Physics studies nature in general. The fundamental basis of physics is based on repeatable observations, which leads to knowledge being built. In semiconductor manufacturing, this translates into a deep understanding of the “related” parameters in addition to the main goal of a specific operation.Creating an IC is a sequential process where hundreds to a few thousand steps are used in sequence from concept/design/verification/manufacturing/tests/packaging to achieve something that works most of the time. This sequential process is a web of interactions that are dictated by first principles. Understanding the relevance of these interactions helps guide the team to the best possible outcome.Working in this manufacturing environment is most effective when first principles are used to build a manufacturing sequence for a given system, but first principles also define the entire system.These constraints are tightly coupled. A change in doping concentration influences electric fields, which in turn affect breakdown behavior. Layout decisions shape parasitic effects, while process variations alter performance and yield. None of these factors operate in isolation. Each propagates through the system; it’s very much like pulling one node of a spiderweb ultimately moves the whole web.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776361522038-simplechips+alain.jpeg" style="width:100%px" class="fr-fic fr-dib" />Alain R. Comeau (M. Sc.), Founder, CEO, and CTO at SimpleChips (Credit: SimpleChips)SimpleChips’ founder and CEO, Alain R. Comeau built his career on a physics foundation. He observed that characteristics of physicists, which are really helpful in the industry, include: <ul><li>Deep understanding of what is possible</li><li>Technical flexibility </li><li>Endless curiosity </li><li>Perspective of technical landscape</li><li>Global picture</li><li>Good sense of observation</li><li>Need for understanding </li></ul><h2>From Physics to Fabrication</h2>Educated in solid-state physics at the University of Sherbrooke, Quebec, M. Comeau entered the industry through yield engineering and process integration at Mitel Semiconductor. Yield engineering focuses on identifying the physical origins of device failure and variation, tracking observed behavior back to underlying mechanisms.Process integration adds another layer of complexity. It requires coordinating multiple fabrication steps—implantation, diffusion, deposition, and etching—while ensuring that the resulting device structure behaves as intended. Each step introduces constraints, and each interaction carries consequences.In this context, problems are rarely localized. A yield issue observed at test may originate from a subtle interaction between process steps several layers earlier. Similarly, a reliability concern may trace back to electric field distribution shaped by layout decisions. Even a performance deviation may reflect material behavior under specific material, thermal or electrical conditions.Addressing these issues requires a mindset that does not isolate variables prematurely but instead traces cause and effect across domains. This is where a physics background becomes operational.Rather than treating fabrication and design as separate concerns, a physics-driven approach connects them through underlying principles. Device behavior becomes something to be understood and anticipated during design and process definition rather than an outcome to be measured only after fabrication.<h2>Bridging Design and Manufacturing</h2>Throughout his career, M. Comeau worked in roles that reinforced this perspective. From process integration and yield problem resolution to technology transfer across fabs and advanced device development, his work consistently involved <a href="https://www.wevolver.com/article/simplechips-a-systems-first-approach-to-high-voltage-ic-design">systems</a> where multiple variables interact simultaneously.Technology transfer, for example, is not a procedural exercise. Moving a CMOS process from one fabrication facility to another involves differences in equipment, materials, environmental conditions, and process control. Even small changes can lead to measurable differences in device behavior and yield. Ensuring consistency requires understanding the physics that governs their outcomes.Similarly, work on silicon avalanche photodiodes or surge protection devices involves managing high electric fields, breakdown mechanisms, and reliability constraints under extreme conditions. These are not problems that can be solved through incremental adjustment alone. They require a clear understanding of how devices behave at their limits.Over time, this kind of exposure builds a distinctive form of expertise, grounded in the ability to interpret behavior, identify root causes, and navigate complex interactions with clarity. Patterns emerge, failure mechanisms become recognizable, and solutions can be evaluated for both immediate effectiveness and long-term stability.This depth becomes particularly valuable at the boundary between design and manufacturing. In many organizations, these domains are separated structurally and operationally:<ul><li>Design teams focus on circuit performance and functionality. </li><li>Manufacturing teams focus on yield, repeatability, and process control. </li></ul>Communication between the two is often iterative, with issues discovered after fabrication and addressed through successive design or process modifications. This separation introduces friction. Design decisions may not fully account for process constraints, and manufacturing limitations may not be visible during early design stages. As a result, development cycles extend, and compromises accumulate.A physics-based perspective aligns both domains through the same underlying principles, allowing constraints to be addressed during design rather than after fabrication. Trade-offs can be evaluated within a unified framework that considers device behavior, process variability, and system requirements simultaneously.At SimpleChips, this integration is treated as a starting point, shaping how projects are approached from the outset. Design choices are considered in direct relation to fabrication realities, with device structures evaluated alongside process capabilities. Reliability and yield are embedded within the design itself, reducing iteration cycles, improving predictability, and reframing how problems are approached.<h2>Experience as a System-Level Advantage</h2>The trajectory of M. Comeau’s work illustrates how this perspective translates into device-level and process-level solutions.In high-voltage device design, controlling breakdown behavior is not a matter of selecting a target voltage and scaling geometry. It requires shaping electric fields across the device so that breakdown occurs in a controlled and repeatable manner. This involves careful definition of junction profiles, doping gradients, and layout structures that distribute field intensity without introducing localized stress points.Several of Alain’s <a href="https://www.simplechips.com/us-patents-and-ip/#:~:text=SimpleChips%20Technology%20Inc.-,PATENTS%3A,-US7701148">patented designs</a> address this class of problem, introducing device structures that manage electric field distribution and improve robustness under high-voltage conditions. Examples include <a href="https://patents.google.com/patent/US7701148B2/en">US7701148B2</a>, “HIGH VOLTAGE CURRENT SPLITTER AND METHOD,” and <a href="https://patents.google.com/patent/US10615132B2/en">US10615132B2</a>, “INTEGRATED CIRCUIT CAPABLE OF OPERATING AT VERY HIGH VOLTAGE AND MЕТHOD OF FABRICATING SAME.”A similar level of system awareness appears in work on avalanche-based components. Devices operating in avalanche regimes depend on stable carrier multiplication under high electric fields. Small variations in material properties or geometry can shift breakdown characteristics, introduce noise, or reduce long-term reliability. Designing for stability in these conditions requires anticipating how the device will behave across variations in process and temperature.This is where fabrication experience becomes inseparable from design.In process integration and yield engineering, problems often present themselves far from their origin. A yield loss observed at electrical testing may trace back to implant conditions, diffusion behavior, or subtle interactions between process steps. Resolving such issues requires reconstructing the sequence of physical transformations that led to the observed behavior.M. Comeau’s work across multiple fabs shows that fabrication and design operate as a single system. Variations in process conditions directly shape device behavior, requiring design decisions to account for how those variations manifest in silicon.Across such domains as high-voltage IC design, avalanche devices, yield optimization, and process transfer, the same method is applied, aligning device structure, process conditions, and operating requirements so that behavior remains stable across variation and over time.This is how experience accumulates as a system.<h2>From Device Physics to Company Design</h2>A physicist’s way of thinking begins with identifying governing mechanisms, understanding how they interact, and following their effects through the system. That perspective extends beyond devices and processes into how the company itself is structured.At SimpleChips, decisions remain closely tied to technical understanding, and the boundary between design and manufacturing is minimized. Project selection reflects the same principle, focusing on problems where depth of understanding can be applied effectively under real constraints.In semiconductor devices, behavior depends on how internal relationships are defined and controlled, from electric field distribution to material properties and process variation. Stability emerges when these elements remain coherent under operating conditions.At the <a href="https://www.wevolver.com/article/business-tai-chi-values-first-model-for-operating-a-deep-tech-company">organizational level</a>, technical capability, resource allocation, and decision-making must remain aligned for the company to operate effectively in a domain where constraints are tightly coupled, and outcomes depend on how well interactions are understood.M. Comeau’s career reflects this continuity. The same approach used to interpret device behavior and resolve fabrication challenges is applied to how the company evolves: grounded in fundamentals, attentive to interactions, and structured to remain coherent under real-world conditions.This is what allows a physicist to move from analyzing devices to building a company without changing the underlying method.The domain changes. The system does not.

Physics is about the web, rather than the spider.

A Physicists Approach to Semiconductor Manufacturing

A technical analysis of 3D printing in healthcare, covering materials, clinical applications, and regulatory considerations for engineers.

3D Printing in Healthcare: Engineering Applications and Impact

Drones and electric vertical takeoff and landing (eVTOL) aircraft are arriving faster than the measurement science needed to certify them. Community noise acceptance is the single largest non-safety barrier to urban air mobility deployment, with EASA's 2021 social acceptance study finding that noise ranked as European citizens' second-highest concern about air taxis, trailing only safety.[1] Yet the acoustic tools and standards that govern conventional aviation were never designed for the noise these vehicles produce.<a href="https://www.head-acoustics.com/">HEAD acoustics</a>, founded in 1986 by Prof. Dr. Klaus Genuit in Herzogenrath, Germany, has spent nearly four decades building expertise in binaural measurement, psychoacoustic analysis, and sound quality engineering. These capabilities, rooted in and refined over decades of automotive noise, vibration, and harshness (NVH) work, address both sides of the aerospace noise challenge: making aircraft quiet enough for communities below and comfortable enough for passengers inside, all under severe weight constraints that limit traditional acoustic treatments.This article explains why conventional noise metrics fall short for this vehicle class, and how HEAD acoustics' measurement, analysis, and simulation tools address each dimension of the problem.<h2>Decibels Alone Don't Capture How Annoying Drone Noise Really Is</h2>Understanding why drones sound so much worse than their decibel readings suggest requires looking beyond the numbers. A-weighted sound pressure level (dBA), the default metric in most noise regulations, cannot predict how annoying multirotor drone noise actually is. Multiple independent studies — at institutions including KAIST, Empa in Switzerland, RWTH Aachen, and Delft University of Technology — converge on the same conclusion: at equivalent sound exposure levels, multirotor drone noise is perceived as substantially more annoying than jet aircraft, helicopter, or road vehicle noise.[2]The reason lies in the acoustic signature. Multirotor drones produce strong tonal components at the blade-passage frequency and its harmonics, with multiple rotors operating at slightly different RPMs creating complex beating and modulation patterns.[3] Researchers at the Swiss Federal Laboratories for Materials Science and Technology (Empa), working in their 22-loudspeaker AuraLab, found that tonality, sharpness, and loudness were the psychoacoustic parameters most strongly linked to drone noise annoyance.[4] Meanwhile, the Effective Perceived Noise Level (EPNL) used in aircraft certification caps its tonal correction at 4.0 dB — a ceiling that researchers at RWTH Aachen have shown is inadequate for multi-rotor tonal content.[5] The Aures tonality method, standard in automotive sound quality, significantly outperforms EPNL for these complex sounds.[6]In short, the aerospace industry needs metrics that reflect how people actually hear and respond to drone noise, not just how loud it is on paper. That is the domain in which HEAD acoustics has operated for decades.<h2>HEAD Acoustics' Psychoacoustic Answer</h2>HEAD acoustics has been quantifying subjective sound perception since its founding. The company's ArtemiS SUITE software platform includes dedicated psychoacoustic analysis modules that calculate the very parameters research identifies as most critical. The basic module covers Zwicker loudness, sharpness, fluctuation strength, and tonality in accordance with DIN 45681. The advanced module builds on the Sottek Hearing Model — a proprietary approach that models the human auditory system's signal processing — to deliver refined analyses of loudness, roughness, and tonality.The Sottek Hearing Model is particularly relevant for aerospace. Its tonality analysis detects disturbing tonal components even in sounds with rapidly changing characteristics, such as those from electric motors — making it directly applicable to the modulated tonal signatures of multirotor drones. This capability contributed to the development of ECMA-74, the international standard for automatic detection and classification of tonal components. ArtemiS SUITE further implements standards including DIN 45631/A1, ISO 532-1, and ECMA 418-2, providing a standards-compliant foundation for emerging aerospace noise evaluation.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776353886198-Tonality_vs_time+%281%29.png" width="624" height="468" class="fr-fic fr-dii" />Tonality analysis of a drone flight sequence captured with ArtemiS SUITE. The spectrogram reveals tonal patterns — blade passage frequency harmonics, RPM shifts, and Doppler effects — that conventional dBA metrics cannot capture. Image credit: HEAD acoustics.<h2>Capturing What Passengers and Communities Actually Hear</h2>Knowing which psychoacoustic parameters matter is only useful if engineers can capture them accurately in the environments where drones and eVTOL aircraft actually operate. That requires measurement hardware designed for human-centric sound capture.<h3>Binaural Field Measurement</h3>HEAD acoustics' approach starts with capturing sound the way humans actually perceive it. The company's HMS and HSU families of artificial head measurement systems record binaurally, using anatomically accurate ear replicas that preserve the spatial cues our auditory system relies on to assess sound. This matters because the perceived annoyance of a sound depends heavily on its spatial properties, something monaural microphones cannot capture.<a href="https://www.head-acoustics.com/products/data-acquisition/squadriga-iii/" target="_blank" rel="noopener noreferrer">The SQuadriga III</a> mobile frontend brings laboratory-grade, 8-channel measurement to the field. For aerospace, HEAD acoustics developed an additional variant, the SQuadriga III-V1, which offers the ability to eliminate the lithium-ion batteries to comply with aviation regulations and air freight restrictions in flight-test environments.[7] These tools capture both exterior community noise during flyovers and interior cabin sound quality during flight, the two critical measurement scenarios for this vehicle class.<a href="https://www.head-acoustics.com/products/data-acquisition/squadriga-iii/" target="_blank" rel="noopener noreferrer"><img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776353899552-SQuadriga_III_Plane_Kabel+%281%29.jpg" width="624" height="261" class="fr-fic fr-dii" />The SQuadriga III mobile frontend delivers laboratory-grade, 8-channel sound and vibration measurement in the field, including inside aircraft cabins during flight. Image credit: HEAD acoustics.</a><h3>From Measurement to Design Decisions</h3>Capturing accurate acoustic data is the starting point, but the true benefit comes from converting that data into actionable design improvements. Measurement data feeds directly into ArtemiS SUITE for analysis, but HEAD acoustics' workflow extends into active design support. The PreSense NVH simulator lets engineers experience and evaluate sound designs in a virtual environment before physical prototypes exist. In the automotive domain, PreSense already underpins active sound design workflows, enabling interactive tuning, variant comparison, and jury panel validation, all without building hardware prototypes. This approach maps directly to eVTOL cabin sound tuning, where prototypes are especially expensive and scarce.The SoundSeat multimodal playback system adds another dimension by reproducing not just audible sound but the coupled vibrations passengers feel through their seat. This is an important distinction: passenger comfort in eVTOL aircraft depends on the full sensory experience, not just what reaches the ears. HEAD acoustics' SQala jury testing software then supports structured listening studies — the same methodology that Wisk Aero uses at Boeing's Critical Listening Facility[8] and that NASA employs in its VANGARD studies, exposing hundreds of participants to recorded eVTOL sounds across multiple U.S. cities.[9]<h2>Tracing Vibration From Rotors to Airframe</h2>Acoustic perception is one half of the equation. The other is understanding where problematic noise and vibration originate in the structure itself, and that is where vibro-acoustic diagnostics come in.<h3>The Drone Development Example</h3>During drone development, sound and vibration serve as early warning indicators of hidden structural problems. Engineers first establish a baseline by characterizing how the motor, mounts, and frame behave dynamically. If a design change — such as switching to lighter motor mounts — introduces a new 'buzz,' recapturing the structural dynamics can reveal a resonance that could lead to future failure: loosening bolts, deformation, or motor breakdown. This feedback loop enables informed decisions that improve both reliability and performance — and it is exactly the workflow HEAD acoustics' vibro-acoustic diagnostic chain supports.<h3>Transfer Path Analysis</h3>To diagnose the root cause of such issues, engineers need to trace how vibration travels from its source through the structure to the point where it becomes audible. Transfer Path Analysis (TPA), the established technique for tracing automotive powertrain noise through body structures, applies directly to eVTOL motor-to-cabin vibration paths. HEAD acoustics' ArtemiS SUITE TPA modules support methods including indirect force determination, effective mount transfer functions, and airborne attenuation determination. The TPA Project calculates individual noise contributions from each path. It generates a complete vibro-acoustic model, with results that can be synthesized in the time domain and fed directly into PreSense for virtual prototyping. The company's Binaural Transfer Path Analysis (BTPA) approach further enables troubleshooting of specific noise issues, as they would actually be perceived. It was originally developed for isolating automotive problems, like inverter whine and rotor-order noise, and it transfers naturally to eVTOL applications.<h3>Modal Analysis and Order Tracking</h3>Once vibration paths are mapped, the next step is to identify the structural resonances that amplify noise and to separate the contributions of individual rotors. The modal analysis project identifies structural resonance frequencies and mode shapes using the LSCF method with stability diagrams and MAC matrices — essential for ensuring drone and eVTOL frames do not operate near resonance. The companion impact measurement module supports the roving hammer and accelerometer methods for structural characterization. Order tracking, standard in automotive powertrain development, maps directly to rotor-order analysis. eVTOL vehicles with six to twelve propellers create complex order spectra where individual rotor contributions must be separated, and run-up measurements identify resonance crossings during motor spin-up.[10]<h3>Sound Source Localization</h3>Visual localization provides clarity when a noise problem has been identified, but its physical source remains unclear. The HEAD VISOR acoustic camera, paired with the VMA V array of up to 120 MEMS microphones, uses beamforming and AI-based Neural Deconvolution to localize noise sources in real time. Neural Deconvolution is currently the only commercially available real-time deconvolution method, dramatically improving the precision of source identification. For drone development, engineers can visually pinpoint which motor, mount, or structural component is responsible for a problematic sound — replacing trial-and-error troubleshooting with targeted diagnosis.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776353911709-Visor_3Images.jpg" width="624" height="351" class="fr-fic fr-dii" />The VMA V acoustic camera in base configuration (left), extended with spiral arms for up to 120 MEMS microphones (center), and rear view showing the integrated display (right). Image credit: HEAD acoustics.<h2>Automotive NVH Expertise Finds a Natural Second Home in Aerospace</h2>The capabilities described above, psychoacoustic evaluation, binaural measurement, transfer path analysis, modal analysis, and sound source localization, were all originally developed for the automotive industry. The reason they translate so directly to aerospace is rooted in physics. Electric motors produce tonal frequencies governed by powertrain topology and speed. Inverter switching noise creates torque ripple that manifests as propeller tonal emissions in drones, the identical mechanism that drives EV drivetrain tonal noise. And the absence of combustion-engine masking makes subtler sounds prominent — the same "unmasking" challenge that reshaped automotive NVH when EVs arrived.HEAD acoustics' ArtemiS SUITE offers over 150 analysis methods, including the complete psychoacoustic, TPA, and order-tracking toolchain for both community noise assessment and cabin sound quality assessment. The company has already moved toward aerospace: the battery-free SQuadriga III-V1 was developed for flight testing, and HEAD acoustics exhibited at AERO 2019, demonstrating aircraft interior noise optimization and headset testing. Regulatory urgency reinforces this timing. EASA's NPA 2025-03, published in August 2025, proposes the most comprehensive VTOL noise certification rule to date, targeting adoption by 2027.[11] The FAA is expanding its Rules of Particular Applicability framework, with six drone-specific certifications issued in 2023.[12] These developments are locking in measurement methodologies within two to three years.<h2>Bridging the Measurement Gap Before the Window Closes</h2>The convergence of electric propulsion, urban operations, and immature noise standards has created a measurement science gap in aerospace that automotive NVH expertise is positioned to fill. Psychoacoustic metrics, not decibels, will determine whether communities accept drones and air taxis overhead. The vibro-acoustic diagnostic chain that automotive engineers rely on daily transfers almost directly to eVTOL development. And the regulatory window is open but closing.<a href="https://www.headacoustics.com/">HEAD acoustics</a>' distinctive advantage lies in its psychoacoustic depth — the very capability that independent research consistently identifies as most lacking in current aerospace noise assessment. The tools, methods, and expertise to solve aerospace's acoustic challenges already exist. They have been proven across nearly four decades of automotive sound-quality engineering. Now they are taking flight. Check them out at <a href="https://vtol.org/forum-and-events/annual-forum-and-technology-display/technical-paper-schedule" target="_blank" rel="noopener noreferrer">the Future of Vertical Flight, May 5-7, 2026</a>! <hr /><h3>References</h3><ol><li>EASA (2021). Study on the societal acceptance of Urban Air Mobility in Europe. Available at: <a href="https://www.easa.europa.eu/sites/default/files/dfu/uam-full-report.pdf">easa.europa.eu/sites/default/files/dfu/uam-full-report.pdf</a> </li><li>Woodcock, J. et al. (2025). Human response to eVTOL drone sound: an online listening experiment exploring the effects of operational and contextual factors. Frontiers in Acoustics, 3. Available at: <a href="https://www.frontiersin.org/journals/acoustics/articles/10.3389/facou.2025.1624669/full">frontiersin.org/journals/acoustics/articles/10.3389/facou.2025.1624669/full</a> </li><li>Raza, W. and Stansbury, R.S. (2025). Noise Prediction and Mitigation for UAS and eVTOL Aircraft: A Survey. Drones, 9(8), p.577. Available at: <a href="https://www.mdpi.com/2504-446X/9/8/577">mdpi.com/2504-446X/9/8/577</a> </li><li>Kawai, C. et al. (2024). Short-term noise annoyance towards drones and other transportation noise sources: A laboratory study. The Journal of the Acoustical Society of America, 156(4), pp.2578–2595. Available at: <a href="https://pubs.aip.org/asa/jasa/article/156/4/2578/3316980/Short-term-noise-annoyance-towards-drones-and">pubs.aip.org/asa/jasa/article/156/4/2578/3316980/Short-term-noise-annoyance-towards-drones-and</a> </li><li>Sahai, A.K. and Stumpf, E. (2013) 'A comparative analysis of the subjective assessment of standard and noise abatement flight procedures', Proceedings of the Inter-Noise Congress and Conference, Denver, CO, pp. 255–270. Available at: <a href="https://ince.publisher.ingentaconnect.com/contentone/ince/incecp/2013/00000246/00000001/art00034">ince.publisher.ingentaconnect.com/contentone/ince/incecp/2013/00000246/00000001/art00034</a></li><li>Torija, A.J. et al. (2019). On the assessment of subjective response to tonal content of contemporary aircraft noise. Applied Acoustics, 146, pp.190–203. Available at: <a href="https://www.sciencedirect.com/science/article/abs/pii/S0003682X1830611X">sciencedirect.com/science/article/abs/pii/S0003682X1830611X</a> </li><li>HEAD acoustics (2021) 'Mobile measurement technology above the clouds — even without battery' [Press release]. Available at: <a href="https://www.head-acoustics.com/news-events/press-releases/press-details/show/mobile-measurement-technology-above-the-clouds-even-without-battery">head-acoustics.com/news-events/press-releases/press-details/show/mobile-measurement-technology-above-the-clouds-even-without-battery</a></li><li>Wisk Aero (2023) 'The science of silence: understanding the impact of noise on passenger comfort in UAM'. Available at: <a href="https://wisk.aero/newsroom/the-science-of-silence-understanding-the-impact-of-noise-on-passenger-comfort-in-uam">wisk.aero/newsroom/the-science-of-silence-understanding-the-impact-of-noise-on-passenger-comfort-in-uam</a> </li><li>Colucci, F. (2023) 'Lost in the noise', Vertiflite, May/June. Available at: <a href="https://evtol.news/news/lost-in-the-noise">evtol.news/news/lost-in-the-noise</a></li><li>Garg, S. et al. (2025) 'Active vibration reduction for eVTOL aircraft through rotor speed modulation', Aerospace Science and Technology. Available at: <a href="https://www.sciencedirect.com/science/article/abs/pii/S1270963825002834">sciencedirect.com/science/article/abs/pii/S1270963825002834</a></li><li>Aviationweek.com. (2025). EASA Proposes Noise Certification Rules For EVTOLs | Aviation Week Network. Available at: <a href="https://aviationweek.com/aerospace/advanced-air-mobility/easa-proposes-noise-certification-rules-evtols">aviationweek.com/aerospace/advanced-air-mobility/easa-proposes-noise-certification-rules-evtols</a> </li><li>Faa.gov. (2022). Noise Certification of UAS/AAM using Rules of Particular Applicability | Federal Aviation Administration. Available at: <a href="https://www.faa.gov/about/office_org/headquarters_offices/apl/aee/noise/uas_noise_certification">faa.gov/about/office_org/headquarters_offices/apl/aee/noise/uas_noise_certification</a> </li></ol>

The aerospace industry's pivot toward drones and electric air taxis has created an urgent noise problem, and the measurement science to address it is migrating directly from automotive NVH labs.

Sound and Vibration Analysis Takes Flight in Aerospace

Attaining design and performance goals for new products can often hinge on material choice. Whether the aim is to make a more sustainable part, optimize a component’s strength-to-weight ratio, or bring costs down, material properties dictate how a part is made and how it will behave. For some applications, existing off-the-shelf materials will do just fine; however, for many high-performance applications, materials with tailored property profiles must be made.Avient, a global materials solutions company, has built a business on this need, specializing in the development of advanced composites and engineered polymer formulations for its clients. The company also understands firsthand the challenges of developing new polymers, including how many cycles of compounding and testing can go into a material before it is ready for validation, the delays associated with multi-party supply chains, and the complexity in bridging material development and application testing. An improved workflow integrated at Avient has enabled its material development teams to address these issues and accelerate their development processes through early validation and real-world application testing. The solution? In-house filament production.  <h2>Bringing Filament Extrusion in-House</h2>Through a structural upgrade—the integration of in-house filament extrusion—Avient is realizing greater independence and agility in its material development workflow. This has enabled it to address common hurdles associated with the development of specialty materials, such as slowdowns related to multiple testing cycles, reliance on external filament suppliers, and a general disconnect between material development and application testing. Previously, new formulations engineered by material developers like Avient would have to be compounded in-house, after which samples would be ordered from external filament manufacturers. From there, the filament would be sent back to them for prototype 3D printing. With long delivery times and feedback needed from both Avient and its customers, material validation was therefore often plagued by inflexible, slow workflows. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776348766435-1776348766434.jpeg" width="624" height="416" class="fr-fic fr-dii" />Avient team with material samplesAvient’s new workflow, enabled by filament extrusion technology, is far more agile and offers a continuous feedback loop that links material formulation, processing, and application. In more specific terms, Avient’s development teams make compounds or pellets of tailored thermoplastics, which can then be converted directly into filament. This filament can then be used to 3D print functional prototypes, test parts, or tooling components to evaluate performance and application fit. This not only keeps validation steps in-house (thus minimizing lead times), it also enables Avient to make on-the-fly changes to material formulations based on application requirements and keep its customers in the loop from an early stage.By integrating filament production, the company has shown that it is possible to (a) create filament test samples without relying solely on external partners and (b) demonstrate real-world applications at an early stage. This has led to faster development and validation and—as we will see in more detail—greater confidence from customers who can see and test prototypes from early on in the material validation cycle.<h2>Tactile Evaluation with Density-Modified Plastics</h2>One of Avient’s product lines, Gravi-Tech™ Density-Modified Formulations, is focused on density-modified polymers, which are known for having similar characteristics to metal—in terms of mechanical properties as well as look and feel. High-performance polymers are increasingly being used for metal replacement applications, as they meet desired properties while offering advantages like weight reduction and lower material costs. In one metal replacement use case, Avient was developing Gravi-Tech™ Density-Modified Formulations for a luxury application. In the development process, the Avient team was able to produce filament and directly 3D print prototypes—in this case, a premium bottle cap—that matched the design of the customer’s product. The 3D printed prototype, which featured custom branding and a unique design, enabled Avient to not only assess the material’s properties but also to show the client how the product would look and feel. In other words, rather than iterate this prototype further down the material development road, Avient was able to test this real-world application early to understand its perception by end users before committing to downstream manufacturing routes. From the customer’s perspective, they were able to interact with a prototype that had high-fidelity visual, tactile, and functional properties far before production was initiated, giving them confidence in the material validation process. <h2>Real-World Testing of Thermally Conductive Materials</h2>In-house filament production has also been valuable in other cases. In the development of Therma-Tech™ Thermally Conductive Formulations, for example, Avient was able to evaluate both material properties and part design. The part in question was a heat sink that required excellent heat transfer, dissipation, and cooling properties.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776348766443-1776348766443.jpeg" width="486" class="fr-fic fr-dii" style="width:598px;height:784.574px" />Close-ups of filament and printed samplesWith a material validation workflow that included both filament extrusion and 3D printing, Avient was able to effectively test its material formulations and a range of geometries to find the best combination for form and functional behavior. Being able to test the Therma-Tech™ Thermally Conductive Formulations’ heat-dissipating properties at an application level supported faster and earlier learning, before Avient and its client progressed to tooling and full-scale production. Put another way, the heat sink use case demonstrates how specialty materials, like thermally conductive polymers, can be validated through real part designs. This combines two previously separate workflows into one, bridging material development and functional performance requirements from an early stage. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776348766463-1776348766463.png" width="624" height="404" class="fr-fic fr-dii" />Gravi-Tech™ Density-Modified Formulations (left) and Therma-Tech™ Thermally Conductive Formulations (right) functional partsThe Gravi-Tech™ Density-Modified Formulations and Therma-Tech™ Thermally Conductive Formulations use cases are just the tip of the iceberg. Avient has leveraged its in-house filament production to enhance the validation of many materials and applications, including automotive lighting parts and various plastic components for metal replacement.<h2>De-Risking Material and Application Validation</h2>By integrating filament extrusion as a step in its validation processes, Avient has experienced a range of benefits. For one, the ability to turn material compounds into 3D printable filament in-house has reduced the company’s reliance on external filament suppliers, particularly during early validation. This capability has also made it possible to prototype parts from new material formulations, effectively enabling application fit and form testing earlier in the development cycle. Not only a boon for making validation processes more efficient, this also supports internal and custom-facing discussions, creating a continuous feedback loop.  Crucially, all this can be done before any investment in tooling is required, allowing both material formulations and product designs to be optimized and finalized before committing to full-scale production. This can be particularly valuable in metal replacement applications, where an early understanding of how material properties and part design converge can actually de-risk material substitution. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776348766475-1776348766475.jpeg" width="440" class="fr-fic fr-dii" style="width:706px;height:927.65px" />Avient Gravi-Tech™ Density-Modified Formulations at 3devo’s boothThe impacts are not only being felt by Avient: other material developers can and are reaping these same benefits by harnessing in-house filament production and 3D printing prototypes for application-led validation. <h2>Conclusion</h2>Filament extrusion systems, such as those developed by Dutch company 3devo, are giving material developers greater independence and control over their validation processes. As Avient demonstrates, filament extrusion facilitates early validation and application-driven testing in the development of special material formulations. This not only offers internal benefits, in terms of reducing reliance on external filament suppliers and accelerating material testing, it also changes the dynamic between material developers and their customers. As Avient has found, the ability to produce filament and rapidly iterate 3D printed prototypes has led to better communication and transparency with its clients, who can better assess application performance, weight reduction, and manufacturability before heading into full-scale production.To explore the full case study and see how this approach can apply to your workflow, visit: <a href="https://www.3devo.com/case-studies/avient">https://www.3devo.com/case-studies/avient</a>

Advanced polymer developer Avient brings 3devo filament production in house for streamlined material testing and validation.

Bridging Material and Application Development with In-House Filament Production

Practical and architectural insights into IoT gateways, including edge versus cloud placement, protocol translation, data aggregation, hardware designs, and deployment patterns for industrial IoT.

IoT Gateway Architecture: Edge vs. Cloud, Protocol Translation, and Deployment Patterns

<h3 id="isPasted">Robots are becoming increasingly capable in vision and movement, yet touch remains one of their major weaknesses. Now, researchers have developed a miniature tactile sensor that could give robots something much closer to a human sense of touch.</h3><blockquote>Most existing tactile sensors are either too bulky, too fragile, too complex to manufacture or unable to accurately distinguish between normal and tangential forces.&lt;cite&gt;Professor Tawfique Hasan&lt;/cite&gt;</blockquote>The technology, developed by researchers at the University of Cambridge, is based on liquid metal composites and graphene – a two-dimensional form of carbon. The ‘skin’ allows robots to detect not just how hard they are pressing on an object, but also the direction of applied forces, whether an object is slipping, and even how rough a surface is, at a scale small enough to rival the spatial resolution of human fingertips. Their <a href="http://www.nature.com/articles/s41563-026-02508-7" target="_blank">results</a> are reported in the journal Nature Materials.Human fingers rely on multiple types of mechanoreceptors to sense pressure, force, vibration, and texture simultaneously. Reproducing this level of multidimensional tactile perception in artificial systems is a significant challenge, especially in devices that are both small and durable enough for practical use.“Most existing tactile sensors are either too bulky, too fragile, too complex to manufacture or unable to accurately distinguish between normal and tangential forces,” said <a href="https://www.eng.cam.ac.uk/profiles/th270">Professor Tawfique Hasan</a> from the <a href="https://www.graphene.cam.ac.uk/" target="_blank">Cambridge Graphene Centre</a>, who led the research. “This has been a major barrier to achieving truly dexterous robotic manipulation.”<iframe frameborder="0" height="315" src="https://www.youtube.com/embed/rGr28Xm1bu4?si=LPNkRQpF63gVLGeM" width="100%"> </iframe>To overcome this, the research team developed a soft, flexible composite material, combining graphene sheets, deformable metal microdroplets, and nickel particles, embedded in a silicone matrix.Inspired by the microstructures found in human skin, the researchers shaped the material into tiny pyramids, some as small as 200 micrometres across. These pyramid structures concentrate stress at their tips, enabling the sensor to detect extremely small forces while maintaining a wide measurement range.The result is a tactile sensor sensitive enough to detect a grain of sand. Compared with existing flexible tactile sensors, the new device improves size and detection limits by roughly an order of magnitude.The sensor can also distinguish shear forces from normal pressure, a capability that allows it to detect when an object begins to slip. By measuring signals from four electrodes beneath each pyramid, the sensor can mathematically reconstruct the full three-dimensional force vector in real time.In demonstrations, the team integrated the sensors into robotic grippers. The robots were able to grasp fragile objects, such as thin paper tubes, without crushing them. Unlike conventional force sensors, which rely on prior information about an object’s properties, the new system adapts in real time through slip detection.At even smaller scales, microsensor arrays could identify the mass, geometry, and material density of tiny metal spheres by analysing force magnitude and direction. This opens the door to applications in minimally invasive surgery or microrobotics, where conventional force sensors are far too large.Beyond robotics, the technology could have significant implications for prosthetics. Advanced artificial limbs increasingly rely on tactile feedback to provide users with a sense of touch. Highly sensitive, miniaturised 3D force sensors could enable more natural interactions with objects, improving control, safety, and user confidence.“Our approach shows that bulky mechanical structures or complex optics are not required to achieve high-resolution 3D tactile sensing,” said lead author <a href="https://faculty.ustc.edu.cn/yuanguolin/en/lwcg/207132/list/index.htm">Dr Guolin Yun</a>, a former Royal Society Newton International Fellow at Cambridge, and now Professor at the University of Science and Technology of China. “By combining smart materials with skin-inspired structures, we achieve performance that comes remarkably close to human touch.”Looking ahead, the researchers believe the sensors could be miniaturised even further, potentially below 50 micrometres, approaching the density of mechanoreceptors in human skin. Future versions may also integrate temperature and humidity sensing, moving closer to a fully multimodal artificial skin.As robots increasingly move out of controlled factory environments and into homes, hospitals, and unpredictable real-world settings, such advances in touch could be transformative — allowing machines not just to see and act, but to truly feel.A patent application has been filed through <a href="https://www.enterprise.cam.ac.uk/">Cambridge Enterprise</a>, the University’s innovation arm. The research was supported by the <a href="https://royalsociety.org/">Royal Society</a>, the <a href="https://www.royce.ac.uk/">Henry Royce Institute</a>, and the <a href="https://aria.org.uk/">Advanced Research and Invention Agency</a> (ARIA). Tawfique Hasan is a Fellow of Churchill College, Cambridge. Reference: Guolin Yun et al. ‘<a href="https://www.nature.com/articles/s41563-026-02508-7">Multiscale-structured miniaturized 3D force sensors</a>.’ Nature Materials (2026). DOI: 10.1038/s41563-026-02508-7

Researchers have developed a miniature tactile sensor that could give robots something much closer to a human sense of touch.

Graphene-based artificial skin brings human-like touch closer to robots

Artificial intelligence (AI) now makes it alarmingly easy to manipulate photos, videos and audio recordings. Whether it is fabricated statements attributed to politicians or misleading images from crisis zones, social media and online platforms are already flooded with so-called deepfakes. The consequences for society and democracy are serious: an increasing number of people are being deceived by such forgeries or are beginning to mistrust even credible sources. Researchers at ETH Zurich have developed a new sensor technology that directly addresses this issue. The concept involves cryptographically signing images, videos or audio signals within a sensor chip at the exact moment they are captured. This signature allows for verification that the data genuinely originate from a camera or recording device, indicates when it was captured, and ensures that it has not been tampered with. “If data is signed the moment it is captured, any later manipulation leaves traces,” explains Fernando Cardes, who co-developed the technology. He is a research associate at Andreas Hierlemann’s Professorship of Biosystems Engineering within the Department of Biosystems Science and Engineering (BSSE) in Basel. “To manipulate the data, the chip would have to be physically attacked, requiring a massive technological effort so that the mass generation of manipulated content for social media platforms would be practically impossible,” he adds. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776333872274-image.imageformat.1286.329827099.jpg" style="width:100%px" class="fr-fic fr-dib" />How the technology works: a real-world event (1) is recorded by a camera whose sensor chip generates both the image data and a cryptographic signature at the moment of capture (2). Once stored in a public register (3), the signature can later be used to verify that the recording is authentic and has not been altered (4). (Graphic created using AI: Felix Franke / ETH Zurich)<h2>Simple verification thanks to a public register </h2>The signatures generated by the sensor could be stored by camera manufacturers in a publicly accessible, immutable ledger (e.g., a blockchain). This approach would enable anyone to verify the authenticity of the data in question at any time by comparing the chip’s signature stored in the ledger with the original data and confirming its source. “As such, it is barely of any relevance whether a person or the technology involved in data processing and transmission is trustworthy,” explains Felix Franke, who co-developed the chip at ETH Zurich and is now a professor at the University of Basel. “Trust in digital content is eroding. We wanted to create a technology that gives people a way to verify whether something is genuine.” In principle, the technology can be incorporated into any type of sensor or camera. In the future, social media platforms could automatically verify whether the content is genuine as soon as it is uploaded. Where this does not happen, journalists, researchers or public authorities could authenticate content themselves using simple tools. <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776333888604-image.imageformat.1286.1418804681.jpg" style="width:100%px" class="fr-fic fr-dib" />The sensor chip is a prototype that demonstrates technical feasibility.  (Image: Caroline Arndt Foppa / ETH Zurich)<h2>The threat of deepfakes identified at an early stage </h2>The idea behind the sensor chips originated as a side project at the Bio Engineering Laboratory at ETH Zurich. Long before AI systems like ChatGPT became a subject of public debate, the laboratory had been working on highly sensitive sensors to measure electrical signals from living cells. The interdisciplinary team also had the necessary expertise to incorporate additional cryptographic functions directly into the sensor chips. “The danger posed by deepfakes was foreseeable,” recalls Franke. Therefore, as early as 2017, the plan was devised to develop a sensor the data of which could not be manipulated without detection. The chip described in a paper published today is a working prototype and demonstrates technical feasibility. Further steps are still required prior to commercial deployment. Nevertheless, the researchers are confident that, with current technologies and processes, the chip can be developed into a working, market-ready product. They have therefore filed a patent application. “We are currently exploring how to reduce costs for camera and sensor manufacturers should they wish to incorporate the new technology into their chips,” reports Cardes. The work was financially supported by the Swiss National Science Foundation (SNSF) and the State Secretariat for Education, Research, and Innovation (SBFI) through the SwissChips initiative.

Researchers at ETH Zurich have now developed chip technology that enables verification of the authenticity of sensor data, including images or videos.

Chips designed to help identify deepfakes

Battery life has quietly become one of the defining constraints of modern electronics. Wearables are expected to run for days, asset trackers must operate for years without maintenance, and industrial sensors are deployed in places where replacing a battery is both inconvenient and expensive. At the same time, these devices are doing more than ever, from running radios to sampling sensors, executing edge algorithms, and transmitting data across long distances.This growing performance envelope has placed unprecedented pressure on power design. In many cases, the weakest link is not the battery itself, but the way that battery energy is converted, regulated, and delivered to the system.<h2>The Real Challenge: Dynamic Loads, Not Static Specs</h2>Portable and IoT devices rarely operate under steady conditions. Instead, they move between two extremes:<ul><li>Long periods of microamp-level standby</li><li>Sudden, high-current bursts when radios transmit or processors wake</li></ul>This “bursty” load profile creates a fundamental engineering challenge. A converter optimized for heavy load may waste energy during standby. A converter tuned for light load may struggle to respond quickly when the system wakes. If voltage droop occurs during a transmit burst, the MCU may reset. If switching noise leaks into analog domains, performance can degrade.Power conversion in portable systems is not simply about stepping voltage down or up. It involves maintaining stability, efficiency, and signal integrity across radically different operating states.An effective portable converter must therefore satisfy three conditions simultaneously:<ol><li>High efficiency from microamps to peak load</li><li>Fast transient response to sudden current changes</li><li>Compact size with controlled EMI behavior</li></ol>Semtech’s approach to portable power conversion is built around these realities.<h2>Step-Down Conversion: Making the Most of a Li-Ion Cell</h2>In battery-powered systems built around single-cell Li-ion or Li-polymer batteries, the core challenge is efficient voltage reduction. The battery voltage varies across its discharge curve, yet the system rails require stable regulation. A synchronous buck converter is typically the most efficient way to translate that energy.Semtech’s SC194A, for example, is designed specifically for single-cell battery applications. It delivers high efficiency across both light and moderate loads, which is essential in devices that spend much of their lifetime asleep. Integrated MOSFETs reduce external component count, while low quiescent current prevents the converter itself from becoming a hidden drain on battery life.For designs that demand even tighter space constraints, integration becomes more important than raw output current. The SC202A addresses this by embedding the inductor directly inside the package. This dramatically reduces board footprint and simplifies layout. With approximately 500 mA output capability, it is well suited to compact IoT nodes, wearables, and tracking devices where every square millimeter matters.In both cases, the objective is to maintain high efficiency during standby while remaining responsive during wake events.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776256948367-SEO+ARTICLE+EMBED.png" style="width:100%px" class="fr-fic fr-dib" /><iframe id="JotFormIFrame-260843278583366" src="https://form.jotform.com/260843278583366" frameborder="0" style="min-width:100%;max-width:100%;height:800px;border:none" scrolling="YES" class="fr-draggable"> </iframe><h2> </h2><h2> </h2><h2> </h2><h2> </h2><h2> </h2><h2> </h2><h2> </h2><h2>Step-Up Conversion: Extracting Energy to the Last Millivolt</h2>Not all systems operate above their battery voltage. Some require rail elevation, such as generating a stable 3.3 V or 5 V output from a nearly depleted primary cell. This is where synchronous boost converters play a critical role.Semtech’s SC120 provides a wide input range, starting as low as 0.7 V, with startup capability around 0.865 V. This allows systems to operate even as batteries approach end-of-life. Rather than shutting down prematurely, the converter continues to extract usable energy.With efficiencies up to 94% and features such as automatic power-save mode and ultra-low shutdown current, the SC120 is engineered for low- to moderate-load boost applications. Anti-ringing circuitry helps control EMI, a non-trivial concern when switching frequencies interact with analog subsystems. In practical terms, this means longer runtime from the same battery, not because the battery changed, but because the converter did.<h2>Converter Selection Is a System Decision</h2>Choosing the right portable converter requires more than browsing a current rating table. It demands alignment across four dimensions of the design:<ol><li>Battery type: Li-ion, alkaline stack, or higher-voltage bus</li><li>Load profile: intermittent standby versus sustained current draw</li><li>Transient demands: how quickly the load changes</li><li>Physical constraints: size, EMI sensitivity, and thermal limits</li></ol>For example:<ul><li>A single-cell Li-ion system typically benefits from a synchronous buck topology.</li><li>An alkaline stack driving higher rails requires boost conversion.</li><li>Light, intermittent loads favor pulse-frequency modulation or auto-skip modes to minimize switching losses.</li><li>Sustained heavy loads benefit from pulse-width modulation for consistent behavior.</li><li>Tight transient requirements may call for adaptive on-time or current-mode control to avoid large output capacitors.</li></ul>These trade-offs are interconnected. Increasing switching frequency reduces inductor size but may raise switching losses. Improving transient response may require tighter layout discipline. And reducing EMI may require filtering and loop-area minimization. Converter selection is, therefore, not an isolated choice. It shapes the behavior of the entire power tree.<h2>Layout Is Not an Afterthought</h2>Even the most efficient converter can underperform if poorly placed. In compact portable systems, analog, digital, and power domains often sit within millimeters of each other. The high di/dt loop formed by the switch node, inductor, and output capacitor carries fast current transitions. If this loop area is not minimized, radiated EMI increases and noise may couple into sensitive traces.Short, wide traces on the switch node help contain voltage transitions. Ground shielding and careful copper placement reduce field propagation. Thermal vias and copper pours distribute heat during high-duty transmit cycles.These layout decisions determine whether theoretical efficiency translates into real-world performance. Early EMI filtering and pre-compliance scans further reduce risk. Waiting until the final layout to address EMI typically increases cost and redesign cycles.<h2>Validation Defines Reliability</h2>Portable power converters must be validated across the full operating envelope, not just nominal load. A proper evaluation includes:<ul><li>Efficiency across the full load curve</li><li>Line and load transient response</li><li>Startup and inrush behavior at low battery voltage</li><li>Quiescent current during standby</li><li>Thermal behavior across duty cycles</li><li>Pre-compliance EMI scanning</li></ul>This structured validation ensures the converter supports both battery longevity and system integrity.<h2>Power Conversion as a Strategic Design Lever</h2>In portable and IoT devices, the converter is often viewed as a supporting component. In practice, it defines runtime, stability, thermal margin, and signal quality. Semtech’s synchronous buck and boost portfolio reflects a design philosophy centered on this reality. High efficiency across dynamic loads, integrated solutions that reduce footprint, wide input flexibility, and low quiescent current operation together enable devices to operate longer, run cooler, and behave more predictably.From wearable health monitors to remote industrial sensors, effective power conversion involves both meeting a voltage spec and shaping how the device lives, from how long it runs to how reliably it wakes, and how cleanly it communicates.With battery replacement being costly and downtime unacceptable, portable power converters have become foundational infrastructure. Semtech’s portfolio provides the efficiency, integration, and control needed to support long-term, reliable operation across modern IoT designs.To explore the full architectural approach, including system-level power strategies and ultra-low-Iq regulation for sensitive domains, download the complete Semtech whitepaper, Efficient Power Management for Portable and IoT Devices. It provides deeper technical guidance across all three pillars of portable power design.

Battery life has quietly become one of the defining constraints of modern electronics.

Rethinking Portable Power: Converters That Keep IoT Devices Alive

Understanding this evolution helps us appreciate how far the industry has come and where it’s heading next.In this post, we explore the significant milestones that have shaped modern manufacturing, from early mechanisation to today’s smart factories.<h2>1760s – The First Industrial Revolution: The Birth of Mechanisation</h2>The first major leap in manufacturing began in Britain during the 1760s. Innovations such as the spinning jenny and James Watt’s enhanced steam engine paved the way for mechanised production. These breakthroughs led to a shift in manufacturing from small, artisanal workshops to large-scale factories, particularly in the textile and iron industries.This era marked the beginning of industrial-scale production, laying the groundwork for all subsequent revolutions.<h2>1870s – The Second Industrial Revolution: Speed, Scale &amp; Efficiency</h2>By the 1870s, manufacturing entered a new age of rapid expansion. Electricity became widespread, and mass production techniques took hold. The introduction of interchangeable parts and assembly line manufacturing, famously refined by Henry Ford in the 1910s, dramatically improved productivity.Meanwhile, the growing networks of railroads and telegraph systems enabled manufacturers to move goods and communicate more quickly than ever before, thereby creating more interconnected supply chains.<h2>1940s–1950s – Post-War Industrial Boom: Automation Begins</h2>In the post-war decades, factories experienced a surge of innovation. Early computers and numerical control (NC) machines initiated the first wave of automation, enhancing production processes by making them more consistent and repeatable.At the same time, the rise of television and radio advertising drove mass consumerism, increasing demand for manufactured goods and pushing industries to scale further.<h2>1960s–1970s – Electronics &amp; Early Automation: Robots Enter the Scene</h2>The 1960s and ’70s saw electronics play a major role in shaping manufacturing. Programmable Logic Controllers (PLCs) have become the backbone of factory control systems, offering flexible and reliable automation.In 1961, General Motors introduced the world’s first industrial robot into its assembly line, marking a historic moment. Robotics quickly became synonymous with advanced automotive manufacturing.<h2>1980s – The Digital Revolution: Computers Transform Production</h2>As personal computers became widely available in the 1980s, manufacturing began to embrace digitalisation. Tools such as Computer-Aided Design (CAD) and Computer-Aided Manufacturing (CAM) revolutionised product development, speeding up design cycles, improving accuracy, and enabling more complex engineering.This decade laid the digital groundwork for modern automated production.<h2>1990s–2000s – Lean, Globalised Manufacturing: Efficiency Meets Scale</h2>The final years of the twentieth century ushered in a wave of efficiency-focused strategies. Lean manufacturing principles gained widespread adoption, aiming to reduce waste and increase value.Globalisation also reshaped manufacturing landscapes, with production shifting to international hubs for cost efficiency. Early Enterprise Resource Planning (ERP) systems emerged, integrating operations across departments and locations for the first time.<h2>2010s – Smart Manufacturing &amp; Industry 4.0: Connected Factories</h2>The 2010s marked the rise of smart manufacturing. With the explosion of smartphones, IoT devices, and industrial sensors, manufacturers gained real-time visibility into their operations.Industry 4.0 emerged, defined by cyber-physical systems, automation, AI, data exchange, and advanced analytics. Factories became smarter, more connected, and increasingly self-optimising.<h2>2020s – Advanced Manufacturing &amp; Sustainability: The Future Takes Shape</h2>Advanced technologies dominate today’s manufacturing landscape. Additive manufacturing (3D printing) has moved from prototyping to mainstream industrial use. Digital twins, AI-powered optimisation, and ultra-fast 5G connectivity enable real-time decision-making on an unprecedented scale.Sustainability has also become a core priority, with industries focusing on green energy, circular economy principles, and reducing environmental impact.<h2>Conclusion: From Steam to Smart Factories</h2>The evolution of manufacturing is a story of continuous transformation. Each revolution has built upon the last, driving us from steam-powered machinery to fully connected, intelligent manufacturing ecosystems.As we look ahead, the future promises even greater innovation. With AI, automation, and sustainability at the forefront, the next chapter of manufacturing is already being written.

Manufacturing has undergone extraordinary change over the past few centuries. What began as manual labour in small workshops has transformed into a globally integrated, highly automated ecosystem powered by digital technology and real-time data.

The Evolution Of Manufacturing

Radar has long been a foundational sensing technology in defense and critical infrastructure. It excels at detecting objects at distance, operating reliably in darkness, smoke, fog, and adverse weather where optical systems fail. Yet traditional radar systems can determine where an object is, but not necessarily what it is. Addressing this challenge is central to Physical AI, where systems must interpret sensor data and act on it instantly in real-world environments.The rapid proliferation of Unmanned Aerial Vehicles (UAVs) has introduced a new class of low-cost, highly mobile aerial threats. Commercial drones are widely available, inexpensive, and increasingly capable. At the same time, birds, debris, and environmental clutter continue to populate the skies. Distinguishing between a harmless bird and a hostile drone is no trivial problem. It is an operational requirement.<a href="https://brainchip.com/leadgen-radar-platform-registration/" target="_blank" rel="noopener noreferrer">BrainChip’s Radar Reference Platform</a> addresses this gap by embedding intelligent, low-power classification directly into the radar signal chain. Built around the Akida™ neuromorphic processor, the platform demonstrates how micro-Doppler analysis can enable real-time identification and classification of flying objects without relying on GPUs or cloud infrastructure, bringing Physical AI into operational radar systems.<h2>The Identification Gap in Modern Radar Systems</h2>Traditional radar systems detect motion and range using reflected radio waves. They provide position, velocity, and sometimes size estimations. While this can be good enough for many applications, in today’s aerial threat landscape, detection alone is insufficient.Security operators face three compounding challenges:<ol><li>Proliferation of UAVs introduces unpredictable aerial activity</li><li>Limitations of cameras and RF scanners in poor weather, night conditions, or spoofed environments</li><li>False alarm fatigue caused by radar systems unable to distinguish drones from birds or debris</li></ol>Cameras depend on line-of-sight visibility, while RF scanners depend on signal emissions that can be spoofed or suppressed. Radar remains the most robust sensing modality across conditions. However, without classification capability, operators must manually interpret signals or escalate unnecessarily.This results in wasted time, misallocated resources, and in defense scenarios, potentially expensive countermeasures deployed against low-cost targets. The missing capability is, therefore, intelligent classification at the sensor edge.<h2>Micro-Doppler: Extracting Identity from Motion</h2>When an object moves, radar detects Doppler shifts caused by its velocity. But complex objects, such as drones, produce additional fine-grained motion signatures. Rotating propellers create distinct micro-Doppler patterns embedded within the radar return.These micro-motions act as a fingerprint. Bird wingbeats differ from propeller rotations, debris exhibits irregular motion, and fixed-wing drones differ from quadcopters. The information is present in the signal, but extracting and interpreting it in real time remains an engineering challenge.Micro-Doppler classification requires continuous processing of high-resolution radar returns. Conventional AI hardware, particularly GPU-based systems, can perform this task but at a significant cost in power, size, and thermal complexity. Such architectures introduce cooling requirements, increased SWaP, and system-level constraints that limit field deployment.<h2>The Akida Radar Reference Platform</h2>BrainChip’s Radar Reference Platform combines:<ul><li>The Akida AKD1500 neuromorphic co-processor</li><li>A compact FMCW radar module</li><li>A micro-Doppler classification model optimized for <a href="https://brainchip.com/introducing-tenn-revolutionizing-computing-with-an-energy-efficient-transformer-replacement/">Temporal Event-Based Neural Networks (TENNs)</a></li><li>Raspberry Pi-based system integration</li></ul>This reference design demonstrates real-time classification of flying objects using neuromorphic computing at the edge, providing a validated blueprint for intelligent radar integration.<img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzc2MTc2Mjk2NjIzLUJyYWluQ2hpcCBSYWRhciBSZWYgUGxhdGZvcm0gTW9ja3VwLmpwZyIsImVkaXRzIjp7InJlc2l6ZSI6eyJ3aWR0aCI6OTUwLCJmaXQiOiJjb3ZlciJ9fX0=" style="width:611px" class="fr-fic fr-dib" />Image of the Akida Radar Reference Platform<h3>Event-Driven Processing for Radar Signals</h3>Akida processes radar data using TENNs and spatiotemporal CNN models, designed to capture both spatial and temporal features in dynamic signals. The architecture focuses on extracting meaningful temporal patterns from the signal without continuous, power-intensive computation.The input to the system is a time-frequency spectrogram generated from Frequency Modulated Continuous Wave (FMCW) radar data. Within this representation, micro-Doppler signatures appear as distinct textures formed by periodic micro-motions, such as propeller rotations or wingbeats. These patterns can be learned and recognized efficiently by models optimized for temporal signal interpretation.Micro-Doppler features manifest as repeating, time-dependent patterns, making them well-suited for TENN-based and spatiotemporal model architectures. This approach enables:<ul><li>Sub-2W total system power consumption</li><li>Low-latency inference directly on-device</li><li>Elimination of GPU dependency</li><li>Fanless, compact system design</li></ul>The result is a compact radar platform capable of classification within constrained power and thermal envelopes.<h2>Smart Classification Without Cloud Dependency</h2>A critical advantage of the Radar Reference Platform is complete on-device operation. All classification occurs locally with no reliance on cloud connectivity. This has significant implications, including:<ul><li>Comms-denied environments: The system operates independently without network access.</li><li>Data privacy and security: Sensitive radar data does not leave the device.</li><li>Low latency: Classification decisions occur immediately at the sensor.</li></ul>For defense and infrastructure monitoring, such autonomy is essential. In contested environments, transmitting data may reveal position. A system that can classify threats silently and locally preserves operational security while maintaining real-time responsiveness.<h2>Strategic Applications</h2>The Radar Reference Platform serves as a blueprint for deployments requiring real-time classification of motion signatures:<ul><li>Aerial Object Classification for Drones and UAVs: Micro-Doppler signatures enable reliable distinction between drones, birds, and debris in real time. This allows defense systems to classify targets early and avoid unnecessary countermeasures.</li><li>Defense and Early Threat Detection: On-device object identification supports real-time awareness in comms-denied environments. Systems operate silently, without data transmission, preserving operational security while maintaining responsiveness.</li><li>Automotive In-Cabin Sensing and Human Activity Detection: Radar can capture fine human motion patterns such as breathing or posture. This enables occupancy detection and health-related monitoring without cameras, preserving privacy.</li><li>Marine and Autonomous Platform Awareness: Radar-based classification helps distinguish between vessels, obstacles, and environmental motion in low-visibility conditions, supporting safer autonomous navigation.</li><li>Drone Countermeasures: Detection of propeller micro-Doppler signatures enables targeted response, reducing false alarms and improving response accuracy.</li></ul><h2>Ultra-Low SWaP for Portable Deployment</h2>Size, weight, and power are central constraints in field systems. The Akida-based radar platform delivers ultra-low SWaP characteristics:<ul><li>Compact FMCW radar module</li><li>Fanless architecture</li><li>Sub-2W total system power consumption, including Raspberry Pi 5, radar module, and Akida processor</li><li>Portable, man-deployable design</li></ul>This makes the system suitable for mobile early warning deployments, forward-operating environments, or temporary infrastructure setups. Conventional GPU-based solutions require active cooling and higher power budgets. The neuromorphic approach fundamentally reduces SWaP constraints while maintaining real-time classification capability.<h3>Reliability in Harsh Environments</h3>Radar inherently operates through smoke, fog, dust, and storms. By integrating classification directly into the radar platform, the system maintains robustness and performance.Unlike camera-based systems, it does not depend on visibility. And unlike RF detection systems, it does not rely on emitted signals that can be suppressed or spoofed. The reference design demonstrates validated, low-latency classification in a fully integrated stack. This reduces integration risk for product teams looking to embed intelligent radar into their solutions.<h2>A Blueprint for Intelligent Radar Products</h2>The Radar Reference Platform is not a standalone product. It is a complete, validated design that combines hardware, software, and AI modeling into a reproducible architecture ready for product adaptation.For defense contractors, infrastructure providers, and OEMs, it provides:<ul><li>Proven micro-Doppler classification capability based on TENNs</li><li>Integration framework around the AKD1500</li><li>Clear path from reference to deployment</li></ul>BrainChip’s ecosystem can adapt this design to specific product requirements, accelerating time to market while preserving low-power advantages.<a href="https://brainchip.com/leadgen-radar-platform-registration/" target="_blank" rel="noopener noreferrer"><img src="https://images.wevolver.com/eyJidWNrZXQiOiJ3ZXZvbHZlci1wcm9qZWN0LWltYWdlcyIsImtleSI6ImZyb2FsYS8xNzc2MTc1NTEwOTUyLTE3NzYxNzU1MTA5NTIucG5nIiwiZWRpdHMiOnsicmVzaXplIjp7IndpZHRoIjo5NTAsImZpdCI6ImNvdmVyIn19fQ==" width="602" class="fr-fic fr-dii" style="width:800px;height:197.75px" />BrainChip’s Radar Reference Platform Webinar April 20, 2026</a><h2>Redefining What Radar Can Do</h2>Radar has always provided visibility. The addition of intelligent classification transforms it into an autonomous decision-support system. Instead of reporting “an object detected at 300 meters,” intelligent radar can report “a quadcopter approaching.” That shift reduces false alarms, conserves resources, and increases operator confidence.By combining compact FMCW radar with neuromorphic processing, BrainChip’s Akida enables micro-Doppler classification at the edge, without GPUs, without cloud dependency, and without high power overhead. This moves radar systems beyond detection toward real-time understanding at the edge.For an exclusive deep dive into BrainChip’s Radar Reference Platform, check out the upcoming Webinar, April 20th: <a href="https://brainchip.com/leadgen-radar-platform-registration/">https://brainchip.com/leadgen-radar-platform-registration/</a>

Adding Intelligence to Radar with the BrainChip Real-Time Micro-Doppler Classification

Intelligent Radar Classification with Neuromorphic AI

The message is consistent: a new era of flexibility where the power to choose (and switch) providers is finally in the hands of the customer. On paper, every new IoT application hitting the market today should be built on this standard.But are they? The initial specifications were released in May 2023 by the GSMA, and the final version has been stable since June 2024. That’s plenty of time to see SGP.32-ready devices flooding the market, but where are these devices? Why don’t we see them everywhere yet? Is something still holding the industry back?Let’s unravel the SGP.32 story and find out if the standard is truly production-ready. <h3>SGP.32 Explained</h3>To understand the obsession, we have to look at how we got here. Consumer eSIMs are already a reality; the latest US iPhones don’t even have a physical SIM slot. For a human, scanning a QR code to download a profile is a quick and seamless process.But IoT devices are different. They are often resource-constrained, "headless" (no screen or camera), come in high volumes, and need to be managed via the cloud. The consumer standard (SGP.22) simply didn't fit these requirements. This led to the birth of SGP.32, featuring an architecture redesigned around three core elements:<ol><li>The Device: Running the IPA (IoT Profile Assistant) which resides inside the SIM (IPAe), or inside the device (IPAd).</li><li>The Orchestrator: The eIM (eSIM IoT Remote Manager), a cloud component which directly communicates with the IPAs.</li><li>The SIM database: The SM-DP+ (Subscription Manager Data Preparation Plus), the server where the actual SIM profiles live.</li></ol> <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776159366689-SGP.32.jpg" class="fr-fic fr-dib" />SGP.32 architecture (simplified)In essence, the IPA on the device "checks in" with one or multiple eIMs at configurable intervals to see if there are any pending instructions, like download, activate, delete or disable a SIM profile. If a download is pending, the device pulls the new profile from the SM-DP+ and, once enabled, switches connectivity without any physical intervention. Simple, right? But does it actually work in a messy, multi-vendor environment? <blockquote>Kigen is proud to be involved in the tests done by IoT Stars to showcase that the SGP.32 technology is mature. It delivers a true freedom and distinction between the hardware choices (plastic or soldered, eSIM form factor, grade) and the digital profiles which can be loaded on the eSIM and enabling the required connectivity or features offered by the various connectivity providers. Kigen stands ready to help the IoT devices makers of tomorrow with a wide ecosystem (notably with our Secure with Kigen <a href="https://kigen.com/secure-with-kigen-partner-modules/" target="_blank">module partners</a>).</blockquote>– Loic Bonvarlet SVP Ecosystem &amp; Marketing, Kigen <h3>Putting SGP.32 to the Test</h3>Beyond the PR and the slide decks, we wanted to validate if SGP.32 was ready for the rigors of production. We teamed up with several partners to put their SGP.32 readiness to the test, including <a href="https://kigen.com/" target="_blank">Kigen</a> (eSIM and eIM provider) and the connectivity providers: <a href="https://www.zariot.com/" target="_blank">ZARIOT</a>, <a href="https://soracom.io/" target="_blank">Soracom</a>, and <a href="https://onomondo.com/" target="_blank">Onomondo</a>.Five different plastic SIMs with eUICC profiles were used, each containing an IPAe connected to Kigen’s eIM. The SIMs were attached to five different host devices:<ul><li>Nordic Semiconductors’ <a href="https://www.nordicsemi.com/Products/Development-hardware/nRF9151-DK" target="_blank">nRF9151DK</a></li><li>DPTechnic’s <a href="https://www.dptechnics.com/en/products/walter.html" target="_blank">Walter</a> with Sequans Monarch 2</li><li>SG Wireless’ <a href="https://sgwireless.com/the-f1-starter-kit/" target="_blank">F1 Kit</a> with the F1 Smart Module</li><li>Murata <a href="https://www.murata.com/en-eu/products/connectivitymodule/lpwa/overview/lineup/type-1sc" target="_blank">Type 1SC</a> EVK2</li><li>Actinius <a href="https://www.actinius.com/icarus" target="_blank">Icarus</a> with the nRF9160</li></ul> The test was conducted to validate whether the SIMs would migrate between three different connectivity providers in real-time.<img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776159439644-IoTStars-SGP32.png" style="width:100%px" class="fr-fic fr-dib" />IoT Stars SGP.32 test environment with the integrated APIs of Kigen, Soracom, Onomondo and ZARIOT <img src="https://s3-us-west-1.amazonaws.com/wevolver-project-images/froala%2F1776159467011-download-profile.png" style="width:100%px" class="fr-fic fr-dib" />Adding a new profile to a SIM by initiating the "Download" operation  <blockquote>Soracom has always believed that dependence on a single connectivity provider only limits IoT deployments, and SGP.32 holds the promise to remove that limit at last. Independent, multi-vendor validation like the testing IoT Stars organized confirms the standard delivers in real-world conditions. That's why we built our Connectivity Hypervisor to enable profile orchestration with SGP.32 at the foundation.</blockquote>— Kenta Yasukawa, CTO and Co-Founder, Soracom <h3>The Verdict: The Standard Works.</h3>The results? The SGP.32 standard is ready for mass adoption by the industry. The tests demonstrate that the ecosystem has successfully moved past the "lab experiment" phase as devices could migrate between vendors seamlessly. As the SGP.32 standard proved to work in a real-world setting, enterprises can start embedding the remote SIM provisioning standard into their next-generation applications. <blockquote>Tests like this are exactly what the industry needs right now. The whole SGP.32 standard and promise of Remote SIM Provisioning only works when it's implemented correctly, and the more we can prove that in real multi-vendor environments, the easier it becomes for companies to actually make the decision to build on it. Adoption accelerates when there's trust backed by evidence and not just promises.</blockquote>– Henrik Aagaard, CTO &amp; Co-founder, Onomondo <h3>Join the live demo!</h3>Let’s move beyond another article claiming an "eSIM revolution" and look at how it works in real life.Join the <a href="https://app.livestorm.co/iot-stars/sgp32" target="_blank">live webinar</a> on April 28, 4pm CEST / 7am PST to watch SGP.32 in action. We won’t just show the technology; we’ll discuss the real-world implications, including the impact on power management and battery life.The demo includes downloading a profile onto a live SIM and reconfiguring an entire fleet remotely. See it for yourself and decide if you're ready to make the switch.<a target="_blank" href="https://app.livestorm.co/iot-stars/sgp32" id="isPasted" rel="noopener noreferrer">Join the webinar</a> <blockquote>It’s great to be involved in this independent real-world demo and prove SGP.32 works in the wild. We would encourage IoT manufacturers, integrators and solution owners to give it a try, and feel free to take advantage of ZARIOT high-touch technical support program for early adopters. Our engineers can quickly help you address any teething problems and allow you to experience the advantages that SGP.32’s flexibility brings at the earliest opportunity.</blockquote>– Justin Alcorn CCO, ZARIOT <hr />by <a href="https://www.linkedin.com/in/laurensslats/" target="_blank">Laurens Slats</a> laurens@iotstars.com

The First Real-World Remote SIM Provisioning Test

The SGP.32 Reality Check

This article presents a comprehensive engineering guide to Hall Effect Sensors, covering theoretical foundations, device structures, sensor types, specifications, applications, and practical integration strategies for modern electronic system design.

Comprehensive Technical Guide to Hall Effect Sensors

Training a large artificial intelligence model is expensive, not just in dollars, but in time, energy, and computational resources. Traditionally, obtaining a smaller, faster model either requires training a massive one first and then trimming it down, or training a small one from scratch and accepting weaker performance. Researchers at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), Max Planck Institute for Intelligent Systems, European Laboratory for Learning and Intelligent Systems, ETH, and Liquid AI have now developed a new method that sidesteps this trade-off entirely, compressing models during training, rather than after.The technique, called <a href="https://arxiv.org/abs/2510.02823">CompreSSM</a>, targets a family of AI architectures known as state-space models, which power applications ranging from language processing to audio generation and robotics. By borrowing mathematical tools from control theory, the researchers can identify which parts of a model are pulling their weight and which are dead weight, before surgically removing the unnecessary components early in the training process."It's essentially a technique to make models grow smaller and faster as they are training," says Makram Chahine, a PhD student in electrical engineering and computer science, CSAIL affiliate, and lead author of the paper. "During learning, they're also getting rid of parts that are not useful to their development."The key insight is that the relative importance of different components within these models stabilizes surprisingly early during training. Using a mathematical quantity called Hankel singular values, which measure how much each internal state contributes to the model's overall behavior, the team showed they can reliably rank which dimensions matter and which don't after only about 10 percent of the training process. Once those rankings are established, the less-important components can be safely discarded, and the remaining 90 percent of training proceeds at the speed of a much smaller model."What's exciting about this work is that it turns compression from an afterthought into part of the learning process itself,” says senior author Daniela Rus, MIT professor and director of CSAIL. “Instead of training a large model and then figuring out how to make it smaller, CompreSSM lets the model discover its own efficient structure as it learns. That's a fundamentally different way to think about building AI systems.”The results are striking. On image classification benchmarks, compressed models maintained nearly the same accuracy as their full-sized counterparts while training up to 1.5 times faster. A compressed model reduced to roughly a quarter of its original state dimension achieved 85.7 percent accuracy on the CIFAR-10 benchmark, compared to just 81.8 percent for a model trained at that smaller size from scratch. On Mamba, one of the most widely used state-space architectures, the method achieved approximately 4x training speedups, compressing a 128-dimensional model down to around 12 dimensions while maintaining competitive performance."You get the performance of the larger model, because you capture most of the complex dynamics during the warm-up phase, then only keep the most-useful states," Chahine says. "The model is still able to perform at a higher level than training a small model from the start."What makes CompreSSM distinct from existing approaches is its theoretical grounding. Conventional pruning methods train a full model and then strip away parameters after the fact, meaning you still pay the full computational cost of training the big model. Knowledge distillation, another popular technique, requires training a large "teacher" model to completion and then training a second, smaller "student" model on top of it, essentially doubling the training effort. CompreSSM avoids both of these costs by making informed compression decisions mid-stream.The team benchmarked CompreSSM head-to-head against both alternatives. Compared to Hankel nuclear norm regularization, a recently proposed spectral technique for encouraging compact state-space models, CompreSSM was more than 40 times faster, while also achieving higher accuracy. The regularization approach slowed training by roughly 16 times because it required expensive eigenvalue computations at every single gradient step, and even then, the resulting models underperformed. Against knowledge distillation on CIFAR-10, CompressSM held a clear advantage for heavily compressed models: At smaller state dimensions, distilled models saw significant accuracy drops, while CompreSSM-compressed models maintained near-full performance. And because distillation requires a forward pass through both the teacher and student at every training step, even its smaller student models trained slower than the full-sized baseline.The researchers proved mathematically that the importance of individual model states changes smoothly during training, thanks to an application of Weyl's theorem, and showed empirically that the relative rankings of those states remain stable. Together, these findings give practitioners confidence that dimensions identified as negligible early on won't suddenly become critical later.The method also comes with a pragmatic safety net. If a compression step causes an unexpected performance drop, practitioners can revert to a previously saved checkpoint. "It gives people control over how much they're willing to pay in terms of performance, rather than having to define a less-intuitive energy threshold," Chahine explains.There are some practical boundaries to the technique. CompreSSM works best on models that exhibit a strong correlation between the internal state dimension and overall performance, a property that varies across tasks and architectures. The method is particularly effective on multi-input, multi-output (MIMO) models, where the relationship between state size and expressivity is strongest. For per-channel, single-input, single-output architectures, the gains are more modest, since those models are less sensitive to state dimension changes in the first place.The theory applies most cleanly to linear time-invariant systems, although the team has developed extensions for the increasingly popular input-dependent, time-varying architectures. And because the family of state-space models extends to architectures like linear attention, a growing area of interest as an alternative to traditional transformers, the potential scope of application is broad.Chahine and his collaborators see the work as a stepping stone. The team has already demonstrated an extension to linear time-varying systems like Mamba, and future directions include pushing CompreSSM further into matrix-valued dynamical systems used in linear attention mechanisms, which would bring the technique closer to the transformer architectures that underpin most of today's largest AI systems."This had to be the first step, because this is where the theory is neat and the approach can stay principled," Chahine says. "It's the stepping stone to then extend to other architectures that people are using in industry today.""The work of Chahine and his colleagues provides an intriguing, theoretically grounded perspective on compression for modern state-space models (SSMs)," says Antonio Orvieto, ELLIS Institute Tübingen principal investigator and MPI for Intelligent Systems independent group leader, who wasn't involved in the research. "The method provides evidence that the state dimension of these models can be effectively reduced during training and that a control-theoretic perspective can successfully guide this procedure. The work opens new avenues for future research, and the proposed algorithm has the potential to become a standard approach when pre-training large SSM-based models."The work, which was accepted as a <a href="https://arxiv.org/abs/2510.02823">conference paper</a> at the International Conference on Learning Representations 2026, will be presented later this month. It was supported, in part, by the Max Planck ETH Center for Learning Systems, the Hector Foundation, Boeing, and the U.S. Office of Naval Research.

Training AI models is costly, forcing a trade-off between compressing large models or accepting weaker performance from smaller ones trained from scratch.

New technique makes AI models leaner and faster while they are still learning

Work from Carnegie Mellon University may someday help doctors measure skin health as precisely as a blood test. Researchers in the university’s Robotics Institute (RI) are turning that possibility into reality with new software that works with a compact imaging device to bring the skin’s microscopic landscape into clearer focus. “When skin texture becomes a measurable signal rather than a subjective one, dermatology can move toward truly quantitative assessments, helping clinicians to track disease progression and treatment response over time,” said Akhil Padmanabha, a 2025 RI Ph.D. graduate and lead researcher on the project. “With that foundation, skincare efficacy can be evaluated with the same rigor as medical tests.”The team’s software interprets subtle variations in the skin’s surface by locating wrinkles and comparing them to nearby high points to estimate depth. In doing so, it transforms otherwise imperceptible skin texture changes into data that can be both visualized and measured. Paired with a handheld imaging device, the software creates detailed 3D maps of the skin’s surface, revealing fine features with newfound consistency. The work could be used in both clinical and cosmetology settings. It could aid in diagnosing and monitoring textural changes in some skin conditions such as eczema and psoriasis and help evaluate treatment efficacy. It could also give insights into skin aging, sun damage and the effectiveness of skincare products.“From a medical standpoint, these technologies have the potential to fundamentally change how we assess tissue in a noninvasive and objective way,” said <a href="http://dermatology.pitt.edu/people/sonal-choudhary-md" target="_blank" rel="noreferrer noopener">Dr. Sonal Choudhary</a>, a dermatologist at UPMC who worked on the research. “In dermatology and other clinical fields where visual inspection and palpation are still highly subjective, high-resolution tactile sensing could improve diagnostic accuracy; enable earlier detection of disease; and support more personalized, data-driven care. It opens the door to better remote assessment and training.”The compact sensor, 3D-printed probe and newly designed elastic gel work together to measure fine skin surface features.Current portable 3D imaging technologies typically lack the resolution or reliability needed to work consistently on soft surfaces like skin and across different parts of the human body. These gaps in technology make it difficult to accurately measure skin texture.To address this challenge, the RI team adapted a tactile imaging technique known as <a href="https://www.gelsight.com/" target="_blank" rel="noreferrer noopener">GelSight</a> that creates detailed 3D maps by observing how surfaces deform when pressed against an object. The team expanded GelSight technology to measure fine skin surface features using a compact sensor, a 3D-printed probe and a newly designed elastic gel that conforms to the skin. The system improves on the accuracy of currently available commercial devices.The custom elastic gel reduces how much skin deforms when the sensor is pressed against it. It has three layers: a flexible base that supports the structure, a powder coating that helps the sensor capture surface slopes, and a skin-safe layer that covers it all. The layered design allows the device to create precise 3D maps of the skin while maintaining gentle contact.“We attached our custom gels to a GelSight sensor and integrated the sensor with a 3D-printed probe that can be used on skin,” Padmanabha said. “We then trained algorithms to estimate 3D geometries from 2D images and to identify and analyze microstructures on the skin surface rather than relying on the naked eye.”To further validate their probe and skin pattern detection methods, the team conducted a human study with participants who had no diagnosed skin disorders. The goal was to measure skin texture across multiple areas of the body and proceed to evaluate how those measurements changed after the application of an over-the-counter moisturizer. The system detected consistent, measurable reductions in wrinkle depth, demonstrating an accurate sensitivity to everyday skin interventions.“Finding the right material to match the compliance of skin turned out to be an interesting challenge for this project,” said Carmel Majidi, CMU’s Clarence H. Adamson Professor of <a href="https://www.meche.engineering.cmu.edu/index.html" target="_blank" rel="noreferrer noopener">Mechanical Engineering</a> and a part of the research team. “We wanted something gel-like that could conform to the features and texture of the skin but also have enough elastic integrity to convey these features to our optical tracking system in a repeatable and reliable manner. In the end, we came up with a material system that had just the right balance.”

CMU researchers develop software that turns subtle skin changes into trackable data

A New Way to Track Skin Health

Learn how current sensing amplifiers, delta-sigma modulators, and Hall-effect sensors improve precision, PWM rejection, and motor control in humanoid robots.

Generics

Latest Posts