Plextek

Thermal AI models need less training data when you use transfer learning - but where does this actually matter? The transfer learning approach for thermal AI applies wherever colour cameras fail: >> Autonomous vehicles need all-weather sensing - fog, smoke, darkness. >> Industrial monitoring requires temperature-based fault detection. >> Search and rescue operations depend on thermal signatures. >> Wildlife monitoring benefits from heat detection. Each application faces the same training data bottleneck. Automated label generation using sensor fusion makes thermal AI practical across domains. The methodology adapts to any scenario where you can collect synchronised colour-thermal image pairs under good lighting conditions. https://bit.ly/45oPdHp

We're proud to support the UKESF's Girls into Electronics programme, which is back for 2026! These are free one-day events hosted by 15 universities across the UK, including Cambridge, Southampton, Edinburgh and Imperial College London, running throughout June and July. They're open to girls aged 14 to 18, either individually or as part of a school group, with over 500 places available nationwide. The programme gives participants hands-on experience of electronics engineering, insight into where a career in the sector can take them, and a chance to see themselves in roles that are still significantly underrepresented by women. If you know someone aged 14 to 18 who might be interested, whether that's a daughter, niece, student or mentee, do share this with them. Places are filling up quick! Find availability and book here: https://bit.ly/4dL8VBq

4D imaging radars now measure range, velocity, azimuth and elevation simultaneously. Larger antenna arrays with more transmitters and receivers enable higher resolution outputs that approach camera-like detail whilst retaining radar's all-weather capability. Even these advanced radars benefit from fusion with complementary sensors, particularly as AI-enhanced algorithms become capable of processing the increased data volumes in real time. Evaluating advanced radar technologies for your application? This article by Dr Damien Clarke examines how 4D imaging radars and AI-driven fusion algorithms are reshaping environmental sensing: https://lnkd.in/eKC9-cxK

When a project ends, we don't disappear. Our clients consider us their trusted design and long-term delivery partner for projects with multiple phases and work packages. This approach has built lasting relationships and enabled us to support products throughout their entire lifecycle. From TRL 1 research through to TRL 9 manufacturing and ongoing support, Plextek offers complete end-to-end capability. We complement your design team or take on complete projects, delivering technically challenging solutions while minimising risk. Our comprehensive approach means nothing gets lost in translation between strategy, design and engineering. We deliver results at pace, within strict quality controls, and with your business goals firmly in focus. Let's talk about your next project!

Most RF engineers are trying to make things smaller. But some are asking a different question: what if you didn't need that part of the system at all? That's the approach Vlad Lenive brings to his work at Plextek. With 30+ years across space payloads, SATCOM, phased arrays and cellular infrastructure, he's seen what happens when you stop treating antennas, filtering and beamforming as separate blocks and start combining them -- filtenna concepts where performance comes from the geometry itself, not from added components. Entire stages of the RF chain disappear and the SWaP savings follow. We sat down with Vlad to talk about what makes HAPS harder than satellite, how he solved an LO pulling problem that turned out to have nothing to do with the circuit, and why the hardest part of any project is usually figuring out what the real problem actually is! Read the full interview: https://bit.ly/4nfMQOI

We’re pleased to see Plextek recognised as Technology Company of the Year 2026 last night, with our approach to integrating radar and communication systems highlighted by the Cambridge Independent Science and Technology Awards. This follows Plextek receiving a King’s Award for Enterprise in Innovation earlier this week. What a great week of recognition for the team here at Plextek and the work we deliver. 🏆Congratulations to everyone involved. https://bit.ly/4f6Aard

🏆 Announcing: Plextek has won a King's Award for Enterprise: Innovation 2026 🏆 This is the highest business honour in the country and it means a huge amount to us. The award recognises the PLX-T60, a millimetre-wave radar we designed and built right here in the UK that is already in orbit, helping satellites detect and manage space debris. We're working with international space agencies and satellite manufacturers on the next generation for debris removal and in-orbit servicing. Our technology is also being developed for the toughest challenges in defence, healthcare, industrial applications and beyond, which says a lot about what a sovereign UK engineering team can do when the design is right. Incredibly proud of everyone who made this happen! >> If you've got a sensing challenge and want to work with an award-winning team, get in touch.

There's a difference between adding a component and integrating a solution. For iFIT, we designed the motor controller as a native part of their system architecture - not a bolt-on module with compromises. This meant understanding their existing infrastructure, communication protocols, power budgets, and thermal constraints. The controller works as if it was always part of the design. Native integration takes longer upfront but eliminates ongoing friction, simplifies manufacturing, and improves reliability. Architecture matters as much as algorithms! https://bit.ly/3Ly8Rd7

We tested the AI model on real transmitters. None were included in the training dataset. The model successfully detected and classified most RF signals despite training entirely on synthetic data. Overlapping signals in time and frequency were handled correctly. Failures occurred when real signals looked substantially different in spectrograms compared to training examples - physics the RF model didn't capture accurately. This validates the core approach whilst highlighting where improved RF modelling would boost performance. A limited set of real measurements can then refine the model further. Synthetic training works when your simulations are good enough! https://lnkd.in/eXBYzZkX

Getting motor control right means understanding what's happening inside the magnetic core. For iFIT's fitness equipment, we ran magnetic flux density and electrical current analysis to pinpoint torque ripple sources. The simulation revealed areas of magnetic saturation - which guided both control algorithm adjustments and confirmed the physical design limits. This simulation-first approach meant we could optimise the controller without multiple hardware iterations. Faster development, lower risk, proven results. Field-oriented control + solid simulation = smooth, reliable motor performance https://bit.ly/4svTFOn