Low Power Consumption Solutions

Your power converter is hitting efficiency limits. But what if the problem isn't your design - it's your semiconductor choice? Most engineers still default to silicon MOSFETs because "they've always worked." Yet these devices are bumping against fundamental physics barriers that no amount of clever engineering can overcome. While silicon MOSFETs max out around 500 kHz switching frequency, gallium nitride devices can push beyond 10 MHz. That's a 20x improvement, enabling smaller inductors and higher power density. The numbers tell a compelling story. In a head-to-head comparison using 400V, 15A devices: • At 200 kHz switching frequency, silicon devices show 40W power loss • SiC devices hit 15W loss at the same frequency   • GaN devices achieve just 8W loss—an 80% reduction from silicon Power factor correction converters, solar inverters, and DC-DC systems all benefit from higher switching frequencies. You can shrink those bulky inductors and transformers that dominate your board real estate. GaN devices need only 22% of the gate charge required by equivalent silicon devices. Less gate charge means faster switching transitions and lower driver power consumption. I used to think GaN was just expensive silicon with better marketing. The cost analysis changed my mind. Yes, individual GaN devices cost more upfront. But when you factor in smaller magnetics, reduced cooling requirements, and higher system efficiency, the total cost equation often favors GaN. The adoption curve reminds me of when MOSFETs displaced bipolar transistors in the 1980s. Initially expensive and exotic, but eventually became standard because the performance advantages were undeniable. Solar installations particularly benefit from this technology. Higher switching frequencies enable smaller filter components while efficiency gains directly boost energy harvest. In data centers, every percentage point of efficiency improvement translates to significant operational savings. What surprised me most was the reverse conduction capability. Unlike silicon MOSFETs that rely on body diodes with recovery losses, GaN devices can conduct in reverse without these penalties, eliminating dead time losses. The manufacturing approach also matters. While SiC requires expensive substrates, GaN devices grow on standard silicon wafers using existing fab infrastructure. This manufacturing advantage should drive costs down faster than expected. Recent developments in isolated gate drivers are addressing adoption barriers. Solutions like those from Allegro MicroSystems integrate bias supplies directly into the driver, eliminating external power rails and simplifying system design while reducing EMI. For engineers working on next-generation clean energy systems, the question isn't whether to consider GaN—it's whether you can afford not to. What's been your biggest challenge in improving power conversion efficiency in clean energy applications?

Carbon Reduction with your BAS? Low-cost building automation strategies can play a significant role in achieving carbon reduction goals by optimizing energy use, improving operational efficiency, and reducing waste. Here are some strategies that can be implemented to help reduce carbon emissions without significant capital investments: Energy Monitoring and Benchmarking: Implement a basic energy monitoring system to track and benchmark energy use across the building. Many energy management systems can be integrated with BAS for minimal cost. Identifies areas of excessive energy consumption, allowing for targeted improvements, reducing waste and carbon emissions. Optimized HVAC Schedules: Use BAS to automate HVAC schedules based on occupancy, seasonality, and operational needs. Turn off or reduce HVAC operations during unoccupied hours or in unused spaces. Reduces energy consumption and emissions from heating, ventilation, and cooling systems. Setpoint Optimization: Adjust temperature setpoints slightly (e.g., increasing cooling setpoints or reducing heating setpoints) within comfortable ranges. Small setpoint changes can lead to significant energy savings over time, reducing carbon emissions from HVAC systems. Demand-Controlled Ventilation (DCV): Integrate sensors that measure CO2 levels in spaces to control ventilation rates dynamically, providing fresh air only when needed based on occupancy. Reduces the energy required for ventilation, cutting down on unnecessary heating or cooling of outdoor air. Lighting Control Systems: Install automated lighting controls (e.g., motion sensors, daylight harvesting) and integrate them with the building automation system to optimize lighting use. Reduced lighting energy consumption translates directly to lower electricity use and carbon emissions. Variable Frequency Drives (VFDs) for Motors: Add VFDs to fans, pumps, and other motor-driven systems, allowing their speed to adjust based on demand rather than running at full capacity. VFDs reduce energy consumption by matching motor speed to actual demand, reducing energy waste and carbon output. Continuous Commissioning: Use BAS data to continuously monitor building systems and performance. Identify inefficiencies and make ongoing adjustments to optimize energy use. Ensures systems are running efficiently, preventing energy waste and emissions over time. Free Cooling (Economizers), Ensure that economizers are properly maintained and optimized to use outside air for cooling when outdoor conditions are favorable. Reduces the need for mechanical cooling, saving energy and cutting emissions. Remote Monitoring and Management: Use remote monitoring and automation tools to adjust system settings and identify energy-saving opportunities without requiring onsite personnel. Allows for better oversight and proactive adjustments, avoiding wasted energy and unnecessary emissions. These strategies, when combined with an ongoing commitment to energy

The Spiking Neural Processor T1, developed by Innatera Nanosystems, is set to revolutionize smart technology by significantly reducing power consumption and enhancing battery life in devices like smart lightbulbs, doorbells, and smoke alarms. Inspired by the architecture of the human brain, this neuromorphic chip represents a breakthrough in artificial intelligence (AI) hardware. Key Features: 1. Brain-Inspired Design: • The chip mimics how the brain detects and processes patterns through “spiking” neural networks. • It processes sensor data in real-time, enabling AI functionality directly on the device. 2. Energy Efficiency: • Current smart devices rely on cloud computing for data processing, which is power-intensive and requires a constant internet connection. • The Spiking Neural Processor T1 eliminates this dependency, drastically reducing power consumption by performing AI computations locally. 3. Enhanced Functionality: • By analyzing sensor data in real-time, the chip can clean and process data more efficiently, leading to faster and more accurate responses. • This capability could enable a new generation of smart devices with extended functionality and autonomy. 4. No Internet Dependency: • Removing the need for cloud-based processing reduces latency and enhances privacy, as sensitive data remains on the device. Implications for Smart Technology: • Battery Life Boost: • Devices equipped with this chip could experience significantly longer battery life, making them more practical and cost-effective for consumers. • Sustainability: • The reduced energy demands align with global efforts to minimize electronic waste and promote environmentally friendly tech solutions. • Broader Applications: • The chip could pave the way for smarter, more independent devices across industries, including healthcare, security, and IoT (Internet of Things) ecosystems. Availability: • The Spiking Neural Processor T1 is expected to hit the market in 2026, with potential to transform the landscape of power-efficient, AI-driven smart devices. This development underscores the growing trend of integrating AI directly into hardware, moving away from cloud dependence and towards more sustainable, efficient, and secure solutions.

Is AI's Growth Sustainable? How to Make Generative Applications Greener. The rise of generative AI tools like ChatGPT and others has been remarkable, but their environmental impact is often overlooked. The data center industry, housing these systems, accounts for up to 3% of global greenhouse gas emissions, with energy consumption doubling every two years. Hyperscale cloud providers like Amazon AWS, Google Cloud, and Microsoft Azure play a significant role in powering these models, leading to major carbon footprints. Understanding the carbon footprint lifecycle of AI models is crucial. Large generative models consume extensive energy during training, and fine-tuning can be a more energy-efficient option. Inference sessions, though less energy-intensive, involve many more sessions, contributing to ongoing energy consumption. Efforts to reduce energy usage include employing less computationally expensive approaches like TinyML and using large models only when significantly valuable. To make AI greener, companies can use existing models from providers instead of creating new ones. Fine-tuning existing models on specific content domains consumes less energy and provides more value. Utilizing energy sources from carbon-friendly regions and monitoring carbon emissions can significantly reduce AI's environmental impact. Reusing models and resources, incorporating AI activity into carbon monitoring, and encouraging green AI practices are crucial steps in promoting sustainability. 1. Prioritize Fine-Tuning: Instead of training new generative models from scratch, focus on fine-tuning existing models for specific content domains. Fine-tuning consumes less energy and provides more value to businesses. 2. Explore Energy-Conserving Methods: Adopt energy-conserving computational approaches like TinyML for processing data. TinyML allows running ML models on low-powered edge devices, significantly reducing energy consumption. 3. Re-use and Open Source Models: Opt for reusing open-source models instead of creating new ones. Recycling tech can lower the carbon impact of AI practices and reduce the need for energy-intensive model development. 4. Monitor Carbon Emissions: Include AI activity in carbon monitoring practices to understand the carbon footprint of AI-related operations. Share footprint numbers to make informed decisions about AI partnerships. 5. Choose Green Energy Sources: Select cloud providers and data centers that prioritize environmentally friendly power resources. Running AI models in regions with carbon-free energy sources can significantly reduce operational emissions. Have you already considered the impact of using compute-heavy applications on our planet? Are you tracking the impact of compute in your sustainability report? #genai #aivalue #sustainableai #sustainability

The idea of submerging computer servers in a liquid coolant to cut data center energy consumption by 70% is a breakthrough in sustainable tech innovation. Traditional cooling systems consume significant energy, but with non-conductive liquid coolants, it's possible to safely dissipate heat while keeping electrical circuits dry and operational. This method optimizes thermal management, capturing all the generated heat and drastically reducing the need for conventional fans and chillers. Sandia National Laboratories approach could set a new standard for energy efficiency in data centers, making them greener and more cost-effective. Florian Palatini ++

A couple of months ago, I was designing a battery-powered door sensor and noticed a leakage issue that would compromise the 2-year battery life target. The sensor used a Flex-14 magnetic switch and a simple pull-up on the input. But the switch stays closed most of the time, so the pull-up was leaking current continuously.. A 100 kΩ pull-up at 3.3 V wastes around 33 microamps 24/7, and for a device that needs to run 2 years on a small battery, that leakage alone is too much. So I changed the approach. Instead of tying the pull-up to 3.3 V, I connected it to a GPIO pin. When I need to read the door state: 1- Set the GPIO HIGH (acts like 3.3 V) 2- Read the input 3- Set the GPIO LOW → zero leakage 4- Go back to sleep This simple change saved tens of microamps, making the battery target achievable without the need for extra components. Small detail --> Big impact. #LowPowerDesign #IoTDesign #BLE #PCBDesign #BatteryPowered

“Low-power” modes can burn more energy if you enter and exit them too often. A small example 👇 (illustrative, order-of-magnitude) Assume a microcontroller at 1 MHz, 3.3 V • Active power ≈ 280 µW • Sleep power ≈ 165 µW • Enter + exit sleep energy ≈ 50 nJ Idle gap = 100 µs Stay active: 280 µW × 100 µs ≈ 28 nJ Sleep: 165 µW × 100 µs ≈ 16.5 nJ 50 nJ transition ≈ 66.5 nJ Sleep burns > 2× energy in this regime. Break-even idle time: 50 nJ / (280 − 165 µW) ≈ 435 µs If idle gaps are shorter than ~0.4 ms, sleep becomes an energy trap. Same effect exists in phones. Frequent power-state toggling triggers: • CPU voltage and frequency transitions • Apps and background tasks pausing and resuming • Cellular, Wi-Fi, and Bluetooth radios shifting power states • OS scheduler reshuffling threads and timers Those transitions cost energy. Low power isn’t a mode. It’s a decision. #Semiconductors #VLSI #ChipDesign #LowPowerDesign #PowerManagement #SoCDesign #EmbeddedSystems #ComputerArchitecture #BatteryLife

At IoT Stars 2025 in Nuremberg, we unveiled the RAK11160 and shared the story behind why we built it. The idea came from a clear pattern we kept seeing: More than 60% of our RAK3172 users were also integrating ESP32 into their designs, just to add MQTT, OTA updates, or BLE provisioning. That meant extra components, extra cost, and more power consumption. That’s exactly what inspired our team at RAKwireless  to build the RAK11160. This isn’t just another LoRaWAN module. It's our first dual-core wireless MCU, blending: - STM32WLE5 for LoRa + ultra-low-power sensing - ESP32-C2 for on-demand WiFi and BLE These two cores don’t run in parallel. Instead, the STM32 manages system logic and keeps power consumption low by fully shutting down the ESP32, only waking it for BLE provisioning, WiFi-based OTA, or MQTT uplinks. This setup is great for #smartfarming , since the devices can stay asleep for days, send soil or crop data with LoRaWAN, and only use Bluetooth when a technician is close by. Read the blog: http://bit.ly/3ZARjkv Check out RAK11160: https://bit.ly/3TscMbF #IoT #RAK11160 #LoRaWAN #LowPowerDesign #STM32 #ESP32

🌟 Excited to share our recent research on energy-efficient hardware design, addressing the critical need for sustainable computing as Generative AI’s energy demand grows exponentially. 🌟 We’ve introduced a new low-power computing paradigm called Hybrid Temporal Computing (HTC). In HTC, data for multiplication are encoded in a hybrid format combining temporal data and traditional bitstream data. The temporal data concept is inspired by recently proposed temporal computing (or race logic), where information is encoded as time delays or waveforms. Temporal computing can significantly reduce energy consumption, as it requires only a single switch to represent data transition but with limited applications due to waveform restrictions. The illustration below demonstrates how HTC handles multiplication and addition efficiently. Our initial findings indicate that HTC can achieves remarkable power savings compared to state-of-the-art stochastic computing methods, with small accuracy trade-offs in applications such as DCT/iDCT and digital filter accelerators. This work will be published in coming  ASPDAC’25! 📄 https://lnkd.in/gMAuEXDH Thanks my student Maliha Tasnim, Sachin Sachdeva and Yibo Liu for their hard works. 🚀 #HybridTemporalComputing #TemporalComputing #EnergyEfficientAI #ASPDAC25

A Practical Solution to Meet Data Center Energy Demand: Rather than expanding generation and transmission capacity to meet the rapidly growing energy demand of data centers, I propose here a more efficient and resource-saving alternative. This approach involves optimizing the design of a Solar PV-Battery Energy Storage (BES) system to supply 80-85% of the daily energy requirements of a data center, while limiting grid dependency to a maximum of 20%. This hybrid system significantly reduces the need for large-scale infrastructure upgrades. Here’s an illustrative example I designed for a 1 GW data center in Saudi Arabia: - Solar PV System: 3.9 GWdc / 3.52 GWac - Battery Energy Storage (BES): 3 GWac / 5.6 GWh - Transmission Line Capacity: 200 MW (20% of the load) The system configuration, as shown in figure, is an AC-coupled system. The PV-BES management system is programmed to ensure that the load power drawn from the grid never exceeds the transmission line capacity of 200 MW. To validate this design, I conducted a full-year simulation with a 5-minute time step for a specific location in Saudi Arabia. Results demonstrated that the State of Charge (SOC) of the battery system never dropped below 15%. The system was designed with the PV and BES capacities approximately three times the load to provide additional power and energy redundancy, achieving an optimal balance between reliability and cost-effectiveness. This optimized hybrid system represents a sustainable and scalable solution to meet the increasing energy demands of data centers while minimizing grid strain and infrastructure costs. Another potential solution involves deploying Battery Energy Storage (BES) systems and data centers adjacent to existing utility-scale PV plants. This approach leverages already-developed infrastructure, optimizing the utilization of renewable energy while minimizing additional land use and transmission requirements.