放射線耐性チップがCERN加速器に電力を供給

高放射線環境向けエレクトロニクスの構築

Electronics are at the core of almost any technology invented in the past decades. As the world digitalizes even more processes and data, this is becoming more true every day.

しかし、いくつかの環境では、標準的なエレクトロニクスは追いつくのが困難です。その一例が粒子加速器です。

一方で、粒子加速器は秒間に膨大なテラバイトのデータを生成するため、超高効率の電子部品が必要です。もう一方で、生成される放射線量が電子システムを乱す傾向があります。

Scientists at the CERN in Switzerland were faced with this dilemma. At the LHC particle accelerator at CERN, the world’s largest, radiations were emitted, making measurement difficult.

「標準的な商用部品をテストしましたが、すぐに壊れてしまいました。放射線が強すぎました。機能するものが欲しいなら、自分たちで設計しなければならないと悟りました。」Rui (Ray) Xu, a Columbia Engineering PhD student

The first chip of this kind was developed in 2017 and put to the test in 2022 for the ATLAS experiments. ATLAS is the largest particle detector ever built, at 46 meters (150 feet) long and 25 meters (82 feet) in diameter.

The detectors contain 100+ million sensitive electronic channels to record the particles produced by the collisions. It contains many sub-detectors, each playing a separate role, to detect at the same time photons, electrons, muons, pions, etc.

出典: ATLAS

A second chip, the data acquisition ADC, has recently passed its final tests and is now in full production. It is fully described in a recently published paper¹ in the journal IEEE Explore, under the title “A Radiation-Hard 8-Channel 15-Bit 40-MSPS ADC for the ATLAS Liquid Argon Calorimeter Readout”.

放射線がエレクトロニクスに与える影響

Since the dawn of electronics, it has been known that radiation tends to damage electronic components and/or make them give out the wrong data.

Among the many effects radiation can have, the most problematic ones can be quickly listed:

• トランジスタの電圧変動により、誤ったデータが出力されたり、トランジスタ自体が破壊されたりする。
• メモリ部品のビット（0 と 1）が反転する。
• 集積回路の電気的または熱的な焼損。
• 光検出器や光放出素子が即座に破壊されるか、寿命が短くなる。

This is a serious problem in high-radiation environments, like space, medical accelerators (radiation therapy, radiography), or nuclear facilities.

One option to solve the issue is simply to use enough shielding, putting the electronics part behind a protective layer, usually water or a heavy element like lead, depending on the type of radiation.

Another option is redundancy and error correction. If a component is in multiple copies, or a program runs several times, an error in only one of them can be detected and subsequently ignored.

The last option is to build electronics systems that are naturally resistant to radiation, which is the only option for electronic systems that have to be directly exposed to radiation, like the detectors of a particle accelerator.

放射線耐性エレクトロニクスの構築

商業的実現性

The problem the CERN engineers and scientists faced is that off-the-shelf components simply can’t survive the harsh conditions inside the accelerator.

At the same time, the market for radiation-resistant circuits is too small to entice investment from commercial chip manufacturers.

「最先端の計測機器の開発は我々の成功に不可欠です。産業界はその労力を正当化できず、学術界が介入せざるを得ませんでした。」John Parsons – Professor of physics and leader of the Columbia University team working on the ATLAS detector.

In this specific case, the researchers needed to develop analog-to-digital converters (ADCs). These devices’ task is capturing electrical signals produced by particle collisions inside CERN’s detectors and translating them into digital data that researchers can analyze.

This is done through a device called a liquid argon calorimeter, which converts particle collisions into an electronic signal.

Columbia’s ADC chips convert these delicate analog signals into precise digital measurements, capturing details that no existing component could reliably record.

過酷な条件

The researchers carefully chose and sized components and arranged circuit architectures and layouts to minimize radiation damage, as radiation shielding is not realistic in the particle detector.

Not only that, but they had to take into account that the electronic boards in question are inaccessible during operation and can be accessed for maintenance at most once per year.

The radiation levels the components will experience in a 12-year operating lifetime are typically encountered by satellites in geostationary orbit.

Temporary errors can be tolerated, but permanent damage cannot be accepted, as it would hinder the work of all the research projects requiring ATLAS.

実証済み半導体製造技術の再利用

Reinventing how to produce semiconductors was not going to be a viable path to create a useful device within a reasonable budget and time frame.

So the researchers used commercial semiconductor processes validated by CERN for radiation resistance and applied innovative circuit-level techniques.

A key decision in that respect was to rely on older, tried and tested lithography methods, using a commercial, triple-well 65-nm CMOS process for production of the ASIC custom chip (Application-Specific Integrated Circuit).

This 65 nm process is known to be inherently radiation hardened.

Another design choice was to minimize the components not directly present on the chip, reducing the risk of errors by integrating the chip’s internal clocks, memories, etc.

出典: IEEE Explore

However, calibration calculations are done off-chip to prevent radiation-induced errors in the calculation that would give erroneous data.

They also looked at capacitors, which can be overcharged by the ionizing effect of radiation.

Metal-insulator-metal (MiM) capacitors are naturally 30x-80x thinner than a more conventional metal-oxide–metal (MoM), while also being half the size, reducing the surface potentially hit by radiation and high-energy particles.

出典: IEEE Explore

最終チップ設計とテスト

The final chip is an electronic design specifically designed to be optimal against radiation, instead of high-speed, ease of manufacturing, or boosted performance like commercial products.

In total, 45,617 of these chips will be used in the ATLAS detector.

出典: IEEE Explore

Eighteen devices were characterized for analog performance; further validation of long-term analog accuracy and an extensive radiation testing campaign was undertaken.

All results indicated that the chips would perform well in the environment of the ATLAS detector.

Still, no matter how hardened, these levels of radiation will cause some errors and problems in any electronic systems. So the researchers then built digital systems that automatically detect and correct errors in real time.

Double- and triple-bit errors, which are more problematic, are detected by periodically reading back all memory registers and comparing them to the initial programming. Any measurement taken when such double and triple errors occur is also discarded.

結論

This research project will enable the advanced analysis of high-energy particles generated by the LHC.

It will also be a vital component of a major upgrade of the accelerator with the “High Luminosity LHC” (HL–LHC), an upgrade intended to boost the luminosity of the LHC by 10x.

For example, the High-Luminosity LHC will produce at least 15 million Higgs bosons per year, compared to around three million from the LHC in 2017.

出典: CERN

It is likely that later CERN’s projects, like the Future Circular Collider (FFC), with first experiments starting in the mid-2040s, will also require similar or even more advanced radiation-proof electronics.

Lastly, this sort of project, funded through academic budgets in fundamental physics, can be an inspiration for a commercial version of radiation-proof electronics.

As mankind is looking to explore deep space, including potentially permanent lunar and Martian bases, or asteroid mining, more durable and radiation-proof electronics will be very useful.

先進的センサーへの投資

CEVAはセンサー企業であり、CERNと提携して同機関のアルゴリズムを活用し、センサーの効率と消費電力の改善を図っています。CEVAのソリューションとIP（200件の特許）は18億台のデバイスに組み込まれています。

The company’s solutions are used by many of the leading electronic brands worldwide.

出典: CEVA

The main application of the collaboration between CEVA & the CERN is “Edge AI”, or artificial intelligence applications deployed on devices away from the data centers (the cloud) and closer to the consumers (the edge).

It may not be surprising to see particle physics algorithms being reused in AI applications, as neural networks were, for example, used in finding the Higgs boson particle. Analyzing particle accelerator data needs to be done on-site instead of in the cloud, due to the sheer volume of data produced very quickly.

CEVA helped CERN create new compression algorithms that can be used in future experiments and will be able to integrate this new technology into its products.

“Thanks to our collaboration with CERN, we were able to develop an innovative approach that enables the networks to run up to 15x faster compared to 16-bit baseline models.

It’s enhancing network speed and reducing energy consumption by up to 90% while maintaining comparable accuracy.”

Olya Sirkin – Senior Deep Learning Researcher at Ceva

This is but one of CEVA’s technological progresses, with the company active in wireless connectivity, sensors (vision, audio, motion), and neural network algorithms.

出典: CEVA

CEVA greatly benefits from the combined trend of 5G connectivity (including satellite 5G) and IoT (Internet of Things) with embedded AI solutions, both for industrial and home solutions. It is also a leader in WiFi 6 solutions and has a leading position in WiFi 7.

出典: Ruije

As a software and IP company, CEVA is well-known among engineers and is often missed by investors interested in the IoT and 5G sectors.

It can be an interesting company at the very edge of technological progress in data processing and edge AI, as illustrated by CERN’s selection of it to help with some of the most complex data analyses ever performed by mankind.

最新のCEVA（CEVA）株式ニュースと開発

参照研究：

^{1. Rui Xu; Jaroslav Bán; Sarthak Kalani; Chen-Kai Hsu; Subhajit Ray; Brian Kirby. A Radiation-Hard 8-Channel 15-Bit 40-MSPS ADC for the ATLAS Liquid Argon Calorimeter Readout. IEEE Explore. 28 2025年5月. pp 180 – 199 DOI:10.1109/OJSSCS.2025.3573904}