The development of emerging AIoT applications has driven many domestic chip manufacturers to develop AIoT system-on-chip (SoC) targeting specific application scenarios. In preparation for the upcoming 2021 International AIOT Ecological Development Conference, the analyst team of ASPENCORE’s “Electronic Engineering Magazine” has categorized these AIoT SoCs into 10 categories based on technology and application, selecting 3 representative domestic chip manufacturers and their representative products for each category. At the end of the article, we present these 30 companies’ AIoT SoC chips in the form of voting options; please choose your favorite manufacturer and chip model for each category.
AIoT: When Artificial Intelligence (AI) Meets the Internet of Things (IoT)
The Internet of Things (IoT) is a technology that helps us reconstruct an interconnected society and world, but Artificial Intelligence (AI) is the real driving force behind the IoT’s full potential.
IoT devices collect, transmit, and process information about interactions between people, between people and things, and between things through the internet. It is estimated that by 2025, there will be 42 billion IoT-connected devices globally. As the number of devices increases, the massive amount of data will also grow exponentially. This requires the intervention of AI, utilizing its learning capabilities for IoT connectivity.
The Internet of Things Requires Three Key Enabling Technologies:
Artificial Intelligence (AI) – Enables devices to learn, reason, and process information like humans in a programmable system;
5G Network — A fifth-generation mobile communication network with high speed and near-zero latency, suitable for real-time data processing;
Big Data —Massive data collected and processed from numerous IoT connection sources.
Currently, these interconnected devices mainly communicate through short-range and low-speed wireless connections such as Bluetooth, WiFi, and NB-IoT, but 5G networks will gradually penetrate the IoT, changing how we interact with devices at home and in the workplace.This constitutes what we call AIoT, so what are the main sub-application markets of AIoT?
AIoT will mainly impact four major application areas: wearable devices, smart homes, smart cities, and smart industries.
1. Wearable Devices
Wearable devices like smart bracelets and smartwatches can continuously monitor and track user preferences and habits, positively impacting both the medical field and fitness. According to Gartner’s data prediction, the global wearable device market is expected to exceed $87 billion by 2023.
2. Smart Homes
Smart homes add networking and intelligence to appliances, lighting, and electronic devices, making life more convenient and improving energy efficiency through seamless connectivity. The smart home market is expected to grow at a compound annual growth rate (CAGR) of 25% from 2020 to 2025, reaching $246 billion by 2025.
3. Smart Cities
Smart cities innovate and evolve, improving public safety, traffic, and energy efficiency through AI and IoT technology. The application of AI in urban traffic control has become clearer. In some of the world’s most congested cities, intelligent traffic management systems (ITMS) are used to make real-time dynamic decisions regarding traffic flow.
4. Smart Industry
From electronics manufacturing to mining, traditional industries are relying on digital transformation to improve efficiency and reduce human error. From real-time data analytics to supply chain sensors, smart devices will assist factories in preventive maintenance and real-time monitoring. Gartner predicts that by 2022, over 80% of industrial IoT projects will adopt AI. The fusion of AI and IoT will unleash greater potential, accelerating AIoT innovation and leading us into a future of interconnected everything.
The future of AIoT technology and applications will mainly manifest in the following aspects: edge computing, voice AI, visual AI. Current smart thermostats, smart appliances, and manually driven/assisted driving vehicles will evolve into home robots and autonomous vehicles; current smart speakers and voice-activated devices with LCD screens will upgrade to natural language processing and voice verification electronic payment methods;current facial recognition, object detection, and 4K resolution video will develop into edge video analysis and ultra-8K resolution display.
The integration of AIoT is becoming mainstream, and in the coming years, it will rapidly push the boundaries of massive data processing and machine learning.
Top 10 Categories of AIoT SoC Chips and Representative Manufacturers
The operation of AI algorithms, data processing, and application-driven/triggered decision-making of IoT devices require low power consumption, small size, and excellent performance AIoT system-on-chip (SoC). The development of emerging AIoT applications has also driven many domestic chip manufacturers to compete in developing AIoT SoCs targeting specific application scenarios. We have categorized these processor chips into 10 categories based on technology and application, selecting 3 representative domestic chip manufacturers and their representative products for AIoT device developers to reference. We also present these 30 companies’ AIoT SoC chips in the form of voting options at the end of the article; please choose your favorite manufacturer and chip model for each category.
TWS Bluetooth SoC
-
Hengxuan Technology BES2300
-
Jieli Technology AC693N
-
Zhongke Lanxun AB5377T
The Hengxuan Technology BES2300 is a smart Bluetooth audio chip that uses 28nm technology and BGA packaging. It integrates RF, PMU, CODEC, high-performance CPU, and embedded voice AI on a single chip, supporting smart voice and hybrid active noise cancellation. This fully integrated adaptive active noise cancellation solution supports dual-mode Bluetooth 5.0, LBRT low-frequency forwarding technology, and third-generation FWS fully wireless stereo technology, dual microphones, etc. Additionally, it supports external heart rate sensors, accelerometers, and eMMC flash memory, allowing external storage devices to play music.
The Jieli Technology AC693N Bluetooth series chip supports Bluetooth audio 5.0 + BLE 5.0, with unique features including: master-slave switching, intelligent charging case, ultra-small packaging, lossless music, ultra-low power consumption, MESH networking, convenient configuration tools, and high integration with minimal external components. Standard features include: dual-way calling, AI intelligence, Gensor tapping, analog digital silicon microphone, intelligent feedback, HIFI sound quality, OTA upgrades, built-in charging management, and constant current and voltage.
The Zhongke Lanxun AB5377T is a Bluetooth 5.0 TWS earphone master control chip with built-in Flash, featuring a 32-bit high-performance CPU with DSP capability, flexible GPIO pin configuration; compatible with Bluetooth 5.0 and BLE specifications (QDID: 115952); audio CODEC with 16-bit mono DAC and 16-bit mono ADC; packaged in SOP16.
WiFi/Bluetooth Dual-Mode SoC
-
Unisoc V5663
-
Broadcom BK7231U
-
Liangshengde W800
The Unisoc V5663 is an AIoT solution that supports Wi-Fi 5 + Bluetooth 5 + MCU, characterized by advanced technology, powerful computing power, safety and reliability, rich applications, and ease of development, widely applicable in smart homes, healthcare, children’s education, beauty and health, warehousing logistics, and many other application scenarios. Its feature parameters include: support for PSA-M security certification; uses RTOS to achieve dual-microphone VAD/voice wake-up interrupt audio processing; supports Wi-Fi 5 / Bluetooth 5.0 / highly integrated single-chip MCU; supports ARM CM33 Combo connection chip, using ARM Cortex M33 ARMv8-M architecture, supporting TrustZone; peripheral interfaces include USB2.0/3.0, eMMC, PWM, PDM, IIS, etc.; dual-band Wi-Fi 2.4/5G, MU-MIMO; supports Bluetooth AOA AOD + Wi-Fi RTT fusion positioning; BT5.0: dual-mode, long-distance, Mesh, AoD; supports 802.11ac 1X1 2X2, etc., offering a rich product lineup.
The Broadcom BK7231U Wi-Fi SoC chip complies with the 802.11b/g/n 1×1 protocol with an output power of 2.17dBm; supports 20/40 MHz bandwidth and STBC; supports Wi-Fi STA, AP, and Direct modes; supports Bluetooth 5.1 protocol with -90dB sensitivity and 20dBm output power; has an internal MCU; maximum frequency of 120MHz; 256Kbyte data RAM; built-in 2MB FLASH, supports transparent download; 6-channel 32-bit PWM; multiple program downloads with JTAG interface; full-speed USB host and device; 50MHz SDIO and SPI interfaces, supporting master-slave mode; supports two I2C interfaces; supports two high-speed UARTs; 6-channel 32-bit PWM; microphone signal amplification; built-in multi-channel ADC; supports 8-bit DVP image sensors; 32-byte eFUSE.
The Liangshengde W800 chip is a secure IoT Wi-Fi/Bluetooth dual-mode SoC chip. It supports 2.4G IEEE802.11b/g/n Wi-Fi communication protocols; supports BT/BLE dual-mode operation, supporting BT/BLE4.2 protocol. The chip integrates a 32-bit CPU processor, built-in UART, GPIO, SPI, I2C, I2S, 7816, and other digital interfaces; supports a TEE security engine, various hardware encryption and decryption algorithms, built-in DSP, floating-point operation unit, supports code security permission settings, built-in 2MB Flash memory, supports firmware encrypted storage, firmware signing, secure debugging, secure upgrades, and other security measures to ensure product security features. It is suitable for various IoT fields such as smart appliances, smart homes, smart toys, wireless audio and video, industrial control, and medical monitoring.
NB-IoT SoC
-
Yichin Communication EC616
-
Xinyu Information XY1100
-
Nolink Technology NK6010
The EC616 is a low-cost, low-power, high-integration NB-IoT SoC chip developed by Shanghai Yichin Communication, supporting 3GPP R13/R14 NB-IoT standards.The EC616 single chip integrates RF transceiver, PA, and power management, significantly reducing the number of peripheral devices.Its feature parameters include:Processor core Cortex-M3; frequency band 663MHz-2200MHz; supports 2.2~4.5V wide voltage input; PSM sleep state current as low as 0.8uA; IDLE power consumption (DRX 2.56s) 0.11mA; built-in PA output power 23dBm; operating temperature -40°C ~ +85°C.The EC616 is suitable for wireless meter reading, smoke detection, smart street lights, smart logistics, asset tracking, smart firefighting, smart parking, smart homes, wearable devices, and industrial and agricultural applications.
The XY1100 NB-IoT chip developed by Xinyu Information Technology is an SoC integrated with CMOS PA, which has now obtained chip certification from the China Mobile TSP platform. The XY1100 feature parameters are as follows: full-band support 690-960MHz, 1.71-2.2GHz; supports 3GPP standards R13 and R14; supports different power consumption modes: deep sleep, normal sleep, standby, low-power working mode; deep sleep state current <1 microamp, receiver state current about 20 milliamps; adopts SDR architecture, flexibly supports optimized effective physical layer receiver algorithms; integrates various peripheral interfaces, including UART, I2C, SPI, uSim, GPIO, etc.; operating temperature range -40℃~+85℃; packaged in QFN52 6mm*6mm.
The NK6010, the first chip of the NK6K series SoC developed by Nolink Technology, is a low-cost, low-power NB-IoT SoC designed for narrowband connections of ultra-large IoT devices (such as smart meters, wearable devices, asset trackers, and industrial sensors). The NK6010 SoC integrates RF front-end (RFFE), RF transmission/reception (RF TxRx), NB-IoT communication baseband, application processor (AP), and power management unit (PMU). The NK6010 supports all NB-IoT frequency bands, targeting major global operators, ensuring smooth and rapid certification of any NB-IoT commercial network device. In addition, the NK6010 SoC also integrates an ultra-low-power GNSS subsystem, supporting GPS / Beidou / Galileo / GLONASS global navigation systems, ensuring highly accurate device tracking and positioning globally.
RISC-V SoC
-
Espressif ESP32-C6
-
Bluesilicon BL602/BL604
-
Telink TLSR9xxx
The ESP32-C6 is Espressif’s first 32-bit RISC-V SoC integrated with Wi-Fi 6 + Bluetooth 5 (LE), featuring ultra-low power consumption and high cost-effectiveness, significantly improving the Wi-Fi transmission efficiency of IoT devices, providing secure and reliable connectivity. Currently, the ESP32-C6 has passed Wi-Fi Alliance certification. The ESP32-C6 features a RISC-V 32-bit single-core processor with a clock frequency of up to 160 MHz, built-in 400 KB SRAM, 384 KB ROM, and supports multiple external SPI, Dual SPI, Quad SPI, QPI flash. The ESP32-C6 has 22 programmable GPIO pins, supporting ADC, SPI, UART, I2C, I2S, RMT, TWAI, and PWM.
The Bluesilicon BL602/BL604 chip’s wireless subsystem includes 2.4G radio, Wi-Fi 802.11b/g/n, and Bluetooth LE 5.0 baseband/MAC design. Its microcontroller subsystem includes a low-power 32-bit RISC-V CPU, high-speed cache, and memory. The power management unit can control low-power modes. Its peripheral interfaces include SDIO, SPI, UART, I2C, IR remote, PWM, ADC, DAC, ACOMP, PIR, etc., supporting flexible GPIO configuration. The BL602 has a total of 16 GPIOs, while the BL604 has a total of 23 GPIOs.
The Telink TLSR 9 series integrates a 32-bit RISC-V MCU (Crisp D25 core), with a standard version running at a maximum speed of 96MHz, supporting 5-stage pipelines, with a computing power of 2.59 DMIPS/MHz, CoreMark score of 3.54/MHz, and also integrates DSP extended instructions and floating-point computation modules, facilitating the development of audio algorithms and sensor algorithms. This high-performance SoC chip series is mainly targeted at wearable devices and various IoT applications.
Smart Wearable SoC
-
Furukawa FR8000
-
Lingwei BX2418
-
HuaPu Microelectronics CMT4502
Furukawa MicroelectronicsFR8000 series low-power Bluetooth SOC mainly targets the wearable device market, supporting external PSRAM to run wearable GUI, supporting Free-RTOS and RT-Thread and persimmon GUI. Its feature parameters include: support for BLE5.1;
CortexM3 96MHz main frequency; built-in 256KB/512KB Flash to store user programs and data; built-in 150KB ROM to store BLE Profile & GATT, LM, LC; built-in 64KB SRAM to store running data; receiving sensitivity -98dbm, Tx Power 10dbm, External PA; built-in power management module PMU (including charger maximum 200mA, LDO can output 100mA); PWM*8, complementary output with hardware dead zone control; rich peripherals, supporting DMA/QSPI*2/SPI/UART/I2C/I2S/ADC*8, etc.
Lingwei BX2418low-power Bluetooth chip complies with BLE 5.0 standard, supports 1Mbps/2Mbps rate switching, and supports SIG Mesh, meeting BQB/SRRC/FCC/CE testing standards. Its built-in MCU has 4 buffers, SWD debugging interface, AHB/APB bus clock rate up to 96MHz. It has 14 GPIOs, 5 PWM outputs, and 2 10-bit ADCs.
HuaPu Microelectronics CMT4502is an ultra-low-power IoT Bluetooth wireless communication chip, equipped with a 32-bit ARM Cortex-M0 processor core, with 138K SRAM/Retention SRAM, featuring ultra-low power consumption, high performance, and wireless multi-mode capabilities, supporting security, applications, and wireless updates with BLE functionality. CMT4502 supports BLE5.0 high-speed data throughput, including BLE 2Mbps protocol and length extension functions; supports various features of Bluetooth Sig-Mesh protocol: Friend/LowPower/ProxyRelay nodes, capable of providing complete protocol support for BLE5.0.
Edge AI Processors
-
Rockchip RK3399 Pro
-
Horizon SUNRISE (Xuri) 3
-
Hangzhou Guoxin GX8010
The Rockchip RK3399 Pro is a high-performance processor with a dual Cortex-A72 + quad Cortex-A53 big.LITTLE CPU architecture, with a maximum frequency of 1.8GHz, supporting 4K 60fps VP9 and 4K 10bits H265/H264 video decoding, with rich peripheral interfaces. This edge AI processor has a built-in neural network processor (NPU) with a computing power of up to 3.0TOPs, supporting PCI-e/USB3.0/RGMII, capable of performing structured recognition and analysis of video from multiple cameras and network cameras. RK3399Pro integrates Rockchip’s NPU developed in the fields of machine vision, speech processing, and deep learning, achieving impressive performance with typical deep neural network models like Inception V3, ResNet34, VGG16, significantly enhancing performance. RK3399Pro is suitable for AI applications in smart driving, image recognition, security monitoring, drones, voice recognition, etc.
The SUNRISE (Xuri) 3 is a next-generation low-power, high-performance AI processor launched by Horizon for AIoT scenarios: it uses Horizon’s Bernoulli 2.0 architecture AI engine (BPU), providing an equivalent computing power of 5TOPS. The new BPU architecture greatly enhances support for advanced CNN network architectures and significantly reduces the bandwidth consumption of AI computations on DDR. The Xuri 3 series includes two chips, X3M and X3E; X3M mainly targets the 8M smart front-view market and edge computing, providing 5TOPS AI equivalent computing power; X3E mainly targets the 5M smart front-view market, providing 3TOPS AI equivalent computing power. Coupled with Horizon’s “Tiangong Kaiwu” AI development platform, it greatly simplifies the algorithm development and deployment process, thereby reducing the commercialization cost of AI products.
The Hangzhou Guoxin GX8010 chip is an embedded SoC chip designed specifically for artificial intelligence and IoT applications, uniquely designed with a multi-core heterogeneous architecture, integrating proprietary NPU neural network processors for voice signal processing, DSP processors, and other modules, with image processing capabilities, featuring high intelligence, low power consumption, and full integration. Its core goal is to make various products smart in daily life, suitable for edge AI applications such as smart speakers, children’s robots, and story machines.
Smart Voice SoC
-
Allwinner Technology R329
-
Zhichun Technology WTM2101
-
Shencong Semiconductor TH2608
The Allwinner Technology high-integration AI voice chip R329 is an intelligent voice processor developed by the Arm China team using AIPU technology, with hardware configuration: dual-core A53 1.5GHz, dual-core HIFI4 400MHz (2MB SRAM), Zhouyi 0.256T AIPU; integrating multiple audio ADCs and DACs; built-in DDR design. The R329 can provide high-integration application solutions for smart speakers and smart homes. The R329 integrates Arm China’s “Zhouyi” AIPU, allowing for complete programmability of artificial intelligence applications and good energy efficiency, providing ample AI computing power and low-power applications for voice AI applications. In terms of interfaces, the R329 integrates multiple audio ADCs and DACs, providing higher quality audio output solutions and multi-microphone voice pickup solutions for smart speakers.
The WTM2101 from Zhichun Technology is the first SoC chip based on Nor Flash technology that integrates storage and computation, with ultra-low power consumption making it particularly suitable for low-power wake-up, voice recognition, noise reduction, and other scenarios, such as TWS earphones, watches, and other smart wearable devices. The WTM2101 measures 2.9*2.6mm, with a peak computing power of 50Gops, an energy efficiency ratio of 15Tops/W, and a maximum storage capacity of 1.8M for neural networks. In addition to its storage-computation integration function, this chip also includes a RISC-V core, audio ADC, power management, and rich interfaces. In smart audio applications, the WTM2101 features wake-up functionality, with Zhichun’s VAD outperforming existing market solutions. Additionally, in voiceprint recognition, this chip can implement voiceprint recognition functionality in wearable devices such as watches and earphones.
The TH2608 chip developed by Shencong Semiconductor has achieved a technological leap from co-processor to main control chip, embedding a master control CPU based on the Cortex-M ARM v8 instruction set, combined with internal SRAM and external XIP Flash resources, as well as peripheral interface resources, providing customers with intelligent voice solutions in various application scenarios. This chip also integrates a flexible and configurable neural network accelerator unit NPU, significantly improving energy efficiency. By deploying various general networks such as DNN/TDNN/FSMN/CNN, the TH2608 can fully support data formats INT8/4/1. Additionally, the TH2608 offers various flexible SIP solutions for multi-chip integration, supporting the encapsulation of different types of PSRAM and Flash, and can integrate wireless communication modules such as Wi-Fi and Bluetooth. With the built-in power management module (PMU), it can strongly support the most streamlined BOM solutions for customers. This chip enables cloud + chip full-link voice interaction, supporting full-duplex, whole-house intelligence, and multi-language far-field calls.
Smart Visual Processors
-
Yizhi Electronics SH516
-
Aixin Technology AX630A
-
Lingsheng Technology JA310
The Yizhi Electronics SH516 chip is an ultra-low-power smart visual chip that uses TSMC 22nm technology, integrating a second-generation self-developed NPU, supporting audio and video collaborative analysis and processing, aimed at smart education and smart home products, efficiently supporting object classification recognition, text recognition, face detection recognition, skeleton analysis, gesture expression behavior analysis, and other intelligent visual application scenarios. The performance and technical specifications of this chip are as follows: single-core RISC CPU, 0.6T computing power, built-in ISP, display processing, supporting image encoding and decoding; rich peripherals including USB, SDIO, UART, I2C, PWM, ADC, audio, video input and output; built-in 128MB DDR3L, supporting SPI NAND, NOR, EMMC, TF Card as boot media; supports Linux/RTOS systems; integrates rich basic algorithms for people, vehicles, and objects; advanced manufacturing process, high integration SOC architecture, making it an ultra-low-power visual analysis platform.
The Aixin Technology AX630A artificial intelligence visual processor is a high-end edge-side, end-side AI chip, with a power consumption of around 3W, computing power reaching 57.6TOPS @800MHz 2w4f, equivalent computing power of 28.8TOPS @800MHz INT4, processing thousands of images per second. This chip supports a maximum resolution of 4K@60fps, supports 20 channels of 1080p decoding or 8 channels of 1080p encoding, and various detection and recognition algorithms, with high resolution, excellent picture quality, strong encoding and decoding, high-definition display, ultra-low power consumption, and strong computing power advantages.The AX630A uses TSMC 12nm technology and is suitable for various IoT fields such as smart wearables, security monitoring, intelligent transportation, smart manufacturing, smart cities, smart retail, smart communities, smart homes, and IoT devices.
The Lingsheng Technology JA310 is an AIoT SoC aimed at visual computing, suitable for smart security, machine vision (robotics/industrial cameras/ADAS assisted driving), facial recognition, video live streaming, and smart displays (smart interactive terminals). JA310 feature parameters are as follows: CPU: Arm Cortex-A55 @1.5GHz x4; ISP: single-channel 4K@30fps; dual-channel 2K+2K@30fps; video encoding: 4K/1080p/720p @30fps; video decoding: 4K @30fps; display: 1080p @60fps; NPU: 2 TOPS (1152 INT8 MAC); memory: supports DDR3/L/4/LP4 (32bit); flash: supports SPI NOR/NAND/eMMC; I/O: Ethernet (RGMII&RMII/MII), I2Sx5, USB2.0x2, MIPI/LVDS/PWM/GPIO/UART/I2C/SPI/SDIO; wireless peripherals: compatible with mainstream Wi-Fi/4G/5G modules/other peripherals for BSP-level integration; manufacturing process: Samsung 11LPP; packaging: 15x15mm 493 Ball.
Smart Sensor SoC
-
Airspace (Shanghai) Intelligent Technology AT5820
-
Zhuhai Plin Chip SPT50
-
Xinhai Technology CSA37F61
The Airspace (Shanghai) Intelligent Technology AT5820 is a presence-sensing SoC that can detect breathing using 5.8GHz microwave radar, integrating a high-performance radar transceiver and 32-bit MCU on a single chip. The MCU not only processes internal radar signals but can also replace part of the upper computer system with Open CPU. Thanks to the chip’s radar sensing performance and signal processing capabilities, the AT5820 can detect human movement while also detecting micro-movements and even breathing signals, achieving true human presence sensing. Other features of this chip include support for hanging at a height of 15m, sensing radius >6m; integrated PLL phase-locked loop, automatically locking frequency to 5.8GHz ±75MHz; complies with FCC/CE/SRRC and other wireless certification standards; industrial-grade features, supporting high and low-temperature (-40 degrees to 85 degrees) reliability testing.
The Zhuhai Plin Chip SPT50 series SoC is designed specifically for TWS earphones, featuring in-ear detection, pressure detection, touch detection, and swipe detection, with ultra-high sensitivity and ultra-low power consumption. The high-sensitivity capacitive in-ear detection can achieve light-sensing performance, while the pressure detection function can realize single-click, double-click, multi-click, and other functions. The chip feature parameters are as follows: supports 1 in-ear detection channel, 2 pressure (or touch) detection channels; external main control power supply (typical VCC=3.3V, recommended IO power supply); built-in LDO voltage regulation circuit; can configure detection channel parameters through a single bus interface; supports output formats: GPIO direct push, serial output; low power consumption: 3.5uA in non-in-ear state, 4uA ~ 12uA in in-ear state; packaging: DFN-8, SOT23-6.
The Xinhai Technology pressure touch chip CSA37F61 adopts resistive micro-pressure strain technology, supporting border touch interactions for products like mobile phones/headphones, such as sliding, gestures, pressing, etc., providing users with a more convenient operating experience. This chip has 8 differential input channels, 1 Force Touch AFE, supports I2C/UART communication, and can reconstruct the ForceTouch algorithm. Its pressure touch parameters are as follows: panel thickness support: <1.0mm; trigger force ≤100g; operating environment: -25℃~85℃; module total thickness: ≤0.18mm.
Smart Home SoC
-
Shiqing Intelligent AT5055
-
Juchip ATS3607D
-
Fengjia Micro PHY6202
The Shiqing Technology AT5055 is an end-side smart visual chip based on RISC-V architecture, equipped with a self-developed RV32 IMCF TM800 processor @800MHz; storage 32-bit DDR2/DDR3, supporting secure boot; with various peripherals. In terms of AI capabilities, it is equipped with Timesformer Bumblebee 600V processor, completing whole-network neural network processing, including pre-processing and post-processing, without additional CPU assistance, reducing scheduling and communication overhead. The AT5055 has u-level standby power consumption, coupled with RTOS’s quick-start system, supporting various battery-powered devices; it supports multiple wake-up modes including human voice (VAD) wake-up, RTC wake-up, GPIO wake-up, etc., supporting multi-level wake-up to adapt to different scene wake-up requirements. Its packaging specification is QFN128, 12mm x 12mm, encapsulated DDR.
The Juchip Technology ATS3607D is a smart voice interaction SoC designed for emerging applications such as smart homes, smart offices, and conference speakers. This chip adopts a high-performance heterogeneous dual-core architecture, integrating the VAD module, with built-in large-capacity DDR. In terms of design solutions, it is compatible with many mainstream voice pre-processing algorithms and post-processing sound effect algorithms. The operating system can use RTOS, enabling smart voice interaction products with fast startup, low standby power consumption, and high voice interaction performance. Its feature parameters are as follows: CPU Cortex-A5 840MHz; DSP CEVA X2 350MHz; Embed Memory 656KB SRAM +32MB DDR; ADC 6*ADC; MIC Array 4 – 8; LCD CPU; Audio Decode MP3, FLAC, APE, OGG, AAC, Opus; PEQ&DRC 14 Bands & 2 DRC+Limiter; packaging QFN68 (8*8mm).
The Fengjia Micro PHY6202 Bluetooth chip is a high-performance, highly flexible ultra-low-power chip. It supports multiple communication protocols such as low-power Bluetooth, Bluetooth 5.0, Bluetooth MESH, ZigBee, Google Thread, Mist, and is aimed at smart terminal peripheral devices, smart voice, smart home, smart manufacturing, smart transportation, and other applications. The PHY6202 uses a 32-bit ARM Cortex-M0 core, with 128K ROM, 138K SRAM, and 256/512K Flash. Its built-in 2.4GHz transceiver supports low-power Bluetooth (BLE), ANT, and proprietary 2.4GHz protocol stack. The PHY6202 can operate as an independent system or as a slave device to the host MCU, providing Bluetooth functionality through I2C/UART interfaces.
Voting: Please choose your favorite manufacturer and chip model from each of the 10 categories below