Alexey Brodkin

If you are an analog or RF IC designer spending countless hours creating reports by capturing screenshots of plots and tables from Cadence ADE Assembler and manually placing them into Excel spreadsheets or PowerPoint presentations, there is now a better solution. MDOC is the first commercially available tool designed to automate this workflow, reducing repetitive reporting tasks by up to 80%.

61

For the data rate 10Gbps, it means the signal has around 100ps to present zero or one. For example at 12.4Gbps, around 56ps is the usable moment considering rising/falling time, jitter, noise. 56ps is the time it takes for light to travel just 16.8mm, which is a distance shorter than the diameter of a coin. If we equate one human second to a single computer clock cycle, the absolute time of one second for a human would be equivalent to 196.6 years for a computer (CLK 6.2GHz, 161.29ps). We live in a world where we deal with time scales of tens of picoseconds at the molecular level. Below timing capture shows HBM3E memory subsystem running at 12.4Gbps at nominal voltages.

2

My book on RISC-V Vector Extension (RVV) support in LLVM is published. It presents a concrete, engineering-level analysis of how LLVM implements vectorization for RVV. https://lnkd.in/etzmsXtq Thanks Apress and Miriam Haidara

350

7 comentários

𝐖𝐡𝐚𝐭 𝐈𝐟 𝐘𝐨𝐮𝐫 𝐀𝐜𝐭𝐢𝐯𝐚𝐭𝐢𝐨𝐧 𝐅𝐮𝐧𝐜𝐭𝐢𝐨𝐧 𝐈𝐬 𝐐𝐮𝐢𝐞𝐭𝐥𝐲 𝐒𝐭𝐞𝐞𝐫𝐢𝐧𝐠 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧? We spend a lot of time debating architectures - Transformers vs CNNs, depth vs width, scaling laws, attention variants. But one of the most overlooked design choices is the 𝐚𝐜𝐭𝐢𝐯𝐚𝐭𝐢𝐨𝐧 𝐟𝐮𝐧𝐜𝐭𝐢𝐨𝐧. 𝐑𝐞𝐋𝐔 became the default because it solved 𝑣𝑎𝑛𝑖𝑠ℎ𝑖𝑛𝑔 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡𝑠 and made deep networks trainable at scale. It’s simple, efficient, and works well in most cases. But it also zeroes out all negative inputs. In domains like audio modelling, where signals naturally oscillate around zero, this can discard meaningful information. Half the waveform can effectively be muted. In several deep speech and audio enhancement models, switching from 𝐑𝐞𝐋𝐔 𝐭𝐨 𝐄𝐋𝐔 has led to more stable training at higher learning rates. Why? Because ELU allows controlled negative outputs, helping keep activations centered around zero and improving gradient flow across many layers. The improvement per layer is small. But across 40–60 layers, those small differences compound and influence convergence behavior. Activation functions don’t just add non-linearity. They shape 𝐬𝐢𝐠𝐧𝐚𝐥 𝐝𝐢𝐬𝐭𝐫𝐢𝐛𝐮𝐭𝐢𝐨𝐧, 𝐠𝐫𝐚𝐝𝐢𝐞𝐧𝐭 𝐝𝐲𝐧𝐚𝐦𝐢𝐜𝐬, and 𝐮𝐥𝐭𝐢𝐦𝐚𝐭𝐞𝐥𝐲 𝐭𝐡𝐞 𝐨𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐩𝐚𝐭𝐡 𝐲𝐨𝐮𝐫 𝐦𝐨𝐝𝐞𝐥 𝐟𝐨𝐥𝐥𝐨𝐰𝐬. Sometimes, performance gains aren’t about bigger models - they’re about better geometry. #DeepLearning #MachineLearning #NeuralNetworks #AI #ModelOptimization #ArtificialIntelligence

6

Preparing for RF interviews requires more than memorizing formulas; it demands a clear understanding of how real circuits behave under high-frequency conditions. Concepts like dB scaling, impedance interaction, linearity, and nonlinearity are not just theoretical; they directly influence how RF systems perform in practice. This article is structured to help you build that intuition efficiently. It revisits the core principles of RFIC design that interviewers consistently focus on signal representation, system behavior, and distortion mechanisms while connecting them to practical scenarios you are likely to be questioned on. https://lnkd.in/gzq8pkkb

1

Characterization is not just testing. It’s about understanding the silicon at its core, uncovering hidden risks, and ensuring the product works under every possible scenario. Over the years, I’ve learned: 🔹 Every failure has a story — and understanding it is where innovation happens. 🔹 Small timing or voltage shifts can impact millions in yield. 🔹 Characterization engineers are the unsung heroes of reliable products. I’m curious to hear from others: What’s the most unexpected memory failure you’ve debugged? How do you see memory characterization evolving in India and global companies? Let’s start a conversation about bridging design, characterization, and real-world reliability. #MemoryDesign #VLSI #Semiconductor #Characterization #SiliconValidation #EngineeringLeadership #IndiaTech

16

Overheating FPGA systems don’t just reduce performance, they shorten product lifespans. The solution? Start with smart PCB design. Here are 3 simple design tactics that make all the difference. #pcbdesign #thermalmanagement #systemdesign

17

🚦 Embedded Systems: An Unknown City with Its Own Language, Map, and Customs Imagine stepping into a completely unknown city. New streets. Unfamiliar signs. Local dialects. Different customs. That’s exactly how it feels when you enter the world of embedded systems for the first time. Here’s a tour of this mysterious city, with analogies to help you understand why each part of embedded learning is crucial 👇 🗣️ 1. Language: Learn the Local Tongue (Assembly) and the Common Tongue (C/C++) In any city, locals might speak a native dialect. In embedded systems, this "local language" is Assembly. It’s harder to learn but closer to the machine, much like how local dialects reveal the culture of the city. Even if you're not fluent, learning a bit of Assembly helps you understand performance bottlenecks, interrupt handling, or startup code. But you don’t need to be a local to survive. The common language of this embedded city is C or C++. Just like English is spoken in many parts of the world, C/C++ is portable, powerful, and widely used across microcontrollers and processors. Knowing both gives you flexibility and confidence. 🗺️ 2. The Map: Understand the City's Layout (Schematic, Datasheet, and Memory Map) You wouldn’t drive through an unknown city without a map. In embedded systems, the schematic and memory map are your guide. Want to know where the GPIO is? Check the schematic. Wondering where to place your ISR vector table or bootloader? Refer to the memory map. The datasheet is your city directory—don’t just read it, live in it. Without understanding these, you’re just wandering around, hoping not to crash into something. 🧠 3. Customs: Respect the City’s Rules (Memory Management, Power, Timing) Every city has its rules and etiquette. In embedded systems, these are: Memory management: There's no forgiveness for leaks or overflows here. One mistake, and your system crashes—like offending someone important in town. Power constraints: You must know when to turn off peripherals or switch to sleep mode. Energy is precious, like water in a desert city. Timing: Real-time response isn’t optional—interrupts and timers are the clock towers of the city. Miss one, and your whole system misses a beat. If you understand and respect the customs, you’ll thrive. If not, you’ll keep running into bugs, like breaking invisible laws. 🚀 4. Explore with a Guide First (RTOS, Bootloaders, Protocols) Just like tourists use guides or translators, early embedded learners benefit from understanding higher-level constructs like: RTOS – A helpful city tour bus that keeps you on schedule. Bootloader – Your immigration officer, deciding what enters the system and when. Protocols (UART, SPI, I2C, CAN) – The local transportation system that moves information reliably. Knowing how to hop on and off these systems makes navigation easier. Embedkari Systems(OPC) Pvt. Ltd. 🏙️ Ready to explore the embedded world? #CProgramming #MemoryManagement #RTOS #IoT #Embedkari

41

1 comentário