#Edge_AI
Explore tagged Tumblr posts
electronicsbuzz · 3 months ago
Text
0 notes
phonemantra-blog · 2 years ago
Link
As part of Kickstarter The unusual concept of a modular mini-PC Khadas Mind has ceased to be a concept. The company has launched on Kickstarter and is ready to begin shipping some of the modules this year.  [caption id="attachment_53208" align="aligncenter" width="780"] Khadas Mind[/caption] Not much has changed since the concept was announced in July. Khadas Mind is still a very compact mini-PC that can be greatly expanded through modules. The mini-PC itself has dimensions of 146 × 105 × 20 mm with a mass of 450 g. Inside, you can choose either Core i5-1340P or Core i7-1360P, 16–32 GB of RAM, an SSD with a capacity of 512 GB, and there is also a slot for a second storage, Wi-Fi 6E, Bluetooth 5.3, USB 3.2 Gen2 (x2), HDMI 2.0, RJ45, USB-C (x2) and SD card slot. To connect those same modules, a proprietary Mind Link connector is used. It also has its own 5.55 Wh battery, which is needed for hot-swap modules.   Khadas Mind is now available for pre-order [caption id="attachment_53209" align="aligncenter" width="680"] Khadas Mind[/caption] As part of a crowdfunding campaign for a PC, they ask for 600 or 800 dollars, depending on the modification. Retail prices will be $800 and $1,100 respectively.  The base module is the Mind Dock, which expands the set of ports and adds a convenient physical volume control.   Mind Graphics is an external graphics card. This is the RTX 4060 Ti 8GB. The unit has its own set of ports and speakers. For a mini PC with a docking station, they are now asking for $ 930, and for a kit with a video card, you will have to pay at least $ 1,500.   https://youtu.be/xWjuh9XoMqo The equipment also includes the Mind xPlay module, which is a screen with a keyboard and a stand. Connecting to a mini PC turns the whole thing into a laptop. True, for some reason, there are no details about this module, and it is missing from the roadmap. It is only in the plans for now, like the Mind Studio Display monitor. 
0 notes
renatoferreiradasilva · 21 days ago
Text
Projeto Completo da Placa CM5 HyperModule: Arquitetura Soberana para Computação Modular de Alta Performance
(Baseado em documentação técnica, pesquisas acadêmicas e padrões industriais)
🔍 1. Objetivo Estratégico
Desenvolver uma placa-mãe modular de código aberto para o Compute Module 5 (CM5) que supere dispositivos comerciais como MacBook Air M2 em:
Desempenho Computacional (26 TOPS/W com Hailo-8)
Resiliência Térmica (operar a -20°C a 70°C sem throttling)
Soberania Tecnológica (100% reparável com peças impressas em 3D)
⚙️ 2. Especificações Técnicas da Placa
Núcleo Computacional
Componente Especificação Finalidade SoC Principal Broadcom BCM2712 (4x Cortex-A76) Processamento central Coprocessador IA Hailo-8 M.2 (26 TOPS/W) Inferência neural offline Memória LPDDR5 8GB + FRAM/MRAM para failover Preservação de estado durante falhas
Subsistemas Críticos
Módulo Tecnologia Inovação Térmico Heatpipe de grafeno + PCM magnético Dissipação passiva até 15W TDP Energia Baterias LiFePO4 + Supercaps Maxwell Hot-swap em <5s sem perda de dados Segurança TPM 2.0 + STM32H7 secure enclave Autodestruição física do firmware
🛠️ 3. Design Avançado da PCB
Parâmetros Chave
Característica Especificação Ferramenta de Validação Substrato Rogers 4350B (4 camadas) Simulação Ansys HFSS Impedância 85Ω ±5% (PCIe Gen4) TDR Teledyne LeCroy Dissipação Microcanais capilares integrados Testes em câmara Weiss WK11-340 Conectores Mecânica de travamento magnético Certificação MIL-STD-810H
Topologia de Alimentação
graph LR USB_C[USB-C PD 100W] --> MPPT[Conversor GaN 20A] MPPT --> BMS[BMS Inteligente] BMS --> Supercaps[Buffer Supercaps] Supercaps --> SoC[BCM2712 + Hailo-8]
🔒 4. Protocolos de Segurança e Failover
Arquitetura Zero-Trust
// Pseudo-código do Boot Seguro void secure_boot() { if (tpm_verify(EFI_SIGNATURE) == VALID) { load_os(); } else { stm32h7_fallback(); // Enclave seguro if (physical_tamper_detected()) { destroy_firmware(12V_GPIO_PULSE); // Autodestruição } } }
Failover Energético
Tempo de Transição: < 3ms (via FRAM/MRAM)
Mecanismo:
def power_failover(): if voltage < 3.3V: save_state_to_fram() switch_to_supercaps()
🧪 5. Metodologia de Desenvolvimento
Fases Críticas
Fase Ações-Chave Entregáveis F1 Projeto esquemático (KiCad) + Simulação SI/PI Layout otimizado para PCIe Gen4 F2 Prototipagem rápida (JLCPCB) + Montagem 3x placas funcionais F3 Validação térmica (-20°C a 70°C) Relatório FLIR + dados de throttling F4 Testes de campo (Amazônia com Suzano Foundation) Métricas de autonomia/resiliência F5 Certificações (IEC 62368-1, MIL-STD-810H) Documentação para produção em escala
📦 6. Repositório Técnico Completo
/cm5-hypermodule/ ├── hardware/ │ ├── kicad/ # Projeto completo da PCB │ ├── 3d_models/ # Chassis e dissipadores │ └── bom.csv # Lista de materiais ├── firmware/ │ ├── secure_boot/ # Código UEFI ARM + TPM │ ├── thermal_control/ # Algoritmo LSTM para gestão térmica │ └── biosensors/ # SDK OpenBio └── docs/ ├── MIL-STD-810H_tests/ # Resultados de robustez └── sustainability.pdf # Análise de ciclo de vida
🌐 7. Impacto Estratégico
Inovações Disruptivas
Soberania Tecnológica:
100% projetado com ferramentas open-source (KiCad, TensorFlow Lite)
Produção descentralizada via redes de fab labs
Sustentabilidade Radical:
Chassis em bioplástico reforçado com fibra de cânhamo
Blockchain para rastreamento de materiais
Caso de Uso Real
# Monitoramento ambiental na Amazônia from hypermodule.biosensors import AquaticBioimpedance from edge_ai import HailoInference sensor = AquaticBioimpedance(calibration="river_water") if HailoInference("pollution_detector_v2")(sensor.read()): lora.send_alert(gps.get_coords(), encryption="AES-256")
🔬 8. Validação e Certificação
Teste Padrão Aplicado Resultado Obtido Integridade PCIe Gen4 IEC 61000-4-21 0 erros em 72h @ 60°C Autonomia Workload misto 14h @ 4W (50Wh) Resistência Mecânica MIL-STD-810H Método 514 Sem falhas @ 20G vibração
0 notes
amit-onetech · 4 years ago
Link
MicroAI ™ Atom A self-learning AI Platform that lives on the smart asset, providing: Predictive Maintenance features that ensures optimized asset health scores, uptime, and productivity. Security enablement to ward off would-be hackers and other cryptic technological attacks against prized assets and asset data.
 Features of Micro AI Atom Advanced AI/ML Platform for Smart Asset & Machine Optimization.
Optimize the performance and ROI of your mission-critical devices and machines with cutting-edge AI/ML technology.
Improve OEE: Use predictive analytics to improve Overall Equipment Effectiveness, reduce Maintenance Costs, and avoid Costly Downtimes.
Protect: Protect your critical assets from cyber-attack via continuous monitoring of all asset data and associated workflows---AI-driven approach to asset security.
Evolve: AI/ML technology that enables unsupervised learning---your smart assets continuously evolve through a process of self-learning and self-correction.
Optimize: MicroAI™ Atom is compact enough to run on any microcontroller units, bringing you advanced AI technologies to where you need it.
  MicroAI™ Atom brings a new approach to smart asset management. MicroAI™ Atom has new technology that delivers:
Flexibility & Scalability: Deploy on a single asset (MCU) or scale across an entire asset ecosystem.
Speed & Agility: Data processed locally—not in the cloud—provides rapid, real-time, data-sampling.
Security: Real-time visibility into asset health and self-learning that enables rapid threat-response.
Cost: Asset model training occurs directly on the asset MCU—no need for costly new hardware.
  Atom – Edge MicroAI ™
Atom can be deployed as a singular entity on a single device. Real-time, cost-effective, insights delivered in a tiny footprint.
0 notes
scalper · 5 years ago
Link
 Topaz LabsはWindowsおよびmacOSに対応したパソコンのアプリケーションを開発、ライセンス販売しているソフトウェアベンダーで、AIを利用した写真/動画編集ツールを多数提供している。その1つである「Gigapixel AI」は、解像度
0 notes
electronicsbuzz · 3 months ago
Text
0 notes
electronicsbuzz · 3 months ago
Text
0 notes
electronicsbuzz · 5 months ago
Text
0 notes
electronicsbuzz · 7 months ago
Text
0 notes
electronicsbuzz · 7 months ago
Text
0 notes
electronicsbuzz · 8 months ago
Text
0 notes