

Characterising highly complex and dense modern integrated circuits (IC) is challenging, often made more complex by the presence of analogue and digital circuitry within the same device. Many designers have to push design rules to the absolute limit to meet design criteria and / or gain a competitive advantage – this necessitates highly accurate physical layer testing.

Power is critical in many applications as functionality increases and IC size decreases. Understanding power consumption accurately in different operating modes is critical to predicting battery lifetime in modern portable applications. Portable devices increasingly include communications, requiring RF test to be brought into the mix.

Since design rules are pushed to the absolute limit, highly accurate physical layer testing is needed.

Microlease test equipment

This broad-based testing requires a myriad of test equipment that may have a limited lifespan. In fact, the more specialist or niche an instrument is, the greater the likelihood that it will be used for a short time to characterise or verify a new semiconductor and then become redundant, tying up cash and requiring maintenance, storage and insurance – all of which drain the R&D budget.

Microlease as your tester

A simplistic ‘one size fits all’ approach of simply purchasing necessary test equipment is, at best, expensive and most likely unsuccessful. Successful, productive, companies are reconsidering the way they manage test equipment and leading semiconductor organisations of all sizes have already benefitted greatly from working with third party test equipment management specialists.

A one size fits all approach is expensive and most likely unsuccessful.

One of the leading specialist test equipment companies is Microlease. With over forty years of experience in supplying expertise and test equipment to a wide variety of companies on a global basis, Microlease clearly understands the needs of the industry. With the ability to offer a wide range of services including sales, leasing, used equipment and asset management, Microlease can provide a flexible, bespoke solution whatever the project needs. Their LEO Asset Management software offers large (and small) teams the ability to track availability and utilisation of equipment, wherever it may be.

Delivering a Digital Twin

Today’s Industrial Internet of Things unfolds before our eyes as businesses leverage new and rapidly evolving technologies. The latest concept is the digital twin, a solution that is far more than a product: It is the outcome that industry demands.

By Ansys

The Internet of Things (IoT) has leap-frogged from consumer applications that facilitate mere interaction and collaboration. Industry leaders like General Electric (GE) extended this connectivity to operating machines. The resulting Industrial Internet of Things (IIoT) enables commercial organizations to engage with large complex machines — wind and gas turbines, jet engines, locomotives — to improve performance, reduce downtime and accelerate new product development. But it doesn’t stop there. Today’s cost models for sensor technology, internet connectivity, and simulation and analytics enable connectivity to not only highly complex, capital-intensive machinery but to almost every piece of equipment in operation.

Data and the industrial internet

The IIoT, in practice, is best used to determine or suggest an action: For example, instruct a wind turbine to tilt its rotors for optimum wind exposure. First, sensor data collected from assets are added to all other available digital information. A dashboard combines this information with the equipment’s real-time and expected-performance data to produce descriptive analytics, which can be mined to forecast potential failures and schedule maintenance. The final step is optimization, which considers individual assets in all their configurations along with systems of assets to arrive at multiple solutions. The objective is to optimize a very complex ecosystem around a particular asset. The very rich models describing structure, context and behavior of industrial assets are called digital twins.

There is a cost for this improved performance: The IIoT manages huge amounts of data, extracting information and gaining actionable insights through big data analytics and deep learning. For security, and also to manage the quantity of data, some data is stored and processed locally “at the edge.” Other functions are performed on data in the cloud. This hybrid edge-to-cloud approach helps manage the quantity of data and allows the best computational approach to be taken for different types of objectives while maintaining safety and security of operation and protecting a company’s valuable IP.

Getting started with a digital twin

A digital twin begins with a basic model that describes the asset. For example, a wind turbine model could include PLM system information with details on materials and components, a 3-D geometric model, a simulation model that predicts expected behavior based on physics algorithms, or recommendations from analytics created using machine-learning techniques. The model also can include service logs of maintenance, and defect and solution details, capturing the entire life cycle of the asset.

Initially the digital twin represents a class of assets — in this example, a wind turbine of type x. This generic twin must be individualized for a specific wind turbine on a particular farm. Consider that the machine has operated for five years, enduring weather specific to its location, running among 50 other turbines. So the entire wind farm must be modeled. Each turbine is similar but different based on its position or experience (wind direction, maintenance record, wake effects). In the end, the twin’s rich digital representation contains its past and present condition moment by moment. The future of a specific wind turbine, in this case, is codified in that digital twin.

Digital twins provide accurate operational pictures of assets right now. There is a significant business value in identifying underutilized devices, so analyzing twin information can lead to optimal usage. For example, GE Power leveraged a digital twin to get 5 percent more output from a wind farm without making wholesale changes. The team optimized the turbines to changing wind conditions and orchestrated the interaction of individual twins on-site. One insight seemed counterintuitive: In specific scenarios, shutting down some turbines improved output compared to running all turbines. By predicting potential problems in a fine-grained way, operators can schedule maintenance to minimize service disruption. Once the information is codified across a system of assets, the team can take that knowledge and turn it into actions that will obtain the desired outcomes.

Building a twin model at the outset is the key to creating a rich set of applications that produce asset-related outcomes — not, as some think, just developing a dashboard for equipment operator decision-making.
A full-featured twin makes it easier to develop and deploy applications later. The physics, analytics and simulation information within the model pave the way for machine learning; many digital twins linked together produce a mass of actionable industrial knowledge.

GE Digital leverages ANSYS simulation in its digital-twin use cases, so the two organizations immediately benefit from collaboration

Platforms support the industrial internet

The latest IIoT challenge is how to make such sophisticated technology user-friendly so end-users (who are manufacturers and engineers, but not programmers) can solve business problems.

To that end, GE has developed the Predix® platform to connect industrial equipment, analyze data and deliver real-time insights.
Predix is an aggregation of microservices that are useful in building, deploying and managing industrial internet applications. Customers consult with GE Digital on business problems, such as increasing a wind farm’s output or optimizing a gas turbine that services a variable-power infrastructure. Within a few weeks, these organizations assemble an initial solution to address the problem.

GE also uses the Predix platform internally to optimize its own production processes and build more efficient solutions for customers.

Digital twins can be practically applied in almost every industry

Simulation and the digital twin

For decades, GE has gathered data on many assets, such as jet engines. Combining such data with statistical models predicts what is likely to happen and when — but it falls short of determining why and how it happens. Adding in physics-based simulation is the final step to gleaning this additional insight. GE’s Predix can overlay data with simulation in an industrial context that operates as a common data model. Simulations can be run on-site or in the cloud at scale — pushing models to the edge then bringing insights they create back to the cloud. Complete integration requires connecting to the customer’s PLM system, linking in CAD data and other valuable information recorded in enterprise systems.
A digital twin that centers on a common model and incorporates many information sources enriches knowledge.

GE Digital leverages ANSYS simulation in its digital twin use cases, so the two organizations immediately benefit from collaboration. ANSYS software’s greatest value is in bringing together different aspects of simulation, so it helps designers completely think through their designs. Because a simulation model demonstrates how the assets should work, the twin approach shows exactly when operation is amiss. Digital twins take simulation results out of the design studio and into real life to provide immediate feedback on one asset or many. Soon the technology will enable optimizing an individual asset in the field; furthermore, it could be deployed throughout an asset’s entire life cycle.

The future

Digital twins can be practically applied in almost every industry: transportation, energy, manufacturing, aviation and more. Companies already are saving millions of dollars by bringing together data, simulation, platform, cloud-based functions and machine learning. Organizations can only imagine the future benefits as the digital twin concept grows more prevalent.

Full SystemVerilog support with Sigasi Studio 3.5

semiconductor industry

Most SystemVerilog users have a love/hate relationship with SystemVerilog. This hardware description and verification language is really powerful, but also really complex. And it does not protect you at all from making mistakes. Therefor SystemVerilog users will benefit even more from the assistance that Sigasi Studio offers, than their VHDL colleagues. Sigasi Studio helps you focus on what is really important, the design, and makes the compiler understand your intentions.

By Sigasi

Based on Sigasi’s experience with providing excellent VHDL support, the Sigasi development team knows how to tackle most challenges in providing a good development environment for SystemVerilog: immediate feedback about syntax errors, autocomplete, open declaration, and so on. Because Sigasi Studio understands what SystemVerilog means, you get very accurate feedback.

Immediate and accurate feedback

The biggest technical challenge was providing good support for SystemVerilog’s Preprocessor. The preprocessor does a textual transformation of SystemVerilog source files. So it can completely rewrite the code that goes into the actual compiler. This preprocessor is stateful and depends on the compilation order. That makes it difficult to keep track of what exactly is going on. To remedy this, Sigasi Studio provides features to easily inspect or preview preprocessed code:

  • Source code that is excluded with the Preprocessor is automatically grayed out in the editor.
  • The result of Macro’s can be easily previewed in an addition view, or simply by hovering your mouse over the macro.
  • Syntax errors are immediately reported

Another example of how Sigasi Studio helps SystemVerilog users, is “include files”. A typical pattern in SystemVerilog is to include sources into other source files (with the Preprocessor). In this case the included file can see everything in the scope of the including file. With an ordinary editor, you have to think about all of this yourself. With Sigasi Studio however, this information is available whenever you require it. For example during autocompletes.

And the nice thing is, this does not require any additional setup. Again, this allows you to focus on the real job: getting your design ready, or making sure it is well tested.

Free trial

Would you like to try Sigasi Studio on your own SystemVerilog designs, you can request a full trial license on our website This enables you to try all Sigasi Studio’s features on your own projects and feel how Sigasi empowers you. And please tell us all about your experience.

ACE electronics buys Viscom 3D AOI

One of the main components of Industry 4.0 is quality inspection and process control. ACE electronics recently invested in an AOI, an Automated Optical Inspection system with advanced features. This will enable them to continue producing reliable and complex PCB’s.

By ACE electronics

 After an intensive testperiod ACE electronics N.V. located in Diest (B) has selected the Viscom 3D AOI S3088 Ultra. Decisive factors such as the unique  3D and software features where complemented by the robust construction of the machine, the high measuring accuracy and exceptional image quality.

“The selection of an AOI machine is not easy and the demo’s that we received from different manufacturors were relative comparable.” says Johan Lieten , Sales & Logistics Manager at ACE electronics . “But when we looked in more detail into the specifications and the software of the S3088 Ultra machine we started to clearly see the differences. Also the results of the more in depht tests were very impressive.”

Viscom is the European marketleader of inspection systems and preferred supplier for customers that focus on complex PCB’s in high tech industries. The last few years the company has been strongly focussing on the development of their new software platform vVision. This allows the user to very simply program, fine-tune and operate the machine, regardless if they are producing small or large production batches.

“Our customer portfolio is very wide, and we produce a combination of mid-volume batches for all types of end-users and markets. This makes fast programming and the detection of all defects very crucial for us.” says Mr. Lieten. “And same as our customers request from us, we also request one indispensable feature when purchasing an AOI machine: reliability.”

As our customers request reliability from us, we request the same from an AOI machine.

The combination of the orthogonal top camera, 8 angle view camera’s and the digital 3D fringe projector guarantees an unprecedented accuray, and is all Made in Germany. Smd-Tec is in the Benelux region the local support for installation, training and service. But also the proximity of Viscom to the Benelux region gives the customer the opportunity to be in direct contact with the manufacturor. This interaction has a double benefit as the customer can count on the highest level of support, and Viscom gets direct feedback to allow even more customer required features in their systems.

“We are very happy with the investment that ACE electronics has made and therewith clearly emphasizing the commitment to their customers.” concludes Tom Van Tongelen, CEO at Smd-Tec. “With this system they will without any doubt have the ability to offer the highest available quality.