Eric Mitchell

In power electronics R&D, the measurement setup is as important as the design. For capturing switching current waveforms (loss analysis, snubber currents, capacitor ripple), a shielded Rogowski probe like CWTMini50HF helps reduce dV/dt pickup while keeping bandwidth high.

3

𝗠𝗶𝗰𝗿𝗼𝗻 𝗷𝘂𝘀𝘁 𝗰𝗼𝗻𝗳𝗶𝗿𝗺𝗲𝗱 𝘁𝗵𝗮𝘁 𝗶𝘁𝘀 𝗲𝗻𝘁𝗶𝗿𝗲 𝟮𝟬𝟮𝟲 𝗛𝗕𝗠𝟰 𝘀𝘂𝗽𝗽𝗹𝘆 𝗶𝘀 𝗮𝗹𝗿𝗲𝗮𝗱𝘆 𝗴𝗼𝗻𝗲. 𝗡𝗼𝘁 𝗻𝗲𝗮𝗿𝗹𝘆 𝘀𝗼𝗹𝗱. 𝗚𝗼𝗻𝗲. 𝗔𝗻𝗱 𝘁𝗵𝗲𝘆 𝗰𝗮𝗻 𝗼𝗻𝗹𝘆 𝗺𝗲𝗲𝘁 𝗵𝗮𝗹𝗳 𝘁𝗵𝗲 𝗱𝗲𝗺𝗮𝗻𝗱 𝘁𝗵𝗲𝘆 𝗮𝗿𝗲 𝘀𝗲𝗲𝗶𝗻𝗴. Let that sink in. The product just entered high-volume production this quarter. HBM4 for Vera Rubin delivers 2.8 TB/s of bandwidth. That is 2.3x the throughput of the previous generation. Twenty percent better power efficiency. The specs are real and they matter. But the supply story is the one that actually moves infrastructure decisions right now. Samsung showed up at GTC this week with HBM4E. Not HBM4. The next step after that. Sixteen gigabits per pin, 4.0 TB/s, hybrid copper bonding that stacks 16 or more layers while cutting heat resistance by more than twenty percent. They are already telling you where this is going. Here is what the memory supercycle actually means for the industry. The bottleneck in AI infrastructure is no longer the GPU. Jensen said it at GTC. Satya Nadella has been saying it for two years. The constraint is the full system: power, memory bandwidth, and interconnect. HBM is the memory layer that makes accelerators worth building at all. Without it you have a very expensive processor waiting for data. With Micron's entire allocation under binding contract and only 50 to 66 percent of core customer demand being met, the companies that locked in supply early have a structural advantage going into 2027. The ones that did not are now designing around a constraint they cannot simply buy their way out of. This is what a real supply-driven market looks like. Not a chip shortage in the 2021 sense. A precision shortage at the exact layer of the stack where every next-generation AI system depends. Samsung is already signaling HBM4E. The cadence is accelerating. Each generation locks in faster than the last. #AI #Semiconductors #HBM #AIInfrastructure #Memory #DRAM #DataCenter #NVIDIA #Compute

7

The amount and complexity of electronics for automotive applications have grown quickly and are projected to continue growing. 👏

6

We are putting focus on the TJA1465 CAN SIC transceiver this month. This device improves CAN communication by reducing signal ringing and enabling higher data rates. The selective wake-up function is also a good way to make sure energy isn’t wasted when the transceiver doesn’t need to be active. You can read more about it here: https://lnkd.in/eeBTfNuf

6

Join Matt Sullivan from our automotive systems team for an exciting webinar: "Unlock the future of SDVs: Integrating TI's remote-controlled edge solutions" where he'll explore how automotive in-vehicle networking is evolving to meet tomorrow's challenges. 🚗 Discover the ways TI's innovative solutions – including Ethernet, CAN, LED drivers and motor drivers – can help you reduce system complexity and cost, while simplifying software management and updates, enabling easy addition of new features to SDVs and scaling of software across vehicle platforms. Specific learnings will include: • The differences between traditional and remote-controlled edge nodes. • The unique capabilities and design considerations of remote-controlled edge protocols, CAN FD Light and 10BASE-T1S. • System level understanding of a remote-controlled edge from transceiver to actuator. Sign up here https://lnkd.in/gXu9GJBx ​

6